L‘mERtQ

2.007

This is to certify that the
dissertation entitled

THE EFFECTS OF LOCAL RADIO STATION GROUP
OWNERSHIP ON STATION PERFORMANCE

presented by
Heather Elizabeth Polinsky

has been accepted towards fulfillment
of the requirements for the

PhD degree in Mass Media

 

 

 

fli/ 549

MatI or Professor’ 3 Signature
2T mg 51/ 962:5

Date

 

MSU is an Affinnative ActiorVEqual Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DaIeDue.indd.p.1

THE EFFECTS OF LOCAL RADIO STATION GROUP OWNERSHIP
ON STATION PERFORMANCE
By

Heather Elizabeth Polinsky

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Mass Media Program

ABSTRACT
THE EFFECTS OF LOCAL MARKET RADIO STATION OWNERSHIP
CONCENTRATION ON STATION PERFORMANCE
By
Heather Elizabeth Polinsky

The passage of the Telecommunication Act of 1996 eliminated national radio
station ownership limits and considerably relaxed local station ownership caps compared
to previous ownership restrictions. After the passage of this act a considerable number of
radio station owners immediately moved to consolidate stations into local and national
station groups. Media policymakers have been concerned that increases in media
ownership consolidation and concentration not decrease market format diversity.
However, past researchers of format diversity in the radio industry have not directly
considered audience measurements in their investigations of the effect of station
ownership concentration. This study uses station audience share as a measure of station
performance to examine the effects of increases in ownership concentration between
1995 and 2000.

The results of this study find that the number of programming formats
broadcasting in a market (market formats) is positively related to the number of radio
stations serving the market (market stations) and that the percentage of market population
listening to radio is positively related to the number of market stations. This study
provides support for Rogers & Woodbury’s (1996) argument that increases in radio
ownership concentration will not affect the number of market formats or the size of a

market’s radio audience. A format’s market audience share appears to be responsive to

the number of market stations broadcasting the format. These results suggest that
increasing the number of stations expands the market radio more significantly than
increasing the number of formats.

This study finds that all else equal, as local groups increased their numbers of
stations froml995 through 2000, stations in the groups become increasingly more
efficient at attracting listeners than independent stations. These finding suggest a public
interest benefit to station consolidation. Finally, this study finds that the length of time a
station broadcasts in a market and the length of time a station programs the same format
are statistically significant and positive predictors of station audience share, but the
length of time a station is under the same ownership is not significantly related to station

audience share.

TABLE OF CONTENTS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LIST OF TABLES ....................................... V

LIST OF FIGURES _ _ _ viii

INTRODUCTIONWW _______ 1

CHAPTER ONE

REVIEW OF LITERATURE 6
Radio Format Diversity and Local Market Structures 6
History of Radio Ownership Regulation 10
Effects of Ownership Concentration on Format Diversity 13
Audience Satisfaction in the Radio Market 16
Effects of Ownership Concentration on Station Performance 21

CHAPTER TWO

METHODOLOGY 26
Ownership Concentration Effects on Market Formats, Stations and Audiencew26
Group-Ownership Effects on Station Performance 32
Format, Broadcast and Ownership Consistency Effects on Station
Performance 42

CHAPTER THREE

DATA _____________________ 47

CHAPTER FOUR

RESULTS AND DISCUSSION 53
Ownership Concentration Effects on Market Formats, Stations and Audiencew53
Group-Ownership Effects on Station Performance 67
Format, Broadcast and Ownership Consistency Effects on Station
Performance __83

CHAPTER FIVE

CONCLUSIONS ........................................................... 86

REFERENCES___ 92

 

 

iv

LIST OF TABLES

TABLE 1

Local Ownership Caps Set by the Telecommunications Act of 1996 ,,,,,,,,,,,,,,,,,,,,,,,,,

TABLE 2
Breakdown of Station Data by Year and Frequency Band

 

TABLE 3
List of Markets Used to Estimate Equation (13) ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

 

TABLE 4
Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the

percentage of market radio listeners in 1995_

TABLE 5
Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the

percentage of market radio listeners in 1996

 

TABLE 6
Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the

 

percentage of market radio listeners in 1997

TABLE 7
Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the

percentage of market radio listeners in 1998

 

TABLE 8
Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the

percentage of market radio listeners in 1999

 

TABLE 9

Two-Stage Least Squares Regression Results for Equations (3) and (4)
estimating the factors affecting the number of market formats and the
percentage of market radio listeners in 2000”” _______

12

49

52

59

59

60

61

62

..63

TABLE 10

Partial Regression Results for the Within Format Effects Examined with

Equations (5) presenting the significance of the number of station broadcasting

a format on the market format audience for each sample year 66

TABLE 11

Regression Results for Equation (6) Estimating the Effects of Group Size

(linear and squared number of group stations variables) and Other Variables

on Station Audience Share in 1995 and 1996 72

TABLE 12

Regression Results for Equation (6) Estimating the Effects of Group Size

(linear and squared number of group stations variables) and Other Variables

on Station Audience Share in 1997 and 1998 73

TABLE 13

Regression Results for Equation (6) Estimating the Effects of Group Size

(linear and squared number of group stations variables) and Other Variables

on Station Audience Share in 1999 and 2000 _74

TABLE 14

Regression Results for Equation (7) Estimating the Effects of Group Size

(linear, squared and cubed number of group stations variables) and Other

Variables on Station Audience Share in 1995 and 1996 _ ".76

TABLE 15

Regression Results for Equation ( 7) Estimating the Effects of Group Size

(linear, squared and cubed number of group stations variables) and Other

Variables on Station Audience Share in 1997 and 1998 =77

TABLE 16

Regression Results for Equation ( 7) Estimating the Effects of Group Size

(linear, squared and cubed number of group stations variables) and Other

Variables on Station Audience Share in 1999 and 2000 _ 78

TABLE 17
Regression Results for Equation (12) Estimating the Effects of Group Size

(spline variables) and Other Variables on Station Audience Share in 1995
and 1996 80

TABLE 18
Regression Results for Equation (12) Estimating the Effects of Group Size

(Spline variables) and Other Variables on Station Audience Share in 1997
and 1998 81

vi

TABLE 19

Regression Results for Equation (12) Estimating the Effects of Group Size

(spline variables) and Other Variables on Station Audience Share in 1999

and 2000 .................. 82
TABLE 20

Regression Results for Equation (13) examining the effects of the length of time

a station has broadcast in a market, has broadcast the same format and has been
owned by the current owner on station audience share in 2000_ 85

vii

LIST OF FIGURES

FIGURE 1
Line Plot Representing the Relationship between Group Size (linear and squared
number of group stations variables) and Station Share from Equation (6) Resultsm71

FIGURE 2

Line Plot Representing the Relationship between Group Size (linear, squared,

and cubed number of group stations variables) and Station Share from

Equation ( 7) Results ......................... ........................... 75
FIGURE 3

Line Plot Representing the Relationship between Group Size (spline variables)

and Station Share from Equation (12) Results 79

viii

INTRODUCTION

The level of ownership concentration in the broadcast industry has long been a
concern for policymakers and station owners. Under the authority of Congress in 1927,
the Federal Radio Commission was put in charge of broadcast regulation to preserve the
public interest, and in 1934 the agency was expanded to the Federal Communications
Commission (FCC). The FCC has historically operated under the assumption that local
broadcast audiences are best served by access to the largest possible number of different
owners supportable by available spectrum because diversity of ownership is assumed to
translate to diversity of broadcast programming types or formats. During the first six
decades of the radio industry, the FCC administered broadcast regulations under a public
trustee model. This policy approach assumes that enforcement of radio station ownership
limits will result in greater format diversity (Berry & Waldfogel, 1999; Glasser, 1984;
Rogers & Woodbury, 1996). The FCC uses format diversity as one measure of the public
interest. The FCC imposed limits on local and national station ownership to allow more
owners to have access to the scarce broadcast Spectrum resource and to restrict the market
power and influence held by any one station owner.

In the 19803, the FCC and Congress began to shift from the public trustee model
of broadcast regulation to a marketplace model and the perceived need for stringent
national and local market ownership limits began to wane. Under the marketplace model,
policymakers trust market forces to provide the optimum ownership structure to serve the
market audience with suitable format diversity. This logic led to the most liberal

relaxation of station ownership limits to date through the passage of the

Telecommunications Act of 1996. The Telecommunications Act of 1996 eliminated
national radio station ownership limits and considerably relaxed local ownership caps.
After the passage of this act a considerable number of radio station owners immediately
moved to consolidate stations into local and national station groups. The number of
national commercial radio station owners decreased thirty-four percent from 5,133
owners in March, 1996 to 3,408 owners in March, 2002, yet the number of commercial
radio stations saw a modest 5.4 percent increase during the same time period (Williams &
Roberts, 2002).

The relaxation of local market ownership caps allows station ownership
consolidation that some policymakers and industry representatives argue will allow local
station groups to operate with greater efficiency through the sharing of station resources.
(Albarran & Pitts, 2001; Berry & Waldfogel, 2001; Chambers, 2001; Ekelund, Ford, &
Koutsky, 2000; Wirth, 2001 ). Station consolidation has led to increases in ownership
concentration. Policymakers are concerned that increases in ownership concentration not
decrease format diversity and are hopeful that increases in ownership concentration will
increase in format diversity. Most past research finds as ownership concentration
increases, market format diversity has remained statistically unaffected. (Berry &
Waldfogel, 2001; Drushel, 1998; FCC, 2000, 2001; Rogers & Woodbury, 1996;
Williams, Brown, & Alexander, 2002; Williams & Roberts, 2002). But, some researchers
argue that there is statistical evidence of a weak effect indicating format diversity is
increasing in smaller markets where ownership concentration is increasing (FCC, 2001;

Williams & Roberts, 2002).

Empirical evidence supports the claim that increases in ownership concentration
have not been detrimental to market format diversity. However, the effects of ownership
concentration on station performance has received comparatively little attention (Drushel,
1998; Ekelund et al., 2000; Rogers & Woodbury, 1996). If there are efficiencies to
ownership consolidation in the radio industry, then station group-ownership may result in
increased station performance. These group efficiencies should result from sharing
station personnel, services equipment and office space. Station groups that can share
resources should be able to streamline costs and use these cost savings to develop higher
quality programming which in turn should attract a larger potential audience (Albarran &
Pitts, 2001; Compaine, 2000; Gomery, 2000; Wirth, 2001). This study uses audience size
estimated by a commercial audience measurement service as a measure of station
performance. Examining station audience provides further insight into the possible
effects of increases in ownership concentration on a station’s capacity to better serve the
market audience.

Napoli (1997) suggests that most program diversity research is limited by
considering ‘diversity of exposure’ to different program types without considering the
‘diversity of consumption’ of different program types. The diversity of programming
available to audiences (format diversity) is what Napoli (1997) terms ‘diversity of
exposure,’ and the mix of audience members that decides to watch or listen to particular
program types is termed ‘diversity of consumption.’ Because media audiences can
consume only those types of programs that are available, measures of these two concepts
are interdependent. Napoli (1997) finds a high correlation between the percentages of

network television program types available and the percentages Of audience that chooses

to watch the different program types. Napoli (1997) concludes either network executives
are highly attuned to audience preferences or the network audience watches television
according to the programming available to them. Napoli (1997) suspects both processes
are occurring in the network television industry.

AS suggested by Napoli (1997), simply counting the radio formats available in a
market is only a measure of the formats available to the market audience. This measure of
format diversity does not consider market audience consumption of the available formats.
A listener chooses which of the available stations to listen to by deciding which of the
stations will best satisfy his or her format preferences. Counting the formats available to
listeners assumes that listeners are aware of all the market format and station choices
available to them. Listeners may not immediately be aware Of all of their station or
format choices; it may take time for listeners to find the station that best serves their
preferences. Furthermore, listeners may not be responsive to more format choices.
Having more format choices may not sway listeners from their primary format
preference. Considering station audience as a measure of station performance provides
further insight into the effects of changes in format availability and changes in ownership
concentration on audience consumption of radio.

Chapter One reviews the prior research pertinent to this study. Chapter Two
presents and describes the methods used to study three effects of local ownership
concentration on station and market performance. The first section of Chapter Two
presents the analytical framework for a study of the effects of local market ownership
concentration on the relationships between the number of formats in a market (market

formats), the number of stations in a market (market stations), and the size of the radio

audience in a market (market audience) . This investigation is a replication and extension
of a study by Rogers & Woodbury (1996) who argued that increases in ownership
concentration would not affect the number of market formats or the percentage of a
marke’st population listening to the radio. They also examined the responsiveness of the
audience for a format to changes in the number of other market formats available to
listeners and number of stations broadcasting the different formats. They concluded that
changes in the number of other market formats available to listeners and number of
stations broadcasting the different formats had little affect on the audience choosing to
listen to a particular format. The next section of Chapter Two describes the analytical
framework employed for an investigation of the effects of local station group size on
station performance. Local group size is measured by the number of stations in the group
and is used to examine whether adding stations to a group improves station performance
presumably through group efficiencies. The third section of Chapter Two presents the
analytical approach employed to investigate the effects of the length of time a station has
been broadcasting a signal in a market, the length of time a station has broadcast its
current format, and the length of time a station has been under its current owner on
station performance.

Chapter Three describes the data collected for these investigations. The results for
these investigations are presented and discussed in Chapter Four and Chapter Five

presents the conclusions from this study.

CHAPTER ONE

REVIEW OF LITERATURE

Radio Format Diversity and Local Market Structures

Broadcast policymakers have historically believed that greater diversity of
ownership provides a greater amount of format diversity; therefore broadcast ownership
policies have promoted competitive media market structures (Albarran & Pitts, 2001;
Bates, 1993; Drushel, 1998; Ekelund et al., 2000; Gomery, 2000; Kattenmaker & Powe,
1994; Wirth, 2001). Theoretical and empirical research has scrutinized the effectiveness
of competitive media markets at providing optimum format diversity to serve market
audiences. Steiner (1952) suggests that monopoly station ownership could provide
greater market format diversity, because a multi-station monopolist will not cannibalize
the audience of one station by programming the same format on another one of its
stations. Thus the monopolist, who owns a local group of stations, will program a
different format on each of its market stations resulting in greater format diversity for a
given number of stations (Steiner, 1952).

The amount of format diversity produced by a monopolist compared to
competitive ownership with a constrained number of market stations is dependent on the
distribution of audience format preferences. When market audience preferences are
highly skewed toward one particular format, multiple owners in the competitive market
will program this popular format. Competitive station owners maximize profits by
attempting to capture the same listener preferences from a larger, majority taste audience

instead of differentiating their formats to capture minority listener preferences that have a

smaller potential audience. Market format diversity decreases because multiple
competitive radio stations are programming the same format. In this instance of highly
skewed audience format preferences, a monopoly will produce a higher degree of format
diversity with fewer stations than a competitive market structure.

On the other hand, when market listener preferences are more evenly distributed
across multiple formats, the format diversity produced by competitive ownership should
be equal to the format diversity produced by monopoly ownership with the same number
Of stations (Steiner, 1952). Because the market’s audience is not skewed in favor of any
one format, competitive station owners will attract similarly sized audiences with
different formats. There is no benefit to multiple stations broadcasting one particular
format until there are a large number of stations in the market. Competitive stations
owners will not duplicate a format broadcast by another market station until doing so will
attract a larger potential audience than broadcasting a unique format.

It is important to note that Steiner’s program choice model demonstrates that
monopoly provides equal or greater format diversity compared to competitive ownership
with the same number of stations in the market, for most audience preference
distributions. But, Steiner (1952) assumes that listeners will only consume one preferred
program type and listeners will choose not to use the medium if their favorite program
type is not available. This assumption is relaxed by Beebe (1977) who adds secondary
audience preferences to the Steiner program choice model. If listeners will listen to other
formats when their favorite format is not available, then Beebe’s program choice model
illustrates that in most cases of audience preference distributions, competitive ownership

will provide as much if not more format diversity than monopoly ownership. The

monopolist will only program as many formats as it takes to get the vast majority of radio
listeners in the market to tune into a station. If most listeners in a market will listen to one
popular format when no others are available, then the monopoly owner will maximize
profits by only broadcasting this popular format on one station in the market while
leaving its remaining stations dark. Yet, with a competitive market in this situation, many
stations in the market may duplicate this popular format because everyone in the market
finds some appeal to this format. However, competitive owners in this situation are more
likely to program other formats to appeal to the first choice preference of listeners in the
market. Under Beebe’s program choice model, competitive ownership creates more
format diversity than the monopolist in most cases where secondary format preferences
are considered.

Spence & Owen (1977) consider how cross-format variation in program costs and
differences in advertiser valuations of format audiences affect a market’s program
offerings. Many formats that ought to be broadcast in a market on welfare grounds
because they generate positive surplus are not broadcast because they do not produce
enough advertising revenue to cover costs. Formats that are more expensive to produce
are less likely to be programmed, regardless of market structure. Furthermore if the
audience attracted by certain formats is not highly valued by advertisers, these formats
will not generate sufficient revenues to cover costs and will not be programmed,
regardless of market structure. The effects of market format losses due to high program
costs or low advertiser valuation of format audiences may create social welfare losses
because these audiences are left under served by radio stations. These social welfare

losses are greater in media markets that are advertiser supported than media markets that

have some degree of consumer pay support (Spence & Owen, 1977). Media that depend
on advertiser financial support will provide formats that best meet the advertiser’s
interests, not necessarily those that best address audience format preferences.

Many of the articles reviewed for this study are inspired by program choice
models of the type developed by Steiner (1952) and Beebe (1977) (e. g. Berry &
Waldfogel, 2001; Chambers, 2001; Williams et al., 2002; Williams & Roberts, 2002;
Wirth, 2001). Like Steiner (1952) and Beebe (1977), empirical research on the radio
industry has assumed (implicitly or explicitly) constant station program costs and
constant advertising prices.

Many program choice models compare competitive and monOpoly market
structures. Program choice theory finds advantages and disadvantages to both monopoly
and competitive market structures for providing greater format diversity depending on
market conditions (Beebe, 1977; Owen & Wildman, 1992; Spence & Owen, 1977;
Steiner, 1952). Proponents of the public trustee model for broadcast regulation have
favored ownership policies that have maintained highly competitive local broadcast
markets for most of the twentieth century believing this would produce the greatest
format diversity. Proponents of the marketplace model of broadcast regulation point to
program choice theory to question the effectiveness of a competitive market structure in
providing greater format diversity than more concentrated markets (Aufderheide, 1999;
FCC, 2000, 2001; Napoli, 2003). A monopoly market structure in the radio industry is
legally prohibited by antitrust law and FCC regulation. However, policymakers have
allowed station ownership consolidation to increase in radio markets because theory and

empirical evidence does not support a claim that more concentrated markets result in less

format diversity (Drushel, 1998; Ekelund et al., 2000; FCC, 2000, 2001; Rogers &

Woodbury, 1996; Williams et al., 2002; Williams & Roberts, 2002).

History of Radio Ownership Regulation

Local ownership caps were placed at one AM station and one FM station since the
Communication Act of 19341 and national radio station ownership limits were formally
placed at seven AM stations and seven FM stations (known as the Rule of Seven) in
1953.2 Station ownership was limited in the early days of radio because of radio
spectrum scarcity issues. More people wanted to own stations than the allotted spectrum
would allow. Later, arguments for station ownership limits shifted to preserving a
marketplace of ideas and preserving a competitive market structure (Aufderheide, 1999;
Compaine, 2000; Gomery, 2000; Kattenmaker & Powe, 1994). The FCC shifted from the
public trustee model of broadcast regulation to the marketplace model in the 19805 and
began deregulating the broadcast industry. During this period of deregulation, the FCC
relaxed national station ownership limits in 1984 from the Rule of Seven to the Rule of
Twelve. Under the Rule of Twelve a station owner could own twelve AM stations and
twelve FM stations nationally, but could still only own one AM station and one FM
station in a single market (Drushel, 1998).

In the late 19803 and early 19905, many industry observers described the radio

broadcast industry as in financial trouble and debate over relaxing both local and national

 

' FM radio technology was introduced to the commercial radio market in 1933, before this local ownership
caps were one AM station per owner (Orlik, 1997).

2 The most stations a national network at that time owned were seven AM stations. The national limit was
put in place to prevent the networks from increasing their station ownership (Drushel, 1998). Both the Rule
of Seven and subsequent Rule of Twelve also applied to the radio and television station ownership. Under
the Rule of Seven, one owner could own one AM station, one FM station and one TV station in a local
market and no more than seven AM, seven FM and seven TV stations nationally.

10

ownership limits intensified. National and local ownership limits were seen as
anticompetitive because they worked against natural market forces that would generate
economic efficiencies and increase consumer welfare (Albarran & Pitts, 2001; Bates,
1993; Drushel, 1998; Ekelund et al., 2000; Gomery, 2000; Kattenmaker & Powe, 1994;
Wirth, 2001). An FCC study observed that between 1989 and 1990 AM station profits
decreased by fifty percent and FM station profits by thirty-three percent (as reported in
Drushel, 1998). In 1992, under pressure from radio broadcasters, the FCC raised national
ownership limits to eighteen AM stations and eighteen FM stations and relaxed local
ownership caps to four stations (with no more than two stations in one frequency band) in
markets with fifteen or more stations and to three stations (with no more than two stations
in one frequency band) in markets with less than fifteen stations. In 1994, the FCC
slightly increased national ownership limits to twenty AM and twenty FM (Drushel,
1998; Ekelund et al., 2000; Wirth, 2001).

According to a paper presented at the 1997 Broadcast Education Association
Convention, “these policy changes unequivocally initiated a radio industry trend, with
more than fifty percent of commercial radio stations engaging in some form of
consolidation between 1992 and 1996” (Hagin, 1997 as quoted in Wirth, 2001, p.252).
The logic for increasing station limits outside local markets was to increase station
resource efficiency, so station group-owners could generate better quality programming
in multiple markets. Stricter ownership caps were preserved inside local radio markets to
maintain local competition and diversity of ownership voices which in turn was thought
to increase format diversity. The 1992 and 1994 increases in local station ownership caps

did not end the debate over further relaxation of local ownership caps.

11

Proponents Of the marketplace model of broadcast regulation accomplished a
policy victory with the passage of the Telecommunications Act of 1996. National station
ownership limits were eliminated and local market ownership caps were considerably
relaxed and set by the FCC according to the number of market stations:

Table 1

Local Ownership Caps Set by the Telecommunications Act of 1996

 

 

 

Commercial market stations Station ownership cap

45 or more stations 8 stations with no more than 5 in one band
30 to 44 stations 7 stations with no more than 4 in one band
15 to 29 stations 6 stations with no more than 4 in one band
Fewer than 14 5 stations with no more than 3 in one band

and not more than 50% of stations in the market

 

The Telecommunications Act Of 1996 gives the FCC the authority to set ownership caps
as it sees fit to serve the public interest but also mandates that the FCC periodically
investigate whether changes in ownership concentration are detrimental to the public
interest. The parameters Of the public interest are also determined by the FCC as overseen
by Congress (Aufderheide, 1999). Station group-owners that have purchased stations that
would collectively control more than fifty percent of the local radio advertising revenue
have been subject to antitrust review by the Department of Justice (Drushel, 1998;
Ekelund et al., 2000; Gomery, 2000).

The elimination of national radio ownership limits has led to the creation of

several large, national station groups with hundreds of stations across the country in

12

dozens of local markets ("Clearly, it's clear channel", 2000). Since the passage of the
Telecommunications Act of 1996, ownership concentration levels have increased in local
radio markets based on the Herfindahl-Hirschman Index (HHI) of market concentration
(Drushel, 1998; Ekelund et al., 2000; FCC, 2000, 2001). The FCC has used format
diversity as the primary measure of the public interest standard. Yet, most investigations
find that increases in ownership concentration have not statistically affected market
format diversity (Drushel, 1998; Ekelund et al., 2000; FCC, 2000, 2001).

In a 2003 report, the FCC began to consider outlet diversity and viewpoint
diversity as a measure of the public interest. The FCC has acknowledged that having only
a handful of owners controlling radio stations in a market might unduly limit the number
of independent viewpoints expressed. Since only a few group-owners control most of the
stations in many markets, if a group-owner decides not to carry local information or
news, then market listeners’ choices for local information sources are significantly
decreased. However, program or format diversity is still a primary public interest
standard for the FCC and was found to remain unaffected by increases in ownership

concentration (FCC, 2003).

Effects of Ownership Concentration on Format Diversity

Because policymakers use local market format diversity as one measure of the
contribution to the public interest by the radio industry, the FCC has been concerned that
increases in ownership consolidation not reduce market format diversity. An FCC report
in January, 2001 finds that the relaxation of radio market ownership restrictions by the

Telecommunications Act of 1996 has indeed led to increases in radio station ownership

13

consolidation. The number of radio station owners declined by twenty-two percent
between March, 1996 and March, 2000. However, the number of formats per local radio
market has remained relatively constant despite the increase in local market HHI
measures of ownership concentration. A couple of studies claim statistically weak
evidence that increases in ownership concentration increase format diversity in smaller
markets (FCC, 2001; Williams & Roberts, 2002). Rogers & Woodbury (1996) provide an
explanation for why increases in ownership concentration may not have a significant
effect on format diversity which is consistent with their own empirical findings.

Rogers & Woodbury (1996) find the number of market stations has a statistically
significant positive effect on the number of market formats. Program choice theory posits
that the number of formats present in a market should be positively correlated with the
number of stations broadcasting in the market. When the number of stations is small, then
the possible number of viable formats in the market is also constrained, regardless of the
economic market structure or market audience characteristics. According to Rogers &
Woodbury (1996), the best way to increase the number of formats Offered in a market is
to significantly increase the number of stations broadcasting in the market. Furthermore,
Rogers & Woodbury (1996) find the percentage of the market audience listening to radio
is positively related to the number of market formats. Rogers & Woodbury (1996) posit
that more radio formats expand the number of people that find satisfying programming.
But, it is also possible that stations program greater format diversity when there is a
larger portion of the population that is willing to listen to the radio. These two
conclusions are consistent with Napoli’s (1997) argument that audiences listen in

proportion to the programming available and station programmers are attuned to market

14

audience preferences. Rogers & Woodbury’s (1996) finding does not identify causality in
the relationship, but does Show that markets with more radio formats have greater
percentages of their populations listening to radio, all else equal.

Rogers & Woodbury (1996) argue that changes in ownership concentration
should not affect radio market format diversity because changes in ownership
concentration will not significantly change the number of stations broadcasting in a
market, nor change the population characteristics of the market. When the number of
market stations and audience characteristics remain fairly constant, format diversity
should remain constant regardless of the market structure. Because the Rogers &
Woodbury (1996) study was conducted on data before the passage of the
Telecommunications Act of 1996 and before radio ownership concentration began to
significantly increase, the question remains whether increases in local market ownership
concentration affect the number of market formats or the percentage of the market’s
population listening to the radio. If Rogers & Woodbury (1996) are correct, adding a
group ownership concentration variable to their equations using data after the passage of
the Telecommunications Act Of 1996 should not affect number of market formats or the
percentage of a market’s population listening to radio.

This study will replicate Rogers & Woodbury’s (1996) equations with the
addition Of an ownership concentration variable. This study uses market data from 1995
through 2000 because this time period covers one year before the passage of the
Telecommunications Act of 1996 through the height and decline of station sales that
occurred after the passage of the Act. According to Change & Ki (2004), the majority of

station sales occurring after the passage of Telecommunications Act of 1996 happened

15

between 1996 and 1998. Stations station sales significantly declined after 1998 (Chang &
Ki, 2004). Conducting this investigation should help to clarify whether Rogers &
Woodbury’s (1996) argument that ownership concentration does not affect format

diversity or audience size is correct.

Audience Satisfaction in the Radio Market

A station must be recognized as a facilitator of communication between the
market’s advertisers and consumers or audience, however radio audiences may not
welcome radio advertising as it interrupts the programming they desire to listen to. Both
advertisers and audiences demand a service from a radio station. The advertisers demand
access to a sizable number of market consumers that will potentially buy their products
and services. Advertisers generally determine the value of a station’s audience by the
number and demographics of the listeners. Consumers demand programming that
satisfies their format preferences. Radio stations strengthen audience loyalty by
differentiating their products through programming (or format) decisions. A radio station
will attempt to gain a competitive edge over other market stations through format
differentiation. A radio station adjusts its programming so the attracted audience is
differentiated from other stations’ audiences in size and demographics (Wildman &
Cameron, 1982). Radio station product differentiation occurs with both the format being
broadcast and the composition of audience that listens.

Program choice theory assumes stations know viewers’/listeners’ preferences
among program formats (Beebe, 1977; Owen & Wildman, 1992; Spence & Owen, 1977;

Steiner, 1952). In reality, the true or actual distributions of market audience preferences

16

can never be known, which also means that the optimum format diversity for serving the
preferences of the market’s audience is also unknown. But, a sample of radio listeners’
choices from the available market stations can be measured through commercial audience
rating services. Most medium—to-large market radio stations make many of their business
operating decisions based on the measured audience ratings for their radio markets.3
Station executives using audience ratings to make format decisions are not basing these
decisions on audience preferences, but estimated audience choices. Effective radio sales
executives can use station brand identity and reputation to entice advertisers to buy
airtime (Ditingo, 1995; Norberg, 1996). However, audience ratings provide stations with
a tangible measure for the value of station airtime. Radio station executives and radio
advertisers have come to trust audience ratings as an objective measure of airtime value;
therefore audience ratings are an accepted measure of the size and other characteristics of
actual station audiences (Napoli, 2003).

Station programming that fails to generate enough advertising revenue to cover
station operation and programming costs must be adjusted to create programming that
advertisers will support. Competitive stations that have success attracting a large,
lucrative audience that generates high advertising revenue may see their programming
strategies mimicked by other stations in the market. Format duplication in competitive
markets is predicted by program choice models. Steiner and Beebe program choice
models assume constant program costs and advertiser value of radio listeners which
easily translates to larger audiences generating larger station revenues and profits. These

models also predict that competitive station owners will program a format with a large

 

3 Arbitron is the leading audience ratings service for radio. Arbitron routinely surveys audience listening
habits in medium to large media markets, but only surveys smaller media markets upon the request of local
stations (Arbitron Inc., 2002).

17

prospective audience until this audience has been divided into such small pieces that
programming another format will generate higher profits. Stations that are successful
attracting a large audience in a market should expect to see other stations in the market
attempting to duplicate their success by programming the same format.

The Steiner and Beebe program choice models assume that listeners will value
multiple offerings of the same program type as perfect substitutes, which means that
consumers should receive the same benefit from listening to any station broadcasting an
identical format. Because stations with identical programming formats are perfect
substitutes in the Steiner and Beebe models, it is also assumed that the stations offering
the same format will evenly split the portion of the audience that prefers to listen to that
format (Beebe, 1977; Napoli, 1997; Owen & Wildman, 1992; Steiner, 1952). These
assumptions simplify program choice models and make them more tractable, but these
assumptions do not reflect actual media audience behavior (Berry & Waldfogel, 1999;
Napoli, 1997). An examination of Arbitron ratings reveals that radio stations with the
same formats in a market do not evenly split the audience choosing to listen to that
format. In reality, stations broadcasting the same format are not perfect substitutes, yet
this does not mean these stations do not have a degree of substitutability.

Broadcasting identical formats does not mean broadcasting identical program
schedules. Radio stations differentiate their programming formats by using different on-
air personalities and station imaging to give their stations a unique and identifiable sound
with market listeners. These differences help a station create brand identity with market
listeners that separates it from market competitors (Ditingo, 1995; Norberg, 1996). When

considering music formats, radio stations also differentiate their programming by

18

choosing to broadcast different playlists of songs in the same genre of music. This has led
to some formats (mainly music formats) splintering into fragmented forms of the original
format category.

Berry & Waldfogel (2001) and Williams, Brown, & Alexander (2002) both
investigate the overlap of new songs added to weekly playlists by stations programming
fragmented formats as reported in radio industry trade journals. Both studies find that the
majority of new songs that are added to station playlists each week are different for
stations programming different format fragments. Both studies also find that there is a
small degree of overlap between new song lists of fragmented formats under the same
format category. For example, under the adult contemporary (AC) general format
category, stations in the Hot AC and Soft AC fragments may not add identical lists of
new songs to their playlists each week, but may have a song or two that appear on both
lists each week.

Based on these results, one may conclude there is a very low degree of
substitutability. between fragmented forms of a general format category. However, both
of these studies are limited by the fact that they only considered the new songs added by
the stations each week. Stations only reveal the new songs added to playlists to trade
journals, so these studies do not consider a comprehensive list of songs that were played
on each station for the week. Stations do not only play new songs, but also program a
significant amount of recurrent and older songs (Williams et al., 2002). Williams et al.
(2002) speculate that if comprehensive playlists were considered, there would be a higher
degree of substitutability between the fragmented formats of a common general format

category than between different general format categories. A format substitutability

19

hypothesis is difficult to test because actual audience preferences are unknown. Audience
ratings aggregate audience listening habits and do not follow the listening habits Of
individual listeners. Audience ratings do not reveal whether listeners of one station might
also tune into another station with a similar format or tune into a station with a different
format. Yet, if some degree of homogeneity between audiences in different radio markets
is assumed, audience format substitutability can be examined through a comparison of
audience listening choices between markets with different mixes of radio formats.

Rogers & Woodbury (1996) investigate what they term ‘within format’ effects
and find that market format audiences are not responsive to changes in the number of
stations broadcasting a format or to the number of stations broadcasting other formats.
These results suggest that increasing the number of stations broadcasting a format does
not expand the market audience for that format (Rogers & Woodbury, 1996). This
unresponsiveness suggests a high degree of substitutability between stations that
broadcast the same general format. Program choice theory predicts that increasing the
number of stations broadcasting a format will not expand the audience choosing to listen
to the format, because the audience listening to the format will treat all stations
broadcasting the format as perfect substitutes.

Rogers & Woodbury (1996) find the number of other formats present in the
market does have a significant negative effect on the market audiences for some formats.
These results indicate that some format audiences are responsive to changes in format
diversity but other format audiences are not. These findings are consistent with Beebe’s

suggestion that secondary program preferences will affect market format diversity.

20

Certain formats will draw larger audiences when the number of market formats is
smaller.

It is worth investigating whether adding a group ownership concentration variable
to Rogers & Woodbury’s (1996) within format equation has an effect on the
responsiveness of market format audiences. If increases in ownership concentration have
affected the number of formats available in the market, then there may also be an effect
on the responsiveness of format audiences to changes in the number of market formats
and to changes in the number of stations broadcasting the format. However, if Rogers &
Woodbury (1996) are correct, increases in ownership concentration will not affect the
number of market formats or the responsiveness of format audiences to changes in format

availability.

Effects of Ownership Concentration on Station Performance

Few studies have specifically examined group-ownership effects on station
performance. Ekelund, Ford, & Koutsky (2000) study the effects of ownership
concentration on radio station sales prices. They utilize a regression model for which
station sale price is a function of station characteristics (revenue, FCC classification,
transmitter power wattage), market characteristics (market radio revenue, number of
competing television stations, newspapers and cable penetration), local market
concentration (Market HHI for revenue and rating shares), and ownership characteristics
(number of station owned locally and nationally). Ekelund, Ford, & Koutsky (2000) test
both a market collusion hypothesis and a station group efficiency hypothesis.

“If increases in market concentration increase the level of market power for all
stations, then the concentration or collusion hypothesis is supported. . .if only large

21

firms (group owned) have higher profits, then higher profits reflect the superior
efficiency of large firms” (Ekelund et al., 2000, p. 169).

Their measure for station market power is station sale price. Ekelund, Ford, & Koutsky
(2000) find strong support for their efficiency hypothesis, and their collusion hypothesis
receives only weak support but cannot be ruled out. The more stations a group owns the
higher price they are willing to pay for stations, which supports the efficiency hypothesis.
Increases in market concentration have a statistically weak effect (p < .1) on station sales
price, which modestly supports their collusion hypothesis. Ekelund, Ford, & Koutsky
(2000) conclude that increases in station sales prices are more likely a result of station
group efficiencies associated with station group consolidation than a result of rising
station value due to overall increases in local market concentration or collusion.

Many researcher posit that local station group-owners should benefit from
resource efficiencies such as sharing station office space, personnel, airtime sales and
station promotion strategies, and programming and production assets (Albarran & Pitts,
2001; Berry & Waldfogel, 2001; Chambers, 2001; Ekelund et al., 2000; Wirth, 2001).
One result of sharing resources among a group’s stations is thought to be an increase in
the quality of station programming (Albarran & Pitts, 2001; Chambers, 2001; Wirth,
2001). To date, radio research has not examined group-ownership effects on station
audience as a measure of station performance. If group—ownership results in station
performance efficiencies, then adding one more station to the group should result in
larger audience for all stations in the group, all else equal. Examining the marginal effect
of group size on station audience is one way to examine the effects of group ownership

on station performance.

22

Other factors that may affect station audience and performance also need to be
considered. Berry & Waldfogel (1999) examine the social inefficiencies of the free-entry
condition of radio markets. This study tested the hypothesis, “if stations [with the same
format] are identical, then listeners will be indifferent between stations, and entry will not
expand the market” (Berry & Waldfogel, 1999, p. 402), which is based on Steiner’s
(1952) assumption that stations broadcasting identical formats are perfect substitutes.
Following this logic, if a new radio station enters the market and duplicates an existing
format, then the new station should steal audience share from incumbent stations and not
expand the market audience listening to radio. Berry & Waldfogel (1999) find that under
a free-entry condition new entrants have little effect on the audience share of incumbents,
whether the new entrants duplicate the formats of incumbent stations or enter the market
with unique formats. Furthermore, a new station entering a local market has little to no
effect on the percentage of radio listeners in the market and does not expand the audience
listening to radio. According to these results, new station entrants attract insignificant
audiences. In essence, this study questions whether there is any public interest benefit to
adding new stations to a radio market. Berry & Waldfogel (1999) conclude that free-entry
into radio markets causes a welfare loss because large fixed costs of entry make it
difficult for new station entrants to recoup initial investments.

While Berry & Waldfogel’s (1999) initial conclusions may be correct, they do not
consider that it may take time for stations to find an audience. New station entrants
cannot immediately gain audience share because they need to build a brand image and
reputation with a market audience (Albarran & Pitts, 2001; Ditingo, 1995; Gomery, 2000;

Norberg, 1996). Listeners must become aware of a new station and make an effort to

23

listen to a new station. This process of growing listener awareness and adoption will take
time. Therefore free-entry into a radio market may generate welfare gains if new station
entrants are able to build loyal audiences that can generate enough advertiser revenue to
cover costs. Incumbent stations in a market may have a natural advantage, because the
potential station audience has greater awareness of their existence in the market. A larger
share of the market audience is likely to be aware of and listen to a station the longer that
station broadcasts in the market, all else equal.

Further support for a longitudinal component to station performance can be found
by examining station format changes. Radio programming formats are not a static
characteristic of radio stations. A station can change its format with minimal operational
costs but with substantial costs to station reputation in the market and to station brand
identity with the market audience. Stations do not take format changes lightly because
they risk losing their current audience with no guarantee that the new format will recoup

the lost audience. Romeo & Dick (2001) find:

“that format changes are often not an effective means of improving station
performance. Larger radio groups have been more successful in using format
changes to generate performance improvements at individual stations, but the data
indicate that these groups make format changes only for stations whose
performance is well below average, and tend to make changes at the time of
purchase” (p. 25).

According to Romeo & Dick’s (2001) findings, the costs of format changes seem to be
lower for stations in a larger station group, where the loss of station revenue from a
format change can be supported by the profits of the other stations in the group.
However, regardless of station ownership circumstances, there are instances when station
performance may be so poor that the risks involved in changing a station’s format seems

no worse than remaining with the current format (Romeo & Dick, 2001). There is at least

24

a perceived, if not actual, benefit to broadcasting a particular format over a period of time
to build audience trust and loyalty and in turn a larger station audience.

Conclusions can be drawn from both Berry & Waldfogel (1999) and Romeo &
Dick (1999). Berry & Waldfogel’s (1999) study presents evidence that station audience is
affected by the length of time a station broadcasts in a market. Stations that have served a
market longer tend to have larger audiences than other stations in the market. Romeo &
Dick’s (2001) study raises the possibility that station audience share is affected by the
length of time a station broadcasts the same format. However, these hypotheses have not
been empirically tested. Furthermore, a large number of stations changed ownership in
the first years after the passage of the Telecommunications Act of 1996 (Chang & Ki,
2004). Anyone connected to the radio industry is aware that changes in station ownership
are generally accompanied by changes in station personnel. The station personnel
turnover after a change in station ownership may affect constancy of station
programming. However, station ownership is not an obvious characteristic when listening
to a station. In fact, most listeners do not know or care to know who owns a particular
radio station; listeners are most interested in the station’s programming or format
(Ditingo, 1995; Norberg, 1996). The effect that the length of time an owner has owned a
station on station audience size is not clear and should be tested. So, the research
questions asked are: do the length of time a station has been broadcasting in a market, the
length of time a station has been broadcasting its current format, and the length of time a

station has been owned by its current owner, affect the size of the station’s audience?

25

CHAPTER TWO

METHODOLOGY

Ownership Concentration Effects on Market Formats, Stations and Audience

This first set of equations replicates Rogers & Woodbury’s (1996) study and adds
an ownership concentration measure to their equations. These regression equations
examine the responsiveness of the number of market formats to changes in the number of
market stations, the responsiveness of market audience to changes in the number of
market formats, and the responsiveness of market format audiences to changes in format
availability. Rogers & Woodbury (1996) first tested two hypotheses, “(i) that the number
of formats in a radio market is positively related to the number of stations, and (ii) that
listenership or audience is positively related to the number formats” (p. 83), using the
following functional relationships:

Formats =f(Stations. Pop. X). and (1)

PercentPop = g(Formats, A9. (2)
where, Formats represent the number of formats present in the market; Stations
represents the number of stations broadcasting in the market; Pop represents the potential
number of radio listeners in the market; PercentPop represents the percentage of the
market population that listens to radio; and, X represents market demographic variables
(percentage of market population that is African-American, percentage of market
population that is Hispanic, per capita income and average population age for the

market’s population).

26

Considering program choice theory, the number of formats in a market should be
a function of the number of market stations, the market population listening to the radio,
and the format preferences of the audience. Differences among market populations are
approximated by demographic variables that are presumed to approximate differences in
market radio listener preferences. The percentage of the market population that listens to
the radio (PercentPop) is a proxy for the market audience that finds desirable
programming on the radio. Program choice theory predicts that the percentage of market
population listening to the radio should be a function of the number of market formats.
AS the number of formats in the market increases, a larger audience should find
programming that is compatible with their format preferences. The percentage Of
audience that uses radio may also be a function of the demographics of the market
population.

In equation (I), the number of stations and the number of radio listeners explain
the number of market formats and in equation (2), the number of market formats explains
the percentage of the market audience listening to radio. Therefore the number of market
formats is the dependent variable in equation (1 ) and is an independent variable in
equation (2). Both equations (1) and (2) use an audience variable as an independent and
dependent variable, but these audience variables are not the same measure of radio
listeners. Equation (1 ) uses the number of radio listeners in the market, which is a
function of market size, as an independent variable, but equation (2) divides the number
of radio listeners by the total population to form the dependent variable for the second
equation. The percentage of radio listeners is not a function of market size but a ftmction

of population characteristics and preferences for radio. The two equations are

27

simultaneous because the number of market formats and audience measures are used in
these equations both endogenously and exogenously. Simultaneous equations by their
nature present endogeneity. Two-stage least squares regression is a preferred method for
correcting endogeneity in simultaneous equations (Kennedy, 1998).

Two-stage least squares regression employs instruments to correct endogeneity
problems that arise when explanatory variables are correlated with the error term of a
regression equation. Instruments are variables that are correlated with the explanatory
variables but not with the error term of the regression equation. The first stage of a two-
stage least squares regression creates instrumental variables by regressing the explanatory
variables on the instruments. The instrumental variables resulting from the first stage are
used as the independent variables in the second stage to estimate the original regression
equation. These two stages correct for the endogeneity in the original equations
(Kennedy, 1998; Wooldridge, 2003).

Rogers & Woodbury (1996) identified per capita income, percentage of the
population that is African-American and percentage of the population that is Hispanic as
market characteristics that should affect audience use of the radio medium. Rogers &
Woodbury (1996) also included an average age of the market population variable in their
equations, a variable that was not used in this replication because a consistent data source
for all markets could not be found. The instruments used by Rogers & Woodbury (1996)
are: the number of market formats, number of market stations, market population, market
average income, percentage of market population that is African-American, percentage of
market population that is Hispanic, market average radio station wage, and number of

market FM stations. Market average radio station wage is not used as an instrument in

28

this replication, because a data source for this variable could not be found. However,
results of the two-stage least square regression with the available 1995 data for this study
are statistically similar to the results of the Rogers & Woodbury’s (1996) study, which
also uses 1995 data. Therefore, the available data for this replication of the Rogers &
Woodbury’s (1996) study are accepted as adequate.

Rogers & Woodbury (1996) argued that significant changes in ownership
concentration due to the relaxation of ownership limits would not affect their findings. If
so, increases in ownership concentration should not affect the number of market formats
or percentage of population listening to radio. These hypotheses could not be tested by
Rogers & Woodbury (1996) because their study was conducted on market data before the
passage of the Telecommunications Act of 1996. This replication of the Rogers &
Woodbury (1996) study determines whether changes in ownership concentration
occurring after the Telecommunications Act of 1996 alters the basic relationships Rogers
& Woodbury (1996) found with their study by adding a variable for local ownership
concentration to their regression equations.

Two different measures of local ownership concentration are employed. First, the
audience shares for all owners owning more than one station in a market are summed to
calculate the first ownership concentration variable (Group Mk! Share). This provides a
percentage of radio audience listening to group-owned stations in a market. The second
ownership concentration variable (Audience HHI) is a sum of squares of all owner
audience shares to construct an HHI for market ownership concentration. The ownership
concentration variables are added to the two-stage least square regression equations used

in the Rogers & Woodbury (1996) study.

29

Another change to the original Rogers & Woodbury (1996) equations adds a
market station variable to equation (2), which tests the relationship between the
percentage of the population listening to the radio and the number of formats present in
the market. Without the inclusion of the number of market stations, the equation assumes
that it is only the number of formats that expands the percentage of people that find
satisfaction from radio. This also assumes that stations broadcasting similar formats are
seen as perfect substitutes by listeners. If the format differentiation between stations
broadcasting the same formats expands the audience, then the number of stations in the
market will also predict the percentage of the market population that finds satisfying
programming from radio.

These equations are estimated with data for each year from 1995 through 2000 to
include variation in market ownership concentration since the passage of the
Telecommunications Act of 1996. Each of the ownership concentration variables is run in
a separate two-stage least squares regression for each data year. Rogers & Woodbury
(1996) estimate the equations in double log form to account for possible nonlinearities.
The same form is used in this study and the full specification of equations (1) and (2),

respectively, are:

InflVumber of Mkt Formats) = a + ,8, In(Number of Listeners)
+ £2 In(Number of Mkt Stations) + ,83 In(Mkt Per Capita Income)
+ ,64 In(Mkt % African -American) + ,65 In(Mkt % Hispanic)
+ fl61n(Mkt Ownership Concentration Measure) (3 )

In(Percent of Mkt Listeners) = a + ,6, lnflVumber of Mkt Formats)
+ ,8; In(Number of Mkt Stations) + ,83 In(Mkt Per Capita Income)
+ 5.; In(Mkt % African-American)+ ,85 In(Mkt % Hispanic)
+ ,86 In(Mkt Ownership Concentration Measure). (4)

30

This study also examines what Rogers & Woodbury (1996) called ‘within format’
effects. The within format effects equation examines the responsiveness of the audience
listening to a specific format to changes in the number of stations broadcasting this
format, changes in the number of other formats available in the market, and changes in
the average number of stations broadcasting other formats in the market. Essentially, this
equation is estimating the responsiveness of a forrnat’s audience to changes in the
forrnat’s availability. The number of market formats is used as an independent variable in
this equation; therefore this equation is simultaneous with equation (3 ). The same
instruments used to correct for endogeneity between equations (3 ) and (4) are used in this
two-stage least squares regression.

In(Mia Audience Share for Format 10 = a
+ ,8, In(Number ofOther Mkt Formats)
+ £2 In(Number of Mkt Stations with Format X)
+ ,8 3 In(A vg Number of Mkt stations for other formats)
+ ,6.) In(Number of Mkt stations)
+ ,65 In(Mkt Per Capita Income)
+ ,86 In(Mkt % Afiican-American)
+ fl71n(Mkt % Hispanic)
+ ,63 In(A udience HHI) (5)
Equation (5 ) is also run on data from each year from 1995 through 2000 to
encompass variation in local market ownership concentration since the passage of the
Telecommunications Act of 1996. Equation (5) is run on data for each of the general
format categories: Adult Contemporary, Rock, Black, Country, Contemporary Hits,
Oldies, Classical, Jazz, Easy Listening, Religious, News/Talk, and Hispanic. This
requires running seventy-two two-stage least square regressions to test each Of the twelve

format categories over the six sample years. The ‘Other’ format category is not estimated

because this is not a structured format with a specific target audience.

31

This equation tests format audience responsiveness to the number of stations
broadcasting the format in the market. If increasing the number of market stations
broadcasting a format has a positive relationship to the percentage of the market audience
listening to the format, this indicates that multiple offerings of the same format are not
perfect substitutes to market audiences. This counters one of the basic assumptions in the
Stein er (1952), Beebe (1977) and other program choice models, but also indicates that

listeners find satisfaction from more choices within a particular format.

Group-Ownership Effects on Station Performance

For this next set of regressions, a station’s share of its market’s audience is used
as a measure of station performance. Station audience share is the estimated percentage
of market listeners tuned to a particular station. The primary justification for using station
audience share as a measure of station performance is the reasonable assumption that
listeners choose to listen to the station from which they receive the most benefit and
satisfaction (N apoli, 1997). Station audience share is a measure of station performance
with a connection to social welfare and with a connection to the station’s financial
interests because station audience size is highly correlated with station revenue (Berry &
Waldfogel, 1999). Advertisers want to reach the largest possible audience with their radio
advertising message; therefore advertisers are generally willing to pay a higher price for
access to a larger audience, all else equal. Given that the sum of all a market’s stations’
audience shares is constrained to one hundred percent, using station audience share
compares a station’s performance relative to all other stations broadcasting in the same

market. Stations with larger audience shares are assumed to be providing more satisfying

32

programming compared to the stations with smaller audience shares in the same market.
This study examines factors affecting the magnitude of station audience share.

Broadcast policymakers have been concerned that increases in ownership
concentration not reduce local market format diversity because this has been the primary
policy measure of social welfare and the public interest for radio programming (Berry &
Waldfogel, 1999, 2001; Drushel, 1998; Ekelund et al., 2000; FCC, 2000, 2001; Williams
et al., 2002; Williams & Roberts, 2002). There is evidence that increases in ownership
concentration, measured by the HHI, have not affected local format diversity (Drushel,
1998; FCC, 2000, 2001), but to date the effects of ownership concentration on station
audience share have not been investigated. The first question to be examined here is does
the number of local stations owned by a group-owner (group size) affect station audience
share, all else equal?

One result of station consolidation is expected to be resource sharing between
local stations in a group (Albarran & Pitts, 2001; Berry & Waldfogel, 2001; Chambers,
2001; Ekelund et al., 2000; Wirth, 2001). Resource sharing is expected to improve the
quality of station programming for all stations in the group, because the stations can pool
their talent to produce more effective sales efforts, news, productions, and programming
(Albarran & Pitts, 2001; Chambers, 2001; Wirth, 2001). If group efficiency translates
into more effective programming, then does adding more stations to a group increase
station audience share for all stations in the group, all else equal?

The first equation used to examine the relationship between station audience share
and local group size uses the number of other stations in the group to measure group size.

This relationship between group size and station audience share could be nonlinear.

33

There are several statistical methods to control for nonlinearity. This study begins by
adding a squared variable for group Size to test for nonlinearity in the relationship.
Station Share = a + B, (Group — 1) + ,8; (Group - [)2 + flX + s (6)
where, Station Share is the station’s audience share, Group — 1 is the number of other
stations in the group, X is the vector of other station, format and market variables that
may effect station audience Share, and e is the error term. Group size is measured by the
number of other market stations with the same owner as the station on the left side of the
equation, (Group — I). If the owner of the station on the left side of the equation only
owns this station in the market (an independent station), then (Group — 1) equals zero.
The square of (Group — I) is used to test the linearity the relationship. If the coefficients
for linear and squared variables are both statistically significant then a nonlinear
relationship is indicated. If equation (6) indicates a statistically significant nonlinear
relationship, then employing a cubic variable in the equation tests for a more complex

curvilinear relationship:

Station Share = a + ,8, (Group — l) + ,8; (Group — [)2 + ,63 (Group — [)3
+ AX + 8 ( 7)

Group size is a continuous independent variable. If the relationship between a continuous
independent variable and dependent variable has numerous points at which the
curvilinear relationship changes direction or slope, these polynomial functions may not
be adequate to effectively estimate the relationship (Marsh & Cormier, 2002).

Using a spline function is another method to estimate a curvilinear relationship
between a continuous independent variable and a dependent variable. “Spline functions
are a device for approximating the shape of a curvilinear stochastic function without the

necessity of pre-specifying the mathematical form of the function” (Suits, Mason, &

34

Chan, 1978, p. 132). Spline functions are used to test for changes in slope at discrete
intervals for an independent variable. The relationship investigated with a spline function

can be specified with the following set Of equations:

Station Share = a + ,8, (Group; — 1) + flX + e, if Group _>_ 1 station
Station Share = a + ,6; (Group; — I) + flX + e, if group 2 2 stations
Station Share = a + ,83 (Group 3 — 1) + flX + e, if group 2 3 stations

Station Share = a + ,8); (Group; — 1) + flX + e, if group 2 k stations. (8)
Each successive equation in equation set (8) estimates the linear relationship between
group size and station audience share that results from adding one more station to the
group from an independent station to the largest group size, while holding constant other
variables that may affect station audience share. Comparing the coefficient estimates for
each of the equations will give a sense of the changes occurring in the relationship at each
level of group size.

Spline regression analysis provides a device for combining the equations in set (8)
into one equation. A spline regression estimates linear approximations of the relationship
between an independent variable and a dependent variable for given intervals Of the
independent variable. Points where the linear relationship may change are referred to as
spline knots. Estimating the slope between each spline knot, or spline variable, produces
a piecewise estimate of the shape of the curvilinear relationship between a continuous
independent and a dependent variable with one equation (Marsh & Corrnier, 2002).

Because the effects of group size on station audience share have not been
previously tested and theory does not identify precise points where the relationship is
likely to change, a spline variable for each group size two through nine or more stations

is estimated. The spline variables for station group size are calculated by first creating a

35

dummy variable (01,) for each level of group size. Each successive spline dummy
variable removes the Observations that have fewer stations than its threshold value
(Marsh & Cormier, 2002). For example, D 3, a dummy variable for a station group with
three or more stations, would equal one if the station group has three or more stations and
would equal zero if the station group has less than three stations. The spline variable is
calculated by multiplying the dummy variables by the continuous independent variable
(Marsh & Cormier, 2002; Suits et al., 1978). In this case, the spline variables are created
by multiplying the number of the group’s other stations, (Group J), by the dummy
variables. A dummy variable is not required for the first level (an independent station).
The spline regression equation is specified as follows:
Station Share = a + ,8; (Group — I) + ,6; [(Group — I) D;]

+ 72 [ (Group - 1) D2] + 33 [ (Group — 1) D3] + 73 [ (Group - 1) D3

+ + 138 [(Group - 1) 08] + 78 [(Group - 1) 0.9] + flX + 8, (9)
where y is the error variance for adding each successive level of’the continuous variable

to the equation to combine the equations in set (8). Let ,6; + y; = 6, then:

Station Share = a + ,6, (Group - I) + 5; [(Group — I) D;]
+ 63 [(Group — 1) D3] + + 68 [(Group— 1) D8] + flX + e. (10)

In this case, the first level of station group size is an independent station, where
Group - 1= 0. Therefore, the independent station effect will be captured in the intercept
of the regression equation and becomes the default case for the dummy variables used in
the equation. Because there is no theoretical basis to indicate where the relationship
between group size and station audiences may change, this spline regression equation
uses all levels of station group size, two through nine or more. Using all levels of the

continuous independent variable in the equation will result in multicollinearity. This is

36

corrected by removing one level as the default case. The regression equation is specified

as follows:

Station Share = a + ,B/ [(Group —1)D;] + 6; [(Group -— 1) D3]
+ + 67[(Group-1)Dg] +flX+ e. (11)

The coefficient for the first variable in the model represents the slope from the intercept
to the first level of the independent variable. The coefficient for each successive spline
variable is a measure of the change in the slope from previous Spine knot. A statistically
insignificant coefficients indicates a slope that is not statistically different from the Slope
between the previous two spline knots. The coefficients of the spline regression can be
used to plot an estimate of the relationship between the continuous independent variable
and the dependent variable (Marsh & Cormier, 2002).

Spline variables will be calculated for each station group size allowed by the
Telecommunications Act of 1996, two stations through eight stations. An additional
spline variable accounts for stations groups that may be larger than the prescribed eight
station limit (nine or more stations). There are instances where station groups in markets
on the Mexican border have controlling interests in foreign stations that broadcast signals
into American markets. These ownership interests in Mexican stations are not counted by
the FCC against the cap and can set a group’s station ownership over the eight station
cap.4 There are also instances where station groups merge placing local station group
operations above the eight station cap. The acquiring owner will typically take
operational control of the acquired stations when the merger deal is settled, but will wait

to sell Off stations to comply with the eight station cap until after the merger has been

 

" Canadian law prohibits foreign investment in or control of Canadian stations.

37

approved by the FCC .5 Controlling for these instances of ownership above the prescribed
cap will reveal if there is a possible advantage to station ownership over the regulated
cap.

Other variables could potentially affect station audience size and need to be
controlled in these equations. These variables are represented by vector X in equations
(5) through (11). The frequency band (AM or FM) of the station is represented by a
dummy variable for AM stations. Although the AM band was the first frequency band Of
operation in the history of the radio industry, since the late 19705 FM stations have
earned higher revenue and been more popular with radio audiences (Orlik, 1997). For
these reasons, FM stations are expected to have larger audience shares than AM stations,
all else equal. Since frequency band is suspected to affect station audience share, the
percentage of stations in a group that broadcasts on the F M band is included in the
equation. If FM stations have larger audience shares, then having a greater percentage of
FM stations in a local group may affect the audience Shares of other stations in the group.
Therefore the percentage of FM stations in the group is included in the equation to
examine whether stations part of a group with a higher percentage of F M stations have
larger audience shares.

The FCC classifies AM stations by daytime and nighttime station transmitter
wattage and classifies FM stations by tower height and transmitter wattage. These two
sets of classifications are not comparable, because AM and FM signal waves have
different propagation principles. Past radio research has incorrectly used station

transmitter wattage to compare AM and FM stations (e. g. Berry & Waldfogel, 1999;

 

5 Informal conversations with station group managers, station group owners and a station broker indicate
that it can take more than a year for the FCC to give final approval to station sales and station mergers.

38

Ekelund et al., 2000). AM station transmitter wattage determines how far the broadcast
signal will travel but the distance is greatly affected by geographical terrain, climate and
amount of sunlight. FM broadcast signal wave propagation is not as susceptible to
differences in geographical terrain, climate and amount of sunlight as AM signal waves.
Rather an FM station’s signal coverage area is determined by both tower height and
transmitter wattage. FM stations with a lower tower height must use higher transmitter
wattage to cover the same area as an FM station with a higher tower height. (Orlik,
1997). Because of these differences between AM and FM signals, transmitter wattage
does not provide an appropriate comparison between AM and FM station signal
differences or FCC classification differences.

A station signal coverage area variable is a possible way to empirically control for
signal differences between AM and FM stations related to FCC class. The station signal
coverage area estimates the square mileage covered by the station’s daytime signal.
However, I found a signal coverage area variable has a statistically insignificant
coefficient in the equations and was highly correlated with other independent variables,
such as format variables and market size. Furthermore, the Durbin-Watson (D-W)
statistic indicated heteroskedasticity when the signal coverage area variable was included
in the equations. Station coverage area has a low correlation with station share‘5 and
removal of the station coverage area variable from the equation decreased the R2 statistic
slightly, but improved the Adjusted R2 statistic for the equations. Furthermore, removal of
the station coverage area variable eliminated the multicollinearity and heteroskedasticity
problems in the equations. Other studies have not used variables to approximate station

class, beyond station band, when examining the effects on audience share (e. g. Romeo &

 

6r (l6l44) = .08, p < .001, while controlling for data year

39

Dick, 2001); therefore this study only uses the AM station dummy variable to distinguish
between stations in the two frequency bands.

This study uses dummy variables to represent thirteen general format categories
listed by Arbitron: Adult Contemporary, Rock, Black, Country, Contemporary Hits,
Oldies, Classical, Jazz, Easy Listening, Religious, News/Talk, Hispanic, and Other.
Formats in the ‘Other’ category do not fit into another format category and are the default
case. Stations under the same general format category are assumed to have a higher
degree of substitutability between each other than with stations under different general
format categories (Williams et al., 2002). Station format should have an effect on station
audience share because some formats are more popular with listeners than other formats.

Market population is included in the equation to control for market size
differences. Arbitron estimates the market population for persons over the age of twelve
years. This is the population who radio advertisers feel are potential customers and whose
radio habits the radio industry accepts as most important to monitor. The number of
market stations is included in the equation. Adding more stations to a market, all else
equal, should drive station audience shares downward as the potential audience is spread
across more stations. Certain market variables, such as income, education and ethnicity
may be indicators of the percentage of the population that is likely to listen to the radio
(Berry & Waldfogel, 1999; Rogers & Woodbury, 1996). These variables are included in
the equation to account for demographic differences among market populations that may
affect station audience Share. The spline regression equation with all independent

variables is:

40

Station Share = a + B, [(Group — 1) D;] + 6; [(Group -— 1) D3]
+ 63 [(Group- 1) D4] + 64 [(Group— 1)D5] + 65 [(Group - 1) D6]
+ 66 [(Group— 1) D7] + 67 [(Group- 1) D8] + 68 [(Group — 1) D9.]
+ ,89 % of group that is FM stations + [1'10 AM station
+ ,8 ,1 Adult Contemporary Format + ,8]; Rock Format
+ [h 3 Black Format + flu Country Format + ,615 Contemp Hits Format
+ 3/6 Oldies Format + my Classical Format + #18 Jazz Format
+ flip Standards/Easy List. Format + ,B;0 Religious Format
+ 6;] News/Talk Format + ,8;; Hispanic Format
+ fly Number Mkt stations + ,6;4 Arbitron Mkt Population
+ fl;5 Mkt Per Capita Income + ,B;6 Mkt % college education
+ ,B;7 Mkt % African-American + fl;8 Mkt % Hispanic (12)

Local radio market ownership concentration levels began to significantly increase
after the passage of the Telecommunication Act of 1996 (Drushel, 1998; FCC, 2000,
2001). The height of station sales in the radio industry took place between 1996 and 1998
with the most station sales occurring in 1996 (Chang & Ki, 2004). The number of
national station owners also decreased twenty-two percent between 1996 and 2000 (FCC,
2001). A separate regression analysis will be run on data for each year from 1995 through
2000, and a Chow test will be employed to determine whether the relationships tested by
the equation are consistent across the sample years.

Equations (6), ( 7) and (12) are used to estimate the marginal effect of increasing
the number of stations in its group on a station’s audience share. These equations seek to
find whether group efficiencies have a turning point at a specific number of stations. In
other words, is there number of stations beyond which adding additional stations to a
local group seems to stop increasing station performance or begin decreasing station

performance?

41

Format, Broadcast and Ownership Consistency Effects on Station Performance

Generally station revenue and station audience size are highly and positively
correlated (Berry & Waldfogel, 1999). Increasing station audience share, all else equal,
should translate to higher advertising revenue for a station. A basic assumption is that
station executives program their stations to attract the largest possible share of the
audience willing to listen to their format to in turn generate the highest possible revenue.
Changes in programming formats may take time to increase station performance. Romeo
& Dick (2001) find that station format choice is important to attracting an audience and
that changing a station’s format can have an immediate negative effect on a station’s
audience share. Berry & Waldfogel (1999) find that new stations entering a market,
regardless of format, do not immediately attract significant audiences. Considering the
results of Romeo & Dick (2001) and Berry & Waldfogel (1999), station programming
consistency over time may play a significant role in increasing station audience share.

A station must build a reputation in the market for a loyal audience to tune into its
programming. Listeners likely prefer to listen to stations that they can trust to provide
consistently satisfying programming. Audience loyalty takes time to establish. Adoption
of innovation principles predict five stages that listeners deciding to listen to a different
(or new) radio station will likely follow: knowledge, persuasion, decision,
implementation, and confirmation (Rogers, 1995). First, the listener must have
knowledge of the different station choices available. Second, the listener must be
persuaded by some one or some occurrence to explore the different stations available to
him or her. The listener must have some reason to change their radio habits to seek out a

different station choice. Next, the listener decides to change the radio dial and tune into a

42

different station. Fourth, the listener compares the new choice to his or her preferences
and either implements the new behavior or moves to another station. Lastly, the listener
must confirm that the new station is a better choice compared to all known alternatives. If
this confirmation is positive then the listener may change their radio listening loyalty to
this different station.

There are many different reasons why listeners may changes stations.
Furthermore, different listeners will make station changes with different frequency.
Devices, such as station preset buttons on car and home radios, make changing stations
very easy, but these preset buttons must be set to a station. This assumes that listeners
have knowledge of their available choices to set the buttons to stations they believe they
will enjoy. Other radio set devices, such as ‘Seek’ or ‘scan,’ allow listeners to cycle
through the radio spectrum to stations that they might not otherwise be aware of. This
investigation relies of the measurement of station audiences from Arbitron. This service
depends on radio listeners to recall and report which stations the listener has listened to in
a diary. Therefore, listeners that have some degree of loyalty to a station will report
listening to that station in the diary. This does skew the audience measurement data
toward stations which the listener has knowledge and recall, but also indicates the
stations that the listener is likely to turn to first when the radio is turned on. Listener
adoption and station audience loyalty are dynamic (Rogers, 1995). Listeners choose
which station to listen to based on their preferences among different stations’
programming.

Stations that have broadcast the same format for a longer period of time should

have more time to build a larger, loyal audience that reports listening to the station.

43

Therefore, the longer a station has broadcast its current format, the greater the station’s
audience share, all else equal. A station’s format may not be the only factor in a listener’s
radio station choice. Berry & Waldfogel (1999) find that new stations entering a radio
market have little impact on the audience share of incumbent stations regardless of the
format they broadcast. Listeners are more likely to be aware of a station and consider a
station as a possible listening choice if the station has been broadcasting in the market for
a longer period of time. Therefore, stations that have been broadcasting in the market for
a longer period of time should also have a larger audience, all else equal. Radio stations
generally do not promote or identify their station ownership to listeners, but anyone
working in the radio industry will confirm that changes in station ownership are usually
accompanied by changes in station personnel. These changes in personnel may impact
the consistency of station operation which in turn may have an impact on the consistency
Of the station’s programming. Length of time since an ownership change may therefore
affect station audience share and should be tested.

Equations (6), ( 7), and (12) examine the effects of group size on station audience
share. Assuming that the size of a station’s group has an effect on station audience share,
group size should also be included in this equation. Two measures of group size are
included in this equation. First, Group size can be measured by the number of other
stations in the group (Group-1) and, second, group size can be measured by the sum of
the audience shares for the other stations in the group. This second group size variable,
Adjusted Group Share, is calculated by summing all station audience shares for all other
stations in the same local group as the station whose share is the dependent variable.

Independent stations have an Adjusted Group Share equal to zero. Both of these variables

44

are included in the equation to examine whether group size affects the station’s audience
share.

There are other variables that affect station audience size that also need to be
included in this equation. The frequency band of the station is controlled by a dummy
variable for AM stations. FM stations are expected to have larger audience shares than
AM stations, all else equal. Station coverage area (estimated square mileage of area
reached by a station’s daytime broadcast signal) was originally included in this equation
to account for differences related to FCC station class. As with equations (6), ( 7), and
(12), there were multicollinearity and heteroskedasticity problems when the station
coverage area variable was included in the equation. Therefore, station coverage area was
removed from this equation. Twelve dummy variables represent the twelve general
format categories indexed by Arbitron: Adult Contemporary, Rock, Black, Country,
Contemporary Hits, Oldies, Classical, Jazz, Easy Listening, Religious, News/Talk, and
Hispanic. Formats that do not fit one of these categories are considered as ‘Other’
formats and are the default case. Format variables should affect the potential size of the
station’s audience because some formats are more popular with audiences than others.

Market population is included in the equation to control for differences in market
size. Arbitron estimates the market population for persons over the age of twelve years.
This is the population who radio advertisers feel are potential customers and whose radio
use the radio industry has accepted as most important to monitor. Markets with more
radio stations are likely to have stations with lower audience shares because more stations

are vying for the same listeners’ attention. The potential market audience is Spread across

45

more stations. Income, education and ethnicity may influence the percentage of the

population that listens to the radio and are included in the equation.

Station Share = a + ,6, Number of Years Owned by current owner
+ fl; Number of Years Broadcasting in Market
+ )63 Number of Years with Some format Category
+ ,8; (Group-1) + ,85 Adjusted Group Share
+ £6 AM station + by Adult Contemporary Format
+ fig Rock Format + ,89 Black Format + filo Country Format
+ flu Contemp Hits Format + fizz Oldies Format
+ B13 Classical Format + ,6” Jazz Format
+ [315 Standards/Easy List. Format + ,816 Religious Format
+ B17 News/Talk Format + ,613 Hispanic Format
+ ,6 19 Arbitron Mkt Pop + flm % of Pop listening to Mid Stations
+ D20 Mkt Per C apita Income + ,B;, % College Education
+ 6;; Mkt % Afi'ican-American + ,B;4 Mkt % Hispanic
+ ,8;5 Number of Mkt Stations + ,6;6 Number of Mkt Formats
+ )827 Mkt Ownership Concentration Measure. (13)

Collecting data for the number of years a station has been broadcasting the same
format, the number of years since station entry into the market, and the number of years
with the same ownership for each station in a local market is time and labor intensive. To
meet realistic goals for completing this investigation in a timely manner, the number of

markets used to estimate this equation was reduced to fifteen (roughly 350 stations) from

the year 2000.

46

CHAPTER THREE

DATA

Station audience share is the dependent variable for equations (6), ( 7), (12), and
(13), and is used to calculate the format audience share used as the dependent variable in
equation (5 ). Station audience share is calculated with station ratings data compiled by
Arbitron, Inc. and reported in the American Radio Report published quarterly (Duncan,
1995-2000). Arbitron routinely collects quarterly ratings data (Spring, Summer, Fall and
Winter) for roughly 100 radio markets ranging from the first to 130th largest markets by
population (using 2000 rankings). Arbitron collects semi-annual ratings (Spring and Fall)
for nearly 190 radio markets. Arbitron estimates the number of listeners for each station
by average quarter hour (AQH) from 6:00 AM. through midnight, Monday through
Friday. AQH is an estimate of the average number of listeners that have reported tuning
into a station for at least five minutes in each quarter hour during weekdays. A station’s
audience share is calculated by dividing the station’s AQH by the estimated average
number of market radio listeners on weekdays, or people using radio (PUR) (Arbitron
Inc., 2002). Station AQH and market PUR figures are estimated by Arbitron, Inc. and
reported in the American Radio Report (Duncan, 1995-2000).

Arbitron surveys a sample population from each target market. Each selected
person is required to keep a diary of their listening habits for one week during a ratings
period and return the diary to Arbitron. Arbitron reports ratings for stations that receive
what they term a ‘measurable’ audience. For a station to have a measurable audience

means that enough listeners reported tuning into the station for Arbitron to estimate an

47

audience rating for this station meeting the standard for statistical reliability. Stations that
have no reported listeners or very few reported listeners according to the returned diaries
do not receive an audience rating from Arbitron. Only stations with an audience rating
from Arbitron are included in this study, which excludes stations in the target markets
that do not attract enough listeners to receive an audience rating from Arbitron.

Data were collected from the Spring volume of the American Radio Report from
1995 through 2000. Each Spring volume reports ratings data for roughly 180 markets. As
the markets decrease in population the percentage of rated stations from all stations
located in the market area decreases because Arbitron has a smaller sample from which to
estimate their station ratings. Therefore, this study uses medium to large markets with a
population over 150,000 people of the age of twelve or above. Markets were selected for
this study by first looking at the markets from the first through 150th ranked markets in
the Spring 2000 volume of the American Radio Report. One hundred and thirty-one
markets between the first and 150th ranked market were reported in this Spring volume.
All of these markets were reported in the Spring volumes of the American Radio Report
for 2000, 1999, 1998, 1997, 1996 and 1995. All of the markets in each data year had
ratings for at least eighty percent of the stations reported to be broadcasting in the market
by Arbitron. Unfortunately, Duncan Media Enterprises ended publication of the
American Radio Report with the Winter 2001 issue, so this data is not available for
investigations of the radio industry after the 2000 Arbitron annual cycle.

The American Radio Report also provides station format and ownership data for
rated stations in markets routinely surveyed by Arbitron. There were a few instances were

station format and ownership data are missing in the American Radio Report and are

48

supplemented with data from the Broadcasting & Cable Yearbook (Broadcasting &
Cable, 1995-2000), an annual directory of radio and television station addresses with
ownership and format information for most stations. Usable data were collected for a
total of 16,145 stations in 131 markets across the six Observation years. Table 2 breaks
down the observed stations by data year and frequency band:

Table 2

Breakdown of Station Data by Year and Frequency Band

 

Data Year FM Stations AM Stations Total

 

 

1995 1,652 925 2,577
1996 1,699 972 2,671
1997 1,732 964 2,696
1998 1,779 936 2,715
1999 1,814 926 2,740
2000 1,833 913 2,746
Total 10,509 5,636 16,145

 

There are nearly twice as many rated FM stations as AM stations in these markets. The
number of rated stations in these 131 markets increased six percent from 1995 to 2000.
Because the stations included in this data must attract a measurable audience according to
Arbitron, one cannot assume that this increase in stations is solely the result of new
stations entering these markets. This increase in stations only indicates that more stations
attracted enough listeners to receive an audience rating from Abritron in 2000 compared
to 1995.

Market level data for the 131 markets were collected using the American Radio
Report (Duncan, 1995-2000) the Duncan 's Radio Market Guide 1995 through 2000

(Duncan, 1996-2001). Some markets included in this study are geographically close to

49

larger metropolitan areas. These markets have considerably fewer stations than more
isolated markets with the same populations. There are several reasons why these markets
may have a smaller numbers of stations. Competition from the stations in the larger
adjacent market may crowd out stations in the smaller market, or the FCC may have
reserved more channels for the larger market leaving fewer stations for the smaller
adjacent market. However, discovering the reasons behind the smaller number of stations
in these markets is beyond the scope of this study. When these markets were removed
from the dataset, there was little change to the statistical results of the regression
analyses, so these markets were kept in the data set.

Station ownership does consolidate in this dataset for each year from 1995
through 2000. There are 1,043 different owners in the 1995 dataset; 1,030 different
owners in 1996; 866 different owners in 1997; 784 different owners in 1998; 744
different owners in 1999; and 682 different owners in 2000. In this dataset the number of
owners in these 131 markets decreases by thirty-five percent from 1995 to 2000. Average
market ownership concentration does increase in this dataset each year until 2000. Using
the Herfmdahl-Hirschman Index (HHI) for market ownership concentration of station
owner audience shares, the average local market HHI for 1995 is 949.48. The average
HHI rises to 1,179.98 in 1996, and to 1,415.72 in 1997. The average HHI increases more
slowly in 1998 to 1,488.20, and appears to level off in 1999 to 1,519.01, and is virtually
unchanged in 2000 at 1,521.52. These data show that station ownership did consolidate
and that market ownership concentration did increase and appears to have leveled off in

the sample markets during the observation period for this study.

50

There are eleven instances where station groups are over the 8 station local
ownership cap prescribed by the Telecommunications Act of 1996; two instances in
1997, four instances in 1998, three instances in 1999, and two instances in 2000. Jacor,
which was acquired by Clear Channel in 1998, reached the eight station cap in 1996 in
San Diego, California. In 1997, the station group acquired controlling interests in two
stations located in Tijuana, Mexico that broadcast into the San Diego market. The
Jacor/Clear Channel group in the San Diego market accounts for four of these over local
cap instances. The other seven instances were caused by group-owner station
acquisitions. The acquiring group-owners took control of stations to be acquired placing
the group above the cap before the Spring survey period and the American Radio Report
reported the stations as under the ownership of the acquiring owner. After the merger was
officially approved by the FCC, which was after the Spring Arbitron survey period, these
group-owners sold stations to remain in compliance with the ownership cap. These
eleven instances are reflected in equation (12) by a spline variable for station groups with
nine or more stations to see if these groups benefited for an additional boost in station
audience share.

Equation (13) employs Station data for only fifteen markets because of the
complexity of tracking station format over ten years and collecting broadcast operation
start dates and ownership acquisition dates. A list of fifteen random numbers was
generated and matched to the list of market ranks by population in 2000. Format data for
stations in these fifteen markets were collected from 1990 until 2000 resulting in a
number of years with the same format variable with values ranging from one to ten years.

The primary source for the format data is the American Radio Report. When format data

51

were missing in the American Radio Report they were supplemented by format data from
the Broadcasting & Cable Yearbook. The broadcast Operation start date and owner
acquisition date are reported in the 2000 edition of the Broadcasting & Cable Yearbook.
The dataset from these fifteen markets includes 255 FM stations and 132 AM stations for
a total of 387 stations. The following table lists the fifteen markets with some descriptive
data:

Table 3

List of Markets Used to Estimate Equation (13)

 

 

Market 2000 Number of Rated
Market Rank Market Stations
Atlanta 11 24
Baton Rouge 84 19
Dallas-Fort Worth 6 36
Grand Rapids 66 23
Greensboro-Winston Salem 43 21
Greenville-Spartanburg 60 l 8
Louisville 54 26
Minneapolis-Saint Paul 17 21
New York 1 30
Norfolk 38 23
Pittsburgh 22 24
Raleigh-Durham 48 20
Sacramento 27 28
San Antonio 32 32
Syracuse 78 19

 

52

CHAPTER FOUR

RESULTS AND DISCUSSION

Ownership Concentration Effects on Market Formats, Stations and Audience

The first set of regressions estimates equations (3 ), (4) and (5 ) to replicate the
Rogers & Woodbury (1996) study. This replication adds group ownership concentration
variables to the equations to test Rogers & Woodbury’s (1996) prediction that increases
in ownership concentration would not affect the number of market formats or the
percentage of the market population listening to radio. Rogers & Woodbury (1996) find
that the number of formats present in a market is positively related to the number of
market stations, and the percentage of the market population listening to the radio is
positively related to the number of market formats. Two different ownership
concentration variables are used in equations (3 ), (4) and (5 ). The first ownership
concentration variable (Group Mkt Share) sums the group-owner audience shares of all
market owners with more than one station. The second ownership concentration variable
(Audience HHI) is an HHI for all owner audience shares, or a sum of squares for all
station owner audience shares in the market.

Equation (3 ) examines the effects of the number market stations and the number
of market radio listeners on the number of market formats, while controlling for market
demographic variables and ownership concentration. Equation (4) examines the effects of
the number of market formats and the number of market stations on the percentage of the
market population listening to radio, while controlling for market demographic variables

and ownership concentration. All of the two-stage least squares regressions used to

53

estimate equations (3) and (4) were statistically significant (see Tables 4 through 9 on
pages 59-64). Adjusted R2 statistics for regressions testing equation (3) range from .67 to
.80, and Adjusted R2 statistics for the regressions testing equation (4) range from .65 to
.71. Chow Tests reveal no statistically significant differences between the estimated
relationships in equation (3) for the sample years regardless of the ownership
concentration variables used in the equation. Chow Tests reveal statistically significant
differences between the estimated relationships in equation (4) for the sample years when
the first ownership concentration variable is used in the equation (Group Mkt Share),

F (6, 780) = 12.80, p < .05; and when the second ownership concentration variable is used
in the equation (Audience HHI), F (6, 780) = 9.51, p < .05. The Chow Test results indicate
that equation (3) testing the responsiveness of the number of market formats appears to
be less susceptible to longitudinal changes in the data, than equation (4 ) testing the
responsiveness to changes in the percentage of market radio listeners.

The number of stations in the market is a significant positive predictor of the
number of market formats in all but one of the regressions. The number of stations
coefficient was not significant in the 1998 sample year when the equation included the
Group Mkt Share ownership concentration variable. This insignificant result for the
number of stations coefficient in this one equation is most likely due to statistical noise in
the data for this sample year, Since this variable is significant in all Of the other
regressions. Therefore, the Rogers & Woodbury hypothesis that the number of market
stations in positively related to the number of market formats is supported by these data.

The number of stations was added to equation (4) and is a significant positive

predictor of the market radio audience across all of the sample years, regardless of the

54

ownership concentration variable used in the equation. However, when holding the
number of stations constant in this model, the number of market formats is not a
statistically significant predictor of the market audience. The number of market formats is
a statistically significant and positive predictor of the market radio audience when the
number of stations is not included in the model, which is consistent with Rogers &
Woodbury’s (1996) findings. All else equal, markets with a larger number of stations
have a larger number of formats, and a larger number of stations also expands the
percentage of the market population that listens to the radio.

Rogers & Woodbury (1996) find the number of market listeners to be a
significant positive predictor of the number of market formats. In this replication, the
number of radio listeners is not consistently a statistically significant positive predictor of
the number of market formats across all sample years. This change in significance from
the original study may be caused by the addition of the ownership concentration
variables, longitudinal changes in the data or statistical noise. Regardless of the
ownership concentration variable used in the equation, the number of market radio
listeners was not a statistically significant predictor of the number of market formats in
1999 and 2000. Although causality cannot be determined with the analysis used in the
study, the results do indicate that the number of market formats may not consistently be
responsive to the radio audience size across all sample years.

The ownership concentration variables added to the equations (Group Mkt Share
and Audience HHI) are not consistently statistically significant throughout the sample
years. When the ownership concentrations variables are statistically significant in

equation (3 ), there is a positive effect on the number of formats. But, when the ownership

55

concentration variables are statistically significant in equation (4), there is a negative
effect on the market audience. Increases in local market ownership concentration can
have a positive effect on the number of market formats, but can have a negative effect on
percentage of market radio listeners. However, these results are not conclusive since the
effects of ownership concentration are not consistent across all of the equations.

Rogers & Woodbury (1996) find the percentages of market population that is
Afi'ican-American and Hispanic to be statistically significant predictors of the number of
formats and do not find market per capita income, the percentage of market population
that is African-American, nor the percentage of market population that is Hispanic to be
significant predictors of market radio audience. This replication finds per capita income,
percentage of the population that is Afi’ican-American and percentage of the population
that is Hispanic are statistically significant and positive in both equations, but the
statistical significance is not consistent across all the equations and sample years. Market
per capita income, the percentage of market population that is African-American, and the
percentage of market population that is Hispanic are variables used to describe
characteristic differences between market populations. These market characteristic
variables do not have a consistent statistical effect on the number of market formats or
the percentage of a market’s population that listens to radio in this study.

According the results of this investigation, Rogers & Woodbury’s (1996)
hypotheses receive further support. There is a positive relationship between the number
Of market stations and the number of formats and a positive relationship between the
number of formats and the percentage of market listeners. However, when the number of

market stations is included in equation (4), the relationship between the number of market

56

formats and the percentage of market listeners is not significant, but the relationship
between the number of market stations and the percentage of market listeners is
statistically significant and positive. This indicates that diversity of stations may be more
important to listeners than just having diversity of formats. Rogers & Woodbury (1996)
also argue that increases in ownership concentration variables would not affect the
number of market formats or the percentage of listeners. This investigation provides
inconclusive evidence that this argument is correct. Ownership concentration variables

did have a statistically Significant affect in some of the regressions.

57

TABLE 4

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the

factors affecting the number of market formats and the percentage of market radio
listeners in 1995

 

 

 

 

 

 

 

 

 

 

58

1995 Equation (3) Equation (4)
Format Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 1.26 0.53 -2.45 2.45
In(# Mkt Formats) 0.73 0.44 0.15
In(# Mkt Listeners) 0.06 0.03 0.28
In(# Mkt Stations) 0.26 0.08 0.46b 1.86 0.28 0.683
In(Per Capita Income) -0.08 0.05 -0.10 0.62 0.22 0.16b
ln(% Afro-Amer Pop) 0.03 0.02 0.17c 0.14 0.07 0.14c
In(Hispanic Pop) 0.02 0.01 0.1 0.12 0.06 0.14
In(Group Mkt Share) 0.06 0.04 0.13 -0.07 0.17 -0.03
. . Adj R2 = .67 Adj R2 = .70
MM?" S‘at‘sms F(6, 124) = 45.753 F(6, 124) = 50.82“
Variable B SE B Beta B SE B Beta
(Constant) 0.99 0.59 0.98 2.59
In(# Mkt Formats) 0.79 0.42 0.16
In(# Mkt Listeners) 0.07 0.03 0.36c
ln(# Mkt Stations) 0.25 0.08 0.43" 2.00 0.24 0.73a
In(Per Capita Income) -0.08 0.05 -0.10 0.47 0.21 012"
In(% Afro-Amer Pop) 0.03 0.02 0.15° 0.11 0.07 0.11
In(Hispanic Pop) 0.02 0.01 0.1 1 0.06 0.06 0.07
In(A udience HHI) 0.05 0.03 0.13 -0.37 0.13 -0. 1 8b
. . Adj R2 = .67 Adj R2 = .72
M°del Stat‘sms F6,124() = 44.5021 F(6, 124) = 56.098
a‘p<.001 I’p<.01 °p<.05

TABLE 5

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the
factors affecting the number of market formats and the percentage of market radio

listeners in 1996

 

 

 

 

 

199 6 Equation (3) Equation (4)
Format Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 1.14 0.48 -2.65 2.60
Ln(# Mkt Formats) 0.86 0.50 0.18
Ln(# Mkt Listeners) 0.05 0.02 024“
Ln(# Mkt Stations) 0.29 0.07 0.512" 1.68 0.31 0.648
Ln(Per Capita Income) -0.07 0.05 -0.09 0.67 0.23 0.17b
Ln(% Afro-Amer Pop) 0.03 0.01 0.13c 0.17 0.07 0.17c
Ln(Hispanic Pop) 0.02 0.01 0.13 0.1 l 0.06 0.14
Ln(Group Mkt Share) 0.07 0.04 0.15 -0.1 1 0.19 -0.05
Adj R2 = .73 Adj R2 = .65

Model Statistics

F(6.124) = 58.47“

F(6, 124) = 42.01“

 

 

 

 

 

Variable B SE B Beta B SE B Beta
(Constant) 0.93 0.51 -0.14 2.61
Ln(# Mia Formats) 1.01 0.49 021“
Ln(# Mkt Listeners) 0.07 0.03 0.34b
Ln(# Mkt Stations) 0.27 0.06 0.49“ 1.73 0.27 0.66“
Ln(Per Capita Income) -0.08 0.05 -0.10 0.57 0.22 0.15°
Ln(% Afio-Amer Pop) 0.02 0.01 0.11 0.16 0.07 0.16°
Ln(Hispanic Pop) 0.02 0.01 0.13 0.06 0.06 0.08
Ln(Audience HHI) 0.05 0.02 0.15c -0.33 0.11 -0.20b
. . Adj R“ = .72 Adj R“ = .67
NM“ ”“5"“ F(6, 124) = 57.23“ F(6, 124) = 45.85“
“p<.001 6p<.01 cp<.05

59

TABLE 6

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the
factors affecting the number of market formats and the percentage of market radio

listeners in 1997

 

 

 

 

 

1997 Equation (3) Equation (4)
Format Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 0.84 0.64 -l4.10 3.29
Ln(# Mkt Formats) 0.69 0.56 0.15
Ln(# Mkt Listeners) 0.04 0.02 0.18
Ln(# Mkt Stations) 0.31 0.06 0.58a 1.58 0.32 0.623
Ln(Per Capita Income) -0.05 0.06 -0.04 1.77 0.30 0.303
Ln(% Afro-Amer Pop) 0.03 0.01 0.13° 0.16 0.07 0.15c
Ln(Hispanic Pop) 0.02 0.01 0.11 0.16 0.06 0.19c
Ln(Group Mkt Share) 0.10 0.04 0.19b 0.09 0.22 -004
Adj R“ = .80 Adj R“ = .69

Model Statistics

F(6, 124) = 89.078

F(6, 124) = 50.283'

 

 

 

 

Variable B SE B Beta B SE B Beta

(Constant) 0.92 0.66 -1 1.47 3.39

Ln(# Mkt Formats) 0.82 0.54 0.17

Ln(# Mkt Listeners) 0.05 0.02 0.25c

Ln(# Mkt Stations) 0.34 0.06 0.62a 1.66 0.30 0.65al

Ln(Per Capita Income) -0.07 0.07 -0.05 1.63 0.30 0.28al

Ln(% Afio-Amer Pop) 0.02 0.01 0.09 0.16 0.07 0.15°

Ln(Hispanic Pop) 0.02 0.01 0.08 0.11 0.06 0.14

Ln(Audience HHI) 0.04 0.02 0.12 -0.27 0.12 -0.l6°
Adj R“: .79 Adj R“: .71

Model Statistics

F(6, 124) = 84.593

F(6, 124) = 52.97“

 

ap< .001

I’p<.01

°p<.05

60

 

TABLE 7

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the
factors affecting the number of market formats and the percentage of market radio

listeners in 1998

 

 

 

 

 

1998 Equation (3) Equation (4)
Format Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 1.29 0.70 -14.09 3.26
Ln(# Mkt Formats) 1.26 0.49 0.25
Ln(# Mkt Listeners) 0.08 0.02 0.38
Ln(# Mkt Stations) 0.13 0.07 0.25 1.53 0.30 0.59
Ln(Per Capita Income) -0. 10 0.07 -0.08 1.73 0.29 0.29
Ln(% Afio-Amer Pop) 0.03 0.01 0.17 0.11 0.07 0.1 1
Ln(Hispanic Pop) 0.02 0.01 0.14 0.12 0.07 0.15
Ln(Group Mkt Share) 0.14 0.04 0.32 -0.21 0.23 -0.10
Adj R“ = .73 Adj R“ = .70

Model Statistics

F(6,]24) = 60.718'

F(6,124) = 52.451’

 

 

 

 

Variable B SE B Beta B SE B Beta
(Constant) 1.15 0.74 -12.20 3 .45

Ln(# Mkt Formats) 1.31 0.47 0.26
Ln(# Mkt Listeners) 0.08 0.02 0.42

Ln(# Mkt Stations) 0.19 0.06 0.36 1.53 0.26 0.59
Ln(Per Capita Income) -0.10 0.07 -0.08 1.61 0.30 0.27
Ln(% Afi'o-Amer Pop) 0.03 0.01 0.13 0.11 0.07 0.11
Ln(Hispanic Pop) 0.02 0.01 0.11 0.10 0.07 0.12
Ln(Audience HHI) 0.07 0.02 0.22 -0.23 0.13 -0.14

Adj R“ = .72 Adj R“ = .71

Model Statistics

F(6,]24) = 58.06a

F(6,124)= 53.57a

 

ap<.001

Fp<.01

cp<.05

61

TABLE 8

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the
factors affecting the number of market formats and the percentage of market radio

listeners in 1999

 

 

 

 

 

1999 Equation (3) Equation (4)
F orrnat Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 1.29 0.70 -11.65 3.30
Ln(# Mkt Formats) 0.53 0.46 0.1 1
Ln(# Mkt Listeners) 0.04 0.02 0.17
Ln(# Mkt Stations) 0.27 0.07 0.49 1.87 0.30 0.74
Ln(Per Capita Income) -0.04 0.07 -0.03 1.59 0.29 0.28
Ln(% Afro-Amer Pop) 0.03 0.01 0.14 0.14 0.07 0.14
Ln(Hispanic Pop) 0.03 0.01 0.18 0.14 0.07 0.17
Ln(Group Mkt Share) 0.1 1 0.05 0.22 -0.38 0.25 -0.15
Adj R“ = .71 Adj R“ = .69

Model Statistics

F(6, 124) = 54.08’1

F(6, 124) = 48.978

 

 

 

 

 

Variable B SE B Beta B SE B Beta
(Constant) 0.72 0.78 -9.72 3.60
Ln(# Mkt Formats) 0.50 0.45 0.1 1
Ln(# Mkt Listeners) 0.04 0.03 0.17
Ln(# Mkt Stations) 0.33 0.06 0.61 1.80 0.26 0.71
Ln(Per Capita Income) -0.03 0.07 -0.03 1.46 0.31 0.26
Ln(% Afro-Amer Pop) 0.02 0.01 0.11 0.16 0.07 0.16
Ln(Hispanic Pop) 0.03 0.01 0.15 0.13 0.07 0.15
Ln(Audience HHI) 0.04 0.03 0.12 -0.26 0.13 -0.15
. . Ade“=.70 Ade“=.69
M°del Stansms F(6.124) = 52.59“ F(6, 124) = 49.27“
“p<.001 I’ja<.01 cp<.05

62

TABLE 9

Two-Stage Least Squares Regression Results for Equations (3) and (4) estimating the
factors affecting the number of market formats and the percentage of market radio

listeners in 2000

 

 

 

 

 

2000 Equation (3) Equation (4)
Format Model Audience Model
Variable B SE B Beta B SE B Beta
(Constant) 0.01 0.72 -11.74 3.56
Ln(# Mkt Formats) 0.01 0.50 0.00
Ln(# Mkt Listeners) 0.00 0.02 0.00
Ln(# Mkt Stations) 0.40 0.07 0.728 1.91 0.36 0.79a
Ln(Per Capita Income) 0.05 0.07 0.04 1.65 0.32 0.293
Ln(% Afro-Amer Pop) 0.04 0.01 0.18b 0.14 0.08 0.14
Ln(Hispanic Pop) 0.03 0.01 0.15c 0.15 0.07 0.18c
Ln(Group Mkt Share) 0.06 0.05 0.12 -0.27 0.26 -0.12
Adj R“ = .76 Adj R“ = .65

Model Statistics

F(6, 124) = 69.74“

F(6, 124) = 41.22“

 

 

 

 

Variable B SE B Beta B SE B Beta

(Constant) 0.10 0.77 -7.72 3.91

Ln(# Mkt Formats) 0.04 0.49 0.01

Ln(# Mkt Listeners) 0.00 0.02 0.01

Ln(# Mkt Stations) 0.44 0.06 0.795’ 2.02 0.31 0.84a

Ln(Per Capita Income) 0.05 0.07 0.04 1.41 0.33 0.25al

Ln(% Afro-Amer Pop) 0.04 0.01 0.16b 0.14 0.07 0.13

Ln(Hispanic Pop) 0.02 0.01 0.13 0.09 0.07 0.1 1

Ln(Audience HHI) 0.02 0.03 0.05 -0.38 0.15 -0.23c
Adj R“ = .76 Adj R“ = .66

Model Statistics

F(6, 124) = 69.068'

F(6, 124) = 42.88“

 

“p<.001 bp<.01 cp<.05

63

Equation (5 ) examines the responsiveness of market format audience share to
changes in the number of stations broadcasting the format in a market, to changes in the
number of other market formats and to changes in the average number of stations
broadcasting other market formats. Rogers & Woodbury (1996) termed this
responsiveness a ‘within format’ effect. All of the two—stage least squares regressions
testing equation (5 ) are statistically significant with Adjusted R2 statistics ranging from
.16 to .86. There is a wide variation in the statistical significance of the variable
coefficients in the equations, but there is one basic statistically significant trend in the
results.

Table 10 (on page 67) presents the beta weight coefficient results and statistical
significance for the number of stations broadcasting the format in the market (Stations in
Format) and the number of other formats present in the market (Other Formats) in
equation (5 ) with the Audience HHI variable used for the ownership concentration
variable (results were Similar when the Group Mia Share ownership concentration
variable is used in the equation). In the majority of the regressions, the number of stations
broadcasting a format has a statistically significant, positive effect on a forrnat’s audience
share. The Black and Hispanic format audiences were not statistically responsive to the
number of stations broadcasting these formats, but the other format audiences appear to
be responsive to changes in the number of stations broadcasting them. As the number of
stations broadcasting these formats increases the percentages of radio listeners to these
formats also increases, all else equal. These results are not consistent with the
assumptions made in Steiner and Beebe program choice models. The Steiner and Beebe

program choice models assume that stations broadcasting the same format are perfect

64

substitutes. As more stations broadcast the same format in the market, the stations are
assumed to share the audience that prefers to listen to this format and not expand the
audience that listens to this format. These results are more consistent with the
observations of Napoli (1997) who finds increases in program type availability increases
program type consumption in the network television industry. These results also suggest
that stations that broadcast the same format are not perfect substitutes to market
audiences. The availability of more choices (stations) within a format entices more
listeners to listen to a format.

The other variables in equation (5 ) were not found to be consistently and
statistically significant for most formats. Format audiences were not consistently
responsive to the number of other market formats or the average number of stations
broadcasting other formats, with the exception of the News/Talk Format. These results
indicate that changes in the number of other formats do not to affect the share of the
audience choosing to listen to a particular format, in most cases. However, as the number
of formats increase, the audience listening to the News/Talk format appears to expand.
There were some formats that were responsive to market characteristic variables. For
instance, as the percentage of market population that is Afi'ican-American increased,
Black format audiences increased. The same results were found with the Hispanic format
audiences and percentage of population that is Hispanic. Rogers & Woodbury (1996)
were not Specific in describing the results of their “within format” effect investigation,
but they did report a low occurrence of significance of coefficients in their equations.
This investigation finds that the size of most format audiences is responsive to the

number of stations broadcasting a format.

65

TABLE 10

Partial Regression Results for the Within Format Effects Examined with Equations (5)
presenting the significance of the number of station broadcasting the format and number
of other formats in the market on the market format audience for each sample year

 

Format Tested
Adj R2 Range

Beta Weight Coefficients for Each Sample Year

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable 1995 1996 1 997 1 998 1 999 2000

for all years

Adult Contemporary Formats

.36 - .51 Stations in Format 0.51 0.63“ 0.51“ 0.45“ 0.56“ 0.22
Other Formats -0.03 0.03 -0.51 -0.45 -0.31 -0.74

Rock Formats

.47 - .56 Stations in Format 0.56c 0.72“ 0.55“ 0.36 0.44“ 0.60b
Other Formats -0.61 -0.11 -0.51 -0.90° -0.61 043

Black Formats

.69 - .77 Stations in Format 0.06 0.06 0.07 0.03 0.08 0.12
Other Formats -0.89 -0.67 -0.65° -0.78° -0.49 -0.56°

Country Formats

.53 - .58 Stations in Format 0.45 0.61 0.54“ 0.58c 0.31 0.51“
Other Formats -0.66 -0.15 -0.17 0.01 -0.60 -0.18

Contemporary Hits Formats

.30 - .46 Stations in Format 0.58a 0.66a 0.71a 0.91a 0.83“ 0.818
Other Formats 0.49 -0.38 0.27 0.94 0.63 0.71

Oldies Formats

.22 - .38 Stations in Format 1.27“ 0.77“ 0.79“ 1.05“ 0.64“ 0.68“
Other Formats 1.98 0.36 0.46 1.35 0.13 0.24

Classical Formats

.61 - .74 Stations in Format 0.82a 0.79a 0.78a 0.818 0.743 0.76”
Other Formats -1.43° -0.33 0.17 -0.25 -0.04 -1.17b

Jazz Formats

.68 - .80 Stations in Format .89’1 0.838| 0.81al 0.75a 0.86a 0.84a
Other Formats 0.55 0.44 0.09 0.06 0.58“ 0.28

Standards/Easy List Formats

.31 - .47 Stations in Format 0.52c 0.57“ 075“ 0.53“ 0.86“ 0.85“
Other Formats -0.25 -0.23 0.18 0.11 1.04c 1.03c

Religious Formats

.45 - .63 Stations in Format 0.30 0.67” 0.89a 0.63a 0.77‘ll 0.80a
Other Formats -0.93 -0.03 0.47 -0.10 0.29 0.19

News/Talk Formats

.13 - .37 Stations in Format 0.66c 1.40a 1.07a 0.99a 1.01a 1.053
Other Formats 0.66 2.67“ 1.65“ 1.38“ 1.67“ 1.23c

Hispanic Formats

.55 - .73 Stations in Format 0.08 -0.35 0.29 0.10 0.14 0.08
Other Formats -0.06 -0.72 -0.17 0.06 -.30 -0.31

“p<.001 “p<.01 cp<.05

66

Group-Ownership Effects on Station Performance

Tables 11 through 13 (on pages 73-75) present results for equation (6), with a
linear and quadratic group size variable representing the number of other stations with the
same owner in the market (Group-1), and Tables 14 through 16 (on pages 77-79) present
results for equation ( 7), which adds a cubic group size variable to equation (6). These
results reveal that the relationship between the number of stations in the group and station
audience share is curvilinear (see Figure 1 on page 72 and Figure 2 on page 76). The
equation (6) regressions for each sample year from 1995 through 2000 are statistically
significant with Adjusted R‘2 statistics ranging from .29 to .35. The Chow Test statistic is
significant, F (22. 16139) = 10.56, p < .05, indicating statistically significant differences
between the estimated relationships across sample years. The coefficients for the linear
and squared group size variables are significant at a greater than ninety-nine percent
confidence level and have a positive and negative Sign, respectively, for each sample
year. The statistical significance and Sign of the linear and squared group size variable
coefficients in the equation (6) results indicate the relationship between the number of
stations in the group and station audience share is parabolic in shape (Wooldridge, 2003),
as shown in Figure 1 on page 72.

The regression results for equation ( 7) (Tables 14 through 16 on pages 77-79) are
also statistically significant for each sample year with Adjusted R2 statistics ranging from
.30 to .36. The Chow Test statistic is significant, F (23, 16139) = 9.19, p < .05, indicating
statistically significant differences between the estimated relationships across the sample
years. The coefficients for the linear, squared, and cubed group size variables were

significant at a greater than ninety-nine percent confidence level and had a positive,

67

negative and positive Sign respectively for each sample year. As indicated by Figure 2 on
page 76, station audience share increases with each additional station in the group; the
increases in station audience share decrease in magnitude as the group gets larger, all else
equal. Adding the cubed variable for the number of stations in the group changes the
estimated relationship between the number of stations in the group and station audience
size from a parabolic shape in the equation (6) results to a concave shape in the equation
( 7) results. Using higher power polynomial functions may indicate that the curvilinear
relationship between group size and station share has even more changes or bends as the
number of stations in the group increases; therefore, using a spline regression method is
appropriate to estimate this curvilinear relationship between station share and group size.
Figure 3 (page 80) plots the results of the spline variable coefficients for equation (12) for
each sample year 1995 through 2000. All of the spline regressions were statistically
significant with Adjusted R2 statistics ranging from .29 to .38 (see Tables 17 through 19
on pages 81-83). The Durbin-Watson (D-W) statistics for all of the regressions were close
to the 2.00 mark which does not indicate autocorrelation or heteroskedasticity problems
in the data. Chow Test statistics reveal statistically significant differences between
relationships estimated by equation (12) across all sample years, F (32, 16144) = 7.54, p <
.05, indicating that there are statistically significant differences between the relationships
estimated for each observation year. Therefore, an assumption can be made that there are
longitudinal changes in the underlying relationships. Looking at the Figures 1 through 3
shows that the lines representing the relationship between station Share and group size for
each sample year rise on the graph from 1995 through 1997 or 1998 and then fall during

1999 and 2000. All else equal, adding stations to a local radio group increases audience

68

share for all stations in the group. However, the incremental effect of an additional station
on the average share for stations in a group decreases in magnitude as the group gets
larger.

In all of the regression results for equations (6), ( 7), and (12), the AM station
dummy variable and the percentage of FM stations in the group are statistically
significant and negative predictors of station audience share, all else equal. When holding
the other variables in the equation constant, F M stations consistently have significantly
higher audience shares than AM stations in all sample years. Because the FM frequency
band is generally considered to be the more popular band with radio listeners, this finding
makes intuitive sense. However, owning a larger percentage of FM stations in a group
has a negative effect on station audience share. Although FM stations generally have
higher audience shares than AM stations, owning a large percentage of FM stations
appears to decrease station audience share, all else equal.

Different formats have positive and statistically significant coefficients in both
equations in different years indicating that the format of a station does have a systematic
impact on station audience share, all else equal. The Contemporary Hits format is the
only radio programming format that was a statistically significant and positive predictor
of station audience share in all of the equations across all sample years. The Black and
Country formats have a statistically significant and positive coefficient in more than half
of the equations run in this set of regressions. One could surmise from these results that
the Contemporary Hits, Black and Country formats may be more popular formats with

radio audiences during these sample years.

69

The number of market stations has a statistically significant and negative effect on
station audience share, all else equal. It stands to reason that adding more stations to a
market would have a negative impact on station audience share, all else equal. Station
audience share Should decrease as more stations are vying for the same audience
attention, thereby splitting the available audience into smaller pieces. The other market
characteristic variables for the most part were not significant predictors of station
audience share. In a few of the years, per capita income had a statistically significant
negative relationship with station audience share, but this relationship was not consistent

across all sample years.

70

FIGURE 1

Line Plot Representing the Relationship between Group Size (linear and squared number
of group stations variables) and Station Share from Equation (6) Results

 

l+1995 —e—1996 +1997 +1998 +1999—A—20001

 

105 1T-_-, _______ .. _______ _ _______________ . ——————————————————— «v —————

Station Audience Share

 

 

Station Group Size

71

TABLE 1 1

Regression Results for Equation (6) Estimating the Effects of Group Size (linear and
squared number of group stations variables) and Other Variables on Station Audience
Share in 1995 and 1996

 

 

 

 

 

 

1995 1996

Variables B SE Beta B SE B Beta
(Constant) 5.80 0.71 6.32 0.73
No. of Other Grp Stations 2.21 0.19 0.71a 1.44 0.12 0.708|
No. ofOther Grp Stations“ -031 0.04 -053“ -0.14 0.02 -O.48"
% of FM Stations in Group -0.01 0.00 -0.08b -0.01 0.00 -0.10”
AM Freq Band -2.89 0.21 -0.428 -2.82 0.21 -0.428|
Adult Contemp 1.36 0.58 0.13c 1.28 0.61 0.12c
Rock 0.76 0.58 0.08 0.55 0.61 0.06
Black 2.20 0.60 0.17a 1.38 0.61 0.11c
Country 2.14 0.58 0.22“ 1.59 0.60 0.17“
Contemp Hits 2.15 0.60 0.16“ 1.94 0.63 0.14“
Oldies 0.31 0.59 0.03 0.07 0.62 0.01
Classical 0.75 0.79 0.02 0.02 0.77 0.00
Jazz 002 0.68 0.00 -0.57 0.68 -0.03
Standards 1.29 0.60 0.10c 0.83 0.61 0.06
Religious 0.08 0.59 0.01 -0.32 0.61 -0.03
News/Talk 2.03 0.58 0.22b 1.38 0.60 0.15c
Hispanic 0.61 0.61 0.04 0.1 1 0.62 0.01
Number of Mkt Stations -0.09 0.01 -0.178 -0.10 0.01 -0.20a
Arbitrom Mkt Population 0.00 0.00 -0.01 0.00 0.00 0.00
Mkt Per Capita Income 0.00 0.00 -0.03 0.00 0.00 -0.04
% College Education -0.01 0.01 -0.01 0.00 0.01 0.00
% Afiican American 0.01 0.01 0.02 0.01 0.01 0.03
% Hispanic 0.00 0.00 0.00 0.00 0.00 0.01

Adj R“ = .29 Adj R“ = .32
Model Statistics F (22, 2554) = 49.17a F (22, 2648) = 57.653

D-W= 1.95 D-W= 1.90
“p<.001 “p<.01 cp<.05

72

TABLE 12

Regression Results for Equation (6) Estimating the Effects of Group Size (linear and
squared number of group stations variables) and Other Variables on Station Audience
Share in 1997 and 1998

 

 

 

 

 

1997 1998

Variables B SE Beta B SE B Beta
(Constant) 6.81 0.76 6.59 0.73

No. of Other Grp Stations 1.20 0.09 0.78a 1.33 0.09 0.98a
No. ofOther Grp Stations“ -010 0.01 -056“ -011 0.01 -072“
% of FM Stations in Group -0.01 0.00 -0.09a -0.01 0.00 -0.12"
AM Freq Band -2.60 0.20 -0.393 -2.49 0.19 -0.398
Adult Contemp 0.53 0.61 0.05 0.83 0.59 0.09
Rock 0.13 0.61 0.01 0.25 0.59 0.03
Black 1.29 0.61 0.1 1° 1.37 0.59 0.12°
Country 1.00 0.61 0.11 1.20 0.58 0.13
Contemp Hits 1.80 0.63 0.13“ 1.99 0.60 0.16“
Oldies -0. 12 0.62 -0.01 -0.23 0.60 -0.02
Classical -0.41 0.75 -0.01 0.26 0.74 0.01
Jazz -1.02 0.68 -0.05 -0.75 0.65 -0.04
Standards 0.45 0.61 0.04 0.49 0.59 0.04
Religious -0.60 0.61 -0.05 —0.23 0.58 -0.02
News/Talk 0.87 0.60 0.10 0.96 0.57 0.12
Hispanic -0.26 0.62 -0.02 0.07 0.59 0.01
Number of Mkt Stations -0.09 0.01 -0.19a -0.10 0.01 -0.20a
Arbitrom Mkt Population 0.00 0.00 0.00 0.00 0.00 0.02
Mkt Per C apita Income 0.00 0.00 -0.07° 0.00 0.00 -0.06°
% College Education 0.00 0.01 0.01 0.00 0.01 -0.01
% Afiican American 0.01 0.01 0.02 0.00 0.01 0.01
% Hispanic 0.00 0.00 0.01 0.00 0.00 0.01

Adj R“ = .33 Adj R“ = .35

Model Statistics

F(22,2673) = 60.01a

F (22, 2692) = 67.483

D-W= 1.83 D-W=1.95

 

“p<.001 “p<.01 °p<.05

73

TABLE 13

Regression Results for Equation (6) Estimating the Effects of Group Size (linear and
squared number of group stations variables) and Other Variables on Station Audience
Share in 1999 and 2000

 

 

 

 

 

1999 2000

Variables B SE Beta B SE B Beta
(Constant) 6.00 0.64 6.10 0.64

No. of Other Grp Stations 1.09 0.08 0.84a 1.10 0.08 0.89al
No. of Other Grp Stationsz -0.09 0.01 -0.598 -0.09 0.01 -0.6Sa
% of F M Stations in Group -0.01 0.00 -0.10" —0.01 0.00 -0.08"
AM Freq Band -2.24 0.19 -0.36° -2.11 0.19 -0.353'
Adult Contemp 0.84 0.50 0.09 0.66 0.51 0.07
Rock 0.61 0.50 0.07 0.32 0.50 0.04
Black 1.25 0.51 0.12° 0.91 0.51 0.09
Country 1.26 0.50 0.14° 0.95 0.50 0.10
Contemp Hits 2.23 0.52 0.19a 1.91 0.52 0.17"
Oldies 0.03 0.51 0.00 -0.12 0.52 -0.01
Classical 0.40 0.67 0.01 0.20 0.67 0.01
Jazz -0.74 0.57 -0.04 -0.57 0.59 -0.03
Standards 0.51 0.51 0.04 0.45 0.51 0.04
Religious -0. 12 0.50 -0.01 -0.49 0.50 -0.05
News/Talk 0.98 0.49 0.12° 0.54 0.49 0.07
Hispanic 0.06 0.50 0.01 -0.22 0.51 -0.02
Number of Mkt Stations -0.09 0.01 -0.202’ -0.09 0.01 -0.21a
Arbitrom Mkt Population 0.00 0.00 0.00 0.00 0.00 0.01
Mkt Per Capita Income 0.00 0.00 -0.04 0.00 0.00 -0.04
% College Education -0.01 0.01 -0.03 -0.01 0.01 -0.02
% Afi'ican American 0.01 0.00 0.03 0.01 0.00 0.02
% Hispanic 0.00 0.00 0.01 0.00 0.00 0.01

Adj R“ = .34 Adj R“ = .35
Model Statistics F (22,271 7) = 66.283 F (22,2723) = 66.68‘1
D-W=1.91 D-W=1.90
ap<.001 °p<.01 °p<.05

74

FIGURE 2

Line Plot Representing the Relationship between Group Size (linear, squared, and cubed
number of group stations variables) and Station Share from Equation ( 7) Results

Station Audience Share

 

‘L—o—1995 —e—1996 +1997 +1998 +1999 +2000)

 

14T"..-.,_-_-____-_-___._______-________.___-____--__-___-_-______

13.5J‘r —————————————————————————————————————————————————————————— "
l ____________________________________ 0 _______________ __

13T a

121e ....................... - I; ___________ _______________

1157“- ----------------------- A

 

 

Station Group Size

75

TABLE 14

Regression Results for Equation ( 7) Estimating the Effects of Group Size (linear, squared
and cubed number of group stations variables) and Other Variables on Station Audience

Share in 1995 and 1996

 

 

 

 

 

 

1995 1996

Variables B SE Beta B SE B Beta
(Constant) 6.39 0.72 6.72 0.73
No. of Other Grp Stations 3.99 0.45 1.298 2.90 0.31 1.418
No. ofOther Grp Stations“ -0.82 0.12 -140“ -051 0.08 -1.79“
No. of Other Grp Stations3 0.08 0.02 0.37a 0.04 0.01 0.67“
% of F M Stations in Group -0.01 0.00 -0.08b -0.01 0.00 -0.10"
AM Freq Band -2.92 0.21 -0.423 -2.83 0.21 -0.42’il
Adult Contemp 1.30 0.58 0.13° 1.20 0.61 0.11
Rock 0.71 0.58 0.07 0.52 0.61 0.05
Black 2.16 0.60 0.168| 1.32 0.61 0.11°
Country 2.12 0.58 0.22“ 1.55 0.60 0.16°
Contemp Hits 2.12 0.60 0.16“ 1.90 0.62 0.14“
Oldies 0.25 0.59 0.02 0.01 0.62 0.00
Classical 0.70 0.78 0.02 -0.02 0.77 0.00
Jazz -0.07 0.68 0.00 -0.55 0.68 -0.03
Standards 1.27 0.60 0.09° 0.78 0.61 0.06
Religious 0.08 0.59 0.01 -0.32 0.60 -0.03
News/Talk 2.03 0.58 0.22“ 1.34 0.59 0.15°
Hispanic 0.56 0.61 0.04 0.09 0.61 0.01
Number of Mkt Stations -0.09 0.01 -0.17° -0.10 0.01 -0.20a
Arbitrom Mkt Population 0.00 0.00 -0.01 0.00 0.00 0.01
Mkt Per Capita Income ' 0.00 0.00 -0.03 0.00 0.00 -0.04
% College Education 0.00 0.01 -0.01 0.00 0.01 -0.01
% Afi'ican American 0.01 0.01 0.02 0.01 0.01 0.03
% Hispanic 0.00 0.00 0.00 0.00 0.00 0.01

Adj R“ = .30 Adj R“ = .33
Model Statistics F (23, 2553) = 48.178 F (23,264 7) = 56.793

D-W= 1.97 D-W= 1.92
“p<.001 “p<.01 cp<.05

76

TABLE 15

Regression Results for Equation ( 7) Estimating the Effects of Group Size (linear, squared
and cubed number of group stations variables) and Other Variables on Station Audience
Share in 1997 and 1998

 

 

 

 

 

1997 1998

Variables B SE Beta B SE B Beta
(Constant) 7.07 0.76 6.81 0.72
No. of Other Grp Stations 2.24 0.20 1.47a 2.27 0.22 1.67“
No. of Other Grp Stationsz -0.32 0.04 -1.763 -0.32 0.04 -2.01°
No. of Other Grp Stations3 0.02 0.00 0.59‘1 0.02 0.00 0.65a
% of F M Stations in Group -0.01 0.00 -0.103 -0.01 0.00 -0.12‘il
AM Freq Band -2.61 0.20 -0.40° -2.50 0.19 -0.39’1
Adult Contemp 0.41 0.61 0.04 0.75 0.59 0.08
Rock 0.03 0.61 0.00 0.18 0.58 0.02
Black 1.22 0.61 0.10° 1.30 0.59 0.12°
Country 0.95 0.60 0.10 1.15 0.58 0.12°
Contemp Hits 1.68 0.63 0.13“ 1.91 0.60 0.16b
Oldies -0.21 0.62 -0.02 -0.32 0.59 -0.03
Classical -0.42 0.75 -0.01 0.25 0.74 0.01
Jazz -1 .05 0.67 -0.05 -0.78 0.64 -0.04
Standards 0.38 0.61 0.03 0.43 0.59 0.04
Religious -0.60 0.60 -0.05 -0.24 0.58 -0.02
News/Talk 0.78 0.60 0.09 0.87 0.57 0.1 1
Hispanic -0.35 0.61 -0.03 0.04 0.59 0.00
Number of Mkt Stations -0.10 0.01 -0.208 -0.10 0.01 -0.21°
Arbitrom Mkt Population 0.00 0.00 0.00 0.00 0.00 0.02
Mkt Per Capita Income 0.00 0.00 -0.07b 0.00 0.00 -0.07°
% College Education 0.00 0.01 0.00 0.00 0.01 0.00
% Afi'ican American 0.01 0.01 0.02 0.00 0.01 0.01
% Hispanic 0.00 0.00 0.01 0.00 0.00 0.00

Adj R“ = .33 Adj R“ = .36
Model Statistics F (23. 26 72) = 59.61al F (23, 2691 ) = 66.002'

D-W= 1.86 D-W= 1.96

 

“p<.001 “p<.01 cp<.05

77

TABLE 16

Regression Results for Equation ( 7) Estimating the Effects of Group Size (linear, squared
and cubed number of group stations variables) and Other Variables on Station Audience

Share in 1997 and 1998

 

 

 

 

 

 

1999 2000

Variables B SE Beta B SE B Beta
(Constant) 6.23 0.64 6.29 0.64
No. of Other Grp Stations 1.92 0.18 1.48a 1.79 0.18 1.448
No. ofOther Grp Stations“ -0.26 0.03 -1.74“ -0.23 0.04 -165“
No. of Other Grp Stations3 0.01 0.00 0.57a 0.01 0.00 0.50“
% of FM Stations in Group -0.01 0.00 -0.10° -0.01 0.00 -0.09°l
AM Freq Band -2.25 0.19 -0.36a -2.10 0.19 -0.34”
Adult Contemp 0.71 0.50 0.08 0.56 0.51 0.06
Rock 0.49 0.50 0.06 0.23 0.50 0.03
Black 1.14 0.50 0.11° 0.81 0.51 0.08
Country 1.17 0.50 0.13° 0.88 0.50 0.10
Contemp Hits 2.1 1 0.51 0.18a 1.83 0.52 0.17ail
Oldies -0.08 0.51 -0.01 -0.18 0.52 -0.02
Classical 0.35 0.66 0.01 0.19 0.67 0.01
Jazz -0.79 0.57 -0.04 -0.61 0.59 -0.03
Standards 0.42 0.51 0.03 0.34 0.51 0.03
Religious -0. 14 0.50 -0.01 -0.53 0.50 -0.05
News/Talk 0.87 0.49 0.1 1 0.44 0.49 0.06
Hispanic -0.05 0.50 0.00 -0.28 0.50 -0.03
Number of Mkt Stations -0.09 0.01 -0.203 -0. 10 0.01 -0.213'
Arbitrom Mkt Population 0.00 0.00 0.01 0.00 0.00 0.01
Mkt Per Capita Income 0.00 0.00 -0.05 0.00 0.00 -0.05
% College Education -0.01 0.01 -0.03 -0.01 0.01 -0.02
% Afiican American 0.01 0.00 0.03 0.01 0.00 0.03
% Hispanic 0.00 0.00 0.01 0.00 0.00 0.00

Adj R“= .35 Adj R“: .35
Model Statistics F(23,2716) = 65.05a F (23,2722) = 64.938

D-W=1.92 D-W=1.91
“p<.001 I’p<.01 cp<.05

78

FIGURE 3

Line Plot Representing the Relationship between Group Size (spline variables) and
Station Share from Equation (12) Results

 

i+1995—-3—1996+1997—9—1998+1999—A—20001

Station Audience Share

 

 

Station Group Size Spline Knots

79

TABLE 17

Regression Results for Equation (12) Estimating the Effects Of Group Size (spline

variables) and Other Variables on Station Audience Share in 1995 and 1996

 

 

 

 

 

 

1995 1996

Variables B SE B Beta B SE B Beta
(Constant) 5.53 0.71 6.11 0.72
2 Mkt Station Grp Dummy 0.86 0.07 0.39a 0.82 0.07 0.498
3 Mkt Station Grp Dummy -0.21 0.07 -0.11° -0.08 0.07 -0.06
4 Mkt Station Grp Dummy -0.19 0.06 -0.08° -0.25 0.06 -0.16a
5 Mkt Station Grp Dummy -0.20 0.12 -0.04 -0.03 0.07 -0.02
6 Mkt Station Grp Dummy -0.18 0.22 -0.02 -0.06 0.07 -0.03
7 Mkt Station Grp Dummy 0.12 0.24 0.01 «001 0.07 0.00
8 Mkt Station Grp Dummy -0.08 0.08 -0.02
9+ Mkt Station Grp Dummy
% of FM Stations in Group -0.01 0.00 -0.08° -0.01 0.00 -O.lOa
AM Freq Band -2.91 0.21 -0.423 -2.83 0.21 -0.42"
Adult Contemp 1.27 0.58 0.12° 1.18 0.61 0.11
Rock 0.69 0.58 0.07 0.51 0.60 0.05
Black 2.13 0.60 0.1681 1.30 0.61 0.11°
Country 2.12 0.58 0.223 1.53 0.60 0.16°
Contemp Hits 2.10 0.60 0.15a 1.88 0.62 0.14b
Oldies 0.26 0.59 0.02 -0.02 0.61 0.00
Classical 0.71 0.78 0.02 0.00 0.77 0.00
Jazz -0.09 0.68 0.00 -0.56 0.68 -0.03
Standards 1.25 0.60 0.09° 0.77 0.61 0.06
Religious 0.07 0.59 0.01 -0.32 0.60 -0.03
News/Talk 2.01 0.58 0.22b 1.32 0.59 0.15°
Hispanic 0.54 0.61 0.04 0.08 0.61 0.01
Number of Mkt Stations -0.09 0.01 -0.17“l -0.10 0.01 -0.20°
Arbitrom Mkt Population 0.00 0.00 -0.01 0.00 0.00 0.00
Mkt Per Capita Income 0.00 0.00 -0.03 0.00 0.00 -0.04°
% College Education 0.00 0.01 -0.01 0.00 0.01 0.00
% Afi'ican American 0.01 0.01 0.02 0.01 0.01 0.03
% Hispanic 0.00 0.00 0.00 0.00 0.00 0.01

Adj R“ = .30 Adj R“ = .33
Model Statistics F (26, 2550) = 42.753 F (2 7, 2643) = 49.173

D-W= 1.97 D—W= 1.93
ap<.001 I’p<.01 °p<.05

80

TABLE 18

Regression Results for Equation (12) Estimating the Effects of Group Size (spline

variables) and Other Variables on Station Audience Share in 1997 and 1998

 

 

 

 

 

 

1997 1998

Variables B SE B Beta B SE B Beta
(Constant) 6.66 0.76 6.59 0.73
2 Mkt Station Grp Dummy 0.80 0.08 0.611’ 0.61 0.08 0.5111
3 Mkt Station Grp Dummy -0.05 0.07 -0.04 0.14 0.07 0.13
4 Mkt Station Grp Dummy -0.21 0.06 -0. 1 8a -0.12 0.05 -0.12°
5 Mkt Station Grp Dummy -0.06 0.05 -0.05 -0.08 0.04 -0.08
6 Mkt Station Grp Dummy -0.08 0.04 -0.06 -0.16 0.04 -0.15°
7 Mkt Station Grp Dummy -0.10 0.04 -0.06° -0.01 0.04 -0.01
8 Mkt Station Grp Dummy 0.01 0.06 0.00 -0.08 0.04 -0.06°
9+ Mkt Station Grp Dummy -0.11 0.07 -0.03 -0.04 0.05 -0.02
% of FM Stations in Group -0.01 0.00 -0.10° -0.01 0.00 -0.13‘il
AM Freq Band -2.61 0.20 -0.40" -2.50 0.19 -0.39°
Adult Contemp 0.42 0.61 0.04 0.76 0.59 0.08
Rock 0.05 0.61 0.01 0.18 0.58 0.02
Black 1.23 0.61 0.10° 1.30 0.59 0.12°
Country 0.95 0.60 0.10 1.15 0.58 0.12c
Contemp Hits 1.67 0.63 0.12“ 1.92 0.60 0.16“
Oldies -0.20 0.62 -0.02 -0.29 0.59 -0.03
Classical -0.40 0.75 -0.01 0.29 0.74 0.01
Jazz -1.05 0.67 -0.05 -0.78 0.64 -0.04
Standards 0.39 0.61 0.03 0.46 0.59 0.04
Religious -0.56 0.60 -0.05 -0.22 0.58 -0.02
News/Talk 0.79 0.60 0.09 0.90 0.57 0.1 1
Hispanic -0.33 0.61 -0.03 0.06 0.59 0.01
Number of Mkt Stations -0.10 0.01 -0.20“ -0.10 0.01 -0.21"
Arbitrom Mkt Population 0.00 0.00 0.00 0.00 0.00 0.02
Mkt Per Capita Income 0.00 0.00 -0.07° 0.00 0.00 —0.07°
% College Education 0.00 0.01 0.00 0.00 0.01 0.00
% Afi'ican American 0.01 0.01 0.02 0.00 0.01 0.01
% Hispanic 0.00 0.00 0.00 0.00 0.00 0.00

Adj R“ = .34 Adj R“ = .36
Model Statistics F (28. 266 7) = 49.42‘1 F (28, 2686) = 54.443

D-W=1.87 D-W=1.97
“p<.001 “p<.01 °p<.05

81

TABLE 19

Regression Results for Equation (12) Estimating the Effects of Group Size (spline

variables) and Other Variables on Station Audience Share in 1999 and 2000

 

 

 

 

 

 

1999 2000

Variables B SE B Beta B SE B Beta
(Constant) 5.94 0.64 6.03 0.63
2 Mkt Station Grp Dummy 0.54 0.08 0.473 0.58 0.08 0.53”
3 Mkt Station Grp Dummy 0.17 0.07 0.17° 0.03 0.07 0.03
4 Mkt Station Grp Dummy -015 0.05 -016“ -004 0.05 -004
5 Mkt Station Grp Dummy -0.09 0.04 -0.10° -0.14 0.04 -0.15b
6 Mkt Station Grp Dummy -0.04 0.04 -0.04 -0.03 0.04 -0.03
7 Mia Station Grp Dummy -012 0.03 -011“ -0.09 0.03 -008“
8 Mkt Station Grp Dummy 0.02 0.04 0.01 -0.04 0.03 -0.03
9+ Mkt Station Grp Dummy -0.10 0.05 -0.04 -0.13 0.06 -0.04°
% of FM Stations in Group -0.01 0.00 -0.10" -0.01 0.00 -0.09a
AM Freq Band -2.24 0.19 -0.36" -2.10 0.19 034“
Adult Contemp 0.76 0.50 0.08 0.55 0.51 0.06
Rock 0.53 0.50 0.06 0.24 0.50 0.03
Black 1.22 0.50 0.11° 0.80 0.51 0.08
Country 1.21 0.50 0.13° 0.88 0.50 0.10
Contemp Hits 2.16 0.51 0.19a 1.83 0.52 0.17“
Oldies -0.03 0.51 0.00 -0.18 0.52 -0.02
Classical 0.41 0.66 0.01 0.19 0.67 0.01
Jazz 075 0.57 -0.04 -0.61 0.59 -0.03
Standards 0.48 0.51 0.04 0.33 0.51 0.03
Religious -0.09 0.50 -0.01 —0.52 0.50 -0.05
News/Talk 0.90 0.49 0.1 1 0.44 0.49 0.06
Hispanic -0.01 0.50 0.00 —0.27 0.50 -0.03
Number of Mkt Stations -0.09 0.01 -0.20" -0.10 0.01 -0.21°
Arbitrom Mkt Population 0.00 0.00 0.01 0.00 0.00 0.01
Mkt Per Capita Income 0.00 0.00 -0.05 0.00 0.00 -0.05
% College Education -0.01 0.01 -0.03 -0.01 0.01 -0.02
% African American 0.01 0.00 0.03 0.01 0.00 0.03
% Hispanic 0.00 0.00 0.01 0.00 0.00 0.00

Ade“ = .35 Adj R“: .35
Model Statistics F(28.2711) = 53.883 F (28,2 71 7) = 53.54a

D-W=1.93 D-W=1.91
“p<.001 “p<.01 °p<.05

82

Format, Broadcast and Ownership Consistency Effects on Station Performance

Equation (13) examines the effects of the length of time since a station began
broadcasting in the market, the length of time a station has broadcast its current format,
and the length of time a station has been owned by its current owner, on station audience
share. Two regressions using one of the two ownership concentration variables (Group
Mkt Share or Audience HHI) were run. The Group Mkt Share ownership concentration
variable sums the group-owner audience shares of all market owners with more than one
station, and the Audience HHI ownership concentration variable is an HHI for all owner
audience shares, or a sum of squares for all station owner audience shares in the market.
Both regressions are statistically significant with an Adjusted R2 of .51 (See Table 20 on
page 90). The Durbin-Watson statistics for both equations do not indicate any
autocorrelation or heteroskedasticity problems.

The number of years broadcasting in the market and number of years broadcasting
the same format were both significant positive predictors of station audience share,
regardless of the ownership concentration variable used in the equation. This supports the
assumption that it takes time for an audience to find a station and that it takes time for a
station to build audience loyalty, all else equal. The number of years with the same owner
was not a significant predictor of station audience share. Because ownership is not
thought to be a salient station characteristic with listeners, it makes sense that this does
not appear to affect station audience Share. It does not matter who owns a station; as long
as this station has been broadcasting in the market with the same format over a longer
period of time, the station has a better chance of attracting a larger audience share, all else

equal.

83

The number of other stations in the group and group audience share for other
stations in the group (Adjusted Group Share) are added to equation (13) to control for
group size effects on station audience share. The number of stations in the group has a
statistically significant and positive effect on station audience share, all else equal. This
result is consistent with the Group-Ownership Effect regressions (equations (6), ( 7), and
(12)). However, the Adjusted Group Share variable for group Size does not statistically
affect station audience share. All else equal, adding more stations to the group increases
station audience share, but increasing the audience share of the other stations in the group
does not affect station audience share.

Controlling for all variables in the equation, AM station band, market population
and number of market stations are statistically significant negative predictors of station
audience share, regardless of the ownership concentration variable used in the equation.
In this set of regressions, as with the Group-Ownership Effect regressions (equations (6),
( 7), and (12)), FM stations have larger station audience shares than AM stations, all else
equal. These results provide further support for the assertions that F M stations are more
popular with radio audiences than AM stations. As the number of market stations
increases, station audience share decreases, all else equal. As the number of market
stations increases, the station audience share should decrease as market radio listeners are
divided among a greater number of stations. As the population of people over the age of
twelve years in a market increases, station audience share decreases, all else equal. All
else equal, the format of the station does not have a significant effect on station audience
share. The market population characteristic variables and the ownership concentration

variables are not significant predictors of station audience share in these equations.

84

TABLE 20

Regression Results for Equation (13) examining the effects of the length of time a station
has broadcast in a market, has broadcast the same format and has been owned by the
current owner on station audience share in 2000

 

 

 

 

 

Group Mkt Share Audience HHI

Variables B SE Beta B SE Beta
(Constant) 2.27 1.98 2.90 1.90

Years Broadcasting in Mkt 0.03 0.01 0.243 0.03 0.01 0.248
Years with Current Format 0.21 0.03 0.29’1 0.21 0.03 029"
Years with Current Owner 0.01 0.01 0.04 0.01 0.01 0.04
No. ofOther Grp Stations 0.37 0.13 0.29“ 0.37 0.13 0.29“
Adjusted Owner Share -0.02 0.03 -0.08 -0.02 0.03 -0.07
AM Freq Band -2.54 0.34 -0.46a -2.53 0.34 -0.46°
Adult Contemp -0.16 1.41 -0.02 -0.21 1.41 -0.03
Rock 0.26 1.41 0.03 0.22 1.41 0.03
Black 0.80 1.39 0.10 0.74 1.39 0.09
Country -0.85 1.40 -0.1 1 -0.91 1.40 -0.1 l
Contemp Hits 1.43 1.43 0.14 1.38 1.43 0.14
Oldies -0.75 1.44 -0.07 -0.80 1.44 -0.07
Classical -1 . 17 1.76 -0.04 -1.28 1.76 -0.04
Jazz -0.82 1.55 -0.04 -0.85 1.55 -0.04
Standards 0.06 1.40 0.00 0.00 1 .40 0.00
Religious -1.21 1.38 -0.14 -1.28 1.38 -0.15
News/Talk 0.01 1.36 0.00 -0.05 1.35 -0.01
Hispanic —0.02 1.40 0.00 -0.08 1.39 -0.01
Number okat Stations -012 0.04 -024“ -0.1 1 0.03 -022“
Arbitrom Mkt Population 0.00 0.00 -0.13° 0.00 0.00 -0.14b
Mkt Per Capita Income 0.00 0.00 0.01 0.00 0.00 0.01
% College Education 0.00 0.04 0.00 0.00 0.04 0.00
% African American 0.02 0.01 0.06 0.02 0.01 0.07
% Hispanic 0.01 0.01 0.04 0.01 0.01 0.04
Ownership Concentration 0.01 0.02 0.03 0.00 0.00 0.02

Ade“=.51 Ade“=.51

Model Statistics

F(25,338) = 16.151’

D-W= 2.04

F(25,338)=16.11a

D—W= 2.04

 

ap<.001 bp<.01

°p<.05

85

CHAPTER 5

CONCLUSIONS

This study supports Rogers & Woodbury’s (1996) hypotheses “(1) that the number
of formats in a radio market is positively related to the number of stations, and (ii) that
listenership or audience is positively related to the number formats” (p. 83). However,
when the number of market stations is added to equation (4), support for the second
hypothesis changes. All else equal, radio audience (percentage of the market population
listening to radio) is positively related to the number of market stations with statistical
significance but is not affected by the number of market formats. This indicates that that
diversity of station choices may be more significant to increasing radio listeners than
increasing diversity of formats. Radio audience may expand from a larger selection of
stations regardless of the number formats. This assertion is further supported by the
results of the ‘within format’ effect equation (5) results. As more stations broadcast
within a particular format, most format audiences under investigation also expand (with
the exception of the Black and Hispanic format audiences). Having more choice within a
format seems to attract more listeners to the format, in most cases. These results question
the viability of diversity of formats as a measure of the public interest, if the public
interest is defined as the percentage of the market population that finds satisfying
programming from radio.

Rogers & Woodbury’s (1996) further argued that increases in ownership
concentration would not affect these relationships, but this replication did not find

conclusive support for this argument. The ownership concentration variables do not have

86

consistent statistical significance in the equations used to replicate Rogers & Woodbury
(1996), but were significant in a few of the regressions. Increases in the ownership
concentration had a statistically positive effect on the number of formats in a market in
some of the regressions. However, increases in the ownership concentration had a
negative effect on the percentage of market population listening to radio in some of the
regressions. Increases in the ownership concentration may have a positive impact on the
number of formats present in a market, but increases in ownership concentration may be
driving radio audience away from radio. Further research should investigate the effect of
ownership concentration on the number of formats and the size of market radio
audiences, since this investigation did not yield clear results.

Rogers & Woodbury (1996) find a statistically significant positive relationship
between the number of radio listeners and the number of market formats, yet while
positive, this relationship is not statistically significant for all years in this replication and
was not significant for either of the last two sample years. This may be an indication that
the responsiveness of the number of market formats to changes in the Size of the market
audience may be disappearing. This could mean that smaller markets (by population)
may be gaining the same format diversity as larger markets. Previous research found
weak statistical significance for an increase in format diversity in smaller markets after
the passage of the Telecommunications Act of 1996 (FCC, 2001; Williams & Roberts,
2002). Further study on a later time period should be conducted on the responsiveness of
market format diversity to changes in the radio audience to ascertain if these results are

indicating the beginning of a trend or are unique to these observation years.

87

Rogers & Woodbury (1996) do not report a statistically significant relationship
between the number of stations broadcasting a particular format and the percentage of
market audience listening to this format. However, this replication of their within format
effect investigation finds format audience size to be responsive to the number of stations
broadcasting the format. These results counter program choice theory assumptions and
support Napoli’s (1997). According to Napoli’s (1997) results, there should be a high
correlation between the percentages of radio formats available and the percentages of the
radio audience that choose to listen the different formats. As more stations are
broadcasting a particular format, more listeners should report listening to the format. This
argument assumes that every station broadcasting the same format is to some degree
differentiated from every other station broadcasting the format. Format audiences are not
statistically affected by the number of other formats present in a market. These findings
suggest that stations within the same format category are not perfect substitutes.

Future research on format fragmentation and the expansion of radio format
audiences should be conducted. Although it is difficult to get a comprehensive record of
station content to compare and contrast differences in station programming, it appears
that radio audiences find a degree of difference between stations under the same format
category. These differences also appear to be expanding the fOrmat audiences.
Furthermore, increasing the number of stations in the market appears to increase the
percentage of market population listening to radio more significantly than increasing the
number of formats. Future research on format diversity should also consider the number
of stations broadcasting the formats and not just the number of formats present in the

market.

88

The results of this study support a station group efficiency hypothesis with respect
to station performance. As group-owners add more stations to their groups, the group-
owners appear to realize efficiencies that increase station audience Shares. The results of
this study indicate that local station consolidation does appear to provide radio audiences
with more satisfying programming; however the increases in station share decrease in
magnitude with each additional station. Therefore, this investigation finds audience
benefits from station consolidation. Further research should consider other possible
measures of group size and station performance to see if an efficiency hypothesis can be
found for other measures of group ownership and station performance.

One limitation of this group efficiency investigation is that station profit data is
not considered; therefore financial economic efficiencies cannot be explored.
Unfortunately, there are sources for station revenue available, but not station costs.
Station performance is approximated by station audience, which has been found to be
highly correlated with station revenue (Berry & Waldfogel, 1999). However this does not
indicate a correlation with station profit. If it costs more to produce quality station
programming to attract larger audiences, then increasing station revenue with a larger
audience is not necessarily increasing profits. A great deal of radio research assumes
constant programming prices and/or advertising rates (e. g. Berry & Waldfogel, 1999;
Drushel, 1998; Ekelund et al., 2000; Steiner, 1952), because the data to relax these
assumptions are not available. Despite these limitations, if one assumes that controlling
for station frequency band and format controls for some differences in program cost and

advertiser valuation of station audiences, then this investigation does support the

89

hypothesis that, all else equal, adding more stations to a group improves station
performance.

Future research should be conducted on station data from after 2000 to see if the
shape of the relationship between group size and station audience share changes form the
results found by this investigation. Station groups may be able to further increase station
performance efficiencies as more time passes since the significant increase in station
consolidation after the passage of the Telecommunications Act of 1996. Five years may
not be enough time for station groups to realize the full potential of group ownership
efficiencies.

Regardless of ownership consolidation, the number of years a station has served
its market and the number of years a station has been broadcasting its current format are
positive predictors of station audience share. The length of time a station remains with
the same owner does not appear to affect station audience share. Therefore it does not
matter how long a station is owned by the same owner. What is more important to
increasing audience share is that the station broadcast in the market over a longer period
of time and that the station’s format remain consistent over a longer period of time
compared to other stations in the market.

Stations new to a market and stations that change their formats clearly have a
disadvantage in attracting radio listeners and it will take time for these stations to gain a
larger audience. Radio stations that gain audience loyalty are more likely to keep
audience loyalty over time if they do not change their formats, all else equal. Stations that
change their formats will suffer Significant loss of audience that can be recouped with

format consistency over time, which is consistent with the finding of Romeo & Dick

90

 

(2001). The results of this study also indicate that FM stations have an advantage over
AM stations when it comes to audience share. Ownership consolidation has created group
owner efficiencies that increase station audience, but station programming consistency is

also important to increasing station audience share, all else equal.

91

 

REFERENCES

Clearly, it's clear channel. (2000, September 18). Broadcasting & Cable 130. no, 130. 50-
62.

Albarran, A. B., & Pitts, G. G. (2001). The radio broadcasting industry.Boston: Allyn &
Bacon.

Arbitron Inc. (2002). Arbitron radio market report reference guide. Columbia, MD:
Arbitron Inc.

Aufderheide, P. (1999). Communications policy and the public interest. New York: The
Guildford Press.

Bates, B. J. (1993). Station trafficking in radio: The impact of deregulation. Journal of
broadcasting & electronic media. 3 7(1), 21-30.

Beebe, J. H. (1977). Institutional structure and program choices in television markets.
Quarterly journal of economics, 91(1), 12-37.

Berry, S. T., & Waldfogel, J. (1999). Free entry and social inefficiency in radio
broadcasting. Rand journal of economics, 30(3), 397-420.

Berry, S. T., & Waldfogel, J. (2001). Do mergers increase product variety? Evidence
from radio broadcasting. Quarterly journal of economics, 116(3), 1009-1025.

Broadcasting & Cable. (1995-2000). Broadcasting & cable yearbook. New Providence,
NJ: R. R. Bowker Publication.

Chambers, T. (2001). Losing owners: Deregulation and small radio markets. Journal of
Radio Studies, 8(2), 292-315.

Chang, B.-H., & Ki, E.-J. (2004). A longitudinal analysis of mergers and acquisitions
patterns of radio companies in the US. Journal of Radio Studies, 11(2), 194-208.

92

Compaine, B. M. (2000). Distinguishing between concentration and competition. In B.
M. Compaine & D. Gomery (Eds), Who owns the media? Competition and
concentration in the mass media industry (3rd ed., pp. 537-581). Mahwah, NJ:
Lawrence Erlbaum Associates.

Ditingo, V. M. (1995). The remaking of radio. Boston: Focal Press.

Drushel, B. E. (1998). The telecommunications act of 1996 and radio market structure.
Journal of media economics. 11(3), 3-20.

Duncan, J. H. (1995-2000). American radio report: Spring ratings report. Kalamazoo,
MI: Duncan Media Enterprises.

Duncan, J. H. (1996-2001). Duncan ’3 radio market guide. Kalamazoo, MI: Duncan
Media Enterprises.

Ekelund, R. B., Ford, G. S., & Koutsky, T. (2000). Market power in radio markets: An
empirical analysis of local and national concentration. Journal of law and
economics, 43(1), 157-184.

FCC. (2000). Biennial review report: Review of the commission 's broadcast ownership
rules and other rules adopted pursuant to section 202 of the telecommunications
act of 1996 (No. MM Docket No. 98-35). Washington, DC: Government
Printing Office.

FCC. (2001). Review of the radio industry, 2000.Washington, D.C.: Mass Media Bureau
Policy and Rules Division.

FCC. (2003). Report and order and notice of proposed rulemaking in the matter of 2002
biennial regulatory review — review of the commission ’s broadcast ownership
rules and other rules adopted pursuant to section 202 of the telecommunications
act of I 996, cross-ownership of broadcast stations and newspapers, rules and
policies concerning multiple ownership of radio broadcast stations in local
markets, definition of radio markets, and definition of radio markets for areas not
located in an Arbitron survey area (No. FCC 03-127). Washington, DC: Media
Bureau.

Glasser, T. L. (1984). Competition and diversity among radio formats: Legal and
structural issues. Journal of broadcasting, 28(2), 127-142.

93

Rogers, R. P., & Woodbury, J. R. (1996). Market structure, program diversity, and radio
audience size. Contemporary economic policy, 14, 81-91.

Romeo, C. J., & Dick, A. R. (2001). The eflect of format changes and ownership
consolidation on radio station outcomes. Washington, DC: Department of Justice
Economic Analysis Group.

Spence, A. M., & Owen, B. M. (1977). Television programming. Monopolistic
competition and welfare. Quarterly journal of economics, 91, 103-126.

Steiner, P. O. (1952). Program patterns and preferences, and the workability of
competition in radio broadcasting. Quarterly journal of economics, 66(2), 194-
223.

Suits, D. 8., Mason, A., & Chan, L. (1978). Spline functions fitted by standard regression
methods. The review of economics and statistics, 60(1), 132-139.

Wildman, S. S., & Cameron, D. J. (1982). Competition, regulation and sources of market
power in the radio industry (Unpublished manuscript). Los Angeles, CA:
University of California Los Angeles.

Williams, 0., Brown, K., & Alexander, P. (2002). Radio market structure and music
diversity. Washington, DC: Federal Communications Commission.

Williams, G., & Roberts, S. (2002). Radio industry review 2002: Trends in ownership,
format, and finance. Washington, DC: Federal Communications Commission.

Wirth, T. L. (2001). Nationwide format oligopolies. Journal of radio studies, 8(2), 249-
270.

Wooldridge, J. M. (2003). Introductory econometrics: A modern approach. Mason, OH:
South-Westem.

95

11111111311111

1111111

4209