Skip Navigation

Federal Communications Commission

English Display Options

Commission Document

Media Ownership Study 8a-Revised Study

Download Options

Released: July 27, 2011


Local Media Ownership and Viewpoint Diversity in Local Television News



June 2011

Adam D. Rennhoff
Assistant Professor, Jones College of Business
Middle Tennessee State University
P.O. Box 27
Murfreesboro, TN 37132
615-898-2931
rennhoff@mtsu.edu

Kenneth C. Wilbur
Assistant Professor, Fuqua School of Business
Duke University
100 Fuqua Drive, Box 90120
Durham, North Carolina 27708-0210
919-926-8536
Kenneth.Wilbur@Duke.edu
http://kennethcwilbur.com/">Http://Kennethcwilbur.com



This study was funded by the US Federal Communications Commission as part of its
quadrennial media ownership review. We are indebted to Jonathan Levy, Tracy Waldon and
Jessica Almond for providing data and many helpful comments. Wagner Kamakura and Carl
Mela offered valuable suggestions. Any remaining errors are our own.



Executive Summary

This study proposes a theory-driven, market-based measure of viewpoint diversity in local
television news. It then calculates this viewpoint diversity metric using a panel dataset of local
television ratings. Finally, an econometric model is used to determine whether viewpoint
diversity is associated with local media market ownership structure. The estimated elasticities of
viewpoint diversity with respect to media ownership variables are very close to zero.

Introduction

This study was written for the United States Federal Communications Commission (“FCC”) as
part of its 2010 Quadrennial Review of Media Ownership Rules. Its purpose is two-fold.
First, it proposes a new market-based measure of viewpoint diversity in local television
news programs. The market-based approach is desirable because it relies on consumers’ actions
to define viewpoint diversity. However, it is complicated by the fact that consumers’ actions
depend on consumer preferences as well as media content. The key to the approach is to use
local viewing of national news programs to learn about local market preferences. This allows
local media content to be distinguished from local market preferences.
Second, the analysis uses a descriptive regression to relate the proposed viewpoint
diversity index to local media cross-ownership, co-ownership and ownership diversity. The
associations between viewpoint diversity and ownership variables are all found to be very close
to zero.
Section 1.1 gives a brief overview of the Media Ownership Rules that the FCC is
currently reviewing. Section 1.2 discusses the tortured legal history of viewpoint diversity, so
that the reader may understand the proposed definition in context. Section 2 defines the proposed
measure of viewpoint diversity. Section 3 presents the empirical approach of relating viewpoint
diversity to local media market ownership structure, controlling for time-invariant market
characteristics. Section 4 contains the estimation results and section 5 concludes by relating the
empirical results to the rules the FCC is reviewing.

1.1. Media Ownership Rules
Three media ownership rules are relevant to the present analysis. This section gives just a brief
overview of the rules. FCC (2010) ahttp://edocket.access.gpo.gov/cfr_2003/octqtr/pdf/47cfr73.3555.pdf">nd 47 CFR 73.3555 are more expansive.
1


Newspaper/Broadcast Cross-Ownership Rule

: Since 1975, the FCC has restricted the
common ownership of a broadcast station and a newspaper when, roughly speaking, the
station’s footprint contains the newspaper’s distribution area. Waivers to this rule may be
granted when common ownership is judged to be aligned with the public interest. In
2007, the waiver criteria were relaxed so that common ownership would be presumed to
be not inconsistent with the public interest in the 20 largest media markets, so long as the
TV station is not among the four largest in the market and there would be at least eight
post-merger “voices” available in the market.1 Common ownership is still presumed to be
inconsistent with the public interest in smaller media markets unless (1) one of the two
media outlets were “failed” or “failing,” or (2) the joint entity would significantly
increase the amount of news available in the market.

Local TV Ownership Limit

: One entity may own two television stations within the same
market if (1) their signals do not overlap (this case is rare), or (2) one of the stations is not
ranked in the top four stations in the market based on market share, and there are at least
eight independently-owned stations in the market. This second provision essentially rules
out dual station ownership in smaller markets, as they are typically served by fewer than
eight stations.

Local Radio/TV Cross-Ownership Rule

: In markets with at least 20 independently-owned
voices, one entity may own one TV station and up to seven radio stations or two TV
stations and up to six radio stations, subject to the Local TV Ownership Limit. In markets
with 10-19 independently-owned voices, one entity may own up to two TV stations and
up to four radio stations. In markets with 9 or fewer independently-owned voices, an
entity that owns a TV station may not own more than one radio station.

1.2. Viewpoint Diversity
The FCC’s policy objectives are competition, localism and diversity. The FCC’s diversity
policymaking objective is nuanced and sometimes controversial. It is motivated by the
observation that, since the public owns the airwaves on which television and radio signals are
broadcast, the media should serve all segments of the population. The need for regulation is
implied by the well known result that some types of content may be underprovided by a

1 A “voice” may be a TV station, radio station, newspaper or a cable system.
2


competitive market. As the US Supreme Court noted in AP v. United States, “[the First]
Amendment rests on the assumption that the widest possible dissemination of information from
diverse and antagonistic sources is essential to the welfare of the public, that a free press is a
condition of a free society.” The Court justified media ownership regulations to preserve this
freedom, saying “freedom to publish is guaranteed by the Constitution, but freedom to combine
to keep others from publishing is not.”http://supreme.justia.com/us/326/1/case.html"> (326 U. S. 1)
The FCC has operationalized its diversity objective in five ways (FCC 2010):

Outlet diversity

is the number of independently-owned media outlets.

Source diversity

is the availability of media content from a variety of content creators.

Minority and female ownership diversity

is the number of media outlets owned by minority
race/ethnic groups and women.

Program diversity

is the variety of program formats and content provided by the media.

Viewpoint diversity

is the availability of content reflecting a variety of perspectives.

The first three definitions reflect the concept of source diversity, that is, increasing the
number of voices available in the media market. Source diversity is fairly straightforward to
define and measure. The final two definitions reflect the concept of content diversity, that is,
increasing the number of types of programs and opinions that are available in the media market.
Source diversity has sometimes been seen as a standalone policy objective, and it has sometimes
been seen as a means to achieve content diversity.

The purpose of the present analysis is to determine whether media co-ownership, cross-
ownership and ownership diversity within a market are associated with viewpoint diversity in
that market’s television news. It is emphasized that the analysis seeks to examine viewpoint
diversity, not program diversity.

Empirical analysis of viewpoint diversity requires a measure of viewpoint diversity. The
“availability of content” component of the definition is fairly straightforward but a “variety of
perspectives” component is not. What qualifies as a “perspective?” And what constitutes a
“variety” of perspectives?

The current paper is not the first to grapple with the question of how to define viewpoint
diversity. The Newspaper/Broadcast Cross-Ownership Rule was challenged shortly after its
passage in 1975. The Supreme Court upheld the rule, noting that “the regulations, which are
designed to promote diversity of mass media as a whole, are based on public interest goals that
3


the FCC is authorized to pursue.” The court went further to note that “diversity and its
effects…are elusive concepts, not easily defined let alone measured without making quality
judgments that are objectionable on both policy and First Amendment grounds.http://supreme.justia.com/us/436/775/case.html">” (436 U. S. 775;
emphasis added) As McCann (2010) put it, “In other words, the court didn’t require the FCC to
specifically define viewpoint diversity, [it] instead relied on the FCC’s rational judgment based
on experiences.”

In 2003, the FCC relaxed its ownership rules substantially. It eliminated cross-media
ownership regulations in media markets with eight or more television stations, and allowed
newspaper/television/radio cross-ownership in media markets served by four to eight television
stations. This action was justified by an analysis based on the “Diversity Index,” which sought to
measure viewpoint diversity in a manner inspired by the Herfindahl-Hirschmann Index (HHI)
that antitrust authorities use to gauge market competitiveness. The Diversity Index used
consumers’ average time spent with each medium to weight its importance. It then assigned
equal “market shares” to each outlet within each medium and combined those “market shares”
for commonly owned outlets. For example, New York was served by 23 television stations, so
each television station was assigned a “market share” of 4.3% (or 1/23). Finally, based on these
weights and “market shares,” the Diversity Index was calculated using a sum-of-squares
approach similar to the HHI.

The ownership rule relaxation was challenged in court immediately and quickly
overturned. In Prometheus Radio Project vs. FCC, the 3rd Circuit Court was emphatic on its view
of the Diversity Index. It ruled that “the Commission did not justify its choice and weight of
specific media outlets.” Further, “the Commission did not justify its assumption of equal market
shares.” And, “the Commission did not rationally derive its Cross-Media Limits from the
Diversity Index results.” http://law.justia.com/cases/federal/appellate-courts/F3/373/372/474282/">(373 F.3rd 372)

The proposed definition of viewpoint diversity in this paper should be understood in light
of these past difficulties. When this concept was used to justify media ownership restrictions in
the 1970’s, it was not precisely defined. The FCC’s one attempt to measure this concept in 2003
was rejected expeditiously.

Despite these difficulties, it is important to try to measure important policymaking
criteria. Unmeasurable policy objectives lead to inevaluable policies. The next section
undertakes this challenge.
4



2. A Market-Based Measure of Viewpoint Diversity
This section proposes a market-based measure of viewpoint diversity. Section 2.1 explains the
use of a market-based measure. Section 2.2 explores some intuitive properties that any
reasonable measure should exhibit, and it shows the difficulty of separating the viewpoint
diversity expressed in the media from the preferences exhibited by the audience of the media.
Section 2.3 defines the proposed measure formally, and section 2.4 shows how viewership for
national news programs can be used to separate local preferences from local news program
characteristics. Section 2.5 discusses the limitations of the proposed definition.

2.1. Basis for Measurement
In considering the question of how to measure the variety of perspectives offered among a set of
media programs, one might quite naturally start by thinking about conducting a content analysis.
For example, one could use computers or human coders to analyze samples of media content and
encode the perspectives expressed in each sample.

While intuitive, such content-based approaches to diversity measurements face three
difficulties. First, accurate content quantification is quite difficult. Human collection of content
data is typically labor-intensive and subjective, and therefore may be costly, slow or inaccurate.
Computer collection of content data can be performed quickly but may fail to capture aspects of
the content which are important but difficult to quantify. Second, time and cost constraints force
the researcher to decide which aspects of content to encode, and those decisions may be at odds
with the aspects of content that actually matter to consumers. Third, measures which are based
solely on media content cannot predict how different audiences would react to the same content.

Consider a thought experiment to illustrate this final point. Suppose there are two
subjective issues, 1 and 2, and two markets, A and B. Suppose everyone in market A is interested
in issue 1 and everyone in market B is interested in both issue 1 and issue 2. Suppose a news
program in market A uses four minutes of program time to present four perspectives on issue 1.
Suppose a news program in market B uses two minutes to present two perspectives on issue 1
and another two minutes to present two perspectives on issue 2. A content-based measure of
viewpoint diversity might well conclude that the news program in market A exhibits greater
diversity, since more perspectives about issue 1 were expressed. However, from a policy
5


perspective, it might be argued that the two news programs served their markets equally well
given market preferences and time constraints. Yet this conclusion depends on information about
market preferences, and therefore would be very impossible to draw using a purely content-based
measurement of viewpoint diversity.

More generally, this is why viewpoint diversity has proven so difficult to define and
measure. It is subjective, depending as much on the preferences of the audience as on the
contents of the media.

The market-based approach proposed below alleviates all three problems of content-
based diversity measurements. It obviates a burdensome data collection task, eliminates the need
to predetermine what content characteristics are important, and it relies on consumers’ observed
choices which embed market-specific preferences. It should be remembered, however, that the
market-based approach is no panacea. It has limitations of its own, discussed in section 2.5.

By presenting these limitations of content analysis, it is not the authors’ intention to
diminish the validity of content-based measures of viewpoint diversity. To the contrary, content
analysis is a worthwhile and informative exercise. For example, Gentzkow and Shapiro (2010)
invented a brilliant means of avoiding the primary limitations of content analysis, text mining the
Congressional Record to identify Democratic and Republican phrases, then counting their
frequency of use in local newspaper articles. They found that media outlets’ use of political
language typically reflected their customer bases’ preferences.
The view of the authors is that policymakers and judges should consider both content-
and market-based approaches to measuring viewpoint diversity. Each type of approach should be
evaluated with a rational understanding of its strengths and weaknesses and the degree to which
those strengths and weaknesses affect the specific application of the method.

2.2. Intuitive Properties
This section presents a series of thought experiments to motivate and justify the viewpoint
diversity measure.

Suppose two competing television stations, A and B, within a market each offer a local
news program, and suppose that each station has a 50% share of the local news audience. The
following two extreme possibilities are fundamentally different but observationally equivalent.
6


1) The market’s television audience consists of two equally sized segments with polar opposite
viewpoint preferences. The observed 50/50 audience split suggests that each local news program
is tailored to one segment’s preferred viewpoint. This would be consistent with high viewpoint
diversity among the programs provided by the media market.
2) All viewers in the market have the same preferred viewpoint. The two stations both offer this
same preferred viewpoint. Since they offer essentially identical programs, they split the market
again, with half the viewers watching station A and half watching station B. This would be
consistent with low viewpoint diversity among the programs provided by the media market.

The observational equivalence of these extreme possibilities illustrates the primary
difficulty in measuring viewpoint diversity. Audience data on local news viewing alone cannot
provide a measure of viewpoint diversity, since media consumption choices are based on both
viewer preferences and media content. Since the concept of viewpoint diversity is fundamentally
subjective, its measurement must account for the preferences of the group receiving the
viewpoints.

Another thought experiment can show how this difficulty will be resolved. Suppose that
each of the two television stations offers a national news program in addition to its local news
program. Assume that the national news programs offer different viewpoints, that they air in
each of many local media markets, and that each garners a 50% rating nationwide. Now,
consider two subcases of this example.

First, suppose that the national news program on station A garners a 80% share of
viewers in a particular local market, and the national news program on station B garners a 20%
share of viewers in that market. Further, suppose that the two local news programs split the local
audience with a 50% audience share each. It is clear that, relative to the national market, the local
market has a strong, homogeneous preference for viewpoints of the type provided in station A’s
national newscast. Since the two local newscasts split the local market, their content must be
roughly similar, indicating a low level of viewpoint diversity.

In the second subcase, suppose the national news programs on stations A and B split the
local audience, each with a 50% share of local viewers. This indicates that the variety in local
consumers’ preferred viewpoints roughly matches the variety in national consumers’ preferred
viewpoints. Further, suppose the local news program on station A is watched by 80% of the
market while the local news on station B is watched by 20% of the market. This information will
7


tell us that stations A and B are providing programs that contain very different viewpoints. This
is because viewers in the local market exhibit some heterogeneity in viewpoint preferences but
the local news on station B is so far away from the preferred viewpoint of the average consumer
that few consumers watch it.

These examples convey the intuition underlying the market-based measure of viewpoint
diversity. It will weigh dispersion in local market shares for local news programs against
dispersion in local market shares for national news programs. The latter indicates the degree to
which the local market’s preferences differ from the national market, and this will distinguish
between the two observationally equivalent extreme cases discussed at the beginning of this
subsection.

2.3. A Market-Based Viewpoint Diversity Index
This subsection defines the proposed viewpoint diversity index. It shows how to recover this
index from program audience data, and shows that it cannot be empirically separated from the
dispersion in local tastes, as illustrated in the examples above.

Consider a market m that is served by three local news programs, indexed in order of
ascending market share by j = ,
1
3
,
2 . Assume that the programs are differentiated by a single
dimension of viewpoint diversity, as in Hotelling (1929). The range of possible viewpoints can
then be represented by a single horizontal line, and the viewpoint expressed by each program j in
market m may be represented by a point m
x on that line.2 Assume for simplicity that the
j
programs are ordered such that program 1 is closest to the left side of the line and program 3 is
closest to the right side of the line.

Let x represent a point on the horizontal line denoting the preferred viewpoint of viewer
i
i. These points are assumed to be distributed Normal with mean µ and variance 2
σ .3 The
m
m

2 Higher-order viewpoint spaces are not considered because the available data do not allow for the nonparametric
identification of additional dimensions of program differentiation. The single line in the model could be thought of
as the first principal component of a higher-dimensional viewpoint diversity space.
3 Below, an assumption is made that the national distribution of preferred viewpoints is Standard Normal. Neither
assumption is necessary-and-sufficient for the other to hold, but they are compatible if viewers’ preferred viewpoints
are imperfectly correlated with their locations and if the moments of the national distribution are compatible with the
moments of the market-specific distributions, for example, if the weighted sum of market-specific mean viewpoint
preferences is zero, where the weights represent the percentage of the national population contained within each
market.
8


Normal distribution is less tractable than the typical assumption of uniform preferences but is
more realistic. Consumer i gets utility m
u from watching program j ,
ij
m
u = V − |
m
x x | ,
(1)
ij
i
j
where V is the value of watching the news and m
x is the location of the viewpoint expressed in
j
the local news program j . It is assumed that each viewer watches the program whose viewpoint
is closest to her preferred viewpoint. It is also assumed that V is large enough that the market is
fully covered, i.e., that all consumers who want to watch local news watch one of the available
local news programs. This assumption is considered to be the primary limitation of the proposed
approach and is discussed in depth in section 2.5.

A useful theoretical construct is the point at which a consumer is just indifferent between
watching programs 1 and 2, m
xˆ . Setting m
m
u
= u shows that ˆm
x
= ( m
m
x + x ) / 2 . Similarly, the
12
i1
i 2
12
1
2
point of indifference between programs 2 and 3 is ˆ m
x
= ( m
m
x + x ) / 2 .
23
2
3

The proposed Viewpoint Diversity Index D is defined as the difference between these
m
two points of indifference:


m
m
D = xˆ
xˆ
(2)
m
23
12
Figure 1 provides the intuition underlying this measurement of viewpoint diversity. It shows that
the news programs in market m provide less diversity than those in market m' , since they cover
less of the line. Accordingly, the diversity index D is greater than D .
m'
m

A few remarks are made to help explain the Viewpoint Diversity Index. First, the
definition in equation (2) is proportional to the entire span of viewpoints available in the market,
m
m
x x . It is written in terms of the points of indifference because the audience shares must
3
1
sum to one, so the three audience shares in the data really provide only two degrees of freedom.
Writing the diversity index in terms of the two points of indifference makes this fact more salient
and shows that other common dispersion indices, such as a standard deviation based on the three
news program locations, are not advisable in this setting. Second, notice that it is based purely on
station locations. Local market preferences, as represented by distributional parameters µ and
m
σ , do not enter the index. Third, notice that if either station 1 or station 3 changes its location
m
in viewpoint space, the diversity index will change its value. However, it will not change with
9


small movements of station 2 (small enough that station 2 remains between stations 1 and 3),
since this would result in a reallocation of market shares without altering the range of viewpoints
provided by the marketplace. Fourth, the index is independent of the scale of available
viewpoints. That is, the index is the same for ( m
x , m
x , m
x ) =
,
1
(
)
3
,
2
as it is for
1
2
3
( m
x , m
x , m
x ) =
)
13
,
12
,
11
(
. Finally, while the Viewpoint Diversity Index will always be positive,
1
2
3
there are no benchmark values that take on special meaning.

To calculate the viewpoint diversity index, it is necessary to determine the program
locations in the viewpoint space. This may be done by relating the predicted audience shares in
the model to data. The market share of program 1 is given by the probability mass of viewers
whose preferred viewpoints lie to the left of m
xˆ ,
12
m
s = Φ(( ˆm
x − µ ) / σ ) .
(3)
1
12
m
m
where Φ is the standard normal cumulative distribution function. Similarly, the market share of
program 3 is given by the probability mass of preferred viewpoints to the right of m
xˆ ,
23
m
s = 1− Φ(( ˆm
x − µ ) / σ )
(4)
3
23
m
m
Presuming m
s and m
s are available in the data, the points of indifference can be recovered from
1
3
equations (3) and (4) as
m
−1
m
xˆ
= σ Φ (s ) + µ
(5)
12
m
1
m
m
−1
m
xˆ
= σ Φ 1
( − s ) + µ
(6)
23
m
3
m
These can be substituted into (2) to show that the empirical Viewpoint Diversity Index is
D = σ (
1

Φ 1
(
m
s )
1

− Φ ( m
s )) .
(7)
m
m
3
1
Equation (7) shows, formally, the indeterminacy between program dispersions and market-
specific tastes. With data from a single market, it will be impossible to separate the program
locations from the dispersion in market-specific tastes, σ . The next section shows how this
m
problem may be resolved using local audience shares for national news programs.

2.4. Recovering Local Preferences
This section uses local viewership of national news programs to separate local preferences from
local stations’ viewpoint diversity.
10



It is assumed that all three national news programs are available in many local markets
and are indexed with k ∈ ( ,
A B, C) in ascending order of national audience share. It is assumed
that these news programs are differentiated on the same viewpoint scale as the local news
programs. This assumption is not innocuous. If the viewpoint diversity expressed in national
news programs is of a fundamentally different nature than that expressed in local news programs,
then the approach proposed here will not work. The arguments in favor of this assumption are as
follows. First, the national news almost always immediately follows or precedes the local news.
Since the two programs’ audiences mostly overlap, attributes that the audience finds important in
one program may also be the attributes that the audience finds important in the other program.
Second, because these are two news programs, they are likely to share many characteristics in
common, such as the types of stories they cover and the possible styles or slants available in their
coverage of those stories. Third, both local and national news use some of the same publicly
available video footage for some of the stories they cover, so some of the main inputs to the two
types of programs are the same.

For simplicity, assume national news program A is closest to the left side of the line and
program C is closest to the right side of the line. Note that national news program A does not
necessarily correspond to local news program 1, and that the two positions of the national and
local news programs on a particular station need not be correlated.

To anchor the location and scale of preferences, it is assumed that the national
distribution of consumer viewpoint preferences is Standard Normal. Under these assumptions,
the locations of the indifferent viewers for national news programs in viewpoint space are given
by equations (5) and (6) as
ˆ N
1
x

= Φ ( N
s )
(8)
AB
A
ˆ N
1
x
= Φ− 1
(
N
s )
(9)
BC
C
where N
s is the fraction of all national news viewers (in all markets) tuned to the national news
k
program on network k.

Let m
s be the fraction of local news viewers in market m who watch the national news
k
program on local channel k . Since (8) and (9) pin down the points of indifference among
national news programs, equations (10) and (11) relate those locations to the local market shares
of the national news programs:
11


N
−1
m
xˆ
= σ Φ (s ) + µ
(10)
AB
m
A
m
N
−1
m
xˆ
= σ Φ 1
( − s ) + µ .
(11)
BC
m
C
m
Equations (10) and (11) now can be solved for local preference parameters:
ˆ N
x
− ˆ N
x

BC
AB
σ =

(12)
m
1

Φ 1
(
m
s )
1

− Φ ( m
s )
C
A
µ = ˆ N
1
x

−σ Φ ( m
s ) .
(13)
m
AB
m
A
Equations (8) and (9) may be substituted into (12), so that local dispersion in preferences σ is
m
defined in terms of local and national viewing shares of national news programs. This, in turn,
may be substituted into (7) so that the Viewpoint Diversity Index may be expressed purely in
terms of data on local audience shares of local news programs, local audience shares of national
news programs, and national audience shares of national news programs.

2.5. Limitations
The primary limitation of the proposed Viewpoint Diversity Index is that it excludes the idea of
“vertical differentiation” in news programming. Vertical differentiation refers to news program
attributes that all consumers like. For example, it may be that spending more money on special
effects, presenters or set design would lead to higher viewing among all consumers, regardless of
their viewpoint preferences. This extension was considered but found to be infeasible. An outline
of the reasons is given.

First, consider how the diversity statistic in equation (2) is calculated. Two degrees of
freedom in national viewership of national news programs are used to pin down the two points of
indifference between the three national news programs. These two points of indifference are
used, in conjunction with the two degrees of freedom available in local viewership of national
news programs, to pin down two moments of the distribution of local viewpoint preferences.
Finally, all of these inferences are used along with the two degrees of freedom available in local
viewership data of local news programs, to pin down the two points of indifference between the
three local news programs provided in each media market.

In the previous paragraph, it was assumed at every step that each news viewership market
was fully covered. This is why three audience datapoints can pin down two points of
indifference. When the assumption of full coverage is dropped, two things happen. One change
12


is positive from the standpoint of the analysis: an additional degree of freedom is acquired, since
the market share of the “outside option” (not watching television news) may be used in the
analysis. There are now three degrees of freedom, not two. The other change is negative from the
standpoint of the analysis: there are now four parameters to be pinned down, not two. It is still
necessary to pin down the points of indifference among the three news programs, as before. But
it is also necessary to pin down the ranges of unserved viewers on each end of the market.

Figure 2 illustrates this. Viewers to the left of N
xˆ do not watch news, and viewers to the
01
right of N
xˆ do not watch news. However, the data on the market share of the outside option do
30
not distinguish between these two groups.

It would be possible to pin down the fourth point of indifference if an additional
assumption were added to the framework. For example, if it were assumed that the national news
programs have positions that leave symmetric tails of unserved viewers, then the same number of
viewers would lie to the left of N
xˆ as to the right of N
xˆ . This would reduce the number of
01
30
locations to be pinned down from four to three, a feasible task given the three available degrees
of freedom. Or, if it were assumed that the three national news programs were evenly spaced on
the line, then there would only be three locations to pin down. However, both of these
assumptions are at odds with the motivation to undertake the analysis in the first place.

The measure of Viewpoint Diversity has other more obvious limitations. It assumes that
distributions of viewpoint preferences are Normal; assuming a different distribution function
may alter the results. It assumes that programs are differentiated on a single dimension, which
may be overly simple. It assumes that viewers know the locations of each available station.

While the proposed definition of a Viewpoint Diversity Index is far from perfect, it does
seem better than what has been done before, since it may be objectively measured, it separates
viewer preferences from program content, and its underlying assumptions may be clearly
evaluated. The next section shows how the new index is constructed and analyzed using data.

3. Empirical Approach
This section describes the empirical model, estimation and data used to link the Viewpoint
Diversity Index to media market ownership.

13


3.1. Model and Estimation
The model links the Viewpoint Diversity Index to media ownership variables. The model is
designed to fit the available data, which is characterized by the “large N, small T” property
common to many survey panel datasets.
The approach is to estimate a descriptive regression since viewpoint diversity and media
ownership may be driven by common factors. If one adopts the assumption that media ownership
drives viewpoint diversity, a position that has sometimes been taken by the courts, then the
empirical results may be interpreted as causal. However, the analysis here is more cautious and
does not seek to attach causal inferences to the empirical results.
D represents the Viewpoint Diversity Index in media market m at time t
,
1
,
0
{
}
2
mt
(corresponding to 2005, 2007 and 2009). It is constructed from the available viewing data as
presented in section 2.4. x is the vector of ownership variables; variable selection and
mt
definitions are discussed in section 3.3. It is assumed that
ln D
= α +α + x β + ε ,
(14)
mt
m
t
mt
mt
where α represents all market characteristics that may influence the viewpoint diversity
m
provided by the media market, α is a time fixed effect, β is a parameter vector to be estimated
t
and the object of primary interest, and ε captures idiosyncratic shocks that vary across markets
mt
and time periods. The log transformation is used so that parameter estimates may be interpreted
as percentage changes in the viewpoint diversity index. Equation (14) should be thought of as a
moving-average representation that likely includes serial correlation in ε . If the precise form of
mt
the serial correlation were known, equation (14) could equivalently be expressed as an auto-
regressive model with lags of the dependent variable appearing as regressors on the right-hand
side.

The market-specific intercepts, α , in equation (14) are likely to be correlated with the
m
media ownership variables. The panel is too short to estimate these intercepts precisely, so two
standard approaches to estimation, first differencing (FD) and fixed effects (FE), are employed
so that they drop out of the estimating equations. The FD approach lags the dependent variable to
transformation equation (14) into

(ln D
− ln D
) = (α − α
) + (x
x
)β + (ε
− ε
) ,
(15)
mt
mt 1

t
t 1

mt
mt 1

mt
mt 1

14


The FE approach drop time-invariant terms, changing equation (14) into

(ln D
− ln D ) = (α −α ) + (x x )β + (ε − ε ) ,
(16)
mt
m
t
mt
m
mt
m

T

T

T

T
where ln D = T 1
ln D
, α = T 1 ∑ α , x = T 1
x
, and ε
= T 1
ε . When the
m

m

m
t
mt
t
t
t
mt
t
mt
sample contains exactly two time periods, FD and FE provide identical parameter estimates.
When the sample contains more than two time periods, they provide different sets of estimates
and both are provided. FD is more efficient when ε follows a random walk while FE is more
mt
efficient when ε is serially uncorrelated (Wooldridge 2010). Given the likelihood of habit
mt
formation in media usage, FD estimates will be preferred to FE estimates. However, we present
both types of estimates to facilitate comparison.

Two sets of standard errors are presented for each of equations (15) and (16). The
common approach would be to apply Ordinary Least-Squares (OLS) regression to equations (15)
and (16). This is commonly known as the “differences-in-differences” estimate in the case of
equation (15) and the “pooled OLS” estimator in the case of equation (16).

The problem with the OLS approach is that, when serial correlation is present in the
errors, the standard errors of the parameter estimates may be severely biased. This has been
known since Cochrane and Orcutt (1949). Recently, Bertrand, Duflo and Mullainathan (2004)
explored the extent to which this issue affects policy-oriented econometric research. They
generated random treatments in their data and estimated the effects of these “placebo laws” on
female wages. They found that 45% of the placebo treatments’ parameter estimates were
statistically significant at the 95% confidence level. This is quite strong evidence against OLS
estimation of equations (15) and (16). Yet while OLS is not viewed as a desirable model in the
current setting, it is presented in section 4 as a familiar benchmark.

Bertrand, Duflo, and Mullainathan (2004, §IV.E) advocate using clustered standard
errors, showing that this alternative to OLS performs about as well as nonparametric estimation
in monte carlo simulations. The second set of estimates presented below follows this advice. This
allows for autocorrelation in the errors and uses an unstructured “sandwich” estimator to control
for possible correlation among the error terms, as in Arellano (1987).

A word is in order about an estimation technique that is not used. The recent dynamic
panel estimation literature (e.g., Arellano and Bond 1991) has advocated using lags and previous
levels as instruments for endogenous variables. In our application, that would imply using
15


(x
x
) as an instrument for x
and assuming that (x
x
) is uncorrelated with ε .
mt
mt 1

mt 1

mt
mt 1

mt
This exogeneity assumption is problematic in the context of media stations, as it would be in
most industrial organization settings. The valuation of a media outlet such as a television station
or a newspaper is typically calculated as the discounted sum of the station’s future earnings, and
this value influences media outlet’s price. The exogeneity assumption required by the
Arellano/Bond approach would imply that media station owners and potential buyers are either
unable to foresee future market-specific shocks to viewpoint diversity, or that they disregard
those shocks in their media station retention/acquisition decisions. This assumption is not
testable and not considered to be credible. This is the primary reason why this paper takes a
descriptive approach rather than claiming to infer causality.

3.2. Data
This section describes the data, ownership variables and market selection.

3.2.1. Data Description
The dataset contains information about 210 local media markets in each of three time periods
from two sources. Media ownership variables were provided by the FCC. They correspond to
three snapshots in time: December 31, 2005, December 31, 2007, and December 31, 2009.

The second dataset consists of television ratings provided by Nielsen Media Research
Galaxy ProFile. The ratings correspond to the November and May “sweeps” months in the 2005-
06, 2007-08 and 2009-10 television seasons. Nielsen selects participants through geographic
randomization and provides financial incentives to participate. In larger media markets, Nielsen
measures television viewing with PeopleMeters, which record television usage and tuning
continuously and prompt viewers to indicate their presence via remote control once or twice per
hour. In smaller markets, audimeters attached to televisions measure set usage and tuning
continuously. Viewer presence is measured via self-reported diaries. Nonresponsive participants
are removed from the sample quickly. Responsive participants are replaced at regular intervals.

The Nielsen data were inconsistently reported. Many datapoints and some entire market-
month datasets were missing from the data. These issues affected the variable definitions in three
ways. First, five markets (Alpena, Biloxi, Miami, New Orleans and West Palm Beach) were
dropped since a balanced panel could not be constructed for these markets. Second, because the
16


measurement technology is more reliable for households than for demographic groups, the
analysis focuses on household ratings. Demographic group ratings are excluded as these are
more often missing. Third, even in the household-level ratings, about 20% of the possible
observations are missing. Therefore, the local news audience share analysis focuses primarily on
evening news viewing, since this daypart featured the highest percentage of data availability
(94%) and local news programming.

The time window analyzed was 6:00-7:00 p.m. EST, 5:00-6:00 p.m. CST, 5:00-6:00 p.m.
MST and 6:00-7:00 p.m. PST. Virtually every local station in the sample airs a local newscast in
the first half-hour within this window, and airs its affiliated network’s national newscast within
the second half-hour of this window. There were a few markets, such as Spokane, in which local
newscasts did not precede national news but were aired immediately afterwards; in those markets
the time period analyzed started thirty minutes later.

Data on market-level demographics are used in section 4.4, including median household
income, median age, the proportion of Spanish-speaking households, the number of television
stations per capita, the percentages of households with televisions and pay-television service.
These data were collected by the American Community Survey and were provided by the FCC in
conjunction with the media ownership data. They are used to ensure consistency with other
studies in the quadrennial review. It was not clear whether the demographic variables were
defined consistently across the three snapshots in the sample, so 2007 and 2009 demographic
data are not used in the analysis.

The study was undertaken with the understanding that the television viewing data would
contain local viewing of national cable networks. Those data would have provided additional
degrees of freedom and allowed for a more nonparametric diversity metric. However, contrary to
the authors’ repeated inquiries, the data provider did not provide local audience data for national
cable networks.

3.2.2. Media Ownership Variables
This section defines the set of media ownership variables. Ownership variables were chosen
according to their relevance to the media ownership rules, but their number was limited to
prevent multicollinearity from inflating the standard errors of the estimates. Three ownership
17


variables were reliably measured and varied extensively, and therefore are included in the base
set of ownership variables x :
mt
Co-ownedTV: The number of television station parents that controlled more than one television
station in the same media market.
TV/Radio: The number of television stations whose parent controlled at least one radio station in
the same market.
LocalOwnerTV: The number of television stations in the market controlled by entities located
within the market.
Two additional ownership variables are available:
TV/Newspaper: The number of television stations whose parent controlled at least one newspaper
in the same market. This ownership variable exhibits the least variation. It changed in
only one market in 2005-2007, and changed in five markets in 2007-2009.
MinorityOwnerTV: The number of television stations in the market with an identifiable controller
who was a member of a minority race/ethnicity. This variable was only measured reliably
in 2007 and 2009; see Turner (2006) for further discussion.
Unfortunately, TV/Newspaper does not show meaningful variation in 2005-2007, and
MinorityOwnerTV data are not available for 2005. Therefore, these two variables must be
excluded from the base set of ownership variables. However, both can be included in a
regression based on 2007-2009 data alone. Therefore, these two variables are included in an
“augmented” set of ownership variables below.

All ownership variables are defined as count data. Percentage definitions were found to
be misleading, as they are influenced by changes in the base number of television stations in the
market. Small independent TV stations sometimes start or stop broadcasting, which then changes
all cross-ownership and co-ownership percentage variables in the market. However, because
these changes typically occur on the fringe of the TV market, they seldom indicate meaningful
changes in station ownership concentration.

To summarize the ownership variables, TV/Newspaper is relevant to the
Newspaper/Broadcast Cross-Ownership Rule; Co-ownedTV is relevant to the Local TV Multiple
Ownership Rule; TV/Radio is relevant to the Local Radio/TV Cross-Ownership Rule; and
LocalOwnerTV and MinorityOwnerTV are relevant to the impact of ownership diversity on
media market competition and localism.
18



3.2.3. Market Selection
Since the Viewpoint Diversity Index defined in section 2 requires at least three newscasts, and
since multiple newscasts would fundamentally change the definition and implications of the
measure, market selection is an important consideration. Local media markets that did not offer
all three national broadcast networks’ news programs (ABC, CBS, NBC) and local news
programs on those network affiliates were dropped from the analysis. This narrowed the number
of markets included from 205 to 132.

Further, the Viewpoint Diversity Index will be fundamentally different in a market with a
larger number of local newscasts. FOX affiliates provided local newscasts in the evening daypart
in some markets. To gauge the sensitivity of the empirical results to the presence of a fourth local
newscast, the empirical analysis is also performed using the subsample of 99 markets in which
evening news was not available on the local FOX affiliate. This was done to gauge the sensitivity
of the results to the assumption of three local newscasts.

Third, in addition to the ABC, CBS and NBC national newscasts, Spanish-language
networks Univision and Telemundo also offer national news programs. It is unlikely that these
newscasts compete extensively with the English-language national news programs for viewers,
as most viewers are not bilingual, so they are not incorporated into the Viewpoint Diversity
Index. However, their presence in a market could potentially change the dynamics of
competition among the English-language language local newscasts. Therefore, the analysis is
repeated on the subsample of 103 markets in which fewer than 20% of self-identified heads of
household report that English is not their native language.

4. Empirical Results
This section reports the estimation results.

4.1. Viewpoint Diversity Index
The Viewpoint Diversity Index was straightforward to calculate and displays substantial
variation across markets. Table 1 shows the raw data for 2007-2009, so that the reader may
compare the changes in the log of the Viewpoint Diversity Index to changes in the media
ownership variables.
19



The table is sorted in ascending order of the change in log Viewpoint Diversity. Visual
inspection shows that there is little in the way of a relationship between Viewpoint Diversity and
the media ownership variables. Extreme changes in viewpoint diversity at the high end and low
end do not coincide with unusual changes in any of the ownership variables. A similar pattern
was observed in the 2005-2007 data.

4.2. Results: Base Ownership Variables, Full Sample
Table 2 reports estimation results for the base set of three ownership variables in the full sample.
The first three columns report the FD point estimates and two sets of standard errors, one
provided by OLS estimation and one provided by clustered standard error estimation. The second
set of three columns report the FE point estimates, followed by two sets of standard errors.

The set of FD estimates with clustered standard errors is the preferred set of estimates, so
the discussion focuses on these; the other estimates are provided as benchmarks. The final
column in Table 2 displays 95% confidence intervals for the mean elasticity on each effect,
based on the FD parameter estimates and clustered standard errors.

Three results merit discussion. First, media ownership variables and time dummies
explain little of the variation in the Viewpoint Diversity Index. The R-squared indicates that the
media ownership variables and time dummies explain approximately 2.4% of the variation in the
Viewpoint Diversity Index. Second, none of the media ownership estimates is statistically
distinguishable from zero. This is true when considering either FD or FE estimates with either
type of standard error. Third, while the standard errors tend to be larger than the point estimates,
none of the confidence intervals admits any appreciable effect of media ownership variables on
viewpoint diversity. It may be safely estimated that none of the elasticities is greater than 0.03 in
absolute value.

4.3. Results: All Ownership Variables, Limited Sample
Table 3 reports model estimation results for the set of five ownership variables based on the final
two years in the sample. The data were limited to the years 2007 and 2009 because
TV/Newspaper showed almost no variation between 2005 and 2007 and because
MinorityOwnerTV was not available in 2005. Since FD and FE provide identical estimates when
the sample contains just two time periods, only one set of estimates are presented in the table.
20



Model fit is slightly better in the subsample, with an R-squared of 0.036, but all estimates
continue to be statistically indistinguishable from zero. Again, the confidence intervals on the
elasticities exclude the possibility of large effects. All elasticities may be safely estimated to be
smaller than 0.04 in absolute value.

4.4. Robustness Check: Market Selection
The calculation of the Viewpoint Diversity Index assumes that there are exactly three news
outlets within a market. This assumption is violated in two markets where FOX affiliates offered
an evening newscast or in markets where appreciable portions of the population consume
Spanish-language news. To check whether these violations influenced the results, the regressions
in Table 2 were re-run using three subsamples of data: all markets without any FOX affiliate
news; all markets in which less than 20% of the viewing population speaks Spanish as a native
language; and the intersection of these two subsamples.

Table 4 offers the results. The point estimates in all three cases are similar to those in
Table 2. Neither change in the market selection criteria admits the possibility of substantial
effects of media ownership structure on viewpoint diversity.

4.5. Robustness Check: Market Demographics and Year Splits
FD and FE estimation remove time-invariant within-market variation by differencing out market-
specific intercepts. However, if market-specific intercepts can be accurately characterized with
demographic variables, the power reduction due to differencing out the intercepts may outweigh
the benefit of doing so. If this is true, FD or FE estimation may yield imprecise parameter
estimates.

As a robustness check, equation (1) was estimated using a cross-sectional approach
wherein the market intercepts α were replaced with the product of a vector of market
m
characteristics and a parameter vector to be estimated, z φ . The market characteristics included
m
t
the median age, median income, percentage of the population whose native language was
Spanish, a dummy variable indicating whether the local FOX affiliate offered evening news, the
number of TV channels per capita, and the percentages of households with pay-TV service or
any TV service. The estimation is done with clustered standard errors. The effects of both the
market characteristics and ownership variables are allowed to vary with time.
21



The results are in Table 5, broken out by year of the sample. The general conclusion is
that the FD and FE estimation techniques did not cause us to fail to find statistically significant
effects when those effects were actually present. Out of 33 parameter estimates, only two are
statistically significant at the 95% confidence level. They indicate that television penetration was
positively related to viewpoint diversity and that increased television station ownership
concentration was negatively associated with viewpoint diversity. Both parameter estimates are
significant in 2007 but are statistically indistinguishable from zero in 2005 and 2009. As such,
they do not provide robust support for the existence of those effects.

5. Summary and Conclusions
This paper proposed a novel market-based approach to measuring Viewpoint Diversity and used
data from a panel of local media markets to investigate how it is associated with local media
ownership variables. These associations are statistically indistinguishable from zero, and all are
estimated to have elasticities less than .04 in absolute magnitude. Still, the following results may
contribute to the policy discussion on the FCC’s media ownership rules.

Newspaper/Broadcast Cross-Ownership Rule

: Based on the 2007-2009 subsample, the
elasticity of viewpoint diversity with respect to TV/Newspaper cross-ownership is 95%
likely to be less than .01 in absolute value.

Local TV Multiple Ownership Rule

: The elasticity of viewpoint diversity with respect to TV
station ownership concentration is 95% likely to lie in the range [−
,.
02
.
]
01 .

Local Radio/TV Cross-Ownership Rule

: The elasticity of viewpoint diversity with respect to
TV/radio cross-ownership is 95% likely to lie in the range [
]
02
,.
0
.

Ownership Diversity

: The full sample results indicate that the elasticity of viewpoint diversity
with respect to local TV station ownership is 95% likely to lie within the range [−
]
01
,.
02
.
. The 2007-2009 subsample indicated the elasticity of viewpoint diversity with respect to
minority ownership of TV stations is 95% likely to lie within the range [−
]
01
,.
02
.
.
In general, these findings show that under the proposed definition of viewpoint diversity,
variation in television station co-ownership and cross-ownership is generally found to negligible
effects on viewpoint diversity. However, it is important to note that the data are limited to the
degree of media co-ownership and cross-ownership currently allowed under FCC rules.
22


The evidence provided in this report is intended to contribute to the policy debate around
the media ownership rules. It does not provide any conclusive basis for policymaking. This paper
describes statistical relationships without any claims of causality. Its findings are limited by the
range of the available data and the reader is reminded that an absence of evidence is not evidence
of absence.

Works Cited.

Arellano, M. 1987. Computing Robust Standard-Errors for Within-Group Estimators. Oxford
Bulletin of Economics and Statistics, 49,4, 431-434.
Arellano, M., S. Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence
and an Application to Employment Equations. Review of Economic Studies, 58, 2, 277-
297.
Bertrand, M., E. Duflo, S. Mullainathan. 2004. How Much Should We Trust Differences-In-
Differences Estimates? Quarterly Journal of Economics, 119,1, 249-275.
Cochrane, D., G. H. Orcutt. 1949. Application of Least-Squares Regression to Relationships
Containing Auto-Correlated Error Terms. Journal of the American Statistical
Association, 44, 245, 32-61.
Gentzkow, M., J. M. Shapiro. 2010. What Drives Media Slant? Evidence from U.S. Daily
Newspapers. Econometrica, 78, 1, 35-71.
Hotelling, H. 1929. Stability in Competition. Economic Journal, 39, 41-57.
McCann, K. 2010. A Diversity Policy Model & Assessment: Debates and Challenges of [Media]
Diversity. Working paper, Fordham University.
Turner, S. D. 2006. Out of the Picture: Minority & Female TV Station Ownership in the United
Stathttp://www.freepress.net/files/out_of_the_picture.pdf">es. http://www.freepress.net/files/out_of_the_picture.pdf, accessed March 2011.
US Federal Communications Commission. 2010. Notice of Inquirhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-10-92A1.pdf">y. http://hraunfoss.fcc.gov/
http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-10-92A1.pdf">edocs_public/attachmatch/FCC-10-92A1.pdf, accessed March 2011.
Wooldridge, J.M. 2010. Econometric Analysis of Cross Section and Panel Data. Second edition.
MIT Press: Cambridge, MA.


23



Table 1. Raw Data fo0072 2007-2009 Subsample


e
r
s
ity

e
r
s
ity

e
r
s
ity

iv
V
iv
V
iv
V
D
r
T

V
r
D
r
T

V
r
D
r
T

V
r
nt
V
nt
V
nt
V
wne
r
T

wne
r
T

wne
r
T

poi
dT
pape
poi
dT
pape
poi
dT
pape
w
wne
o
w
wne
o
w
wne
o
i
t
y
O

O
wne
ws
i
t
y
O

O
wne
ws
i
t
y
O

O
wne
ws
Vie
al
Vie
Vie
/
Ne

al
/
Ne

al
/
Ne

og
nor
nor
nor
oc
V/Radi
V
og
oc
V/Radi
V
og
oc
V/Radi
V
Mi
L
Co-O
T
T
Mi
L
Co-O
T
T
Mi
L
Co-O
T
T
n L
n
n
n
n
n
n L
n
n
n
n
n
n L
n
n
n
n
n
Television M arket
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Television M arket
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Television M arket
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Chg. i
Juneau, AK
-0.19
0
0
0
0
0
Waco-et al.
-0.02
0
-1
0
0
-1
Chico-Redding, CA
0.02
0
0
0
0
0
Anchorage, AK
-0.14
0
0
0
0
0
Columbus, GA
-0.02
0
0
0
0
0
Norfolk-et al.
0.02
0
1
0
0
0
Harlingen-et al.
-0.13
0
0
0
0
0
Washington, DC
-0.02
0
0
0
0
0
Atlanta, GA
0.02
0
0
0
0
0
Madison, WI
-0.13
0
0
0
0
0
Dayton, OH
-0.01
0
0
0
0
0
Detroit, MI
0.02
0
0
0
0
0
San Angelo, T X
-0.11
1
-1
0
0
0
Charleston-et al.
-0.01
0
0
0
0
0
Hartford-et al.
0.02
0
0
0
0
0
T ampa-et al.
-0.11
0
0
0
0
0
Jackson, MS
-0.01
0
0
0
0
0
Charlotte, NC
0.02
0
0
0
0
0
Denver, CO
-0.09
0
0
0
-1
0
Johnstown-et al.
-0.01
0
0
0
0
0
Sacramento-et al.
0.02
0
0
0
0
0
Spokane, WA
-0.08
0
0
0
0
0
Los Angeles, CA
-0.01
1
0
0
1
0
Albany-et al.
0.02
0
0
0
0
0
Bangor, ME
-0.08
0
0
0
0
0
Shreveport, LA
-0.01
0
0
1
0
0
Memphis, T N
0.02
0
0
0
0
0
Meridian, MS
-0.07
0
0
0
0
0
Greenville,SC-et al.
-0.01
0
0
0
0
0
Kansas City
0.02
0
0
0
0
0
Charlottesville, VA
-0.07
0
0
0
0
0
Montgomery, AL
-0.01
0
1
0
0
0
Monroe-et al.
0.02
0
-1
0
0
0
La Crosse-et al.
-0.07
0
0
0
0
0
Minneapolis - et al.
-0.01
0
0
0
0
0
Minot-et al.
0.02
0
0
0
0
0
Peoria-et al.
-0.07
0
0
0
0
0
Lexington, KY
0.00
0
-1
0
0
0
Duluth-et al.
0.02
0
0
0
0
0
Savannah, GA
-0.07
0
0
0
0
0
Oklahoma City, OK
0.00
0
1
1
0
0
Columbia-et al.
0.03
0
0
0
1
0
Las Vegas, NV
-0.06
0
0
-1
0
0
Knoxville, T N
0.00
0
0
0
-1
0
Mobile, et al.
0.03
0
0
0
0
0
Sioux Falls-et al.
-0.06
0
0
0
0
0
Baton Rouge, LA
0.00
0
0
0
0
0
Springfield-et al.
0.03
0
0
0
0
0
Paducah-et al.
-0.05
1
0
0
0
0
Youngstown, OH
0.00
0
0
2
0
0
Raleigh-et al.
0.03
0
0
0
0
0
Lansing, MI
-0.04
1
0
0
0
0
Chicago, IL
0.00
0
0
0
0
0
Springfield, MO
0.03
0
-1
1
0
0
Fresno-Visalia, CA
-0.04
0
-2
0
0
0
T allahassee-et al.
0.00
0
0
0
0
0
Abilene-et al.
0.04
1
0
0
0
0
Birmingham, AL
-0.04
0
0
0
0
0
Joplin, et al.
0.00
0
0
0
0
0
Dallas-et al.
0.04
0
0
1
0
1
Charleston, SC
-0.04
0
0
0
0
0
Columbia, SC
0.00
0
0
0
0
0
Odessa-et al.
0.04
0
0
0
1
0
Columbus, OH
-0.04
0
-1
0
0
0
Little Rock-et al.
0.00
0
-2
1
0
0
Green Bay-et al.
0.04
0
0
0
0
0
Huntsville-et al.
-0.04
0
0
0
0
0
Augusta, GA
0.00
0
0
0
0
0
Ft. Wayne, IN
0.04
1
0
0
0
0
Houston, T X
-0.04
0
-1
0
0
0
Amarillo, T X
0.00
0
0
1
0
0
Omaha, NE
0.04
0
0
-1
0
0
Pittsburgh, PA
-0.04
0
0
0
0
0
Columbus-et al.
0.00
0
0
0
0
0
Evansville, IN
0.04
0
0
0
0
0
Louisville, KY
-0.04
0
0
0
0
0
South Bend-et al.
0.01
0
0
0
0
0
Ft. Smith-et al.
0.04
0
0
1
0
0
Idaho Falls-et al.
-0.04
0
0
0
0
0
Greenville-et al.
0.01
0
0
0
0
0
Wilkes Barre-et al.
0.05
0
0
0
0
0
Fargo, ND-et al.
-0.04
0
0
0
0
0
Syracuse, NY
0.01
0
0
0
0
0
T ulsa, OK
0.05
0
0
0
0
0
Binghamton, NY
-0.04
0
0
0
0
0
Buffalo, NY
0.01
0
0
0
0
0
New York, NY
0.05
0
1
0
0
-1
Cedar Rapids-et al.
-0.03
0
0
1
0
0
Ft. Myers-et al.
0.01
0
0
0
0
0
T yler-Longview, T X 0.05
0
0
1
0
0
Burlington, VT -et al. -0.03
0
0
0
0
0
Milwaukee, WI
0.01
0
0
0
0
0
T ucson, AZ
0.05
0
0
1
0
0
Corpus Christi, T X
-0.03
0
1
0
0
0
Harrisburg-et al.
0.01
0
0
0
0
0
Greensboro-et al.
0.06
1
0
0
0
0
T raverse City-et al.
-0.03
1
0
0
0
0
Flint-et al.
0.01
1
0
0
0
0
St. Louis, MO
0.06
0
0
1
-1
0
Phoenix, AZ
-0.03
1
0
0
0
0
Nashville, T N
0.01
0
0
0
0
0
Philadelphia, PA
0.06
1
-1
0
-1
0
Baltimore, MD
-0.03
0
0
0
0
0
San Francisco-et al.
0.01
-1
0
0
-1
0
Rockford, IL
0.07
0
0
0
0
0
T opeka, KS
-0.03
0
0
0
0
0
Jacksonville, FL
0.01
0
0
-1
0
0
Santa Barbara-et al.
0.08
0
0
0
0
0
Wichita - et al.
-0.03
0
0
0
0
0
T ri-Cities, T N-VA
0.01
0
0
0
0
0
Bluefield-et al.
0.08
0
0
0
0
0
Des Moines-et al.
-0.03
0
0
0
0
0
Portland-Auburn
0.01
0
0
1
0
0
Austin, T X
0.09
0
0
0
0
0
Providence-et al.
-0.02
0
0
0
0
0
Cincinnati, OH
0.01
0
0
0
0
-1
Lubbock, TX
0.09
0
0
0
0
0
Macon, GA
-0.02
0
0
0
0
0
San Antonio, T X
0.01
0
0
0
0
0
Salt Lake City, UT
0.09
0
0
0
0
0
Orlando-et al.
-0.02
0
0
0
0
0
Wichita Falls, et al.
0.01
0
0
0
0
0
T oledo, OH
0.10
0
0
0
0
0
Indianapolis, IN
-0.02
0
0
0
0
0
Davenport, IA-et al.
0.02
0
0
0
0
0
Marquette, MI
0.12
0
0
0
0
0
Rochester, NY
-0.02
0
0
0
0
0
Champaign-et al.
0.02
0
0
0
0
0
Medford-et al.
0.19
0
0
0
0
0
Grand Rapids-et al.
-0.02
1
0
0
0
0
Roanoke-et al.
0.02
0
-1
1
0
0
Boston, MA
0.27
-1
0
0
0
0



24


Table 2. Estimation Results: Base Ownership Variables, 2005-2009 Sample

First Differences

Fixed Effects

Mean Elasticity

Point

Std. Errors

Point

Std. Errors

95% Conf. Int.

Est.

OLS
Clust.

Est.

OLS
Clust.
(FD, Clust. s.e.)
Logged Diversity Index
LocalOwnerTV
-.006 (.008)
(.007)
-.004 (.007)
(.006)
(-.02,.01)
Co-Owned TV
.004
(.009)
(.007)
.001
(.008)
(.007)
(-.01,.02)
TV/Radio
.012
(.011)
(.009)
.011
(.010)
(.009)
(.00,.02)
Num. Obs.
264
396
R-squared
.024
.553
Year-specific intercept estimates excluded from table for brevity.
** Significant at the 99% confidence level. * Significant at the 95% confidence level.


Table 3. Estimation Results: All Ownership Variables, 2007-2009 Subsample

First Differences

Mean Elasticity

Point

Std. Errors

95% Conf. Int.

Est.

OLS
Clust.
(FD, Clust. s.e.)
Logged Diversity Index
LocalOwnerTV
.008
(.012)
(.007)
(-.01,.02)
Co-Owned TV
.014
(.013)
(.009)
(.00,.03)
TV/Radio
.006
(.020)
(.020)
(-.02,.03)
Minority
-.027 (.016)
(.025)
(-.02,.01)
TV/Newspaper
-.004 (.029)
(.013)
(.00,.00)
Num. Obs.
132
R-squared
.036
Year-specific intercept estimates excluded from table for brevity.
** Significant at the 99% confidence level. * Significant at the 95% conf. level.


25



Table 4. Market Selection Robustness Checks, 2005-2009 Sample

No FOX

No Spanish

No FOX or

Mean Elasticity

Markets

Markets

Spanish Markets

95% Conf. Int.

Point Std.

Point

Std.

Point

Std.

(No FOX or

Est.

Err.

Est.

Err.

Est.

Err.

Span.)

Logged Diversity Index
LocalOwnerTV
-.002 (.008)
-.013
(.008)
-.010
(.007)
(-.02,.00)
Co-Owned TV
.002
(.008)
.002
(.007)
-.003
(.008)
(-.01,.01)
TV/Radio
.014
(.012)
.011
(.010)
.016
(.014)
(.00,.01)
Num. Obs.
198
238
176
R-squared
.019
.036
.035
Year-specific intercept estimates excluded from table for brevity.
** Significant at the 99% confidence level. * Significant at the 95% confidence level.


Table 5. Estimation Robustness Check, 2005-2009 Sample

2005 sample
2007 sample
2009 sample

Point

Std.

Point

Std.

Point

Std.

Explanatory Variable

Est.

Err.

Est.

Err.

Est.

Err.

Median age
.000 (.003)
.000 (.002)
.003 (.002)
Median income
.000 (.000)
.000 (.000)
.000 (.000)
Spanish-speaking population
.039 (.040)
.008 (.062)
-.004 (.051)
Local evening FOX news
-.004 (.012)
.010 (.016)
.027 (.015)
TV channels per capita
-.003 (.003)
-.001 (.002)
-.001 (.002)
Pay TV penetration
-.073 (.104)
-.175 (.126)
.015 (.149)
TV penetration
.087 (.150)
.220 (.108) *
.011 (.079)
LocalOwnerTV
-.001 (.003)
-.002 (.004)
.000 (.004)
Co-Owned TV
-.004 (.005)
-.014 (.007) *
-.008 (.005)
TV/Radio
.005 (.007)
.007 (.009)
.009 (.008)
Year 2005 Intercept
-.045 (.123)
Year 2007 Intercept
-.111 (.164)
Year 2009 Intercept
-.188 (.131)
Num. Obs.
132
132
132
R-squared
.713
.639
.661
** Significant at the 99% confidence level. * Significant at the 95% confidence level.





26



Figure 1. Diversity Index Example

Dm
Market m
m
xˆ
m
xˆ
12
23
m
x
m
x
m
x
1
2
3
Market m'
'
ˆm
'
x
ˆm
x
12
23
m'
x
m'
x
m'
x
1
2
3
Dm'



Figure 2. Uncovered Media Market

watching
watching
watching
watching
watching
no news
news 1
news 2
news 3
no news
N
N
xˆ01
xˆ
N
xˆ
N
xˆ
12
23
30
N
x
N
x
N
x
1
2
3


27


Note: We are currently transitioning our documents into web compatible formats for easier reading. We have done our best to supply this content to you in a presentable form, but there may be some formatting issues while we improve the technology. The original version of the document is available as a PDF, Word Document, or as plain text.

close
FCC

You are leaving the FCC website

You are about to leave the FCC website and visit a third-party, non-governmental website that the FCC does not maintain or control. The FCC does not endorse any product or service, and is not responsible for, nor can it guarantee the validity or timeliness of the content on the page you are about to visit. Additionally, the privacy policies of this third-party page may differ from those of the FCC.