.x; . _ 3 . . .
Qua. . . . . V , . . Eu»!
.7:

. 35...:

z:

. .
.:
:37}

.1...
1...:
.531...

.T? 4.... .9
4455?;

1.? a.
.i 194’:- E.‘ (A.
.3 p .2. x .3. ,

 

 

 

THesK

 

  

LIBRARY
Michigan State
University

      

This is to certify that the
dissertation entitled
A Study of How a Strategy Creating Clusters
of Commonly—Owned Newspapers Affects Prices,
Quality and Profits

presented by

Hugh J. Martin

has been accepted towards fulfillment
of the requirements for

Ph-Do—degree in M

C’Jé—Z‘7/

/ Major professof

Date May 3, 2001

MSUiu-n‘” ,. . . r m , . - 0.12771

 

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
AUB C3103

 

6/01 c./CIRC/DaieDue p65—p 15

A STUDY OF HOW A STRATEGY CREATING CLUSTERS OF COMMONLY-
OWNED NEWSPAPERS AFFECTS PRICES, QUALITY AND PROFITS

By

Hugh J. Martin

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
College of Communication Arts & Sciences

2001

ABSTRACT

A STUDY- OF HOW A STRATEGY CREATING CLUSTERS OF COMMONLY—
OWNED NEWSPAPERS AFFECTS PRICES, QUALITY AND PROFITS

By
Hugh J. Martin

Newspaper companies have adopted a strategy of acquiring newspapers in
adjacent markets. The acquisition of neighboring newspapers creates clusters of
commonly-owned newspapers that previously competed with each other. Newspaper
companies argue clusters allow them to create economies of scale by sharing production
and news gathering resources, resulting in lower costs. However, economic models
suggest this strategy could be intended to deter competition from other newspapers.

Industry data was used to identify all clustered newspapers in 1988 and 1998.
Random samples of clustered newspapers were selected and detailed information about
their markets, their advertising and subscription prices, and their ownership was gathered.
This data was analyzed in comparisons with a randomly selected control group of non-
clustered newspapers. A mail survey was used to gather additional information about
newsroom expenditures and resource sharing at clustered and control newspapers.

Results support industry arguments that clustered newspapers are more likely to
share resources in an apparent attempt to create economies of scale. However, there is no
evidence that cost savings are used to increase spending on news gathering and
publication. Cost savings may instead be used to finance additional acquisitions or to
increase profits.

Results also show there is less competition in markets with clustered newspapers.

Evidence shows demand is less elastic in those markets. Clustered newspapers charge

higher advertising and subscription prices. Clustered newspapers also spend less on
news. Theory suggests less elastic demand would allow clustered newspapers to reduce
quality without risking a disproportionate decline in readers.

Predictions that companies would use savings from clustering to increase profits
received only weak support, however. Predictions about relationships among costs, the
degree of diversification and the degree of inside control at companies that own clustered
newspapers also received only limited support. The data precluded multivariate tests, so
the effects of clustering on these relationships could not be examined.

However, there is evidence that clustering has contributed substantially to the
overall decline in the number of daily newspapers in the United States. At the same time,
the number of clustered newspapers is increasing. The implications of clustering for

newspapers and public policy are discussed.

Copyright by
Hugh J. Martin
2001

 

 

For Peggy and for Catherine.
In loving memory of
Susan S. Martin.

ACKNOWLEDGEMENTS

My thanks to Peggy, for agreeing to give this a try, for her unwavering support,
and for the many times she pitched in when I needed some help. Thanks also to
Catherine for taking valuable time from her advanced study of Harry Potter to put stamps
on envelopes.

Thanks to my father for all he has given me. Thanks also to Mary and Mike
Marchner for standing with us through the years.

I could not have done this without each of you.

vi

 

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix

LIST OF FIGURES ......................................................................................................... xiii

INTRODUCTION ............................................................................................................... 1

CHAPTER 1

MOTIVES FOR MERGER ................................................................................................. 4
Mergers for Market Power ..................................................................................... 10
Effects of Reducing Competition ........................................................................... 18

CHAPTER 2

MODELING THE BEHAVIOR OF CLUSTERED NEWSPAPERS .............................. 2l
Clustering’s Effects on Quality .............................................................................. 21
Clustering’s Effects on Prices ................................................................................ 26
Interactions Between Newspaper Prices ................................................................ 28
Implications for Quality and Prices ................................ . ....................................... 31

CHAPTER 3

HYPOTHESES .................................................................................................................. 34
Expected Relationships Among Variables ............................................................. 35

CHAPTER 4

METHOD .......................................................................................................................... 43
Defining and Identifying Clustered Newspapers ................................................... 43
Information about Newspapers .............................................................................. 46
Information about Competition from Other Media ............................................... 48
Information about Quality and Resource Sharing ................................................. 49
Information about Publicly-Owned Newspapers ................................................... 50
The Profit Measure ................................................................................................ 51
The Cost Measure .................................................................................................. 51
The Diversification Measure .................................................................................. 51
The Measure of Inside Control .............................................................................. 52
Potential Problems with Endogenous Variables .................................................... 52

vii

CHAPTER 5

RESULTS .......................................................................................................................... 55
t-Tests of Differences between Clustered and Control Newspapers ..................... 57
Results from Tests of Association ......................................................................... 64
Results from Regressions ....................................................................................... 70
Regressions Examining Newsroom Spending ....................................................... 86
Comparing Elasticity for Clustered and Control Papers ........................................ 89
First-Difference Regressions ................................................................................. 96
Results of the Mail Survey ................................................................................... 104
Results from Profit and Cost Measures ............................................................... l 18

CHAPTER 6

CONCLUSIONS .............................................................................................................. 122
Implications of the Study ..................................................................................... 129

APPENDIX A

VARIABLES IN THE STUDY ....................................................................................... 134

APPENDIX B

SURVEY LETTER AND QUESTIONNAIRE ............................................................... 138

APPENDIX C

CORRELATIONS FOR VARIABLES USED IN REGRESSIONS .............................. 141

REFERENCES ................................................................................................................ 146

viii

LIST OF TABLES

Table 1: The Population of Clustered Newspapers ............................................................ 55
Table 2: Changes in Clustered Newspapers from 1988 to 1998 ........................................ 56
Table 3: Changes in the Number of US. Newspapers From 1988 to 1998 ....................... 56
Table 4: Why Clustered Dailies Ceased Publication ......................................................... 57
Table 5: Circulation Comparisons for Study Population and Random Samples ............... 58
Table 6: t-Tests of Intercounty Circulation Differences .................................................... 59
Table 7: l-Tests of Differences in Number of Papers Headquartered in Markets ............. 60
Table 8: t-Tests of Differences in Market Share ................................................................ 61
Table 9: t-Tests of Differences Between Prices ................................................................. 63
Table 10: Pearson’s Correlations in 1988 for Clustered Papers ........................................ 65
Table 11: Pearson’s Correlations in 1988 for Control Papers ........................................... 66
Table 12: Pearson’s Correlations in 1998 for Clustered Papers ........................................ 67
Table 13: Pearson’s Correlations in 1998 for Control Papers ........................................... 69
Table 14: 1988 Cluster and Control Regressions, Advertising Price as DV ..................... 72

ix

Table 15:

Table 16:

Table 17:

Table 18:

Table 19:

Table 20:

Table 21:

Table 22:

Table 23:

Table 24:

Table 25:

Table 26:

Table 27:

Table 28:

Table 29:

Table 30:

1998 Cluster and Control Regressions, Advertising Price as DV ..................... 74

1988 Cluster and Control Regressions, Advertising cpm as DV ....................... 76
1998 Cluster and Control Regressions, Advertising cpm as DV ....................... 77
1988 Cluster and Control Regressions, Subscription Price as DV .................... 79
1998 Cluster and Control Regressions, Subscription Price as DV .................... 81
1988 Cluster and Control Regressions, Subscription ppm as DV ..................... 83
1998 Cluster and Control Regressions, Subscription ppm as DV ..................... 85
1988 Cluster and Control Regressions, Wires pm as DV .................................. 87
1998 Cluster and Control Regressions, Wires pm as DV .................................. 88
1988 Elasticity Regressions, LN of Paper’s Market Penetration as DV ........... 92
1998 Elasticity Regressions, LN of Paper’s Market Penetration as DV ........... 94
F irst-Difference Regressions, Ad Price Difference as DV ................................ 97
First-Difference Regressions, Ad cpm Difference as DV ................................. 98
F irst-Difference Regressions, Subscription Price Difference as DV ............... 100
F irst-Difference Regressions, Subscription ppm Difference as DV ................ 101
First-Difference Regressions, Wires pm Difference as DV ............................ 103

Table 31:

Table 32:

Table 33:

Table 34:

Table 35:

Table 36:

Table 37:

Table 38:

Table 39:

Table 40:

Table 41:

Table 42:

Table 43:

Table 44:

Table 45:

Table 46:

Intercounty Differences, Survey Respondents ................................................ 105

Papers in Markets, Survey Respondents .......................................................... 106
Market Share, Survey Respondents ................................................................. 107
Price Differences, Survey Respondents ........................................................... 108
Quality Measures, Survey Respondents .......................................................... 109
Survey Responses to Questions About Resource Sharing ............................... 110
Tests of Differences for Resource-Sharing Questions ..................................... 111
Cluster and Control Regressions, Size of News Staff as DV .......................... 1 12
Cluster and Control regressions, News Staff pm as DV .................................. 114
Cluster and Control Regressions, Column Inches in Newshole as DV ........... 115
Cluster and Control Regressions, Column Inches pm as DV .......................... 1 16
Pearson’s Correlations for Spending and Clusters, Survey Respondents ........ 117
1988 Correlations, Company Performance (Spearman's rho) ......................... 1 19
1998 Correlations, Company Performance (Spearman's rho) ......................... 119
Results of Hypotheses Tests ............................................................................ 123
1988 Pearson’s Correlations for Clustered Papers ......................................... 141
xi

 

Table 47: 1988 Pearson’s Correlations for Control Papers ............................................. 142
Table 48: 1998 Pearson’s Correlations for Clustered Papers .......................................... 143

Table 49: 1998 Pearson’s Correlations for Control Papers ............................................. 144

xii

LIST OF FIGURES

Figure 1. How economies of scale may deter entry into newspaper markets .................... 14
Figure 2. Competition’s effects on reader demand for quality .......................................... 23
xiii

 

INTRODUCTION

The First Amendment of the US. Constitution encourages vigorous competition
in the marketplace of ideas as the best way to find truth in a public debate. Economic
theory argues vigorous competition to supply goods and services efficiently allocates
scarce resources among competing consumer demands. United States newspapers
operate in markets for ideas and markets for goods and services -- but for decades
newspaper competition has declined and the concentration of newspaper ownership has
increased (Lacy and Simon 1993, p. 132; Lacy and Simon 1997, p. 815).

These trends accelerated in the 19908 when daily newspapers changed hands 856
times, including 153 dailies that were sold more than once. Sales of daily newspapers in
the 19905 were higher than sales for the previous two decades (Presstime, 2000). Most of
this activity involved newspaper groups; only 114 independent dailies were sold from
1990 to 1999. Many acquisitions were driven by an increasingly popular strategy as
groups assembled clusters of dailies in markets where adjacent newspapers previously
had multiple owners (Presstime, 2000). Newspaper markets are geographically defined
and newspaper clusters, the industry argues, can reduce production costs, provide
regional coverage for advertisers, and concentrate a group’s newspapers in thriving
markets (Asher 1999; Bass 1999, p. 64-68). In 1999 a journalism review identified “125
major regional concentrations involving more than 400 papers — or well over a fourth

ofthe nation’s dailies” (Bass 1999, p. 66).

However, researchers argue that clusters also stifle the competition that exists
when adjacent newspapers are separately owned (Lacy & Simon, 1997). This
"circulation gerrymandering" (p. 822) may result in a reduction in the quality of news
coverage and an increase in some advertising prices (F u, 2000; Lacy & Simon, 1997).
Lacy and Simon (1997) describe clustering as a strategy with “the potential of promoting
anticompetitive actions across county lines" (p. 822). They call for the US. Justice
Department to investigate whether newspaper sales that create clusters violate anti-trust
laws intended to protect the public by preserving competition.

The popularity of newspaper clusters and their potential effects on competition,
prices, and quality suggest the importance of empirical examination. Lacy and Simon
( 1997), in the only published study of newspaper clusters, concluded clustering is
associated with a reduction in competition between daily newspapers in adjacent
counties. In a dissertation Fu (2000) concluded that when one group owns a newspaper
and its potential competitors, the newspaper enjoys higher circulation and charges more
for a column inch of advertising. A published study by Shaver and Lacy (1999) found
evidence that newspapers compete with other media for some forms of advertising, but
concluded the effect of this competition varies across markets and the type of advertising.

However, much remains to be done. None of the three studies (F u, 2000; Lacy &
Simon, 1997; Shaver & Lacy, 1999) examined how clustering affects the quality of news
coverage. None of the studies examined industry arguments that clusters reduce costs or
how cost reduction might interact with other variables. Only one study included
newspaper cover prices, but not as a dependent variable (Fu, 2000). F u did concluded

newspaper clusters are associated with higher advertising prices. However, the price

measure was not adjusted for circulation and the study did not control for competition
from other newspapers. Shaver and Lacy (1999) measured the effect of radio and
television competition on newspaper advertising linage, concluding “competition varies
by type of advertising and from market to market” (p. 740). However, Shaver and Lacy
had a limited sample of 40 newspapers and did not include a measure for clustered
papers. Therefore, current studies of the relationship between clustering and advertising
prices leave many questions unanswered.

Shaver and Lacy (1999), noting the acceleration of clustering, said there is an
"urgent" (p. 740) need to determine whether it has anticompetitive effects on advertising
and circulation markets. They point out competition is difficult to recreate once it
disappears. This dissertation replicates and extends existing studies by examining how

newspaper clustering affects prices, quality, and profits.

CHAPTER 1
MOTIVES FOR MERGER

Newspapers operate in geographic markets defined by the extent of their
circulation. Newspapers compete when one paper crosses another’s circulation boundary
and enters its rival’s market. The acquisition of one newspaper by an adjacent newspaper
represents a form of “horizontal integration” (Lacy and Simon 1997, p. 817) that
eliminates competitors.

Horizontal mergers can result from either normal business motives or from
attempts to profit from the reduction of competition (Scherer and Ross 1990, p. 160-166).
Firms assembling newspaper clusters cite normal business motives, such as reducing
production costs, as the reason for their strategy (Bass, 1999). However, research
demonstrates clustering also reduces competition (Fu, 2000; Lacy & Simon, 1997).

Horizontal mergers arising from normal business motives include a larger firm’s
acquisition of a smaller firm when the larger firm enjoys lower capital costs. Such
mergers can benefit the smaller firm by making possible otherwise uneconomical capital
investments in new plants or equipment (Scherer and Ross 1990, p. 163-166). Another
example of such motives are horizontal mergers to create economies of scale. Economies
of scale exist when a firm's long-run average costs decline as more goods are produced.
Economies of scope are a special subset of scale economies. When the cost of separately
producing at least two goods is higher than the cost of producing those goods together,

there are economies of scope (Teece, 1980).

There are three kinds of scale economies. The first, product-specific economies,
are associated with the volume of a single product (Scherer & Ross, 1990, p. 97-98).
These economies are available when there are specialized production processes that
require substantial preparation before a production run can begin. If this is the case, the
cost of setting up for a production run may be substantial, so average production costs
decline as production increases. Product-specific economies of scale may also be
available if workers become more efficient as they learn more about how to perform their
individual, specialized tasks. Product-specific economies might be available to
newspapers if creating and editing news and advertising entails substantial initial costs
for labor and specialized machinery, such as computerized typesetters. Lacy and Simon
(1993, p. 72) suggest there are product-specific economies in circulation when the costs
of putting a single truck in service are spread across increasing numbers of newspapers
delivered by that truck. Economies also exist for labor costs when a carrier delivers
increasing numbers of newspapers on the same block. However, Lacy and Simon point
out, if circulation is less concentrated economies dissipate as the carrier has to cover
increasing distances to deliver newspapers.

The second kind of scale economies are associated with the total output of an
entire production plant (Scherer & Ross, 1990, p. 98-99). These may be available if
output increases with the size of equipment used by the plant. However, economies will
only be available if costs of acquiring and operating the larger equipment are
proportional. In other words the ratio of cost to production capacity cannot increase with
size. Newspapers might realize plant-level economies of scale if they can expand their

presses, and their output, without a disproportional increase in associated costs.

Economies of scope result from interactions between these first two kinds of scale
economies. This is because the cost of producing individual products is affected both by
the volume of production and the size of the plant where a product is produced (Scherer
& Ross, 1990, p. 100-102). If a single plant is not producing a large enough volume to
realize all available scale economies, then producing more than one product may result in
economies of scope. For instance, overhead and other costs associated with operating a
press are spread across each day’s press run. However, if a larger press run would not
increase these costs, but instead allow them to be spread across more copies, then
economies of scope would be available. In this case, more than one newspaper might be
printed on the same press.

The third kind of scale economies are associated with the Operation of multiple
plants. This may allow fimis to lower their average production costs if different parts of
the production process can be located in different plants (Scherer & Ross, 1990, p. 102).
Newspaper groups might be able to take advantage of multi-plant economies if, for
instance, they centralize administrative functions such as payroll processing instead of
having each newspaper perform these separately.

Empirical studies beginning with Rosse (1967) have found evidence that
newspapers enjoy scale economies. These studies have reported evidence of what appear
to be plant level economies associated with size (Blankenburg, 1989, p. 100; Ferguson,
1983, p. 645). There also is evidence of what appear to be economies of scope associated
with interactions resulting from the production of “circulation and news at a single plant”

(Dertouzos & Trautman, 1990, p. 12). Thompson (1988, p. 372) found indirect evidence

 

of what appear to be product-specific economies of scale associated with increases in
circulation.

However, Scherer and Ross (1990, p. 164-165) state that using mergers to expand
existing economies of scale is difficult when production facilities already exist. A firm
may instead find it more efficient to take the newly acquired plant out of business.
Horizontal mergers may make it easier to achieve scale economies if a firm can afford to
replace previously separate production facilities with new plants after the merger. This
suggests newspaper groups might make such replacements if capital equipment used by
the acquired newspapers is obsolete. Such economies also can be realized if they are
product specific and the acquiring firm can quickly rearrange production of similar
products (Scherer and Ross 1990, p. 164). In the newspaper industry, this might be
achieved if previously separate production facilities for multiple newspapers are
combined.1

Newspaper companies that create clusters have indeed argued they are trying to
rearrange production processes to gain economies of scale (Asher, 1999; Bass, 1999).
These include economies of scope available from printing more than one newspaper on a
central press or by consolidating the production of news or advertising. What appear to
be attempts to create multi-plant economies include consolidating administrative
functions for multiple newspapers, and offering regional placement of advertising across
clustered newspapers. Cross-promotion of newspapers is another way companies try to

create economies with clusters.

 

' Economies also can be achieved by merging facilities that represent different steps in a manufacturing
process (Scherer & Ross, 1990). However, daily newspapers are likely to own facilities, such as presses
and distribution networks, needed for timely production) and delivery.

If clustering creates such economies of scale, then these horizontal mergers may
be regarded as arising from normal business motives because the merged newspapers are
becoming more efficient than competitors or matching cost advantages enjoyed by
competitors.

However, empirical studies of horizontal mergers in a range of industries provide
weak evidence that post-merger profits increase. Profits decrease in many cases (Scherer
and Ross, p. 170-171). These general studies also show horizontal mergers often fail to
increase efficiency, and that the efficiency of smaller firms may actually be reduced after
they are absorbed by larger firms (p. 172-174). These studies are consistent with
evidence showing the breakup of monopolies results in lower operating costs for
individual business units, a result attributed to so-called X-inefficiency resulting from the
absence of competitive pressure on monopolists.

The degree of similarity between different businesses involved may be another
important factor influencing the success of mergers. Litman and Sochay (1994) report
that economies are difficult to achieve in mergers that created large, diversified mass
media conglomerates. One reason economies are difficult to achieve is that conglomerate
mergers create a diverse set of strategic business units that duplicate, rather than share,
resources. Conglomerate media mergers also tend to diversify and decentralize
operations at the expense of focusing on a single core business (p. 259).

Firms that diversify their lines of business also increase the complexity of
problems facing their managers (Govindarajan & Fisher, 1990; Hill, Hitt, & Hoskisson,
1992; Hill & Hoskisson, 1987). This line of research suggests that if different business

units within the firm do not engage in related activities, managers should focus on

quantifiable performance measures such as production costs. However, when lines of
business are related it may be more important that managers of different business units
coordinate their activities.

Studies show diversified firms with related lines of business are generally more
profitable than unrelated diversification (Markides & Williamson, 1994; Robins &
Wiersema, 1995). However, even related diversification cannot provide sustained
increases in profits if managers are unable to efficiently coordinate the firm’s activities to
take advantage of the potential for economies of scale (Markides & Williamson, 1996).

These studies generally view diversification resulting from mergers from the
perspective of effects on cost. An alternate view argues that companies diversify not to
become more efficient, but to provide a hedge against cyclical changes that may affect
one business and not another. This view argues that newspapers, with their dependence
on retail advertising sales, are particularly sensitive to “changes in the local and national
economies” (Picard & Rimmer, 1999, p. 16). Newspaper companies therefore diversify
into other media because they “are not as dependent on advertising or not as affected by
recessions as the newspaper industry” (p. 7).

This study (Picard & Rimmer, 1999) measured diversification as the percentage
of revenue from newspaper operations at diversified newspaper companies. The study
concluded companies with a higher percentage of revenue from newspapers earned lower
profits before, during and after a recession in 1990-1991. The study also examined
overall costs for these companies. However, the company level measures in the study
suggest the analysis would only indirectly record any effects from economies of scale,

which most likely exist at the level of individual newspapers.

 

These studies leave open the question of whether economies of scale may be
available when a firm creates a cluster of newspapers in the same geographic region
(Asher, 1999; Bass, 1999). The creation of a cluster of newspapers represents a form of
diversification, but not into unrelated businesses. It seems less likely that resources will
be duplicated or that firms with clusters will face the same problems coordinating
production that exist when different types of mass media products -— such as magazines,
movies and outdoor advertising -- are at issue.

Therefore, this dissertation examines whether there is evidence that newspaper
clusters result in lower costs. The previous discussion also suggests that if companies
which own newspaper clusters also diversify into other lines of business, it will be more
difficult to translate such cost advantages into higher profits. However, other variables
will mediate this effect. First, this disadvantage will be more acute if those companies
diversify into unrelated lines of business than if they are in related businesses. Second,
newspaper companies may gain a financial hedge against fluctuations in the retail
advertising market by diversifying. Therefore, this dissertation also examines how
diversification affects the profitability of firms which own clusters of newspapers.
Mergers for Market Power

For decades, daily newspapers have battled an erosion of audience and
circulation. The effort to maintain profits in the face increasing competition from radio,
television, cable television, and most recently the Internet has preoccupied researchers
and members of the industry (Hickey, 1997; Lacy, 1992b; Lacy & Simon, 1993, p. 266;

Meyer, 1995; Moses, 2001). However, publicly-owned newspaper companies still earned

extraordinary profits — substantially above the levels that might be expected in a highly
competitive industry — throughout the 19803 and the early 19908 (Martin, 1998).

This suggests newspaper companies had successfully adopted strategies to
maintain profitability despite the maturity of the industry. Picard and Rimmer (1999)
suggested diversification into broadcasting, cable television, book and magazine
publishing and other industries was intended to preserve profitability by providing a
hedge against cyclical variations in advertising sales. Economic theory suggests that
clustering may be another strategy intended to maintain profits.

Neoclassical theory assumes that markets will be competitive if firms produce a
homogenous product, have identical production costs, and nothing prevents new firms
from freely entering markets if positive economic profits can be earned (Nicholson, 1995,
p. 462-463). Positive profits are defined as any earnings above the amount a business
owner would receive from the alternative of investing capital elsewhere.

In a competitive market, firms keep producing additional goods until the marginal
cost of producing the last unit equals the price of the good (Nicholson, 1995). In
equilibrium, this marginal cost will also equal the average cost of production. If firms
can raise their prices above this marginal cost, then they are earning positive economic
profits. New firms will enter the market and offer their products at lower prices,
restoring equilibrium conditions.

However, if these basic assumptions are changed markets become less
competitive and firms in them can earn positive economic profits without having their
advantage competed away. One way to do this is by differentiating a product so that

consumers distinguish between offerings from different firms. Lacy (1989, p. 41 ; Lacy

ll

& Simon, 1993, chap. 2-3) has discussed how newspapers can differentiate themselves by
altering the news and other information they publish. This is an application of the theory
of monopolistic competition which argues that by catering to differences in consumer
taste, firms can to some degree isolate themselves from the effects of competition and
raise prices above their costs (Scherer & Ross, 1990, p. 32).

Differentiation alone is insufficient to entirely protect a newspaper from the
effects of competition. As Lacy (1989) pointed out, the content of newspapers is defined
by the geographic area where the newspaper circulates. Other newspapers circulating in
the same area can offer plausible alternatives for consumers if they match the content.
Differentiation, in other words, may reduce the importance of prices and cost, but it does
not eliminate them.

The fact that newspapers enjoy economies of scale offers them additional
alternatives for escaping the effects of competition. Keep in mind that economies of
scale mean that average production costs decrease as the number of copies produced
increases. The newspaper able to take most effective advantage of scale economies often
won a battle for survival in cities where there was direct competition. This was because
papers with increasing circulation attracted more and more advertising (Lacy & Simon,
1993, p. 96-99). The advertisers abandoned papers with smaller circulation, which could
only respond by cutting prices and trying to increase quality. However, these tactics
often failed because the larger papers costs were spread across more copies as circulation
increased, so it cost less for advertisers to reach each reader.

The same economies that allow a newspaper to win this kind of competitive battle

can also create a barrier that keeps other newspapers from entering the market by

undercutting the higher prices charged by the now dominant paper. This is illustrated in
Figure 1. This shows that that a newspaper must print a certain number of copies before
it can take full advantage of economies of scale. Let Q be quantity, or the number of
copies printed by a newspaper. Let C be the average cost of printing each copy, defined
as the newspaper’s total production costs divided by the number of copies. If Q1 is the
number of copies printed and sold by the dominant newspaper in a market, its average
cost per copy is C1. However, some consumers are not buying newspapers because the
price is too high; they would be willing to buy them if the price were lowered. A second
newspaper might enter the market by taking advantage of this unmet demand for papers if
it could charge a lower price. The amount of unmet demand is represented by Q2, which
is the number of copies the second newspaper will be able to sell.

The problem facing the second newspaper is that even if it meets the unmet
demand, its average cost per copy of C2 is higher than C1. This is because the second
newspaper cannot publish enough papers to take full advantage of economies of scale and
reach the lowest point on the cost curve. This point is called the Minimum Efficient
Scale (MES). The MES is the smallest number of copies a newspaper can print and still
match the low cost of its competitor.

However, economies of scale alone may be insufficient to deter other newspapers
from entering this market. A single newspaper in a market can be expected to raise
prices above costs, therefore another newspaper could enter the market charging lower
prices and still make enough to cover its costs. Economic theory suggests the mere threat

of such entry could be enough to deter the existing newspaper from raising prices much

13

Newspaper’s long-run average cost

 

 

 

 

(S/Q)
C2 _____
l
I
I
C1 _____ _:_ _ _ _ J
I I
: : Cost curve
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
Q2 MES - Q1 Quantity of copies

Figure 1. How economies of scale may deter entry into newspaper markets.
The Minimum Efficient Scale (MES) is the smallest quantity of newspapers that must be
produced to reach the lowest point on the cost curve.

beyond its costs if such restraint was rewarded by keeping potential competitors out of
the market (Tirole, 1988, p. 308-309).

The threat of such entry is likely to exist in many US. newspaper markets. By
Feb. 1, 1999 only 19 cities had separately-owned daily newspapers in direct competition
(Maddux, 1999, p. xxiii), but this does not mean competition no longer exists. Lacy and
Simon (1993, p. 95, 1 12-115) describe how newspaper competition exists in different
layers. The first layer is regional metropolitan dailies, the second layer is satellite-city
newspapers emphasizing local coverage, the third layer is suburban dailies and the fourth
layer is weeklies, shoppers and specialized newspapers. National dailies are a fifth layer
of competition, and group-owned suburban newspapers make up a sixth layer.

Newspapers in this model are not perfect substitutes, but studies show they do
compete. Lacy (1984) surveyed editors and publishers of southwestern newspapers using
a three-layer model of competition. Respondents reported competition between layers,
with respondents at lower levels reporting competition from metropolitan dailies. A
national study (Niebauer Jr., Lacy, Bernstein, & Lau, 1988) of 900 suburban
communities examined how competition in those communities was affected by
metropolitan newspapers in markets that were competitive, had monopoly dailies, or had
joint-operating agreements. The study (Niebauer Jr. et al., 1988) concluded the central
city market structure had little effect on suburban papers.

Lacy and Davenport (1994) studied the potential for competition in daily
newspaper markets in 1983 and 1988. In 1988 about 44% of all US. counties had at least
one daily from at least two layers of the umbrella model. About 47% of all US. counties

had two or more dailies from the same layer of the umbrella model that year. Both of

 

these percentages had decreased slightly from 1983 (p. 38-39). Still, in 1988 about 22%
of U. S. counties had dailies that potentially could engage in both interlayer and intralayer
competition.

The study (Lacy & Davenport, 1994) also examined concentration in daily
markets using a random sample of 500 counties. Results showed that both daily and
Sunday markets had a high degree of concentration — three dailies tended to dominate
these markets in 1983 and 1988. However, the authors concluded the newspaper industry
“was far more competitive than many have thought, and it was not declining in that
competitiveness” (p. 43).

Another study that examined newspaper penetration in non-metropolitan counties
in Michigan during four years found intense competition between satellite dailies and
weeklies, and noticeable competition between satellite dailies and metro dailies (Lacy &
Dalmia, 1993). The Michigan study was extended to throughout the United States using
a random sample of 381 counties (Lacy, Coulson, & Cho, 2001). This unpublished study
also used three layers, and again found competition between different layers. The study
also concluded umbrella competition varies between metropolitan and non-metropolitan
areas, identifying weekly circulation as an important intervening variable.

These studies of the umbrella model suggest many newspapers not only face the
threat of entry, they feel the pressure of competition. Therefore, newspapers can be
expected to look for strategies that allow them reduce that pressure. One way to do this is
find a way to take advantage of the fact that competing newspapers must be able to lower
their costs enough to match the MES shown in Figure 1. A competing newspaper can

only do this by selling a large enough “lump” of circulation to reach the MES. Clustering

l6

may allow newspapers to take advantage of this problem of “lumpiness” to deter the
entry of new competitors and protect profits.

Lumpiness can become an advantage if transportation costs are a significant
contributor to the final cost of a product (Scherer & Ross, 1990, p. 396-404). If this is the
case, a product’s market tends to be defined within a geographic region. When
transportation costs are important, consumers farther away from a company’s plant must
pay a higher price because transportation costs increase with distance. A competitor can
enter the market by locating in an area where the existing firm’s transportation costs are
high enough to be undercut and still allow the new firm to earn a profit.

Newspapers incur substantial transportation costs when they distribute c0pies to
individual subscribers. If a newspaper extends its circulation to distant areas where the
population is increasing it may be forced to increase subscription or advertising prices to
cover the increased transportation costs. However, a second newspaper that is closer
could enter this growing market with lower prices because its transportation costs are
lower. There is evidence that newspapers benefit from reducing circulation that is costly
to maintain (Blankenburg, 1982). Gannett newspapers in the 19703 reduced circulation
and market share when compared with a control group of newspapers. The strategy was
rewarded with increases in revenues and profits. However, Blankenburg did not measure
the effects of competition on the changes in circulation.

Economic theory suggests firms can solve the problem of transportation costs and
deter the entry of new competitors by locating new plants at the edges of the territories
that existing plants serve (Scherer & Ross, 1990). However, this strategy can only be

effective if the company locating plants throughout a region enjoys economies of scale.

17

Otherwise, new competitors can enter on a very small scale and still compete by lowering
prices. This is because “economies of scale create a lumpiness problem” (Scherer and
Ross, 1990, p. 400).

A regional cluster of commonly-owned newspapers means a new competitor
would have to produce a large, low-cost “lump” of circulation in a market where existing
newspapers are already meeting the demand needed to absorb the lump. This gives the
entering newspaper the option to circulate only to consumers unwilling to pay for the
existing newspapers. There will not be enough of those consumers to allow the entering
newspaper to reach the MES shown in Figure 1.

Clusters of newspapers, in other words, may result from a strategy to deter
competition. Instead of paying high transportation costs to circulate newspapers in more
distant areas, companies acquire existing newspapers that are closer to those areas. The
acquisition of these newspapers reduces transportation costs. The scale economies in
newspaper production also deter the entry of new competitors.-

Effects of Reducing Campetition

There is evidence that the horizontal mergers which create newspaper clusters are
associated with decreased competition. Lacy and Simon (1997) compared commonly-
owned newspapers in adjacent counties to a control group of separately-owned
newspapers in adjacent counties. The commonly-owned newspapers had an average of
6% less penetration outside their home counties (p. 820). F u (2000) examined the home
county circulation of newspapers in clusters, and found a positive association between

circulation and common ownership of adjacent newspapers (p. 32).

Fu also examined how contacts in multiple markets between a group that owns
clustered newspapers and its competitors affected circulation. Theory suggests that if
competitors have asymmetric positions in different markets, they will collude to refrain
from competition because they can retaliate against each other in different markets (p. 15-
16). This is called mutual forbearance. This might take place if the same groups have
newspapers in different markets, but the group with the highest market share in one
market has the lowest market share in the other. Fu found a positive association between
the circulation of clustered newspapers and multimarket contact with competing groups
(p. 32).

Neoclassical economic models suggest that mergers which lower production costs
should also result in lower prices as firms exploit the cost advantage relative to their
competitors. However, if the mergers have reduced or eliminated competition, then the
firm can be expected to increase prices as it exploits market power arising from the lack
of consumer choice.

Recall that in a competitive market new firms enter if prices rise above costs.
Firms in these markets cannot control the price they charge, they must instead accept
whatever price is dictated by conditions in the market. Product differentiation gives firms
some control over prices, but relatively high prices still will convince consumers to
accept the imperfect substitute of another firm’s differentiated product. As the number of
firms in a market decreases the market becomes an oligopoly. Oligopoly is characterized
by a relatively small number of firms that divide a market while keeping prices relatively

high, but still below the levels found in a monopoly. A monopoly sets its own price. The

 

only constraint on a monopolist’s price is whether consumers are willing to pay for its
product.

Newspaper markets have some characteristics of monOpolistic competition, but
tend toward oligopoly or monopoly (Lacy & Simon, 1993, p. 92-94; Litman, 1988;
Morton, 1997). This suggests newspapers have some degree of market power even when
they face competition. Therefore, mergers that eliminate competition will increase
existing market power. Therefore, this dissertation examines both the size and

geographic extent of clusters and how clusters influence prices.

20

CHAPTER 2
MODELING THE BEHAVIOR OF CLUSTERED NEWSPAPERS

Newspapers complicate economic analysis because they are joint products that
provide information, ideas and entertainment to an audience, and then sell advertisers
access to that audience (Lacy & Simon, 1993, p. 5). Newspapers do all this using a
single production process. This makes advertising prices dependent on the newspaper's
ability to attract an audience. However, the ability to attract an audience depends on how
much utility is available from the news and advertising that the newspaper produces.
Clustering ’5 Effects on Quality

The effect of competition on news coverage is generally described by the
financial commitment model first developed by Litman and Bridges (1986). Lacy (1992)
formalized this model as: (1) increased competition leads to an increase in financial
resources devoted to covering news, and (2) increased spending results in an increase in
the quality of news coverage which (3) increases reader utility resulting in (4) increased
circulation and improved market performance. Empirical tests support the financial
commitment model (see, e.g., Lacy, 1992; Lacy, Shaver, & St. Cyr, 1996).

This model suggests that if clustering decreases competition less money will be
spent covering news and quality will decrease. The financial commitment model also
suggests the decline in spending eventually will lead to a decline in audience and a
consequent loss of advertising revenue.

However, two other variables can mitigate this effect. First, if newspaper clusters

reduce production costs these economies may create an option for managers who want to

21

 

maintain quality. Cost reductions could allow those managers to offset declines in
newsroom spending, and in turn limit the reduction in quality.

Second, if demand for a newspaper is relatively inelastic, then a decline in quality
does not necessarily imply there will be a proportionate loss of audience or advertiser
demand for that audience. Elasticity, or the percentage change in circulation divided by
the percentage change in quality, measures the sensitivity of readers to changes in
quality. Elasticity can also be used to determine whether readers and advertisers are
sensitive to changes in price.

Let Q: the quantity, or circulation, of a newspaper. Let S = spending on the
news, or quality. Let P = the newspaper’s cover price. Elasticity of demand is
(AQ/Q)/AS/S) when quality is at issue, or (AQ/Q)/AP/P) when price is at issue. If the
result is less than 1, then demand is inelastic. Readers are less sensitive to changes in
quality or price. If the result is greater than 1, demand is elastic. Readers are more
sensitive to changes in quality or price. If the result is 1, then demand is said to be unit
elastic.

Lacy (1989) modeled how quality affects demand elasticity when newspaper
firms compete. Lacy argued that if newspaper readers expect a minimum level of
acceptable quality, there will be a kink in the newspaper’s demand curve. Demand is
more elastic below the minimum quality level, and less elastic above the minimum
quality level. This means consumers are less sensitive to changes in quality above the
minimum level. This is because higher quality makes the product more useful --

consumers receive more utility from the newspaper.

22

Quality (Newsroom spending)

     

-_———_—————_—-—————— ———————— ———-——

 

 

I
I
I
I
L_
I
I
I
I
I
I
I
I
l
I
I
l
l
I
I
I
l
I
l
I
I
l
I

I
|
|
|
l
I
l
I
l
l
l
I
l
|
l

 

Qc Qm Circulation (Q)

Figure 2. Competition’s effects on reader demand for quality.

Quality (S) is measured as newsroom spending in competitive (c) and monopoly (m)
markets. The demand (D) for quality shifts in a monopoly market because substitute
newspapers are not available. This figure was adapted from Lacy, S. (1989). A model of
demand for news: Impact of competition on newspaper content. Journalism Quarterly,
66(1), 40-48, 128.

23

However, Lacy (1989, p. 46) noted that if only one newspaper serves a market,
the demand curve shifts because of the lack of substitutes and readers accept a lower
level of quality. Figure 2 shows how this shift, to the right and down, lowers the
minimum acceptable level of quality. The minimum acceptable quality for consumers in
a competitive market is Sc, which results in circulation of QC. However, the minimum
acceptable level of quality in a monopoly market, Sm, results in a higher level of
circulation, Qm. This allows the monopoly newspaper to maintain circulation while
reducing quality. This suggests that if newspaper clusters reduce competition, quality
may also be reduced without a proportionate loss of audience.

Martin (1997, p. 15-16) developed a simple model of consumer surplus (V) from
newspaper use as a function of price (P), the utility a newspaper provides to readers (U,
determined by the quality of content) and opportunity cost:

V=U—Pn—Poc (2.1)

In this model Pn is the cover price of a newspaper and P0C is the opportunity cost
of time devoted to reading the newspaper. Opportunity cost is assumed to represent the
value of using a competing form of mass media (Martin, 1997). The first equation can be
rewritten to show high newspaper’s cover price can be set. This is accomplished by
setting V equal to 0, then adding newspaper’s cover price to both sides of the equation.

Rewriting Equation 2.1 then shows the cover price of a newspaper must be

Pn S U — P0C (2.2)
or consumers will not subscribe. This model implies that competition from other media,
which offers alternatives to reading the newspaper, also can affect the elasticity of reader

demand.

24

 

Both models (Lacy, 1989; Martin, 1997) imply that as competition decreases
clustered newspapers can reduce the quality of news without risking a substantial loss of
audience. This is because (a) the consumer cannot switch to other newspapers that may
provide more utility so (b) the relative "price" of the news that still is available has
declined. However, both models also suggest the reduction in quality is constrained
either by (a) some minimum quality level that consumers find acceptable or (b) the
availability of imperfect substitutes such as television or other competing forms of mass
media.

One study indirectly supports this argument. Lacy and Martin (1998) concluded
newspapers owned by a company that emphasized high profit margins at the expense of
news lost substantially more circulation than a control group of newspapers during the
19805. The study controlled for competition, but did not directly measure quality.

Other studies also show competition results in more spending on news or more
space for news. For example, two studies of publicly-owned newspaper companies
(Blankenburg & Ozanich, 1993; Lacy et al., 1996) concluded there is less emphasis on
profit at the expense of newsroom spending if the original owners or their successors
have more control of the company. Lacy, et al. (1996) also concluded these newspaper
companies increase news budgets when newspapers they own face competition. They
found that companies facing competition had a lower return on equity, but were forced to
spend more to keep their newspapers “attractive to readers and advertisers” (p. 339).
Lacy (1988) studied the amount and type of news in 1 l4 newspapers, concluding the

“intensity of intercity newspaper competition influences the percentage of space in a

25

newspaper given to news coverage and coverage of news in the city in which the
newspaper is located" (p. 405).

However, the two studies of newspaper clusters did not directly measure quality
(Fu, 2000; Lacy & Simon, 1997).2 Fu (2000) included independent variables estimating
the effects of clustering on the elasticity of demand associated with a newspaper’s price.
He concluded that newspaper clustering is associated with a 22% reduction in cover-price
elasticity of demand. This suggests that clustered newspapers may be able to
substantially reduce quality without losing circulation.

This discussion suggests that if clustering reduces competition, it also will allow
newspapers to reduce newsroom spending and the quality of news coverage without a
disprOportionate loss of circulation. However, this will only be possible if (a) quality still
is above the minimum acceptable level represented by the kink in Figure 2 or (b)
competition from other media does not significantly reduce the newspaper’s audience.
This dissertation examines spending on news by companies that own clustered
newspapers to determine how quality is affected. Measures used include the size of the
news staff and the space available for news at individual newspapers. Overall spending
on the newspaper division of companies that own clustered newspapers provides a second
level of analysis.

Clustering ’5 Effects on Prices
The previous discussion implies there is an interaction between quality and prices.

Industrial Organization models suggest a fomralization of this problem (Scherer & Ross,

 

2 Fu (2000) measured the average number of pages printed each day, arguing this represented quality
because news, advertising, editing and print quality are highly correlated. However, this fails to account
for newsroom spending or factors such as the ratio of newspaper staff to wire service stories (Lacy and
Simon 1993, p. 64-67).

26

1990, p. 70-71). These models measure the price-cost margin, known as the Lerner
Index:

M = (P — C)/P
In this model P is the price of one unit of a product and C is the marginal cost of
producing one additional unit of that product. In competitive markets, price is equa
the marginal cost of producing an additional unit of output so M = 0. However, as
competition decreases firms can raise prices above marginal cost because consumer
cannot then substitute products from other, lower priced firms. Therefore, the value
the Lerner Index increases as markets become less competitive.

The Lerner index is related to elasticity of demand (Tirole 1988, p. 66). Let
the price a monopoly new3paper would charge, C the change in cost associated with
change in quantity or marginal cost, and e is price elasticity of demand or the Percer
Change in Demand/Percentage Change in Price:3

Pm — C/Pm = 1/8

Equation 2.4 implies that a monopolist's increase of price above marginal cost is
"inversely proportional to the demand elasticity" (Tirole 1988, p. 66). In other worc
demand becomes more elastic and consumers respond to small increases in price wi
disproportionately large decreases in consumption, and the monopolist must accoun'
the effect that high prices have on consumption. This equation also implies the
monopolist always operates in the elastic portion of a demand curve because if e < 1
price-cost margin will be negative -- the monopolist will actually charge a price belt

marginal cost.

 

3 The relationship between price and demand is negative.

27

Interactions Between Newspaper Prices

The application of this analysis to newspapers is complicated by the fact
newspapers actually must decide how to adjust two sets of prices -- prices for advertising
and prices for copies of the newspaper. These prices are charged to two different sets of
consumers -- advertisers and readers.

This means newspapers must set prices based on the demand in two separate
markets. However, demand in these markets interacts. Advertisers are buying access to
the newspaper’s readers, while readers are, in part, buying newspapers to read the ads.
This suggests there is a negative relationship between advertising prices and newspaper
cover prices. If cover prices decrease, the newspaper’s circulation and audience increase.
This, in turn, makes the newspaper more attractive to advertisers, enabling it to raise
advertising prices.

Bucklin, Caves and Lo (1989) constructed one formal model of this relationship
(p. 632-634). Their model suggests that advertising and cover prices must also be set so
the newspaper produces enough revenue to cover its costs. These costs include
producing advertising, gathering the news, and printing and distributing the newspaper.
Therefore, this model (Bucklin et al., 1989) suggests newspapers adjust cover prices
based on “the ability of additional circulation to induce additional advertising [sales],
which feeds back positively to the increase in total revenue resulting from any decrease in
the circulation price” (p. 634).

In other words, decreases in circulation price must be more than offset by
increases in advertising sales or prices. The study (Bucklin et al., 1989) points out that

changes in the demand for either circulation or advertising can reduce the revenue earned

28

 

from both. This is because if advertising sales decrease, then circulation prices must be
raised to make up for the lost revenue, and vice versa. Merrilees (1983) argues this
interaction between audience size and advertising demand explains why monopoly or
duopoly newspapers may want to set their cover prices below the marginal cost of
producing additional newspapers.

This has clear implications for newspaper clusters. As newspaper markets
become less competitive, allowing individual papers to begin acting like monopolists,
there will be a shift toward pricing advertising in the elastic portion of the demand curve.
Another way to think of this is that clustered newspapers will increase advertising prices
until the price-cost margin 2 1. At this point, the elasticity of demand becomes important
in determining the actual magnitude of further increases in price.

This interaction between circulation and advertising also suggests that clustered
newspapers can select from two alternative strategies when setting the level of newsroom
spending, or quality. The first alternative is to reduce costs by reducing spending until
quality is below the minimally acceptable level. As noted earlier, Lacy (1989) suggests
this is the point where demand for news becomes elastic. Therefore, the magnitude of
further spending cuts will depend on how much audience elasticity of demand for news
exceeds 1. Given the importance of audience to advertisers, newspapers adopting this
strategy risk substantial reductions in advertising revenues as readers transfer their
attention to competing forms of mass media.

This suggests a second strategy, which is to maintain the quality of news at a level
where demand is inelastic -- perhaps at the minimum acceptable level represented by the

kink in Figure 2. This, in turn, would reduce the elasticity of demand among advertisers

29

 

because the size of the audience, measured in circulation, does not decline. The clustered
newspaper can then impose even larger price increases on its advertisers.

The second strategy would be moderated by two considerations. First, the cost of
maintaining quality cannot exceed the additional revenue generated by the advertising
price increases. Second, the price increases cannot be so large that advertisers begin to
shift business to other forms of media.

Still, assuming that advertiser elasticity of demand for access to an audience is a
function of audience size, clustered newspapers still have an incentive to operate on the
inelastic portion of their audience demand curve.

This relationship will be complicated by the fact that some advertisers target
certain segments of an audience (Shaver, 1995), so increases in cover price will only lead
to reduced advertising prices if the readers who stop taking the newspapers are
considered desirable. However, this does not negate the general point that interactions
between the demand for advertising and circulation require that newspapers protect their
ability to attract both readers and advertisers. From this perspective, clustering is a
strategy that protects newspapers from competition in both markets so they can maintain
higher prices and profitability.

Recall that Lacy (1989) and Martin (1997) argue a reduction in competition
allows newspapers to maintain audience size even when the level of quality is reduced.

F u (2000) concluded clustering is associated with decreases in cover-price elasticity.
Newspaper clusters may also create economies of scale that reduce production costs.
This suggests that clusters create long-run changes in market conditions -- shifting

demand curves so demand remains inelastic even when quality is below the minimum

30

acceptable level in a competitive market (see Figure 2). Cost curves may shift in a
similar fashion, allowing newspapers to maintain minimum quality while spending less
on production then is required in a competitive market. Clustering may provide
newspapers with a cushion that reduces the effects of short-term fluctuations in the
demand for advertising or circulation. This cushion may allow clustered papers to reduce
newsroom spending while maintaining quality at levels that still result in relatively
inelastic demand. This, in turn, would allow clustered newspapers to maintain higher
advertising prices.

Implications for Quality and Prices

This discussion suggests some general results. First, as clusters are formed,
reductions in quality will be constrained by (a) whether a newspaper wants to create
inelastic demand among its audience and (b) the degree of elasticity resulting from
reductions in quality. However, these constraints will be relaxed in comparison to
competitive markets where newspapers must maintain inelastic demand to hold onto their
audiences. Therefore, the quality of newspapers in clustered markets will be lower than
the quality of newspapers in competitive markets.

Second, as clusters form, there will be interactions between quality and prices.
Clustered newspapers may be able to save costs by reducing quality without risking a
proportionate loss of audience. This suggest that the negative relationship between
advertising and cover prices will be less intense in clustered markets. Newspapers in
clustered markets will have a cushion allowing them to respond to changes in demand for
either advertising or circulation with smaller prices changes than newspapers in

competitive markets.

31

Therefore, this dissertation examines the effect of clustering on both advertising
and cover prices. As noted earlier, the two clustering studies did not examine cover
prices as a dependent variable (Fu, 2000; Lacy & Simon, 1997). However, Lacy and
Simon (1997) found clustering is associated with decreased competition. F u (2000)
concluded clustering is associated with decreased price elasticity and increases in home
county circulation, traditionally the core of a newspaper’s market. These results imply
clustered newspapers may not have to lower cover prices to increase circulation as much
as newspapers in competitive markets do. Therefore, this dissertation includes cover
price as a dependent variable.

F u (2000) did include advertising prices as a dependent variable, concluding that
clustered papers charge more for a standard column inch. This study did not control for
competition from other daily newspapers.4 Fu did include the number of radio and
television stations in a market to control for competition from other media.

However, advertising price studies are sensitive to different measures. Research
that did not adjust for circulation or did not measure the intensity of competition failed to
show expected relationships between competition and prices (Lacy and Simon 1993, p.
108-109). Ad prices can be adjusted to reflect the advertising cost per thousand copies of
a newspaper (cpm). This is the industry standard because it allows advertisers to
compare prices while controlling for differences in circulation. When researchers used
the ad cpm and measured the intensity of competition, they found the expected negative
relationship between prices and competition (Shaver & Lacy, 1999, p. 731). Shaver and

Lacy argue the contradictory results reflect different measures and the complexity of

 

4 The study (Fu 2000) argues advertisers regard each newspaper reader as another potential customer.
Therefore, if newspapers reach separate audiences they will not compete for advertising.

32

advertising markets where cost is only one factor determining where to place ads.
Another factor is advertiser perceptions of ability to reach a target audience (p. 732-733).

Shaver and Lacy (1999) also argue the assumption that newspapers compete with
other media for advertising may explain why the US. Justice Department, with one
exception, has not investigated the potential anti-competitive effects of newspaper
clustering. Shaver and Lacy used a convenience sample of 40 papers which had released
the number of lines of advertising they sold in 1994. The papers also made public their
run of the paper (ROP) advertising lines, a measure of local retail advertising sold during
the year. The study included the advertising cpm and controlled for competition from
other newspapers and from television and radio stations.

The study (Shaver & Lacy, 1999) concluded that newspapers selling ROP ads do
compete with radio and television stations in some markets. However, other dailies
provided the strongest competition for total advertising lines, probably because they are
the best substitutes for classified and insert advertising (p. 739).

Shaver and Lacy (1999) did not include a variable measuring newspaper clusters.
The small, non-random sample in their study also precludes generalization.

The need for more research into the effects of clustering on advertising prices is
apparent. Therefore, this dissertation also examines advertising prices as a dependent
variable. This study examines clustering using both unadjusted advertising prices and

advertising cpm.

33

CHAPTER 3
HYPOTHESES

The theoretical discussion in the previous chapters suggests that newspaper
clusters exist when newspaper in adjacent markets are owned by the same company.
Clusters may allow these firms to take advantage of decreased costs by increasing
earnings without imposing price increases on advertisers or readers. If there are cost
savings in news production, then quality might also be maintained without compromising
earnings. This implies that if there are cost savings associated with newspaper clusters,
firms will prefer to use those savings to increase profits instead of reducing prices or
increasing quality. This is possible because in clustered markets competition is reduced,
so the pressure to use cost savings to gain market share is also reduced.

However, the discussion also suggests that clusters allow newspapers to exploit
economies of scale and effectively deter the entry of other newspapers into clustered
markets. This reduction in competition means clustered newspapers need not maintain
the pre-clustered status quo. Clustered newspapers are less constrained than competitive
newspapers when imposing price increases on advertisers, reducing the quality of news
offered to their readers, or imposing price increases on those readers. The theoretical
discussion suggests clustered newspapers can make some or all of these changes and still
reap the benefits of lower costs and higher earnings.

These general propositions will be examined using four dependent variables:

1. The advertising prices newspapers charge.

2. The cover prices newspapers charge.

34

 

3.

The quality of news coverage measured by newsroom spending and by the

space available for news content.

4.

The profit a newspaper firm earns.

The propositions will be examined using nine independent variables:

1.

2.

9.

The size of a newspaper cluster.

The distance between newspapers in a cluster.

. The extent of a newspaper’s participation in a cluster.

. The cluster’s share of a newspaper market.

. The intensity of competition from other newspapers in the market.
. The intensity of competition from other media in the market.

. Production costs for companies that own clustered newspapers.

. The degree of diversification at companies which own clustered newspapers.

The degree to which insiders, and not the public, have ownership control of

companies that own clustered newspapers.

An additional independent variable, the number of multi-market contacts between

different newspaper companies, was considered for inclusion in this dissertation.

However, the markets in the study proved too complex to fully identify these contacts, so

the variable was dropped from the study.

Expected Relationships Among Variables

The theoretical discussion suggests the following relationships between

independent and dependent variables in this model. These relationships are stated as

explicit hypotheses.

35

The newspaper industry is a mature industry where markets are defined
geographically by the extent of circulation. Transportation costs are significant when
distributing newspapers, but newspapers also enjoy economies of scale. Newspaper
companies assemble clusters of commonly-owned papers in adjacent markets to deter
competition, a strategy that relies on economies of scale to make it difficult for
competitors to enter these markets. Lacy and Simon (1997) found that clustered
newspapers are less aggressive competitors than non-clustered newspapers.

Hypothesis 1. Clustering is a strategy to reduce competition from other
newspapers in markets where clustered newspapers operate.

Hypothesis 1a: Clustered newspapers are less aggressive competitors.

This reduction in competition, in turn, affects prices, quality, profits and costs.
Neoclassical economic models describe a negative relationship between
competition and prices. This prediction is supported by newspaper studies that measure
advertising cpm and the intensity of competition. One clustering study (Fu, 2000) found

a positive relationship between clustering and advertising prices. The theoretical
discussion also suggests a negative relationship between competition and newspaper
cover prices. This relationship, however, is moderated by interactions between demand
for circulation and demand for advertising since advertisers are buying access to the
newspaper’s readers.

Clustering is a strategy to reduce competition and increase market power to set
advertising and cover prices by decreasing the elasticity of demand for advertising and
demand for circulation. The theoretical discussion also suggests that competition for

advertising and circulation is not confined to other newspapers. Other media, such as

36

 

 

 

 

television and radio, compete for some forms of advertising but not for others. These
other media also compete for a newspaper’s audience, thereby affecting circulation
demand.

Hypothesis 2: There is positive relationship between clustering and advertising
prices. This is because clustering reduces newspaper competition, reducing the elasticity
of circulation demand. This makes clustered newspapers less sensitive to advertising
competition from other newspapers and media.

Hypothesis 2a: There is a positive relationship between clustering and
newspaper cover prices. This is again because clustering reduces elasticity of circulation
demand. This makes clustered newspapers less sensitive to competition for audience
from other newspapers and media.

Hypothesis 2b: Clustering moderates the interaction between demand for
circulation and demand for advertising which usually creates a negative relationship
between ad and subscription prices. This negative relationship is weakened because
circulation demand is relatively inelastic.

Hypothesis 2c: If newspapers are not clustered, the negative relationship
between advertising and cover prices is strengthened because circulation demand is
relatively elastic.

Hypothesis Zdz. There is a stronger effect of competition on advertising prices in
markets without clustered newspapers. The demand for these newspapers is more elastic,
so they are more sensitive to advertising competition from other newspapers and from

other media.

37

Hypothesis 2e: There is a stronger effect of competition on subscription prices in
markets without clustered newspapers. This again is because demand is more elastic,
making these newspapers more sensitive to competition from other newspapers and
media.

The financial commitment model predicts a positive relationship between
competition and newsroom spending to increase the quality of content. This model states
that newspapers spend more on quality to increase demand for circulation when they are
threatened by competition. There is empirical support for this model (Lacy, 1992a; Lacy,
Shaver, & St. Cyr, 1996; Litman & Bridges, 1986)

Hypothesis 3: There is a negative relationship between clustering, which reduces
competition, and newsroom spending. This, in turn, reduces the quality of clustered
newspapers because they face less elastic demand.

Hypothesis 33: There is a positive relationship between competition and
newsroom spending in markets without clustered newspapers. These newspapers can
differentiate themselves by increasing quality. The theory of monopolistic competition
suggests differentiation will reduce the elasticity of demand.

The neoclassical models that predict a negative relationship between competition
and prices also suggest what happens when competition is reduced. As markets become
less competitive, newspapers can increase their prices above their costs. This so-called
monopolists’ mark-up, however, will attract new competitors seeking to undercut the
newspapers unless anti-competitive strategies are in place. Clustering is a strategy to

deter competition.

38

Hypothesis 4: There is a positive relationship between clustering and the profits
earned by firms that own clustered newspapers. This is because clustering allows
newspapers to charge higher prices without attracting new competitors into clustered
markets.

The theoretical discussion also states that newspaper companies may form
clusters to reduce production costs and protect profits in a mature industry. Economic
theory suggests cost savings from economies of scale can only be achieved if production
processes are rearranged properly — for instance if more than one newspaper is printed on
the same press. Clustering makes this possible because commonly-owned newspapers
are located adjacent markets. The trade press also reports many companies that own
clustered newspapers are rearranging production processes (Bass, 1999).

Hypothesis 5. There is a negative relationship between clustering and the costs
incurred by companies which own clustered newspapers.

Hypothesis Sa: Clustered newspapers are more likely than non-clustered
newspapers to share resources in an attempt to save costs by creating economies of scale.
This is possible because clustered newspapers are in adjacent markets and can more
easily rearrange production processes.

Clustering occurs on one level of analysis, the individual newspaper’s market.
The diversification of newspaper companies into different lines of business occurs at a
second level of analysis. The literature review discussed conflicting evidence regarding
the effects of diversification. Firms that diversify into different lines of business hoping
to achieve economies of scope often fail to increase profits. This occurs at a higher level

of analysis than clustering does. One reason for the failure of diversified firms to create

39

increased profitability may be that they are more complex, making it more difficult for
managers to efficiently coordinate business activities.

However, one study of newspaper firms found diversification was associated with
increased profits (Picard & Rimmer, 1999). This study included a cost measure, but did
report relationships between cost and diversification or examine potential economies of
scale. This study concluded diversification offers a hedge against fluctuations in
advertising sales, and therefore is associated with higher profits.

This suggests that diversification will interact with clustering, but leaves open the
question of exactly how. As a general rule, however, diversified newspaper firms which
also own clustered newspapers can be expected to take advantage of clusters to increase
profits in markets where the firm faces less competition. If diversification creates
problems managing different lines of business to achieve company-level economies of
scale, these firms will have an incentive to take advantage of clusters to compensate. If
[diversification is a hedge against fluctuations in advertising markets, then the reduction in
competition associated with clustering would provide an additional hedge. In either case,
companies could be expected to take advantage of the lack of competition in clustered
markets by increasing prices and reducing newsroom spending. This will be possible
because demand is less elastic in these markets.

Hypothesis 6: There is a negative relationship between profits and the degree of
diversification at firms that own clustered newspapers. This is because problems
managing multi-business firms outweigh the benefits of hedging against fluctuations in

the advertising market.

40

Hypothesis 63: There is positive relationship between advertising and
subscription prices and the degree of diversification at firms that own clustered
newspapers.

Hypothesis 6b: There is a negative relationship between newsroom spending and
the degree of diversification at firms that own clustered newspapers.

Empirical studies also suggest that pressure to maintain or increase profits at the
expense of quality is affected by whether a company is publicly owned (Blankenburg &
Ozanich, 1993; Lacy et al., 1996). This is because publicly-owned companies attempt to
maintain high stock prices by increasing profit and reducing earnings fluctuations. This
can be accomplished by increasing advertising prices and reducing spending on news,
which in turn reduces quality. This is also accomplished by reducing circulation in areas
where circulation costs are high (Blankenburg, 1982).

However, the degree of public ownership moderates these effects. Companies
where more stock is controlled by the public tend to define quality based on "financial
concerns, not by news values" (Lacy et al., 1996, p. 332). This study found the tendency
to favor profits over quality was lessened at companies where the managers or directors
still controlled substantial portions of the stock.

This suggests pressures to take advantage of clustering to maintain or increase
profitability will be moderated by the degree of inside control at companies that own
clustered newspapers.

Hypothesis 7: There is a negative relationship between profits and the amount of

stock controlled by insiders at firms that own clustered newspapers.

41

Hypothesis 7a: There is negative relationship between advertising and
subscription prices and the amount of stock controlled by insiders at companies that own
clustered newspapers.

Hypothesis 7b: There is a positive relationship between newsroom spending and

the amount of stock controlled by insiders at companies that own clustered newspapers.

42

 

CHAPTER 4
METHOD

The hypothesized relationships were tested by comparing the characteristics and
behavior of clustered newspapers with a control group of papers that are not clustered.
Data was collected for two years, 1988 and 1998, to examine whether these relationships
were consistent. Ten years is long enough to allow for changes in the ownership of
newspapers as new clusters form and old ones are reconfigured. Ten years is also long
enough to allow for changes in capital investments, such as realignment of production
processes to achieve economies of scale, that might alter the economics of clusters. This
long-run comparison of clustered behavior provides a more robust examination of
hypothesized relationships than a single-cross sectional analysis.

Defining and Identifying Clustered Newspapers

Clustered newspapers were defined as commonly-owned newspapers operating in
adjacent markets. The local newspapers included in this study typically have their largest
market share in the county where they are located. These newspapers often circulate in
more than one county and therefore operate in multi-county markets, but this circulation
represents an addition to their core circulation in their home county. Therefore, only the
newspaper’s core market -- its home county -- was used to identify clusters. Newspapers
were considered to be operating in adjacent markets if they were (a) Operating in separate
towns within the same county or (b) in counties with a common boundary (Lacy &

Simon, 1997).

43

This definition excluded commonly-owned newspapers -- typically a morning and
afternoon paper -- in the same town because that represents a form of vertical, not
horizontal, integration. However, commonly-owned morning and evening newspapers
were included in clusters if a third newspaper in an adjacent market was also owned by
the same group. Newspapers that were part of a joint-operating agreement, where
separate companies combine the business operation of two newspapers, were excluded
from the study because they are exceptional. Fewer than five markets with joint-
operating agreements were excluded from the study.

The rules in this definition of clusters were used to identify all clustered
newspapers in the United States during 1988 and 1998. Information about these
newspapers and their markets came from standard industry references published in 1989
and 1999 (American Newspaper Markets Circulation 89-90, 1989-1990; Jessell, 1999;
Maddux, 1999; SRDS Circulation 99, 1999; Taliaferro, 1999; Velez, 1989; West, 1989).
Much of the data in these references is from the year prior to publication.

The clustered newspapers were identified using lists of all newspapers owned by
newspaper groups operating in the United States. These lists are published in the Editor
& Publisher International Year Book. This annual also provides individual listings of all
dailies, including the address and county where the newspaper is located and the name of
its owner.

The lists of group-owned newspapers were first used to locate all of these
newspapers geographically. Maps were then used to identify commonly-owned
newspapers in the same or adjoining counties. The initial location of newspapers and

their counties was made with county maps published by Standard Rate & Data in its

44

comprehensive listings of newspaper circulation in 1988 and 1998. The 1999 edition of
the Editor & Publisher International Year Book also included county maps. However,
the maps in both references are relatively crude, so county boundaries were also checked
against a standard road atlas published by National Geographic. Additional checks were
made with a computerized road atlas, DeLorme Street Atlas USA, Version 7.0.

This procedure identified clustered newspapers in 39 states in 1988 and 42 states
in 1998. However, a preliminary examination of statistics for circulation, annual
subscription prices and the number of wire services showed that the distribution of these
variables had a substantial positive skew. Numerous newspapers had circulation,
subscription prices and wire subscriptions that were more than three standard deviations
from the mean. Newspapers with circulation more than three standard deviations were
eliminated, but the skew remained. In both years, several iterations of this procedure
failed. to eliminate outliers even after 25% of the population was eliminated.

An alternative approach identified outliers using boxplots and normal probability
plots. Normal probability plots were used to examine differences between expected and
observed standardized scores for circulation. The plots were used to eliminate the largest
3% of clustered newspapers in 1988, and the largest 3% of clustered newspapers in 1998.

This procedure defined the 1988 study population as clustered newspapers with a
circulation of 100,000 or less. In 1998, the study population included clustered
newspapers with a circulation of 106,000 or less. There still were outliers apparent on
the plots used in the procedures, but they were judged to be close enough to the other data

points to proceed with drawing a sample.

45

A random sample of 200 clustered newspapers was selected each year from the
study population for further analysis. All clustered newspapers, except those in Los
Angeles, New York, and Chicago, were eligible for inclusion in the sample. Newspapers
in these three cities were excluded because the size and complexity of the markets makes
them exceptional. About three newspapers in the counties where these cities are located
were excluded from the sample. A control group of 200 newspapers was also selected
randomly from the same states where the study newspapers were located. The control
group was selected from state-by-state listings of individual newspapers in the Editor &
Publisher International Year Book.

Information About Newspapers

Information about newspapers in the study came from a variety of sources. Basic
information about the newspaper’s owner, its advertising price, subscription price, and
the number of wire service subscriptions it had came from the Editor & Publisher
yearbook. Detailed information about each newspaper’s circulation and the circulation of
other newspapers in its market was gathered from Standard Rate & Data Circulation
listings.

Standard Rate & Data publishes county-by-county information throughout the
United States. To accurately measure the extent of newspaper competition in the study,
complete data was gathered for all counties where each newspaper circulated, not just for
the home county. This data was broken down to identify the circulation of the clustered
or control newspaper in the study. A second category included all circulation from other
newspapers that were part of a cluster because this does not represent competition for

clustered newspapers. A third category included circulation for all other daily

46

newspapers, which compete with the clustered and control newspapers. A fourth
included weekly newspaper circulation, which was also assumed to compete with the
clustered and control papers.

A fifth category of circulation included commonly-owned newspapers that were
not part of a cluster. For instance, a statewide newspaper might have some circulation in
a distant county where its group owns another newspaper. This commonly-owned
circulation was identified for clustered and control papers, but the variable was dropped
after initial regressions showed it contributed little to the analysis.

The Standard Rate & Data listings also were used to gather county-by-county
information about the number of newspapers headquartered in each county. This
provided an additional measure of competition. These newspapers were categorized as
competitors if they were not owned by the same company as the clustered or control
newspapers. The newspapers were categorized as friendly if they were owned by the
same company that owned the newspaper in the study.

Data about the population, the number of households, retail sales, and average
household income in each county also came from the Standard Rate & Data listings. The
circulation and demographic data for all of the counties were each newspaper circulated
was combined to provide data for the newspaper’s entire market. The average household
income was corrected before it was combined. This was done by multiplying the figure
for each county by the number of households in that county to recover the actual income.
Income was than totaled for all counties in a newspaper’s market and divided by the total

number of households in that market.

47

Information about Competition from Other Media

Information about the number of radio and television stations in each newspaper’s
market was gathered from listings in the Broadcasting & Cable Yearbook. In 1988, this
yearbook was also used to gather information about the number of cable television
subscribers in each newspaper’s market. However, these cable listings were no longer
available for 1998. The 1998 cable data came from the Television & Cable Factbook.

Like newspapers, radio stations, television stations, and cable companies operate
in geographically defined markets. However, the location of these markets is not always
congruent with the location of newspaper markets in this study.5 Therefore, the number
of radio stations in the market was measured as the number of stations in the town where
a newspaper was located. The number of cable subscribers was measured as subscribers
to the cable system serving the town where the newspaper was located. These measures
accurately identify radio and cable companies operating in a market where the
newspaper’s circulation is based.

Television competition was measured as the number of stations in the Area of
Dominant Influence (ADI) that included the newspaper’s home county. Every county in
the United States is assigned to an ADI for a given number of television stations. This is
a more geographically diverse market definition than the one used for radio and cable
television, but it again accurately identifies stations operating in the market where the

newspaper’s circulation is based.

 

5 Differences in reporting the location and markets of different media make it impossible to precisely
determine differences in the boundaries of these markets.

48

Local newspapers are most likely to face significant competition for advertising
from radio, cable and television companies that also operate in the newspaper’s core
market — which is always its home town or county.

A diagnosis of initial regressions using this data showed there were problems with
multicollinearity. This was corrected by (a) eliminating the population and retail sales
variables and (b) using the market share for different media in the study instead of raw
numbers for circulation, cable subscribers, and radio and television stations. Market
share was defined as the household circulation penetration for newspapers. The market
share for other media was also calculated as household penetration for cable subscribers,
radio stations, and television stations. A list of variables used in the dissertation is in
Appendix A.

Information about Quality and Resource Sharing

A mail survey was used to gather information about newsroom spending and
whether newspapers engage in resource sharing. A one-page questionnaire was mailed to
senior news executives at all clustered newspapers with a circulation of 106,000 or less
which were part of the 1998 study population. The questionnaire also was mailed to a
similar number of control newspapers selected randomly from the same states.
Responses to this survey made a second sample used for analysis of the 1998 papers.
This sample, which was separate from the random sample of clustered and control
newspapers for that year, is described in the Results chapter.

The survey questionnaire asked about the size of the newspaper’s news staff, the

ratio of news to advertising, and the size of its newshole. Four additional questions asked

49

whether the newspaper shared newsgathering, administrative and production resources.
A copy of the cover letter and the questionnaire are in Appendix B.

The resource-sharing questions provided information about whether newspapers
are making an effort to realize economies of scale. The questions about news staff and
size of newshole replicated measures of quality that Litman and Bridges (1986) used in
their financial commitment study.6 Information replicating a third variable from that
study, the number of wire service subscriptions at each paper, was gathered from the
Editor & Publisher International Year Book.

Information about Publicly-Owned Newspapers

Information about profits, cost, the degree of diversification and insider control of
stock was available only for publicly-owned companies included in the study. This
information was gathered for these companies when they had newspapers appearing in
the sample of clustered or control newspapers for both 1988 and 1998.

Financial and ownership data about these publicly-owned companies was
gathered from three sources. The first source of financial data was Compustat, a financial
information service owned by Standard & Poor’s. Compustat takes financial information
filed with the US. Securities and Exchange Commission (SEC) and converts it into
easily accessible formats. The second source, used in a few instances when data not
available from Compustat, was SEC files available on the agency’s website. The third
source used to gather information about the percentage of stock owned by company
insiders was The Value Line Investment Survey, a monthly newsletter for investors. The

newsletter calculates these percentages using information about stock ownership filed

 

6 The space measure in the original study was based on data about 134 papers gathered by Media Records,
Inc. (Litman & Bridges, 1986, p. 12).

50

with the SEC. If information was not available from Value Line, it was supplemented
with data from the SEC.
The Profit Measure

Each company’s overall profit was measured as its return on sales (Martin, 1998).

This is a measure of pre-tax profits. The measure is:

[Operating Revenue — Operating Expenses - (Non - Operating Expenses — Non - Operating Income)]

 

Return on Sales = ,
Operating Revenue

Operating revenue comes from the sale of products and services. Operating
expenses include items such as production costs and depreciation. Non-operating income
and expenses include investment income and interest expenses.

The Cost Measure

Costs for each company’s newspaper division were calculated using line of
business data. This is information about the total revenues and costs associated with a
company’s different lines of business. These lines of business are identified using
Standard Industrial Classification Codes. The cost for the newspaper division was
recovered from the line of business data for newspaper publishing. This data includes
revenue from sales associated with newspaper publishing and the profit from that
revenue. The measure is:

Cost = Segment sales — Segment profit (loss)

 

Segment Sales
The Diversification Measure
The financial data for different lines of business also was used to calculate a
diversification index for each company. This index was intended to describe the degree

to which different lines of business contributed to the company’s total revenues. The

51

index was modeled on the Herfindahl-Hirschman Index used to measure the degree of
concentration in an industry (Scherer & Ross, 1990, p. 72-73). This index squares and
then sums the market share of each firm in an industry, so that l = an industry that is a
pure monopoly.

In this dissertation, the index can be thought of as measuring the degree of
concentration, or diversification, among a company’s lines of business. The index was

calculated by squaring and then summing total revenues from each line of business:

N
Diversification Index : 2 Sales for segmenti 2

i=1

This index assigns a value of 1 to a company that only publishes newspapers.
Values less than 1 are assigned to companies with multiple lines of business. Smaller
values indicate the company’s revenues are more diverse and are not dominated by one
line of business.
The Measure of Inside Control

This is a straightforward percentage based on the amount of stock owned by the
company insiders such as managers and members of the board of directors (Lacy et al.,
1996). This information is calculated by The Value Line Investment Survey from SEC
filings detailing who owns the company’s stock.
Potential Problems with Endogenous Variables

Some variables in this dissertation may feed back into each other because their
values are determined simultaneously. If there is feedback between simultaneously
determined variables, the variables are considered endogenous instead of exogenous
(Doran, 1989, chap. 1 1). Regression analysis assumes the independent variables are not

correlated with the errors in prediction. This is because the errors are assumed to be

52

random. This assumption is violated if the dependent and independent variables are
endogenous. Violations of the assumption can alter the coefficients for the independent
variables, leading to incorrect interpretation of the regression slopes.

This problem can be addressed by identifying variables which are not determined
simultaneously with the other independent variables (Doran, 1989). These so-called
instrumental variables have a one-way relationship with the endogenous variable.
Instrumental variables can be used to correct for endogenity using two-stage regression.
In the first stage, instrumental variables are used to predict the value of the endogenous
variables. The second stage uses the corrected values of the endogenous variables in a
regression to produce a corrected result.

This problem may exist in this dissertation. The feedback between subscription
and advertising prices described in the theoretical discussion raises the possibility these
prices are determined simultaneously.

However, two-stage regression was not used in this dissertation for practical and
theoretical reasons. One technique for identifying instrumental variables is to find a
value of the endogenous variable in a previous time period (Norusis, 1994, p. 238). For
instance, the price of advertising in one year is not likely to be determined by the price of
a subscription in the next year. However, this dissertation is not a true-time series study
where such values are available.

In addition, the problem of endogeneity exists at a theoretical level and correctly
identifying instrumental variables requires a formal theoretical explication. This is done

by creating a system of simultaneous equations and then solving them to identify both

53

endogenous and instrumental variables (Greene, 2000, p. 656-657). This is beyond the
scope of this dissertation.

However, two-stage regressions were attempted using demographic variables,
such as households, and the number of radio and television stations, that appeared
intuitively to be instrumental variables. The results made no theoretical sense.

The problem of endogenity was addressed instead by examining correlations
among variables to determine whether the direction of relationships appeared robust in
regressions. This was the case with the price variables. In addition, Ordinary Least
Squares (OLS) regression coefficients in this study are interpreted with caution. Results
significant at less than the .05 level are considered weak. Results significant at the .005
level are considered stronger. These measures should reduce the chance of Type I error

arising from endogenity.

54

CHAPTER 5
RESULTS

Results in Table 1 describe the population of clustered papers in each of the study

years. There was a net increase of 55 clustered dailies from 1988 to 1998. The average

circulation of a clustered newspaper increased by 16% during the decade of the study.

Clustering spread to three more states during the study, and the average number of

newspapers in a cluster declined about 4%. The number of newspaper groups with

clustered newspapers remained stable.

Table 1: The Population of Clustered Newspapers

 

 

Year 1988 1998
Clustered Papers 442 497
Smallest/largest circulation l,299-462,084 768-1,065,54O
Mean circulation 22,1 12 26,310
Circulation Std. Deviation 38,659 67,948
Smallest-largest cluster 2-1 1 2-8
Mean cluster size 3.33 3.26
Cluster Std. Deviation 1.99 1.54
States with clustered papers 39 42
Newspaper groups with clustered papers 81 82

 

However, these figures do not tell the entire story of changes in the number of

clustered newspapers. A comparison of newspapers that were clustered in both years

identified 135 papers that were clustered in 1988 but were not part of a cluster in 10 years

later. Results in Table 2 show how 190 newspapers were added to clusters after 1988,

resulting in the net gain of 55 clustered newspapers.

55

Table 2: Changes in Clustered Newspapers From 1988 to 1998
Total in 1988 Removed from clusters Added to clusters Total in 1998
Clustered U.S. dailies 442 ‘ 135 190 497

Listings in the Editor & Publisher International Year Book were examined to
determine what happened to the 135 papers which left the population of clusters after
1988. Eighty-one, or 60%, of these papers no longer appeared in the daily newspaper
listings.

In other words, the net increase of 55 clustered dailies between 1988 and 1998 is
misleading. A substantial number of newspapers left clusters, but even more newspapers
were added to clusters. At the same time, a substantial number of clustered papers ceased
daily operations.

These complex changes had a substantial effect on the newspaper industry. The
Newspaper Association of America keeps track of the total number of dailies in the
United States. Newspaper Association figures show 153 dailies ceased publication
between 1988 and 1998. Results in Table 3 show the 81 clustered dailies that ceased
publication accounted for 53% of this total.

Table 3: Changes in the Number of US. Newspapers From 1988 to 1998
1988 Total Clustered dailies ceasing Total dailies ceasing 1998 Total

 

publication by 1998 publication by 1998
US. dailies 1,6422‘ 81" 1531‘ 1,489“
Percent dailies in 27% -- -- 33%
clusters
Total clustered 442 497

 

a Figures from Newspaper Association of America.
b Based on analysis of Editor & Publisher International Year Book listings.

This decline in the number of dailies was accompanied by a substantial increase
in the proportion of United States newspapers that are part of a cluster -- one-third of all

newspapers were part of a cluster by 1998.

56

However, these results leave open the question of what happened to the 81
clustered dailies ceasing publication. Their fate was determined by examining reports of
newspapers ceasing publication in Presstime magazine, and similar information in the
Editor & Publisher yearbook. Both publications were reviewed during each year from
1989 to 1999 when information for the previous year would have been published.
Results in Table 4 show two-thirds of the dailies merged with another newspaper or went
out of business. The rest converted to weekly publication. Almost all of these papers
were still in the hands of their 1988 owners when they left the daily rolls.

Table 4: Why Clustered Dailies Ceased Publication

 

 

Total from 1988 to 1998 = 81 N Percent
Out of business or merged with another daily 51 63%
Converted to weekly publication 27 33%
lnforrnation not available 3 4%
Changed owners after 1988 9 1 1%

 

Note: Based on analysis of Editor & Publisher International Yearbook listings
and information published in Presstime.

The fact that only a few of these newspapers changed hands before they ceased
daily publication suggests their owners might have been cutting costs. Closing or
merging some of the newspapers in a cluster may be a way to rearrange production
processes to realize economies of scale. These results support Hypothesis 1 that
clustering reduces competition.

t- Tests of Differences between Clustered and Control Newspapers

. As described in the method section, the population of clustered newspapers in
both years was screened for outliers. Random samples were then selected for further
analysis. The initial screening resulted in the exclusion of 12 clustered dailies with a

circulation of more than 100,000 in 1988. A random sample of 200 newspapers was

57

selected from the remaining 430 clustered dailies. A control group of 200 non-clustered
newspapers was randomly selected from the 36 states where clustered newspapers in the
1988 sample were located.

The screening excluded 15 dailies with a circulation of more than 106,000 in
1998. A random sample of 200 newspapers was selected from the remaining 482
clustered dailies. A control group of 200 was randomly selected from the 36 states where
clustered newspapers in the 1998 sample were located.

The fact that samples for both years included newspapers in 36 states is a
coincidence. The clustered sample in each year was drawn from the entire study
population.

Table 5: Circulation Comparisons for Study Population and Random Samples

 

 

Year 1988 1998
Clustered papers Clustered Control Clusters papers Clustered Control
100,000 or less sample sample 106,000 or less sample sample
Mean circ. 17,195 16,367 18,858 17,193 17,365 19,330
Std. dev. 16,988 16,195 16,902 18,267 18,781 21,089
N 430 200 200 482 200 200

 

Table 5 compares the circulation for newspapers in the samples with the study
populations of clustered newspapers. In 1988, the sample of clustered dailies averaged
about 5% less circulation than the study population. In 1998, there was no circulation
difference between the p0pulation and the sample. The control newspapers were larger
than clustered newspapers in both years. Control papers averaged about 15% more

circulation in 1988, and about 9% more in 1998.

58

Statistical tests in this dissertation were based on information about each
newspaper and its market. As described in the Method chapter, each newspaper’s market
was defined as all of the counties where it had circulation.

The first set of statistical tests examined mean differences between clustered and
control samples of newspapers. The price and market share variables used in these tests
were not distributed normally. However, this does not violate assumptions of the t-test
because “even when samples are taken from a non-normal population, the distribution of

the sample means will be approximately normal for sufficiently large samples” (Norusis,

 

 

l993,p.252)
Table 6: t-Tests of Intercounty Circulation Differences
1988 1998
Clustered Control Clustered Control
Mean number of adjoining counties 1.19 2.44" 1.55 2.38.
where paper circulates
Std. Deviation 2.14 3.79 2.58 3.92
Paper’s mean penetration in .032 .069" .037 .062"
adjoining counties '
Std. Deviation .069 .105 .066 .096
N ' 200 200 200 200

 

* difference between clustered and control significant, p < .05
** difference between clustered and control significant, p < .005

Lacy and Simon (1977) found that clustered newspapers circulated in fewer
counties than newspapers that were not clustered. Results in Table 6 replicate these
findings. Clustered newspapers circulated in fewer counties outside their home county
during both years of the study. Clustered newspapers also averaged about half as much
penetration in counties outside their home county as the control papers did, although this
difference was less pronounced in 1998. These results support Hypothesis 1a that

clustered newspapers are less aggressive competitors.

59

This study also examined differences in the number of other newspapers
headquartered in clustered and control newspaper markets. This was done by counting
(a) the number of competing newspapers headquartered in all counties where each
newspaper circulated and (b) the number of commonly-owned newspapers headquartered
in these counties. This measure examines whether newspapers operate in the core market
of competing newspapers. The measure also examines whether newspapers circulate in
areas where they potentially can coordinate circulation with other papers owned by the

same company.

Table 7: t- Tests of Differences in Number g Papers Headgtartered in Markets

 

 

1988 1998
Clustered Control Clustered Control
Mean ofcompeting dailies 1.77 2.15 1.21 1.59
headquartered in paper’s market
Std. Deviation 2.73 2.92 2.13 2.30
Mean of commonly-owned dailies 0.73 0.00.. 0.51 0.02"
headquartered in paper’s market
Std. Deviation 1.26 0.07 0.76 0.12
N 200 200 200 200

 

‘ difference between clustered and control significant. p < .05
" difference between clustered and control significant, p < .005

Table 7 shows fewer newspapers headquartered in clustered markets for both
years, but the difference is not significant. However, clustered newspapers frequently
circulate in counties where commonly owned newspapers are located. This is rare, but
possible, for control newspapers. Recall that newspapers are clustered if they are located
in the same or adjoining counties. However, non-clustered newspapers could extend their
circulation across adjacent counties to more distant counties where commonly-owned
newspapers are located. One control newspaper did this in 1988, and one control paper

did it in 1998. These results suggest that only clustered newspapers seek opportunities to

60

circulate in markets where commonly-owned newspapers are located. This supports
Hypothesis 1 that clustering is a strategy to reduce competition.

Table 8 examines differences in the average household penetration, or market
share, of newspapers and their competitors. Clustered newspapers had significantly
smaller market shares than control newspapers in 1988 and 1998. Meanwhile, circulation
from other members of the cluster added about 7% to the market share of clustered
papers in both years. This additional circulation was enough for clustered papers to

dominate their markets with 31% penetration in both 1988 and 31% percent penetration

 

in 1998.
Table 8: t-Tests of Differences in Market Share
1988 1998
Clustered Control Clustered Control

Paper’s mean penetration in its .241 .282. .242 .288**
market

Std. Deviation .193 .182 .174 .177
Mean penetration for other .072 NA .067 NA
members of cluster
Std. Deviation .110 NA .097 NA
Mean competing daily penetration .394 .423 .294 .340“
in paper’s market

Std. Deviation .198 .195 .160 .169
Mean weekly penetration in paper’s .41 1 .471 .233 .254
market
Std. Deviation .410 .401 .296 .223
N 200 200 200 200

 

" difference between clustered and control significant, p < .05
‘”' difference between clustered and control significant, p < .005

The average market share of competing daily newspapers was lower in clustered
markets during both years of the study, and the difference was significant in 1999.
Weekly penetration was lower in clustered markets both years, but the difference was not

significant.

 

Overall, Table 8 suggests clustered newspapers are less aggressive competitors
than non-clustered newspapers. This is because clustered papers can rely on other
members of the cluster to increase their market share. There also is evidence that
clustered newspapers may face less competition from other dailies than control
newspapers do. These results support Hypotheses 1 and la that clustering reduces
competition and allows clustered newspapers to be less aggressive.

This dissertation also examines prices at clustered and control newspapers. Prices
are measured two ways. The first measure is the actual advertising or subscription price
listed in the Editor & Publisher International Year Book. The second measure uses
prices that are standardized to control for the newspaper’s size.

Recall from the theoretical discussion that the industry standardizes advertising
prices as the cost per thousand circulation. The advertising cpm allows advertisers to
compare the cost of reaching 1,000 readers at newspapers with different sizes.

This dissertation also uses a standardized subscription price per thousand
circulation. This is not a measure of price paid by readers, but of the circulation cost
faced by the newspaper. Recall from the theoretical discussion that circulation costs
vary. There can be economies of scale in circulation when many subscribers are located
close together, but circulation can be costly when subscribers are distant from the
newspaper or there are only a few subscribers in an area. In either case, subscription
prices should reflect circulation costs. Standardizing these prices to remove the effects of
circulation size should (a) allow comparisons of subscription prices across newspapers

and (b) better reflect other factors affecting the newspaper’s circulation costs. Therefore,

62

this dissertation adopts a standardize measure of subscription price per thousand c0pies,
or subscription ppm.

All advertising prices reported here are based on the price of Standard Advertising
Unit (SAU), which is one standard column inch. A11 subscription prices are for a year’s
subscription.7 Table 9 shows clustered newspapers averaged significantly higher
advertising prices in three of the four comparisons for 1988 and 1998. Clustered
newspapers charged more for an SAU in 1998, but this difference was not significant.
The lack of significance may have been because of the large increase in the variance of
prices charged by control papers that year.

Table 9: t- Tests 9’ Differences Between Prices
1988 1998

 

Clustered Control Clustered Control
Mean price of standard advertising unit $14.47 $1 128 $21.85 $20.61
Std. Deviation $18.29 $7.79 $25.38 $20.26
Mean annual subscription price $75.15 $78.61 $108.97 $109.96
Std. Deviation $22.55 $27.10 $36.13 $38.83
Mean advertising cpm $1.42 $0.80" $1.60 $1.39‘
Std. Deviation $2.22 $0.39 $1.14 $0.66
Mean subscription cpm $9.50 $7.24“ $14.01 $10.73"
Std. Deviation $10.66 $5.49 $16.15 $7.10
N 200 200 200 200

 

"‘ difference between clustered and control significant, p < .05
“' difference between clustered and control significant, p < .005

 

7 If no annual price was listed, the price for the longest available period was multiplied to determine an
annual price. This was usually a monthly or weekly price. A few newspapers only listed a daily price.

63

Table 9 shows the average advertising cpm at clustered papers was 44% higher
than for control papers in 1988. In 1998 this difference was 13%. These results support
Hypothesis 2, that clustered newspapers charge higher ad prices.

Clustered newspapers did not have significantly higher annual subscription prices
in either year of the study. However, clustered papers averaged significantly higher
subscription ppm in both years. Subscription ppm at clustered papers was 24% higher
than control papers in 1989. This difference was 23% in 1999. These results support
Hypothesis 2a, that clustered newspapers have higher subscription prices.

Results from Tests of Association

Regression analysis was used to examine relationships between clustering, prices
and measures of newsroom spending. First, a decision had to be made about which price
variables should be included in which regressions. This was done by examining
correlations between newspaper prices, circulation, and household penetration.

The correlations for clustered newspapers also used three variables measuring
different dimensions of a cluster (see Appendix A for a list of variables in the study).
The first was cluster size, defined as the number of papers in a cluster. The second
variable was the percent of a cluster’s total circulation that came from the paper in the
study. Percent measures the paper’s participation in a cluster. The third variable was the
distance, in miles, from the clustered paper in the study to the nearest other member of
the cluster.8

The correlations in Table 10 show that at clustered papers in 1988 there was a

generally positive relationship between the advertising and subscription prices. This

 

8 Distance was determined with DeLorme Street Atlas USA, version 7.0. The program was used to map
the shortest travel distance between towns where clustered newspapers were located.

64

 

 

 

contradicts the argument in the theoretical discussion that these prices have a negative
relationship because of interacting demand curves.

The unadjusted and standardized prices had different signs when correlated with
the newspaper’s circulation. The unadjusted prices increased with circulation size. The
standardized prices decreased with circulation size, perhaps because of economies of

scale.

Table 10: Pearson '3 Correlations in I 988 for Clustered Papers
Ad price Sub. price Ad cpm Sub. ppm Paper's circ. Paper's pent. Cluster size Percent_p_is_tafi
Ad price 1.000

 

Sub. price .452“ 1000
Ad cpm .610* .143M 1.000
Sub. ppm -. 120 .030 .493" 1.000

Paper’scirc. .412M .431** -.235** -.453** 1.000

Paper’s pent. —.233* -.172** -.225** -.080 -.l78* 1.000

Cluster size .574** .261** .367** -.042 .134 -.l68** 1.000

Percent -.099 .123 -.207** -.077 .238” .146* ‘ -.54l** 1.000
Distance -.220** —.085 -.204** -.084 -.035 .l9l** -.065 -.046 1.000
N 200 200 200 200 200 200 200 200 200

 

" Correlation Significant at the 0.01 level (2-tailed).
"‘ Correlation significant at the 0.05 level (2-tailed)

All prices were negatively correlated with the paper’s household penetration,
although the correlation for subscription ppm was not significant. This shows that prices
decline as a newspaper’s market share increases. This is not surprising since lower prices
should increase the paper’s circulation and also allow it to attract more advertising.

The correlations in Table 10 also show that larger clusters are associated with
higher advertising prices and higher unadjusted subscription prices. However, ad prices

decrease as the distance between clustered papers increases. This may be because

65

increasing distance provides more opportunities for competitors to enter a market. These
correlations support Hypothesis 2 that clustering increases ad prices.

There is a negative correlation between advertising cpm and percent. or
participation in a cluster. This may be because participation increased with circulation
and penetration. However, this does not support Hypothesis 2.

Unadjusted subscription prices also have a positive association with cluster size.
None of the other correlations between subscription prices and clustering variables are
significant, however. The significant correlation offers only limited support for

Hypothesis 2a that clustering increases cover prices.

Table ll: Pearson’s Correlations in 1988 for Control Payers
Ad price Sub. price Ad cpm SubJme Paper's circ. Paper's pent.
Ad price 1.000

Sub. price .573** 1.000

Ad cpm -.329* -.388** 1.000

Sub. ppm —.533** -.269** .819** 1.000

Paper’s circ. .878" .583" -.551** -.603** 1.000

Paper’s pent. -.237* -.239** .069 .l83** -.154* 1.000

N 200 200 200 200 200 200

'” Correlation significant at the 0.01 level (2-tailed)
* Correlation significant at the 0.05 level (2-tailed)

 

Correlations in Table 1 1 show positive relationships between prices, but only
when both prices are either unadjusted or standardized. The relationship between
unadjusted and standardized prices is negative. This same sign relationship between
similar price measures again contradicts the theoretical argument that demand for

circulation and advertising interact negatively. The negative correlations between

unadjusted and standardized prices may result from the different signs these prices have

66

for circulation. Larger papers have higher unadjusted prices, and lower standardized
prices. This is the same relationship found with clustered papers.

For control papers, unadjusted prices decrease with market share, the same
relationship found with the clustered papers. Standardized subscription prices, however,
increase with penetration, an association that was not significant for clustered papers.

The positive correlations between advertising and subscription prices, and
between advertising cpm and subscription ppm, are higher for control papers than for
clustered papers. This supports Hypotheses 2b and 2c that interactions between prices
are weakened at clustered newspapers. However, these interactions are not in the
predicted direction.

Comparisons of price correlations for clustered and control papers show they are
similar in many respects. However, there are some differences. Multivariate analysis
will be needed to sort these out.

Table 12: Pearson’s Correlations in 199§for Clustered Papers
_ Ad price Sub. price Ad cpm Sub. ppm Paper's circ. Paper's pent. Cluster size Percent Distance
Ad price 1.000

 

Sub. price .190” 1.000

Ad cpm -.l45* -.l39* 1.000

Sub. ppm -.3 19** .050 .84l** 1.000

Paper’s circ. .738“ .324“ -.369** -.451** 1.000

Paper’s pent. -.314** -.176* -.017 .067 -.3ll** 1.000

Cluster size —.042 -.046 .220” .256** -.094 -.084 1.000

Percent .183** .132 -.288** -.302** .340** .106 -.589** 1.000
Distance -.1 11 -.102 -.066 -.084 -.010 .182** -.093 .099 1.000
N 200 200 200 200 200 200 200 200 200

 

*" Correlation is significant at the 0.01 level (Z-tailed)
* Correlation is significant at the 0.05 level (2-tailed)

67

Table 12 shows associations between prices changed for clustered newspapers in
1998. The association between similar prices still is positive, but there is a negative
association between unadjusted and standardized prices. This again contradicts
theoretical predictions of a negative association between prices. However, these results
also suggest there is some difference between 1988 prices 1998 prices at clustered papers.
The signs for correlations between 1998 prices are similar to those for the control papers
in 1988.

Table 12 also shows unadjusted prices are positively associated with circulation.
Standardized prices are negatively associated with circulation. There is a negative
association between unadjusted prices and market share. These results are similar to
1988, suggesting again there are economies of scale in circulation.

There is a positive association in 1998 between standardized prices and cluster
size. This supports Hypotheses 2 and 2a that clustering is associated with higher prices.
The associations between distance and prices are not significant in 1998.

However, there is again a negative association between participation in a cluster and
standardized prices. At the same time, there is a positive association between unadjusted
ad prices and participation in a cluster. Meanwhile, participation has a positive
association with circulation, just as it did in 1988. These correlations may reflect the
relationships of price to circulation. However, they both support and fail to support
Hypotheses 2 and 2a.

Table 13 shows the 1998 associations between prices, circulation and penetration
at control papers were generally unchanged from 1988. The correlation between similar

prices is again positive, contradicting theoretical predictions. However, the correlations

68

 

between prices are all larger for control papers with the exception of advertising cpm and
subscription ppm. These results again offer some support for Hypotheses 2b and 2c that
interactions between prices are weakened at clustered newspapers.

Table 13: Pearson ’s Correlations in I 998 for Control Papers
Ad price Sub. price Ad cpm Sub. ppm Paper's circ. Paper's pent.
Ad price 1.000

Sub. price .625" 1.000

Ad cpm -.200** -.320** 1.000

Sub. ppm -.533** -.366** .728** 1.000

Paper’s circ. .902" .630** -.438** -.645** 1.000

Paper’s pent. -.327* -.353** .081 .307“ -.330** 1.000

N 200 200 200 200 200 200

'” Correlation is significant at the 0.01 level (2-tailed)
"‘ Correlation is significant at the 0.05 level (2-tailed)

Tables with correlations for all variables in the multivariate analysis are in
Appendix C. However, the results reported here suggest there is interaction between
demand for circulation and demand for advertising, but this interaction is not negative.
For three of the four groups of papers, the only negative relationship was between
unadjusted and standardized prices. This appears to be because unadjusted prices
increase with circulation — larger newspapers charge higher prices. Standardized prices,
however, decrease with circulation. This may reflect the presence of economies of scale
which lower costs as circulation increases thereby allowing lower prices. In other words,
the negative relationship between unadjusted and standardized prices is unlikely to reflect
the interaction of demand.

So what explains the positive relationship between like prices? Unadjusted prices

are positively correlated with circulation, but this is probably because larger newspapers

69

which operate in larger, more affluent markets, also have higher prices. It’s unlikely that
demand for circulation rises when prices go up, leading in turn to increased demand for
ads and higher advertising prices. A more likely explanation is that ad and subscription
prices are not determined simultaneously. Lewis (1995) discussed a series of studies that
show circulation demand is relatively inelastic. Blankenburg (1982) showed how
Gannett was able to raise its cover prices while cutting marginal circulation, and still
increase both revenues and profits. This suggests factors such as the size of a newspaper
and its costs may be as important as demand for determining subscription prices.

As for the unusual relationships among prices for the 1988 clustered papers,
regression analysis is needed to further illuminate these and other relationships.
However, the generally positive associations among similar kinds of prices suggests
regressions should also use either unadjusted or standardized prices. Mixing different
price measures seems likely to produce misleading results.

Results from Regressions

As discussed in the Method chapter, the descriptive statistics for variables used in
the regressions showed that most had a positive skew. Outliers for these variables were
recoded as three standard deviations from the mean (Tabachnick & F idell, 1996, p. 69).
The Method chapter also reported there was multicollinearity for regressions using
circulation variables. These problems were resolved by using variables for market share
instead.

Examination of scatter plots for the residuals showed linear relationships for all
regressions reported in this dissertation. However, the scatter plots also show the

regressions using untransformed variables frequently have a skewed distribution. In

70

 

 

 

some cases, there is evidence of heteroscedasticity. These problems do no invalidate
regression analysis, but they do weaken it (Tabachnick & Fidell, 1996, p. 138-139).
Other diagnostic statistics, such as Cook’s distance and Mahalanobis distance, also
showed discrepant cases in the 1988 regressions. These cases potentially can have
extraordinary influence on regression results (p. 133-134).

However, attempts to eliminate multivariate outliers and discrepant cases
produced results similar to the attempts to eliminate skewed cases reported in the Method
chapter. Eliminating one set of influential cases simply revealed a new set.

To reduce the chance of Type I error, correlations that are only significant at the
.05 level are considered weak in this dissertation. As suggested in the earlier discussion
of endogenity, multiple measures will be examined to determine whether relationships
appear consistent and robust instead of relying on single, possibly misleading, result.

The variables used in the regressions included the measures for prices and clusters
used in the Pearson’s correlations. The regressions also include households, defined as
the number of households in each paper’s market, and average household income in that
market. These are control variables. The regressions include as independent variables
the market share of competing daily newspapers and weekly newspapers. The market
share for radio stations, television stations and cable companies is also included in the
regressions.

The first set of regressions examined relationships between clustering variables
and advertising prices. These regressions were repeated using the control papers — which

do not have any of the clustering variables — to see how relationships might differ.

71

Table 14: 1988 Cluster and Control regressions, Advertising Price as DV
Cluster (N= 196) Control (N = 197)

 

B Std. Error Beta 1 B Std. Error Beta t
(Constant) -28.261 5.104 -5.537** .029 2.527 .012
Households .000 .000 .297 4.95" .000 .000 .554 7.805“
Avg. household .001 .000 .375 5.261 ** -.000 .000 -.009 -.901
income
Paper’s penetration 7.344 7.592 .108 .967 14.608 3.626 .364 4.029M
Clustered 13.880 8.156 .111 1.702 NA NA NA NA
penetration
Daily penetration -7.195 5.879 -.110 -l.224 -1.498 2.772 -.040 —.540
Weekly penetration -1.346 1.807 -.038 -0.745 -1 .059 .952 -.054 -1.112
Subscription price .1 12 .035 .193 3.155“ .099 .015 .370 6.645“
Size of cluster 2.864 .483 .381 5.929” NA NA NA NA
Percent of cluster’s 13.163 3.952 .217 3331-" NA NA NA NA
c1rc.
Distance to nearest -.056 .034 -.089 -1.643 NA NA NA NA
paper
Radio penetration 7487 9088 .064 0.824 -824 6548 -.011 -.126
TV penetration -1575 24533 -.049 -0.642 -4087 1343 -.228 -3.041 **
Cable penetration -.680 5.531 —.008 -0.123 -6. 166 2.970 -. 139 -2.076*
Adjusted R2 .543 .568
F statistic 18.936” 29.831**

 

"' p< .05, " p< .005

All regression tables give results for both clustered and control newspapers so

they easily can be compared. These tables should be read from left to right, with the data

for clustered papers in the left-hand columns. Data for the same variables from the

control papers is in the right-hand columns.

Table 14 shows a significant positive association between unadjusted ad prices

and the clustering variables in1988. If the OLS coefficients are accurate, the addition of

72

one newspaper to a cluster resulted in a $2.86 increase in ad price. A 1% change in the
paper’s participation in a cluster increased ad prices $13.16. The standardized
coefficients suggest cluster size had more influence on ad prices than any other variable.
These results support Hypothesis 2 that clustering increases ad prices.

The betas for the control papers show households were the most important
predictor of ad prices. However, the OLS coefficient indicates an increase of 1,000
households increases ad prices by less than a cent. Control papers are also sensitive to
competition from television and cable — both of which are negatively associated with
advertising prices. Interpretation of these OLS coefficients is difficult because of the
extremely large value for television. These results offer weak support for Hypothesis 2d
that prices at non-clustered papers are more sensitive to competition.

Large standard errors make the negative association between ad price and
competition from dailies and weeklies unstable for both cluster and control papers. This
is also true for television and cable coefficients for clustered paper.

Table 14 also shows subscription prices positively influence ad prices. If the OLS
coefficients are correct, a one cent change in subscription prices increases ad prices 11
cents at clustered newspapers, and 10 cents at control papers. However, the betas suggest
subscription prices have more influence on ad prices at control papers. These results
offer weak support for Hypotheses 2b and 2c that interactions between prices are stronger
at non-clustered papers.

Table 15 shows the 1998 predictors of unadjusted ad prices. Cluster size and a
newspaper’s participation in a cluster still had positive associations with advertising

prices, but only participation was highly significant. However, the market share for

73

Table 15: 1998 Cluster and Control Regressions, Advertising Price as DV
Cluster (N: 199) Control (N = 195)

 

B Std. Error Beta t B Std. Error Beta t
(Constant) -22.342 8.633 -2.588** 4.947 6.702 .738
Households .000 .000 .672 10.460M .000 .000 .515 8.339M
Avg. household .000 .000 .133 2.149* -.000 .000 -.061 - l .038
income
Paper’s penetration 19.649 10.486 .178 1.874 10.339 8.345 .103 1.239
Clustered 37.025 12.578 .186 2.902" NA NA NA NA
penetration
Daily penetration -11.803 8.965 -.099 -1.317 -15.377 6.342 -.150 -2.425*
Weekly penetration 3.433 4.810 .036 .714 -3.622 3.787 -.967 .335
Subscription price .053 .028 .100 1.922 .188 .025 .432 7.649M
Size of cluster 1.904 .727 .158 2619* NA NA NA NA
Percent of cluster’s 22.993 5.744 .255 4003" NA NA NA NA
crrc.
Distance to nearest .021 .069 .016 .305 NA NA NA NA
paper
Radio penetration -12410 11890 -.082 -1.044 - 16641 10709 -. 121 -l .554
TV penetration -2433 2153 -.079 -1.l30 -2291 2498 -.917 .360
Cable penetration .029 .992 .002 .030 .129 1.516 .085 .932
Adjusted R2 .548 .601
F statistic 19.574" 33.612**

 

* p< .05, " p< .005

other members of the cluster also had a positive, highly significant association with ad

prices. If the OLS coefficients are accurate, 3 1% increase in participation resulted in a

$23 increase in ad price. A 1% increase in the penetration for other members of the

cluster raised ad prices $37.03. However, the betas suggest the most important predictor

0f ad prices in 1998 was households in the market, which also was true for control

papers. These results support Hypothesis 2 that clustering increases ad prices.

74

Control newspapers in 1998 appeared sensitive to competition from other dailies,
but the significance level was weak. This offers only weak support for Hypothesis 2d
that competition has more effect on ad prices in non-clustered markets.

There is also a negative association between ad prices and weekly penetration,
radio penetration, and television penetration at control papers but these are not significant
because of large standard errors. The same is true for competing daily penetration and
radio and television penetration at clustered papers.

There is a highly significant positive relationship between ad and subscription
prices at control papers. The relationship at clustered papers is not significant. This
supports Hypotheses 2b and 2c that clusters weaken interactions between prices.

Table 16 examines predictors of advertising cpm in 1988. There is a highly
significant relationship between cluster size and ad price. The OLS coefficient indicates
the addition of one paper to a cluster increases ad cpm by 15 cents. This result supports
Hypothesis 2 that clustering increases ad prices.

However, the betas suggest subscription ppm is the most important predictor of ad
price in clustered markets. The same is true for control papers. If the OLS coefficients
are correct, a 1 cent increase in subscription ppm increased ad cpm 14 cents for clustered
papers, but just 5 cents for control papers. The betas suggest the influence of subscription
ppm was roughly equal at clustered and control papers. These results fail to support
Hypotheses 2b and 2c that clustering weakens interactions between prices.

There is, however, a weak indication that control papers have to lower ad cpm to
increase market share. Recall that ad cpm represents the cost an advertiser pays to reach

1,000 readers. This relationship makes sense if readers value advertising and newspapers

75

Table 16: 1988 Cluster and Control Regressions, Advertisinggpm as DV

Cluster (N= 196)

Control (N = 197)

 

B Std. Error Beta t B Std. Error Beta t
(Constant) -2.246 .568 -3.952** .401 .097 4.123M
Households .000 .000 .033 .486 .000 .000 .088 1.423
Avg. household .000 .000 .393 5.823“ .000 .000 .068 1.157
income
Paper’s penetration -.236 .859 -.031 -.275 -.421 .163 -.214 -2.588*
Clustered 1.060 .865 .076 1.225 NA NA NA NA
penetration
Daily penetration -.792 .626 -.109 -1.266 -.127 .114 -.069 -l.108
Weekly penetration -.300 .186 —.077 -1.617 -.060 .039 -.062 - l .542
Subscription ppm .140 .011 .719 12.225** .054 .005 .723 11.418“
Size of cluster .151 .049 .180 3.1 16** NA NA NA NA
Percent of cluster’s .298 .403 .044 .740 NA NA NA NA
c1rc.
Distance to nearest -.005 .004 -.072 -l.379 NA NA NA NA
paper
Radio penetration 1251 972 .095 1.287 932 293 .261 3.178”
TV penetration -350 276 -.098 -l.266 56.221 58.248 .064 .965
Cable penetration .197 .585 .022 .338 -.136 .122 -.062 -1.115
Adjusted R2 .588 .702
F statistic 22.473M 52.551 **

 

* p<.05, “‘ p< .005

can increase market penetration by offering more ads. A similar negative association

between market share and ad cpm for clustered papers is not significant, again because of

the large standard error.

Surprisingly, Table 16 also shows a positive and highly significant association

between radio competition and ad cpm at control papers. A similar association for

clustered papers has a large standard error and is not significant. An earlier study of

76

weekly newspapers found a positive association between radio penetration and ad cpm at

Michigan weeklies (Lacy & Dravis, 1991, p. 342). These correlations are puzzling. Still,

results in this study weakly support Hypothesis 2d that prices at non-clustered

newspapers are more sensitive to the effects of competition.

Table 17: 1998 Cluster and Control Regressions, Advertising cpm as DV

B
(Constant) 1.401
Households —.000
Avg. household .000

income
Paper’s penetration —1.554

Clustered —.957
penetration

Daily penetration -.501
Weekly penetration -.242

Subscription ppm .049

Size of cluster -.001
Percent of cluster’s -. 151
circ.

Distance to nearest -.004
paper

Radio penetration 1249

TV penetration 1 17.258

Cable penetration -.016
Adjusted R2 .598
F statistic 23.181"

.377

.000

.000

.449

.515

.364

.194

.005

.030

.235

.003

485

95.547

.040

Cluster (N= 199)
Std. Error Beta

—.050

.044

-.326

-.111

-.097

-.059

.682

-.002

-.039

-.078

.191

.088

—.019

t
3.718"

-.778

.742

-3.463**

-1.858

-1.376

-1.243

10.787M

-.029

-.639

-1.531

2573*

1.227

—.399

Control (N = 195)
Std. Error

B
.938

.000

.000

- l .226

NA

-.474

-.154

.069

NA

NA

NA

817

118

.038

.612

35.128“

.236

.000

.000

.295

NA

.232

.131

.006

NA

NA

NA

386

90.754

.053

Beta

.116

.012

—.346

NA

—.131

-.056

.782

NA

NA

NA

.094

.035

t
3.978"
1.838

.217

-4.l85**

NA

-2.044*
—1 .179
11.857"
NA

NA
NA

2.111*
1.303

.723

 

* p<.05, **p< .005

Table 17 shows a change in influences on advertising cpm in 1998. Clustering

had no significant effect on ad cpm. However, for both clustered and control papers there

77

was a highly significant negative association between ad cpm and market share. The
OLS coefficients indicate changes in penetration had a slightly larger effect for clustered
papers. However the betas suggest the effect was slightly more influential for control
papers. This relationship suggests, as it did in the 1988 regressions, that newspapers are
willing to lower the standardized price to reach 1,000 readers in return for the benefits of
increased market share. Larger market share presumably makes newspapers even more
attractive to advertisers. These results do not support Hypothesis 2 that clustering is
associated with increased ad prices.

However, control papers also had a weakly significant, negative association
between the penetration of competing dailies and ad cpm. A similar negative association
for clustered papers was not significant, possibly because of the relatively large standard
error. This relationship is expected if papers compete with other dailies for market share
by offering lower ad prices. However, it contradicts the stronger and more significant
relationship between a paper’s own market share and ad cpm. At both clustered and
control papers there also was a weakly significant, positive association between radio
penetration and ad cpm. Both the OLS coefficients and the betas suggest this association
had more effect for clustered papers. These results are difficult to interpret and do not
appear to support Hypotheses 2 and 2d that clustered newspapers are less sensitive to
competition.

There is a positive, highly significant association between subscription ppm and
ad cpm for both clustered and control papers. If the OLS coefficients are correct,

increases in subscription prices have a larger effect for control papers. The betas also

78

show subscription prices are more influential at control newspapers. These results

support Hypotheses 2b and 2c that clustering moderates interactions among prices.

Table 18: 1988 Cluster and Control Regressions, Subscription Price as DV

Cluster (N= 196)

Control (N = 197)

 

B
(Constant) 39.556
Households .000
Avg. household -.000

income

Paper’s penetration 43.513

Clustered 26.315
penetration
Daily penetration 2.401

Weekly penetration 13.489
Advertising price .462
2.947

Size of cluster

Percent of cluster’s 25.446
circ.

Distance to nearest .1 13
paper

Radio penetration -51751
TV penetration ~12998
Cable penetration -l6.473
Adjusted R2 .368
F statistic 9.765M

Std. Error

10.386

.000

.000

15.155

16.618

12.013

3.547

.147

1.051

8.067

.070

18133

4906

11.192

Beta

.023

-.009

.371

.122

.021

.222

.267

.227

.243

.103

-.255

-.236

-.116

t
3.650“

.781

-.098

2.871**

1.584

.200

3.803**

3.155**

2.804**

3.154**

1.624

-2.854**

-2.649**

-1.472

B
55.801

-.000

-.000

6.643

NA

17.216

4.981

1.910

NA

NA

NA

-48360

-11314

24.708

.400

15.571**

Std. Error

10.291

.000

.000

16.543

NA

12.085

4.168

.287

NA

NA

NA

28640

5971

13.031

Beta

-.063

-.044

.045

NA

.124

.068

.514

NA

NA

NA

-.179

-.170

.150

I
5.422"

-.659

—.526

.402

NA

1.425

1.195

6.645M

NA

NA

NA

-1.699

-l.895

1.896

 

* p< .05, ** p< .005

The next set of tables examines relationships between clustering and subscription

prices. Table 18 shows that both the size of a cluster and a paper’s participation in a

cluster are positively associated with unadjusted subscription prices. The OLS

coefficients indicate adding 1 paper to a cluster increased the annual price $2.95. A 1%

79

increase in participation increased the annual subscription price $25.45. These results
support Hypothesis 2a that clustering increases subscription prices.

Clustered newspapers also increased subscription prices as their market share
increased. The beta suggests market share was the most influential price predictor. If the
OLS coefficient is correct, a 1% increase in penetration was associated with a price
increase of $43.51 at clustered papers. There was also a positive association between
penetration and price at control papers, but the large standard error made it insignificant.

Advertising prices had a highly significant, positive association with subscription
prices at clustered and control papers. However the OLS coefficients indicate increases
in ad prices have much smaller effects at clustered newspapers. The beta for control
papers, where ad prices were the only significant predictor of subscription prices, is also
larger than for clustered papers. These results support Hypotheses 2b and 2c that
clustering moderates interactions between prices.

However, subscription prices at clustered papers had a highly significant and
negative association with competition from radio and television. The negative
associations between prices and radio and television competition were not significant for
control papers. This suggests readers considered radio and television substitutes for
newspapers in clustered markets, but not in control markets. There are two possible
explanations. First, managers at clustered papers may have set high subscription prices
and readers began switching to other media. Managers at control papers, however, did
not set such high prices because of concern that readers would switch to other media. In
other words, managers at clustered papers overreached. The second possibility is that

readers of clustered papers were responding to decreases in quality associated with

80

decreases in competition. Recall from the theoretical discussion that part of the “price"

of reading a newspapers is the opportunity cost of not using a competing form of media.

In either case, these results fail to support Hypotheses 2a and 2e that competition from

other media has more effect on subscription prices at control papers.

Table 19: 1998 Cluster and Control Regressions, Subscription Price as DV

Cluster (N= 199)

Control (N = 195)

 

B
(Constant) 70.263
Households -.000

Avg. household .000
income

Paper’s penetration 46.240

Clustered 44,683
penetration
Daily penetration 26.474

Weekly penetration 22.852
Advertising price .367
Size of cluster .986

Percent of cluster’s 13.825
circ.

Distance to nearest .002
paper

Radio penetration -63527
TV penetration -8802
Cable penetration 1.632
Adjusted R2 .120
F statistic 3.094M

Std. Error

22.534

.000

.000

27.643

34.166

23.619

12.562

.195

1.946

15.720

.181

31031

5649

2.606

Beta

-.054

.039

.222

.119

.117

.127

.195

.043

.081

.001

-.223

-.151

.044

t

3.314”

-.478

.452

1.673

1.308

1.121

1.819

1.922

.507

.879

.014

-2.047*

-1.558

.626

B
35.380

-.000

.001

19.027

NA

20.383

18.098

1.275

NA

NA

NA

3426

-15715

-6.l71

.466

19.922M

Std. Error

17.303

.000

.000

21.799

NA

16.721

9.808

.167

NA

NA

NA

28096

6424

3.926

Beta

-.074

.204

.084

NA

.090

.102

.566

NA

NA

NA

.122

-.196

-.088

1
2045*

-.883

3.097"

.873

NA

1.246 .

1.845

7.649"

NA

NA

NA

.903

-2.446*

-l.572

 

* p< .05. " p< .005

Table 19 shows these relationships had changed by 1998. The association

between price and cluster size and participation was positive, but not significant because

81

of large standard errors. These results do not support Hypothesis 2a that clustering
increases subscription prices.

The regression for clustered papers in 1999 is generally unstable. The adjusted R2
is .12, compared with .37 in 1989. Radio competition is the only variable significantly
associated with subscription prices, but the significance level is weak.

The regression for control newspapers has more stability; the adjusted R2 is .47.
Ad prices at control papers had a positive and highly significant association with
subscription prices. A similar association at clustered papers was not significant. Given
the small R2 for clustered papers, these results weakly support Hypotheses 2b and 2c that
interactions between prices are stronger at control papers.

Table 20 shows the 1988 relationship between subscription ppm and clustering.
The association between price and cluster size was negative, and weakly significant. The
association between price and penetration from other members of the cluster also was
negative and weakly significant. These results fail to support Hypothesis 2a that there is
a positive association between clustering and subscription prices.

There is also a negative, highly significant association between subscription ppm
and households, income, and a clustered paper’s market share. The OLS coefficients
indicate a 1% increase in market share resulted in a decrease of $15.94 in the annual
subscription ppm. Recall that this price reflects the newspaper’s own circulation cost.
Therefore, the results for clustered papers suggest that when size was held constant
clustering was weakly associated with lower circulation costs. Increases in market size,
income, and a clustered paper’s market share also resulted in lower circulation costs.

These results might reflect economies of scale in circulation. Another possibility is that

82

managers increase penetration in larger, more affluent markets by passing fewer

circulation costs to readers.

Table 20: 1988 Cluster and Control Regressions, Subscription ppm as DV

Cluster (N= 196)

Control (N = 197)

 

B Std. Error Beta
(Constant) 16.717 2.549
Households -.000 .000 -.244
Avg. household -.000 .000 -.252
income
Paper’s penetration -15.940 3.941 -.409
Clustered -8.938 4.106 -. 125
penetration
Daily penetration 3.449 2.996 .092

Weekly penetration 1.621 .888 .081

Advertising cpm 3.213 .262 .627
Size of cluster —.503 .235 —. 123
Percent of cluster’s —2.830 1.921 —.081
c1rc.

Distance to nearest -.022 .018 —.060
paper

Radio penetration 6544 4650 .097
TV penetration 6266 1247 .343
Cable penetration -1.003 2.797 -.021
Adjusted R2 .640

F statistic 27.853“

I
6.559M
-4.08l**

-3.813**

-4044**

-2.177*

1.151
1.827
12.255”
—2.254*

-l.473
—l .231

1.407
5.025**

-.359

B Std. Error Beta
.056 1201
4000 .000 2261
4000 .000 4011
-5.869 1.911 4223
NA NA NA
3490 L333 .143
.555 .460 .043
1559 .662 .566
NA NA NA
NA NA NA
NA ‘ NA NA
8751 3502 .184
2599 663 .221
2901 L428 .100
7m
72.898**

t
.046

—5.060**

-.218

—3.071**

NA

2.618*

1.207

11.418M

NA

NA

NA

2.499“

3.920"

2.031*

 

" p< .05. ** p< .005

Table 20 shows that control papers also enjoy lower subscription ppm when they

have larger market share or larger markets. However, at control papers there is a

positive, weakly significant association between subscription ppm and competition from

other dailies and from cable. The positive association between price and competition

from radio and television is highly significant. For clustered papers there is a positive,
highly significant association between prices and television competition. The effects of
competition from other dailies, and from radio and cable are not significant because of
large standard errors. These results suggest that control papers are less able to ignore the
effects of competition on circulation costs. This may be because they have higher costs
and less “room” for price changes. These results support Hypothesis 2e that competition
has stronger effects on subscription prices in control markets.

Table 20 also shows a positive, highly significant association between
subscription ppm and ad cpm at clustered and control newspapers. If the OLS coefficients
are correct, this effect is larger for control papers. However, the betas show ad prices
have more influence at clustered papers. These results do not support Hypotheses 2b and
2c that clustering moderates interactions among prices.

Table 21 again shows that by 1998 the relationship between clusters and prices
had changed. The sign of the associations between price and cluster size and penetration
had changed. However, the associations were not significant because of large standard
errors. These results do not support Hypothesis 2a that clustering increases subscription
prices.

However, clustered papers again had lower subscription ppm in larger, more
affluent markets. Control papers had lower subscription ppm in larger markets. These
results again suggest there are circulation economies of scale, or that managers pass
fewer circulation costs along in larger markets.

Clustered papers also had higher subscription ppm in markets with more

television competition. A positive association between prices and daily competition had

84

a large standard error and was not significant. However, higher prices at control papers

were associated with competition from television and from other daily papers. These

results support Hypothesis 2e that competition has stronger effects on prices in control

markets.

Table 21: 1998 Cluster and Control Regressions, Subscription ppm as DV

Cluster (N= 199)

Control (N = 195)

 

(Constant)
Households

Avg. household
income

Paper’s penetration

Clustered
penetration

Daily penetration
Weekly penetration
Advertising cpm
Size of cluster

Percent of cluster’s
circ.

Distance to nearest
paper

Radio penetration
TV penetration
Cable penetration
Adjusted R2

F statistic

B
9.803

-.000

-.000

-8.324

7.376

5.825

.145

7.912

.099

-3.819

-.036

1837

4463

.270

.668

31.795**

4.928

.000

.000

5.874

6.610

4.652

2.489

.733

.377

2.993

.036

6300

1178

.514

Std. Error Beta

-.193

-.144

-.124

.061

.080

.003

.564

.014

-.070

-.046

.020

.239

.023

t
1989*

-3.381**

-2.711**

-1.417

1.116

1.252

.058

10.787**

.265

-1.276

-.992

.292

3.786**

.525

B
.118

-.000

—.000

1.133

NA

10.126

2.032

6.231

NA

NA

NA

5632

2584

.434

.727

58.635“

Std. Error

2.333

.000

.000

2.927

NA

2.100

1.240

.525

NA

NA

NA

3695

844

.499

 

Beta

-.257

-.083

.028

NA

.248

.065

.550

NA

NA

NA

.103

.182

.035

t
.050

-5.150**

-1.744

.387

NA

4.821M

1.639

11.857”

NA

NA

NA

1.524

3.059**

.869

 

‘p<.05,” p<.005

Table 21 also shows highly significant, positive associations between ad cpm and

subscription cpm for cluster and control papers. If the OLS coefficients are correct, ad

85

prices had a larger effect at clustered papers. The betas are almost identical for clustered
and control papers. These results do not support Hypotheses 2b and 2c that clustering
weakens interactions between prices.

Regressions Examining Newsroom Spending

Only one measure of newsroom spending, or quality, was available for all
newspapers in the random samples. This was the number of wire service subscriptions.
Regressions using the number of wire subscriptions failed to produce a significant F, and
scatterplots of the residuals showed a non-linear relationship.

However, a standardized measure did produce regressions that were linear and
significant. This measure was the number of wire subscriptions per thousand circulation
or wires pm. This measure again may be though of as holding the size of different
newspapers constant while examining the relationships between variables.

Results for 1988 in Table 22 show a negative, highly significant association
between wires pm and cluster size. There is also a negative, highly significant
association between wire subscriptions and a paper’s participation in a cluster. The OLS
coefficients show an increase of 1 paper in cluster size resulted in a decrease of 0.02 wire
subscriptions per thousand circulation. An increase of 1% in participation in a cluster
decreased the number of wire subscriptions by 0.1 l per thousand circulation. These
results support Hypothesis 3 that clustering reduces newsroom spending.

Table 22 also shows a positive, highly significant association between radio
penetration and wires pm for clustered papers. There are positive, weakly significant
associations between the number of wire services and competition from other dailies and

cable competition. These results show clustered papers had higher newsroom spending

86

Table 22: 1988 Cluster and Control Regressions, Wires pm as DV
Cluster (N= 196) Control (N = 197)

 

B Std. Error Beta t B Std. Error Beta t
(Constant) .130 .062 2095* .013 .037 .728
Households .000 .000 .007 .095 .000 .000 .059 .727
Avg. household -.000 .000 -.1 17 -1.562 .000 .000 .135 1.754
income
Paper’s penetration -.077 .090 -.099 -.851 -.087 .061 -.156 -l.433
Clustered .092 .091 .064 1.01 1 NA NA NA NA
penetration
Daily penetration .165 .066 .221 2510* -.087 .042 -. 168 -2.071*
Weekly penetration -.004 .020 -.009 -.179 -.01 l .014 -.041 -.792
Advertising cpm .0316 .008 .309 4.094“ .033 .027 .118 1.251
Subscription ppm .007 .002 .383 4.728“ .012 .002 .507 4.750"
Size of cluster -.017 .005 -.l98 -3.274** NA NA NA NA
Percent of cluster’s -.1 14 .042 -. 164 -2.698** NA NA NA NA
c1rc.
Distance to nearest —.000 .000 —.051 —.942 NA NA NA NA
paper
Radio penetration 306 102 .228 2.998M 64.993 1 10.238 .064 .590
TV penetration 21.368 29.096 .059 .734 41.382 21.360 .166 1.937
Cable penetration .127 .061 .135 2072* .012 .045 .020 .274
Adjusted R2 .569 .501
F statistic 19.457“ 20.770"

 

* p<.05, " p< .005

in markets with more competition. The positive associations between wires pm and radio

and cable penetration were not significant for control papers, apparently because of large

standard errors. Competition from other dailies had a'negative, weakly significant

association with wire subscriptions at control papers. These results do not support

87

Hypothesis 3a that non-clustered papers spend more on the newsroom in response to

competition.

However, these results do not show if control papers respond to competition by

increasing local news coverage or hiring more staff. The results may also imply clustered

newspapers have more money to purchase wire subscriptions.

Table 23: 1998 Cluster and Control Regressions, Wires pm as DV

(Constant)
Households

Avg. household
income

Paper’s penetration

Clustered
penetration

Daily penetration
Weekly penetration
Advertising cpm
Subscription ppm
Size of cluster

Percent of cluster’s
circ.

Distance to nearest
paper

Radio penetration
TV penetration
Cable penetration
Adjusted R2

F statistic
"' p < .05, ** p < .005

B
.039

.000

-.000

-.059

.151

.031

.025

.042

.006

.002

-.028

.001

116.212

43.416

-.006

.625

24.734**

.069

.000

.000

.081

.091

.064

.034

.013

.001

.005

.041

.001

86.683

16.829

.007

Cluster (N= 199)
Std. Error Beta

.028

-.093

-.068

.097

.033

.033

.230

.436

.025

-.039

.081

.098

.180

-.042

t
.567

.456

-1.626

-.727

1.654

.482

.729

3.247**

5.602**

.465

-.668

1.627

1.341

2.580*

-.904

88

Control (N = 195)
Std. Error

B
.012

.000

-.000

.031

NA

-.029

.012

.051

.007

NA

NA

NA

-9.958

16.751

.009

.606

28.400M

.044

.000

.000

.055

NA

.042

.024

.013

.001

NA

NA

NA

70.004

16.296

.009

Beta

-.035

—.023

.051

NA

-.046

.025

.297

.430

NA

NA

NA

-.012

.077

.053

.273

-.528

-.391

.567

NA

-.685

.520

3.913**

4.749"

NA

NA

NA

-.142

1.028

1.058

Table 23 again shows that in 1998 the effects of clusters had changed. Cluster
size and participation in a cluster had large standard errors and were no longer
significantly associated with wire subscriptions. These results do not support Hypothesis
3 that clustering reduces newsroom spending.

There was a weak positive association between competition from television and
wires pm at clustered papers. Competition from daily newspapers had large standard
errors and was not significantly associated with wires pm for clustered or control papers.
Competition from other media also was not significantly associated with wire
subscriptions at control papers. These results fail to support Hypothesis 3a that control
papers spend more in response to competition.

Prices had a positive, highly significant association with wires pm at both
clustered and control papers in 1998. The OLS coefficients indicate there was little
difference between the effects of prices at clustered and control papers. The betas
suggest prices were equally influential for both groups of papers. These results, which
use standardized prices reflecting costs, may simply show that wire subscriptions are one
element of a newspaper’s costs.

Comparing Elasticity for Clustered and Control Papers

Recall that elasticity is a measure of how sensitive newspaper readers, or
advertisers, are to changes in price or quality. This is expressed as the Percent Change in
Quantity/Percent Change in Price or Quality. If this ratio is larger than I, demand is
elastic. If this ratio is smaller than 1, demand is inelastic. Clustering, the theoretical

discussion argues, reduces the elasticity of demand — it makes both readers and

89

advertisers less sensitive to changes in price or quality. Many of the hypotheses rely on
this general assumption.

Calculus can be used to show that when variables in a regression are transformed
as their natural logarithm, the resulting OLS coefficients represent the elasticities
associated with those variables (Gujarti, 1992, p. 218-222). In other words. the
regression coefficients show whether the dependent variable is elastic with respect to the
independent variables.

The essential argument is that regression measures how changes in the
independent variables affect the value of the dependent variable. However, as was
pointed out in the theoretical discussion, elasticity is not constant (Gujarti, 1992;
Nicholson, 1995, chap. 7). For instance, as price increases, demand eventually becomes
more elastic as readers begin to switch to alternative forms of media. Therefore, the
relationship between price and demand is non-linear.

Mathematically, let Y = a newspaper’s market penetration and X = its cover price.
A simple linear relationship would be:

Y: A +82X (5.1)
where A is some constant and B; is the coefficient of change in Y resulting from a change
in X. A non-linear form of Equation 5.1 is:

Y=AXm (5»
where the exponent for A is 1. In this model there is a non-linear relationship between Y
and X. However, an alternative eXpression of this non-linear model is:

mY=mA+BflnX (5m

90

This expression is equivalent to the first derivative of X because of the rule for
calculating derivatives of exponential equations. The power rule states the derivative of
X" is nX"'1 .

Note that 1nA = B 1 because 1nA = A0 = 1. So Equation 5.3 can be rewritten as:

lnY=B1+Bg 1nX+u (5.4)
where u is the error term (Gujarti, 1992, p. 220).

These equations suggest that elasticities can be measured using transformed
variables by regressing price on a newspaper’s circulation. Multicollinearity prevented
the use of circulation in this dissertation. The natural log of a paper’s market share was
used instead. Initial regressions again showed evidence of multicollinearity, so the log of
households and the log of income were removed from the regressions. One indicator, a
conditioning index above 30, suggested there still were problems with multicollinearity
(Tabachnick & F idell, 1996, p. 87). However, other measures including small variance
inflation factors (Norusis, 1993, p. 355), suggested the regressions should not be unduly
affected.

Before transforming the variables to their natural log, 1 was added to all of the
values for newspaper, radio, television and cable market share. This because these
variables are percentages expressed in decimals, and values less than 1 have a negative
log. This transformation was done to avoid confusion when interpreting signs of
coefficients in the regressions. The natural log of a paper’s market share is the dependent
variable in these double-log regressions. The independent variables include the natural
log of market share for all other papers in the market. However, this does make these

regressions determinant because market share, or circulation penetration, does not sum to

91

1. Some households take multiple newspapers — total penetration in some markets was
more than 200 percent during 1988 and more than 300 percent in 1998.

Results in Table 24 show that in 1988 there was a positive, weakly significant
association between a clustered paper’s market share and its subscription price. The OLS
coefficient indicates a small increase in subscription price was associated with a 5%
increase in market share. Keep in mind that market share is penetration, or a percentage,
so this is a very small increase. The association between subscription price and market

Table 24: 1988 Elasticity Regressions, LN of Paper ’5 Market Penetration as DV

 

 

Cluster (N= 196) Control (N = 197)

B Std. Error Beta t B Std. Error Beta t
(Constant) .105 .091 1.162 .233 .083 2.824“
LN clustered -.374 .072 -.239 -5.l64** NA NA NA NA
penetration
LN daily -.527 .052 -.505 10.078“ -.425 .050 -.422 8.486“
penetration
LN weekly .014 .023 .024 .614 .016 .023 .028 .693
penetration
LN ad price -.002 .011 -.009 -.186 .015 .014 .065 1.112
LN subscription .048 .022 .102 2227* .005 .022 .014 .252
price
LN cluster size -.013 .018 -.040 -.735 NA NA NA NA
LN percent of .042 .051 .044 .823 NA NA NA NA
cluster’s circ.
LN distance to .009 .007 .057 1.349 NA NA NA NA
nearest paper
LN radio 119 65.611 .108 1.817 376 95.883 .287 3.929"
penetration
LN TV penetration 55.986 14.047 .232 3.986” 27.216 20.436 .090 1.332
LN cable .254 .054 .227 4.687** .272 .047 .284 5.837M
penetration
Adjusted R2 .738 .693
F statistic 51.314" 64.427“

 

"‘ p< .05, ** p< .005

92

share for control papers had a large standard error and was not significant. This result
weakly supports the theoretical argument that demand among readers is price inelastic in
clustered markets.

Table 24 also shows a negative, highly significant association between a clustered
papers’ market share and the market share for other members of the cluster. If the OLS
coefficient is correct, a small decrease clustered penetration is associated with a 37
percent increase in the paper’s penetration. This suggests clustered papers can, by
coordinating circulation, control their market share. This again suggests that clustering
substantially affects a newspaper’s elasticity of demand.

However, these results also show a negative, highly significant association
between competition from other dailies and clustered papers’ market share. There is a
similar association between daily competition and market share for control papers. The
OLS coefficients indicate increases in daily competition are associated with larger
decreases in market share for clustered papers. The betas also show the effect of daily
competition is more influential for clustered papers. These results do not support the
general hypothesis that clustering reduces elasticity of demand.

There are positive, highly significant associations between competition from other
media and market share for both clustered and control papers. These results are
surprising, but may indicate that demand for other media increased demand for
newspapers. In other words, newspapers and other media were compliments in 1988.

The cable coefficients are the only associations that are significant for both
clustered and control papers. The OLS coefficients indicate increases in cable market

share have almost identical effects for cluster and control papers. The betas show this

93

effect was more influential for control papers. These results weakly support the argument
that clustering reduces elasticity of demand among readers.

The 1998 results in Table 25 show a change in relationships between price and
market share. The coefficients for subscription price had large standard errors and were
not significant for clustered or control papers. However, there was a negative, weakly
significant association between ad prices and market share at clustered papers. There was
a negative, highly significant association between ad prices and market share at control

Table 25: 1998 Elasticity Regressions, LN of Paper '5 Market Penetration as DV

 

 

Cluster (N= 197) Control (N = 195)

B Std. Error Beta t B Std. Error Beta t
(Constant) .198 .097 2033* .193 .093 2070*
LN clustered -.456 .084 -.286 -5.413** NA NA NA NA
penetration
LN daily -.295 .058 -.452 -8.647** -.424 .053 —.400 7.949"
penetration
LN weekly -.031 .034 -.040 -.926 -.013 .034 -.016 -.383
penetration
LN ad price -.021 .010 -.l 14 —2. 139* -.039 .012 -.207 -3.291**
LN subscription .039 .019 .095 1.965 .045 .023 .1 18 1.936
price
LN cluster size -.000 .018 -.002 -.039 NA NA NA NA
LN percent of .107 .053 .124 2025* NA NA NA NA
cluster’s circ.
LN distance to -.009 .010 -.044 -.949 NA NA NA NA
nearest paper
LN radio 214 63.318 .230 3.389" 248 65.940 .249 3.768M
penetration
LN TV penetration 34.147 9.895 .207 3.541 ** 63.719 12.092 .304 5.270“
LN cable .027 .019 .066 1.444 .045 .018 .105 2443*
penetration
Adjusted R2 .646 .670
F statistic 33.739“ 57.675M

 

* p< .05, ** p<.005

94

papers. The OLS coefficients indicate a small increase in ad prices at control papers was
associated with a 4% decrease in market share. However, the equivalent change at
clustered papers was associated with 2% decrease in market share. The betas suggest ad
prices were more influential at control papers. These results support the general
hypothesis that clustering reduces elasticity.

The 1998 results also show a positive, weakly significant association between
cluster size and market share for clustered papers. There again is a negative, highly
significant association between the penetration of other clustered papers and market
share. These results again suggest clustered papers can control their market share by
coordinating circulation with their counterparts. These results weakly suggest that
increases in cluster size are associated with larger market share. This supports the
general hypothesis that clustering reduces elasticity of demand.

For clustered and control papers in 1998 there again was a negative, highly
significant association between daily competition and market share. If the OLS
coefficients are correct, this effect is much smaller for clustered papers. However, the
betas show the effect of daily competition is more influential for clustered papers. These
results cannot be interpreted to support the hypothesis that clustering reduces elasticity of
demand.

The positive association between competition from other media and newspaper
market share was again evident in 1998. The correlations for radio and television
penetration were highly significant for both clustered and control papers. The OLS
coefficients indicate radio and television penetration had larger effects on the market

share of control papers. The betas show these effects were more influential for control

95

papers. These results support the hypothesis that that clustering reduces elasticity of
demand.
F irst—Diflerence Regressions

The estimation of population parameters using regressions in two different time
periods can be influenced by autocorrelation. Autocorrelation exists when the value of a
variable in one time period is correlated with its value in the next time period (Gujarti,
1992, chap. 12). This problem is similar to heteroscedasticity in cross-sectional data.
Autocorrelation may exist in economic data because economic changes do not occur at a
constant rate. Changes in economic conditions, a recession for example, will alter the
rate of change in retail sales for newspapers.

Autocorrelation may exist and have no effect on the OLS estimations from
regressions. However, autocorrelation may also result in an a variety of errors. For
example, there could be underestimation of variances and standard errors and inflation of
t values (Gujarti, 1992, p. 356). First-difference models can be used to correct for this
problem by transforming the regression so the error term is homoscedastic (p. 366).

This transformation is accomplished by subtracting the value of a variable in one
year from its value in another year. In this dissertation, the first difference regressions
are based on subtracting the 1988 value of each variable from its 1998 value. However,
two caveats are in order. This technique assumes the value of autocorrelation = 1
(Gujarti, 1992). In other words, there is a perfect positive correlation between errors in

the regressions. First difference models also are assumed to not have an intercept.

96

Results in Table 26 show a positive, highly significant association between
changes in advertising prices and changes in cluster size. These results support

Hypothesis 2 that clustering increases advertising prices.

Table 26: F irst-Diflerence Regressions, Ad Price Difference as D V
Cluster (N= 195) Control (N = 193)

B Std. Error Beta t B Std. Error Beta t
(Constant) 2.332 3.383 .689 -4.l91 2.125 -1.972**
Change in .000 .000 .481 7.071 ** .000 .000 .476 6.829“
households
Change in income .000 .000 .284 4.154” -.000 .000 -.055 -.752
Change in paper’s 15.537 12.789 .123 1.215 12.079 7.926 .137 1.524
penetration
Change in clustered 16.860 16.103 .074 1.047 NA NA NA NA
penetration
Change in daily -21.704 12.221 -.168 -1.776 -7.758 6.509 -.092 1.192
penetration
Change in weekly -2.608 3.580 -.040 -.729 -4.418 2.430 -.097 -1.818
penetration
Change in .048 .046 .065 1.050 .184 .027 .308 6.766
subscription price
Change in cluster 5.033 .893 .374 5.634“ NA NA NA NA
size
Change in percent 25.158 7.325 .237 3.434** NA NA NA NA
of cluster’s circ.
Change in distance .002 .069 .002 .034 NA NA NA NA
to nearest paper
Change in radio ~2223 14065 -.013 -.158 -21810 11867 -.171 -1.838
penetration
Change in TV -2123 2542 -.060 -.835 -363 2181 -.013 -.167
penetration
Change in cable -.597 1.183 -.028 -.505 -l.389 1.208 -.O64 -1.150
penetration
Adjusted R2 .491 .466
F statistic 15.441** 19.686”

 

* p< .05, "p< .005

97

For clustered papers, there is a positive, highly significant association between

changes in the number of households and household income and ad prices. These results

may indicate that clustered papers were locating in growing. affluent markets.

Table 27: F irst—Diflerence Regressions, Ad cpm Difference as DV

(Constant)

Change in
households

Change in income

Change in paper’s
penetration

Change in clustered
penetration

Change in daily
penetration

Change in weekly
penetration

Change in
subscription ppm

Change in cluster
size

Change in percent
of cluster’s circ.

Change in distance
to nearest paper

Change in radio
penetration

Change in TV
penetration

Change in cable
penetration

Adjusted R2

F statistic

Cluster (N= 195)

B Std. Error Beta

.938 .232
n000 .000
.000 .000
-L550 L036
—L050 L288
-L810 .981
a312 .282
.070 .008
.351 .071
.792 .561
.003 .006
1323 1126
40.649 214
a029 .096
.419
1L795**

-.050

.337

-.161

-.060

--.183

-.062

.560

.342

.098

.027

.101

.015

-.018

t
4.035M
-.676

4.518**

1.496

-.815

-1.845

-1.107

8.873"

4.967"

1.411

.448

1.175

.189

-.308

Control (N = 193)
Std. Error

B
-.274
.000

-.000

-1.087

NA

-.493

—.131

.066

NA

NA

NA

552

93.464

-.019

.602

33.393“

.066

.000

.000

.250

NA

.206

.075

.006

NA

NA

NA

401

67.378

.037

Beta

.153

-.006

—.346

NA

-.l65

-.081

.738

NA

NA

NA

.122

.096

-.025

t
-4.l76**
2450*

-.099

-4357"

NA

-2.388*

-l.749

11.330”

NA

NA

NA

1.378

1.387

-.509

 

r p< .05, " p< .005

98

Results in Table 27 show a positive, highly significant association between
changes in cluster size and changes in ad cpm. These results support Hypothesis 2 that
clustering increases ad prices.

The association between changes in subscription ppm and ad cpm is positive and
highly significant for clustered and control papers. The OLS coefficients indicate
changes in subscription ppm had the same effect on cluster and control papers. However,
the betas indicate these changes were more influential for control papers. These results
weakly support Hypotheses 2band 2c that interactions between prices are stronger at
control papers.

Results in Table 28 examine the effects of changes on subscription prices. There
are positive, highly significant associations between changes in three clustering variables
and changes in subscription prices. Changes in cluster size, changes in a paper’s
participation in a cluster, and changes in the market share of other members of the cluster
all were associated with higher prices. The fact that subscription prices increase along
with the market share for other members of the cluster is particularly noteworthy. This
suggests clustered papers may zone circulation within markets to avoid competing for the
same readers. These results support Hypothesis 2a that clustering increases subscription
prices.

Results for clustered papers also show positive, highly significant associations
between changes in a paper’s penetration, changes in weekly penetration, and changes in
price. There is a positive, weakly significant association between price and changes in
daily competition. Similar positive associations for clustered papers are not significant,

possibly because of large standard errors. These results are counter-intuitive because

99

Table 28: F irst-Difference Regressions, Subscription Price Difference as D V
Control (N = 193 )

Cluster (N= 195)

 

(Constant)

Change in
households

Change in income

Change in paper’s
penetration

Change in clustered
penetration

Change in daily
penetration

Change in weekly
penetration

Change in ad price

Change in cluster
size

Change in percent
of cluster’s circ.

Change in distance
to nearest paper

Change in radio
penetration

Change in TV
penetration

Change in cable
penetration

Adjusted R2

F statistic
* p <05, *"' p < .005

B
-42.129
.000

.000

66.759

68.296

48.367

21.157

.123

4.101

51.497

.129

-58716

-3 842

3.662

.275

6.681**

4.388

.000

.000

19.809

25.178

19.268

5.481

.118

1.509

11.397

.110

21938

4040

1.863

Std. Error Beta

.084

.040

.397

.280

.241

.093

.229

.364

.080

-.258

-.081

.127

t
-9.600**
.913

.466

3.370"

2.713**

2.510*

3.860”

1.050

2.718**

4.519“

1.172

-2.676**

-.951

1.966

B
-16.465
-.000

.001

23.737

NA

25.585

5.718

1.085

NA

NA

NA

8322

-l3929

-.497

.360

13.043“

Std. Error

5.079

.000

.000

19.311

NA

15.774

5.946

.160

NA

NA

NA

29110

5204

2.948

Beta

-.112

.216

.121

NA

.137

.057

.489

NA

NA

NA

.029

-.229

-.010

t

-3.240**

-1.322

2.758**

1.229

NA

1.622

.962

6.766“

NA

NA

NA

.286

-2.676**

-.169

they suggest changes in prices at clustered papers were helped along by positive changes

in both market share and competition. The meaning of these results is unclear.

100

Results in Table 29 show no significant associations between changes in cluster

variables and changes in subscription ppm. However, there is a positive, highly

significant association between changes in ad cpm and changes in subscription cpm at

clustered and control papers. If the OLS coefficients are correct, these changes had a

Table 29: F irst-Difference Regressions, Subscription ppm Difference as DV
Control (N 73193 )___

(Constant)

Change in
households

Change in income

Change in paper’s
penetration

Change in clustered

penetration

Change in daily
penetration

Change in weekly
penetration

Change in ad cpm

Change in cluster
size

Change in percent
of cluster’s circ.

Change in distance

to nearest paper

Change in radio
penetration

Change in TV
penetration

Change in cable
penetration

Adjusted R2

F statistic

B
-9.204
-.000

-.001

-14.200

-3.162

10.266

3.553

4.313

-1038

-7.075

-.038

581

5019

.622

.440

12.800**

.1777

.000

.000

8.115

10.122

7.735

2.203

.486

.587

4.397

.044

8873

1644

.749

Cluster (N= 195)
Std. Error Beta

-.120

-.357

-.184

-.023

.129

.088

.539

-.126

-.109

-.051

.006

.231

.047

1
-5.179**
-l.683

-4.934**

-1.750

-.321

1.322

1.612

8.873“

1.769

1.609

-.853

.066

3.052“

.830

B
-.639
-.000

-.000

-.120

NA

8.592

1.403

6.163

NA

NA

NA

13491

1102

.565

.700

50.939"

Std. Error

.659

.000

.000

2.521

NA

1.914

.718

.544

NA

NA

NA

3745

646

.356

Beta

-.254

-.067

-.003

NA

.079

.557

NA

NA

NA

.269

.102

.066

t
-.969
-4.889**

-1.240

-.047

NA

4.490“

1.955

11.330**

NA

NA

NA

3.602"

1.707

1.587

 

* p< .05. H p<.005

101

larger effect at control papers. The betas suggest changes in ad cpm were equally
influential at clustered and control papers. These results weakly support Hypotheses 2b
and 2c that interactions between prices are weaker at clustered papers.

There is a negative, highly significant association between changes in income and
subscription ppm at clustered papers. There also is a negative, highly significant
association between changes in households and prices at control papers. These results
may indicate managers lowered the amount of circulation costs passed to subscribers in
more affluent, or larger markets.

There is a positive, highly significant association between changes in daily
penetration and radio penetration and changes in subscription ppm at control papers.

The positive association with changes in daily penetration is not significant for clustered
papers. However, there is a significant positive association between changes in television
market share and subscription ppm for clustered papers. These results may indicate that
managers at control papers passed on more of their circulation costs in response to
changes in competition from dailies and other media. Clustered managers only passed on
more costs in response to changes in competition from other media. These results
support Hypotheses 2a and 2e that clustered newspapers are less sensitive to competition
than control papers.

Results in Table 30 examine changes in one measure of newsroom spending, the
number of wire subscriptions per thousand circulation. There is a negative, highly
significant association between changes in wires pm and changes in a paper’s
participation in a cluster. These results support Hypothesis 3 that clustering reduces

newsroom spending.

102

For control papers, there was a negative, highly significant association between

changes in daily penetration and changes in wires pm. Changes in daily penetration had

Table 30: F irst-Dzflerence Regressions, Wires pm Difference as DV
Control (N = 193 )

Cluster (N= 195)

 

(Constant)

Change in
households

Change in income

Change in paper’s
penetration

Change in clustered
penetration

Change in daily
penetration

Change in weekly
penetration

Change in ad cpm

Change in
subscription ppm

Change in cluster
size

Change in percent
of cluster’s circ.

Change in distance
to nearest paper

Change in radio
penetration

Change in TV
penetration

Change in cable
penetration

Adjusted R2

F statistic
* p <05, " p < .005

B
.015
-.000

-.000

-.117

-.022

.061

-.028

.016

.007

-.011

-.174

-.000

356

~5.707

.000

.564

19.040**

.021

.000

.000

.091

.113

.087

.025

.006

.001

.007

.049

.000

98.888

18.789

.008

Std. Error Beta

-.004

-.145

-.120

-.012

.061

-.056

.162

.535

-.106

-.212

-.043

.270

-.021

.000

t
.714
-.063

-2.135*

-1.282

-.195

.703

-1.151

2.529*

8.184**

-l.670

-3.4l7**

-.806

3.603**

-.304

.004

103

B
.055
.000

.000

-.073

NA

-.l37

-.005

.019

.010

NA

NA

NA

-73.788

42.655

.004

.487

19.311**

Std. Error

.015

.000

.000

.058

NA

.046

.017

.016

.002

NA

NA

NA

89.412

14.979

.008

Beta t
3.592“
.069 .956
,060 .852
-.119 -1.261
NA NA
-.234 -2.950**
-.016 -.299
.099 1.186
.579 6.012**
NA NA
NA NA
NA NA
-.084 -.828
.225 2.848"
.027 .488

the opposite sign for clustered papers, but were not significant because of a large standard
error. Control papers had a positive, highly significant association between changes in
television competition and wire subscriptions. There was a similar association between
changes in radio penetration and wire subscriptions at clustered papers. These results are
mixed, and do not support Hypothesis 3a that control papers are more likely to respond to
competition by increasing newsroom spending.
Results of the Mail Survey

As described in the Method chapter, a one-page questionnaire was used to gather
additional information from 1998 newspapers about newsroom spending and efforts to
share resources. The questionnaire was mailed to the senior newsroom executive or
publisher at 968 newspapers. This list included 482 clustered newspapers from the study
population and 486 control newspapers selected randomly from the same states.

A mailing in June 2000 resulted in 275 returns for an initial response rate of 28%.
A reminder postcard three weeks later, and a second mailing two weeks after that,
produced 164 additional responses, bringing the total to 439. The final overall response
rate was 45%.

Three questionnaires were returned as undeliverable and a fourth questionnaire
was returned but not filled out. This means the effective response rate was 46%.

The response rate for clustered newspapers was lower than the response rate for
control newspapers. There were 189 usable questionnaires returned from clustered
papers, a response rate of 39%. The 249 usable questionnaires returned from control

papers represented a 51% response rate.

104

The responses to the mail survey created new samples of clustered and control
newspapers that were separate from the random samples drawn from the study
population. There was limited overlap between the samples of newspapers that
responded to the survey and the random samples of cluster and control newspapers in
1998. Only 75 clustered newspapers from the random sample also returned mail
questionnaires. This was 38% of the random cluster sample. Only 106 control
newspapers from the random sample also returned mail questionnaires. This was 53% of
the random control sample.

Therefore, the t-tests used with the 1998 random samples were repeated to see if
the survey respondents differed from newspapers in the random samples. The tests used
all 189 clustered newspapers that responded to the survey. Another 189 control
respondents were selected randomly for inclusion in this analysis. This was done to
compensate for the larger number of control respondents because t-tests assume equal
sample sizes. County-by-county circulation data was not available for a few papers in

these tests.

Table 31: Intercounty Differences, Survey Respondents
__ 1998
Clustered Control

 

Mean number of adjoining counties 1.64 3.34**
where paper circulates

Std. Deviation 2.41 4.84
Paper’s mean penetration in .049 .071**
adjoining counties

Std. Deviation .074 .089
N 184 188

 

* difference between clustered and control, p < .05
** difference between clustered and control, p < .005

105

Table 31 shows clustered newspapers responding to the survey circulated in
significantly fewer counties outside their home county than control papers did. The
clustered newspapers also averaged significantly less penetration in those counties.
These results replicate findings for the random sample in Table 6. These results support

Hypothesis 1a that clustered papers are less aggressive competitors.

Table 32: Papers in Markets, Survey Respondents

 

 

1998
Clustered Control

Mean of competing dailies headquartered 1.15 2.05**
in paper’s market

Std. Deviation 2.02 3.36
Mean of commonly-owned dailies .54 .01 **
headquartered in paper’s market

Std. Deviation .75 .10
N 184 188

 

* difference between clustered and control, p < .05
** difference between clustered and control, p < .005

Table 32 shows the average number of competing dailies headquartered in the
clustered markets was significantly less than the number of competing dailies in control
markets. This differs from the finding for the random sample in Table 7 showing no
significant difference in the number of competing dailies. However, Table 32 also shows
the average number of commonly-owned dailies headquartered in the clustered markets is
significantly higher than in clustered markets. This replicates findings for the random
sample in Table 7. The results in Table 32 support Hypotheses l and la that clustering
reduces competition.

Table 33 shows the average market share for clustered newspapers is significantly
less than for control newspapers among survey respondents. The average penetration

from other members of the cluster is about 8 percent. These findings replicate results for

106

the random sample in Table 8. The average market share of competing dailies is higher
for control papers responding to the survey, but the difference is not significant. Weekly
penetration is about the same for clustered and control markets. Results for the random
sample in Table 8 showed the higher penetration of dailies in control markets was
significant. Results in Table 33 offer some support for Hypotheses 1 and la that

clustering reduces competition.

Table 33: Market Share, Survey Respondents
Cluster Control

 

Paper’s mean penetration in its market .241 .285**

Std. Deviation .152 .167
Mean penetration for other members of .079 NA
cluster

Std. Deviation .000 NA
Mean competing daily circulation in .289 .322
paper’s market

Std. Deviation .165 .162
Mean weekly penetration in paper’s .243 .255
market

.204 .233

N 184 188

 

" difference between clustered and control, p < .05
*‘ difference between clustered and control, p < .005

The comparisons in Tables 31-33 show clustered newspapers in the survey, like
their counterparts in the random sample, are less aggressive competitors and operate in
markets where there is less competition from other daily newspapers.

Table 34 shows ad prices are slightly higher in clustered markets, but the
difference is not significant. Standardized subscription prices also are slightly higher in
clustered markets, but the difference is not significant. The results for unadjusted prices

replicate findings for the random sample in Table 9. However, standardized price

107

differences were significantly higher for the random sample. The price comparisons in

Table 34 do not support Hypotheses 2 and 2a that clustering increases prices.

Table 34: Price foferences, Survey Respondents
1998 _,
Clustered Control

 

Mean price of standard advertising unit $23.25 $22.31

Std. Deviation $24.01 $21.07
Mean annual subscription price $1 1 1.94 $1 13.53
Std. Deviation $37.07 $37.84
Mean advertising cpm $1.47 $1.37
Std. Deviation $0.72 $1.01
Mean subscription ppm $1 1.93 $10.32
Std. Deviation $1 1.01 $9.63
N 189 189

 

* difference between clustered and control, p < .05
** difference between clustered and control, p < .005

Table 34 shows clustered papers responding to the survey are less aggressive
about pricing than clustered newspapers in the random sample. This suggests (a)
clustered survey respondents are not representative of the population of clustered
newspapers, or (b) the 1998 random sample is non-representative. The first seems more
likely because respondents to the survey were self-selected and do not differ from the
random sample in terms of market characteristics.

The questionnaire included three different measures of quality, or spending on
news. The first was the size of the news staff, the second was the news/advertising ratio,
and the third was the average number of column inches in the paper’s weekday newshole.

Table 35 shows clustered papers had an average of about 5 fewer staff members

than control newspapers, a difference that is significant. However, when the staff size

108

was standardized to control for the effects of circulation size the difference was not
significant. These results weakly support Hypotheses 3 and 3a that clustered papers

spend less on quality, while control papers spend more.

Table 35: Quality Measures, Survey Respondents

 

 

1998
Clustered Control
Mean size of paper’s staff 22.78 2823*
Std. Deviation 21.75 28.64
Mean size of staff per thousand circulation 1.43 1.38
Std. Deviation 1.39 1.05
N 189 189
Mean news/advertising ratio 1.33 1.33
Std. Deviation .54 .55
N 163 166
Mean column inches in newshole 1707 1822
Std. Deviation 878 1531
Mean column inches per thousand circulation 138.52 125.53
Std. Deviation 120.54 120.83
N 102 1 19

 

* difference between clustered and control significant, p < .05
** difference between clustered and control significant, p < .005
Note: Many reSpondents did not provide information about news/ad ration or newshole size.

Clustered and control papers responding to the survey had identical
news/advertising ratios. The differences in newshole also were not significant. These
results do not support Hypotheses 3 and 3a that clustering reduces spending on quality.

The questionnaire included four yes/no questions about sharing resources to save
costs (see Appendix B). The first was whether the paper shared local newsgathering

resources, such as reporters, with commonly-owned papers. The second was whether

109

copy was produced for multiple papers using a universal desk. The third was whether
administrative expenses, such as payroll, were shared with other papers. The fourth was
whether production resources, such as a press, were shared with other newspapers.

Table 36: Survey Responses to Questions About Resource Sharing

 

 

Cluster N Control N

Yes No Yes No

Shares newsgathering resources 139 49 189 67 131 198
Percent 73.9 26.1 33.8 66.2

Has universal desk 31 156 187 16 180 196
Percent 16.6 83.4 8.2 91.8

Shares administrative expenses 120 67 187 98 97 195
Percent 64.2 35.8 50.3 49.7

Shares production resources 122 65 187 59 138 197
65.2 34.8 29.9 70.1

 

Table 36 reports all available responses to these questions. Results show
substantially more clustered papers shared resources. Almost three quarters of the
clustered papers shared newsgathering resources, and two thirds shared production and
administrative resources. However, only about 17% had a universal desk to edit copy for
different newspapers.

About half of the control newspapers shared administrative expenses. A third or
less shared newsgathering or production resources. Results in Table 36 support
Hypothesis 5a that clustered newspapers are more likely to share resources to create
economies of scale.

A non-parametric Mann-Whitney test, which examines differences in the mean

rank of ordinal level data (Norusis, 1993), was used to determine whether differences in

110

responses to the resources sharing questions were statistically significant. Responses

were coded yes = 1, and no =2, so lower ranks indicate more yes responses. Results in

Table 37 show significantly more resource-sharing among clustered respondents for

every question. These results support Hypothesis 5a that clustered papers are more likely

to share resources.

Table 37: T est of Differences fir Resource-Sharing Questions
Paper’s classification N Mean Rank

 

 

Shares newsgathering resources Cluster 188 153.80“
Control 198 231.19
Total 386
Has universal desk Cluster 187 183.75*
Control 196 199.87
Total 383
Shares administrative expenses Cluster 187 177.93**
Control 195 204.51
Total 382
Shares production resources Cluster 187 157.74**
Control 197 225.50
Total 3 84
* difference between clustered and control significant. p < .05
‘”' difference between clustered and control significant. p < .005
Note: Yes = 1. no = 2. More yes responses result in a lower mean ranking.

The survey responses cannot be generalized to the population of clustered

newspapers because the respondents appear to differ in significant ways. However, there

are significantly more efforts to share resources among clustered survey respondents.

The only area where as many as half of the control papers share resources is

administrative expenses. This is not surprising because processing items such as

paychecks does not require that newspapers locate close to one another. Sharing

111

production or newsgathering resources to achieve economies of scale probably requires

that newspapers be close to each other.

However, if clustering produces Opportunities to create scale economies and save

costs, there is scant evidence from the t-tests these savings are used to increase newsroom

spending among clustered survey respondents. Nor is there any evidence the savings are

passed to advertisers or readers in the form of lower prices.

 

Table 38: Cluster and Control regressions, Size of News Stafl as DV

 

Cluster (N= 181) Control (N = 245 )

B Std. Error Beta t B Std. Error Beta t
(Constant) -l3.200 9.191 -1.436 .645 6.399 .101
Households .000 .000 .178 2.168* .000 .000 .243 4.762“
Avg. household .000 .000 .049 .889 —.000 .000 -.020 -.521
income
Paper’s penetration 5.869 11.611 .044 .505 13.807 8.521 .082 1.536
Clustered 7.017 11.455 .034 .613 NA NA NA NA
penetration
Daily penetration -7.737 8.281 -.063 -.934 -l.525 6.461 —.010 -.236
Weekly penetration 1.566 4.591 .016 .341 -.073 3.610 -.001 -.020
Ad price .610 .081 .551 7.571** 1.016 .072 .675 14.191**
Subscription price .043 .027 .077 1.583 .019 .028 .028 .702
Size of cluster .864 .739 .064 1.170 . NA NA NA NA
Percent ofcluster’s 19.641 5.893 .201 3333’” NA NA NA NA
circ.
Distance to nearest .082 .067 .057 1.229 NA NA NA NA
paper
Radio penetration -20957 14384 -.1 18 - l .457 -14599 11917 -.068 -1.225
TV penetration -4375 3892 -.076 -1.124 —3285 2645 -.060 -l.242
Cable penetration 3.831 2.334 .076 1.641 .047 2.214 .021 .983
Adjusted R2 .671 .785
F statistic 27.381** 90.476**

 

* p<.05, " p<.005

112

 

Regressions were used to further examine how clustering affects the measures of
newsroom spending. All available survey responses were used in each of these
regressions which is why there are far more control cases than clustered cases. Results in
Table 38 show a positive, highly significant association between the size of the news staff
and participation in a cluster. These results do not support Hypothesis 3 that clustering
reduces newsroom spending.

Table 38 also shows a positive, highly significant association between advertising
prices and staff size at cluster and control papers. There is a positive, highly significant
association between households and staff size for control papers. A similar association is
weakly significant for clustered papers. If the OLS coefficients are correct, changes in ad
prices result in larger changes in staff size at control papers. The betas show ad prices
and households are more influential at control papers. These results fail to support
Hypotheses 3 and 3a that control newspapers are more likely to increase newsroom
spending in response to competition.

Table 39 uses a standardized measure of staff size per thousand circulation. This
measure is not significantly associated with clustering. However, there is a negative,
weakly significant association between daily competition and staff pm at control papers.
A similar association for clustered papers has a large standard error and is not significant.
There is a positive, weakly significant association between television penetration and
staff size at control papers. A similar, highly significant association between cable and
staff size exists for clustered papers. These results are difficult to interpret because they
suggest control newspapers have smaller staff pm when daily competition increases.

However, these results do suggest control papers are more sensitive to competition when

113

determining staff size. These results weakly support Hypothesis 3 that clustered

newspapers are less sensitive to competition when determining spending.

Table 39: Cluster and Control Regressions, News Staff pm as DV

(Constant)
Households

Avg. household
income

Paper’s penetration

Clustered
penetration

Daily penetration
Weekly penetration
Ad cpm
Subscription ppm
Size of cluster

Percent of cluster’s
circ.

Distance to nearest
paper

Radio penetration
TV penetration
Cable penetration
Adjusted R2

F statistic
* p <05, ** p < .005

B
.226

-.000

.000

-794

.704

-.219

-.118

.377

-.000

.000

.191

-.001

128

-49.216

.303

.294

6.394**

.420

.000

.000

.528

.461

.341

.183

.085

.006

636

170

.095

Cluster 0‘1: 181)
Std. Error Beta

-.210

.274

-.214

.123

-.064

-.043

.426

-.008

.001

.070

-.019

.026

-.031

.214

t
.591

-2.346*

3.333**

-l.503

1.527

-.642

-.642

4.437“

-.075

.009

.794

-.262

.201

-.289

3.194**

Control (N = 245 )
Std. Error

B
1.606

-.000

.000

-1.968

NA

-.575

—.123

.056

-.002

NA

NA

NA

803

222

.134

.169

5.978**

.226

.000

.000

.321

NA

.225

.124

.068

.008

NA

NA

NA

450

99.470

.078

Beta

-.043

.089

-.699

NA

—.205

-.061

.080

-.030

NA

NA

NA

.213

.231

.115

t
7.108"

-.522

1.233

-6.128**

NA

-2.557*

-.995

.830

-.252

NA

NA

NA

1.783

2.242*

1.722

Regressions using the size of the newshole as a dependent variable had too few

responses to be considered stable. There are 14 independent variables, so 118 to 162 to

cases are required to test individual relationships and multiple correlation (Tabachnick &

114

F idell, 1996). Only 97 cases were available for the clustered papers in regressions using

newshole as a dependent variable. However, there were 158 cases available for control

papers.

Table 40: Cluster and Control Regressions, Column Inches in Newshole as D V
Cluster (N= 97)

Control (N = 158 )

 

B Std. Error Beta t B Std. Error Beta t
(Constant) 341.200 636.666 .533 347 434 .801
Households .001 .000 .161 - 1.160 -.000 .000 -.029 -.307
Avg. household -.002 .009 -.018 -.212 .010 .010 .078 1.057
income
Paper’s penetration 871 866 .138 1.006 1115 587 .193 1.900
Clustered 549 790 .063 .696 NA NA NA NA
penetration
Daily penetration 153 591 .028 .260 -27.039 425 -.005 -.063
Weekly penetration -242 327 -.057 -.793 445 258 .104 1.722
Ad price 22.919 5.244 .529 4.370** 29.838 4.673 .564 6.386M
Subscription price 1.065 1.843 .048 .578 1.508 1.945 .061 .775
Size of cluster 44.348 54.196 .072 .818 NA NA NA NA
Percent of cluster’s 395 394 . 100 1.002 NA NA NA NA
Circ.
Distance to nearest 1 1.963 5.387 .170 2221* NA NA NA NA
paper
Radio penetration -1961401 986478 -.250 -1.988 -1279040 853868 -.154 -1.498
TV penetration -149719 253856 -.059 -.590 -330337 184663 -.162 -1.798
Cable penetration 168 129 .094 1.299 -184 159 -.075 -1. 157
Adjusted R2 .566 .497
F statistic 10.046" 16.630**

 

‘ p< .05, “p< .005

Results in Table 40 show a positive, weakly significant association between

distance to the nearest clustered newspaper and newshole size. Clustered papers might

115

 

 

increase newsroom spending if other members of the cluster are more distant because

increases in quality help decrease reader elasticity of demand. However, the small

number of cases in this regression means these results cannot be interpreted as supporting

a hypothesis. The positive, highly significant associations between ad prices and

newshole at cluster and control papers may reflect the fact that larger newspapers have

higher prices.

Table 41: Cluster and Control Regressions, Column Inches pm as DV
Cluster (N= 97)
Std. Error Beta

(Constant)
Households

Avg. household
income

Paper’s penetration

Clustered
penetration

Daily penetration

B
20.196

-.000

-.000

-l7.269

58.794

43.689

Weekly penetration -56.044

Ad cpm
Subscription ppm

Size of cluster

Percent of cluster’s

circ.

Distance to nearest

paper
Radio penetration
TV penetration
Cable penetration
Adjusted R2

F statistic
* p< .05, ** p < .005

36.955

5.218

-.909

-4.366

.287

25379

40026

7.280

.520

8.503**

84.665

.000

.001

117

92.152

72.057

36.953

18.222

1.284

6.501

45.279

.660

134777

32338

15.435

-.041

-.026

-.024

.059

.070

—.114

.218

.471

-.013

-.010

.036

.028

.138

.035

t
.239

-.368

-.288

-.147

.638

.606

-1.517

2.028

4.062**

-.l40

-.096

.434

.188

1.238

.472

116

Control (N =158)

B
86.704

-.000

-.000

-33.989

NA

-3 8.332

37.817

-8.240

6.937

NA

NA

NA

-14262

6047

-4.538

.380

10.676**

Std. Error

41.547

.000

.001

59.212

NA

40.078

23.871

11.769

1.512

NA

NA

NA

87527

18568

15.059

Beta

—.139

-.041

—.069

NA

-.083

.106

-.070

.651

NA

NA

NA

-.021

.036

-.022

2087*

-l.551

-.175

-.564

NA

-.956

1.584

-.700

4588‘”

NA

NA

NA

_ -.168

.326

-.301

 

Results in Table 41 show a positive, highly significant association between

subscription ppm and newshole for clustered and control papers. This may indicate that

increases in newsroom spending allow managers to pass more circulation costs to their

subscribers, presumably because the newspaper’s quality is higher.

An alternative to these regressions examined Pearson’s correlations between the

newsroom spending measures and clustering variables for survey respondents. These

results are presented in Table 42. Thee is a positive association between staff size and

Table 42: Pearson’s Correlations for Spending and Clusters, Survey Respondents

 

 

News staff size Staff pm News/ad Col. inches Col. Cluster size Percent Distance Clustered pent.
ratio inches pm
News staff size 1.000
N 189
Staff pm .123 1.000
N 189 189
News/ad ratio -.127 -.073 1.000
N 163 163 163
Col. inches .803" .046 -.123 1.000
N 102 102 99 102
Col. inches pm -.345** .631" .166 -.156 1.000
N 102 102 99 102 102
Cluster size -.086 .083 .028 -.076 .171 1.000
N 189 189 163 102 102 189
Percent .269" -.146" -.093 .263" -.295** -.576** 1.000
N 189 189 163 102 102 189 189
Distance -087 -.1 10 -.090 .013 -.028 .000 -.014 1.000
N 189 189 163 102 102 189 189 189
Clustered pent .116 .031 -.007 .107 .000 .040 -.221** -.190“* 1.000
N 184 184 159 100 100 184 184 184 184

 

** Correlation is significant at the 0.01 level (2-tailed)
* Correlation is significant at the 0.05 level (2-tailed)

participation in a cluster. However, this correlation is negative when using staff size

standardized for circulation. This same pattern exists for the average column inches in

117

 

 

 

the newshole. The positive correlations for unadjusted figures may reflect the fact that a
paper’s participation in a cluster increases as the size of the cluster decreases.
Participation, measured as the paper’s percentage of the cluster’s total circulation, also
tends to increase with size. In other words, increased participation suggests larger papers
in smaller clusters which may therefore spend more on the newsroom. However, once
size differences are controlled, there is a negative relationship between clustering and
newsroom spending. These results weakly support Hypothesis 3 that clustering reduces
newsroom spending and news quality.

Results from Profit and Cost Measures

Data for publicly-owned newspaper companies in the study was gathered from
sources in described in the Method chapter. These variables included each company’s
overall profit, and the costs associated with each company’s newspaper division. The
degree of diversification among different lines of business was another variable. Recall
that lower values for this variable indicate more diversification. The degree of inside
control at these companies was the fourth variable in this section of the study.

There were not enough public companies in either 1988 or 1998 to conduct
regression analysis. Analysis of these variables was limited to correlations using
Spearrnan’s rho. This is a non-parametric measure that uses rank-order correlations and
does not require the assumption of a linear relationship. Results for both years are
essentially identical and will be discussed together.

Results in Table 43 show negative associations between newspaper division costs
and the market share of other clustered newspapers in 1988. There is a negative

association in Table 44 between costs and cluster size in 1998. These results support

118

 

 

 

 

 

Table 43: 1988 Correlations, Company Performance (Spearman’s rho)
Profit Cost Inside Divers Cluster Percent Distance Cluster Ad price Sub price Ad cpm Sub ppm

Size pent

Profit 1.000

Cost -.808** 1.000

Inside -.501" .906" 1.000

Divers -.779* .726” .284 1.000

Cluster size .052 -.044 -.029 -.060 1.000

Percent -.157 -.028 .028 .117 -.640"' 1.000

Distance -.306 .087 . 160 .317 -.035 -.009 1.000

Cluster pent. .584" -.594** -.307 -.714** .040 -.120 -.464** .1000

Ad price .585“ -.680** -.537** -.613**‘ .231" .066 -.231** .359" 1.000

Sub price .409“ -.601** -.554*"' -.327 .101 .138 -.067 .207“ .495” 1.000

Ad cpm -.092 .076 -.506* .164 -.032 -.237** -.249** -.065 -.248‘* -.276"‘ 1.000

Sub ppm -.450** .395* -.349 .562" -.132 -.l96** -.028 -.218** —.649** -.189** .780" 1.000

N 36 36 21 36 200 200 200 200 200 200 200 200

 

** Correlation is significant at the .01 level (l-tailed)
* Correlation is significant at the .05 level (l-tailed)
Note: There are fewer cases for company-level variables because many newspapers were not owned by public companies. There are
21 cases for the Divers variable because 1988 information was not available for all public companies.

Table 44: 1998 Correlations, Conwany Performance (Spearman’s rho)

 

 

Profit Cost Inside Divers Cluster Percent Distance Cluster Ad price Sub price Ad cpm Sub ppm
srze pent
Profit 1.000
Cost -.108 1.000
Inside 233-8" -066 1.000
Divers .397M .041 278* 1.000
Cluster size -.l86 -.274* .131 .111 1.000
Percent .155 .071 -.l84 —.133 -.636‘“" 1.000
Distance -. 130 .212 -.109 -.077 -.037 .067 1.000
Clusterpent. .066 .023 .073 .115 .013 -.163** 433.7" 1.000
Ad price -.101 -.269* -.015 -.125 -.109 .242" -.123* .342" 1.000
Sub price .187 -.494" -.O42 .183 -.O63 .169" -.137* .262" .498" 1.000
Adepm .361" .142 .206 .344“. .042 -.254"'* -.015 -.190** -.491** -.325** 1.000
Subppm .322" .087 .028 .413” .095 -.297** .025 -.269** -.810** -.187** .741" 1.000
N 63 63 63 63 312 312 312 309 312 312 311 312

 

** Correlation is significant at the .01 level (l-tailed)
* Correlation is significant at the .05 level ( l-tailed)
Note: There are fewer cases for company-level variables because many newspapers were not owned by public companies.

119

 

 

Hypothesis 5 that companies which own clustered newspapers have lower costs.

There is a negative correlation between profits and a company’s diversification
into different lines of business in 1988. However, this correlation is positive and
significant in 1998. This mixed result does not support Hypothesis 6 that profits are
lower at firms with a higher degree of diversification when those firms own clustered
newspapers.

Unadjusted advertising prices were negatively correlated with diversification in
1988. This correlation was not significant in 1998. The correlation between
diversification and standardized prices is positive in both years. However, the correlation
for ad cpm is not significant in 1988. These results are mixed and only weakly support
Hypothesis 6a that ad and subscription prices are higher at more diversified firms that
own clustered newspapers.

There is a positive relationship between cost and diversification in 1988. This
relationship is not significant in 1998. Since higher values of the diversification variable
indicate a less diverse company, this weakly supports Hypothesis 6b that more diverse
firms spend less on newspapers.

There is a negative relationship between profits and the amount of stock
controlled by insiders in both years. This supports Hypothesis 7 that profits decrease as
inside control increases at firms that own clustered newspapers.

There is a negative relationship between ad and subscription prices and the
amount of stock controlled by insiders in 1988. These correlations are not significant in
1998. These results weakly support Hypothesis 7a that prices are lower at firms with

more inside control when they own clustered newspapers.

120

 

 

 

 

 

 

There is a positive relationship between newspaper division costs and inside
control for 1988. This correlation is not significant in 1998. This weakly supports
Hypothesis 7b that firms spend more on their newspapers when insiders have greater
control.

In 1988 profits were positively associated with penetration from other clustered
papers in a market. This suggests clustering increases profits. However, in 1998 profits
had no significant association with clustering. These results only weakly support

Hypothesis 4 that clustering is associated with higher profits.

121

 

 

 

CHAPTER 6
CONCLUSIONS

Clustering is an increasingly pOpular strategy among United States newspapers,
and results of this study show the strategy profoundly affects the industry. There is less
competition among newspapers in clustered markets than in control markets. More than
half of the United States daily newspapers that vanished during the decade Of the study
were part of a cluster. Clustering is also associated with increases in ownership
concentration among the remaining newspapers. All of these effects are apparent from
the descriptive statistics and from t-tests performed for this dissertation.

Multivariate tests did not always offer strong support for some hypotheses. This
was partly due problems with skewness and other violations of regression assumptions
that weakened these tests. These problems apparently contributed to the instability
apparent in many regressions where OLS coefficients had large standard errors and
relationships were not significant. There may also have endogenity associated with the
price measures in the study.

However, most regressions in this study also produced an R2 of .50 or higher, and
many times this statistic was .60 or higher. This suggests that, despite the lack Of
significance, these regressions explain substantial portions of the variance associated with
variables used in this study.

The instability of individual OLS coefficients may also result from lumping
together markets of various sizes in the tests that were used. Previous studies have found

substantial variation across newspaper markets — papers in larger markets sometimes

122

 

 

 

Table 45: Results of Hypotheses Tests

 

Statistic used

Descriptive

[-1651

Correlation

Regression

1988 1998

First
difference
regression

Mann-
Whitney U

 

 

Clustering‘s effect
On elasticity

On competition
H 1

Hla

On prices
H2

H2a
H2b
H20
H2d
H2e

On quality
H3

H3a

On profits
H4

On cost
H5

H5a
Effects Of
diversification

H6

H6a

H6b
Effects of insider
control

H7

H7a

H7b

 

 

 

2.5

2.5

 

 

 

 

 

123

 

 

 

 

behave quite differently from papers in smaller markets, (see, e.g., Lacy, Coulson, &
Cho, 2001; Lacy, F ico, & Simon, 1989). However, the use of multiple tests in this
dissertation allowed a check on whether any given result was robust.

Table 45 outlines the results Of this approach. The table reports all tests that
supported each hypothesis in the dissertation. The columns listing the type of test report
the number of relevant tables supporting each hypothesis. If more than one statistic in a
table supported a hypothesis, the table still was only counted once. If a table had mixed
results which only partly supported a hypothesis, it was counted as 0.5. Keep in mind
when reading the table that some hypotheses could not be subjected to tests using 1
multiple statistics.

The tests of elasticity for 1988 and 1998 support the general argument in the
theoretical discussion that clustering reduces elasticity Of demand. Theory suggests this
effect results from the decrease in newspaper competition that occurs when clusters are
formed. Theory also suggests that clustered newspapers can take advantage Of reduced
competition to increase advertising and subscription prices. Results show that clustered
newspapers do have higher prices when using unadjusted and standardized measures for
advertising prices. However, price differences are less apparent using this study’s
measure of subscription ppm. These results may indicate that managers at clustered and
control newspapers are reluctant to pass circulation costs to their readers. This would not
be surprising given the importance of market share when trying to satisfy advertiser
demand for reaching potential customers.

The results are not just consistent with predictions that clustered newspapers have

increased market power allowing them to raise prices. The results are also consistent

124

 

with predictions that clustered newspapers are less sensitive to interactions between
advertising and subscription prices. Recall from the theoretical discussion that if
subscription prices affect demand for circulation, then they will feed back into
advertising prices. This is because advertisers buy access to the readers the newspaper
circulates to. One of the most consistent results in this study, found in Pearson’s
correlations, regressions and first-difference regressions, shows this interaction is weaker
for clustered newspapers than for their non-clustered counterparts. Theory suggests this
is because of the reduced elasticity Of demand among readers in clustered markets.

However, theory also predicted a negative relationship between subscription and
ad prices. That was not the case for clustered or for control papers. This may reflect the
measure used. The price of a standard advertising unit is one among dozens of prices that
many newspapers list for a variety Of local and national retail and classified advertising .
However, the SAU price was created to allow buyers to easily make comparisons
between newspapers, which is difficult using other, non-standardized prices.

The consistently positive relationship between advertising and subscription prices
morelikely reflects the fact that circulation demand is inelastic according to previous
studies (see, e.g., Lewis, 1995) and to the 1988 elasticity regressions in this study.
Managers may be reluctant to pass subscription costs to readers, but this inelasticity
means they do not have to lower subscription prices when they raise advertising prices.
This may explain the positive association between the two.

Results suggesting that clustering is associated with higher prices can also be
found in simple correlations, regressions and first-difference regressions. However, the

effects of clustering on price differences were more apparent in 1988 than 1998. The

125

association between clustering and higher advertising prices still was apparent in 1998,
but it was supported by fewer regressions. The association between clustering and higher
subscription prices was no longer apparent in the 1998 regressions.

These results suggest something changed over the course Of the study. By 1998
these differences might have been less apparent because Of two factors. The first was a
decline in competition in control markets that is apparent in Table 8. Average competing
daily penetration dropped from 42% to 34% in these markets; weekly penetration
declined from 40% to 22%. This, combined with a long economic boom that
strengthened demand, may have brought prices in control markets closer to prices in
clustered markets. The data in this study suggest this is the most likely explanation.

However, it’s possible that clustering became less effective as a strategy during
the 1990s. This seems unlikely given the continuing efforts by newspaper companies tO
assemble clusters. The first-difference regressions also show that changes in clustering
variables were positively associated with advertising and subscription prices. This
suggests there may be some changes in the nature of clusters that are not apparent when
aggregating the data across different-sized markets. Another possibility is that the
difference in 1998 reflects random fluctuations in the relationships among variables used
in this study.

In any case, further study is needed to fully understand why the market power
associated with clustering appears to have diminished in 1998. A true time series study
would help answer this question.

This dissertation also used multiple measures to examine clustering’s effects on

quality. These tests were limited by the measures available in the study. However, there

126

is support from t-tests, correlations, regressions and first-difference regressions for the
hypothesis that clustering reduces newsroom expenditures. Theory suggests that
clustering reduces elasticity Of demand, so clustered newspapers can reduce their
newsroom spending, and quality, without risking a disproportionate loss of readers.

There is less support for the hypothesis that non-clustered newspapers must
respond to competition by spending more on news coverage in an effort to differentiate
themselves. Indeed, one surprising result showed control papers have fewer wire
subscriptions in markets where there is more daily competition.

Responses to the mail survey provide strong support for industry arguments that
companies which create clusters share resources to create economies of scale. The
negative correlations between newspaper segment cost and clustering variables also
provide some support for arguments that clustering reduces costs.

However, there is no evidence, using the limited measures available in this study,
that these savings result in lower prices for subscribers or advertisers. There also is nO
evidence these savings are used to hire more news staffers or open up more space for
news. As was pointed out earlier, results instead show clustering is associated with
reduced newsroom spending.

Companies may instead use savings from clustering to increase their profits. The
theoretical discussion suggested this would be the case, and the positive correlation
between profits and clustered penetration in 1988 suggests this possibility. However,
there was no similar correlation in 1998. The overall profit measure for companies which

are sometimes diversified may not capture the relationship. This result may also indicate

127

 

savings from clustering were used to lower other costs or to finance additional
acquisitions in 1998.

The correlations are also mixed with regard to the association between
diversification and profits. There was a negative association between diversification and
profit in 1988, and a positive association in 1998. The result in 1988 supports the
theoretical argument that diversified companies are more difficult to manage effectively.
However, the 1998 result fails to support this argument and is instead consistent with
Picard’s (1999) study suggesting diversification is a hedge against fluctuations in the
advertising market. One possibility is that managers at diversified companies have
learned how to overcome the problems posed by diversification, or perhaps the nature of
diversification has changed. Recall from the theoretical discussion. that profitability is
easier to maintain when companies diversify into related lines of business.

Results of this study also show associations between increased diversification and
higher advertising and subscription prices. In 1988 the correlations also showed more
spending on newspaper divisions at less diversified companies. However, the lack of
multivariate tests make it difficult to say how these relationships are affected by
clustering.

This study also shows a positive association between the amount of stock
controlled by insiders and spending on a company’s newspaper division in 1988. In both
years inside control was negatively associated with profits. These results replicate earlier
findings (Lacy, Shaver, & St. Cyr, 1996). However, the lack of multivariate tests again

makes it difficult to say how these relationships are affected by clustering.

128

Implications of the Study

Overall, this study shows that clustering has substantial effects on the newspaper
industry. Newspaper companies appear to be buying competing newspapers so they can
reduce elasticity of demand and increase their market power. They appear to be using
this increased power to raise prices and reduce newsroom spending. In many cases,
newspaper companies are taking some Of their clustered newspapers out of business,
apparently in an effort to reduce costs.

None Of this is surprising. Competition’s effects on individual businesses are
rarely perceived by owners and managers as beneficial. In the case Of newspapers, this is
because competition gives consumers and advertisers an Option if they believe the price
Of one newspaper is too high, or its quality too low (Lacy, 1992; Litman & Bridges,
1986; Martin, 1997; Nicholson, 1995). This, in turn, forces newspapers to lower their
prices and increase spending at the expense of profitability. Competition increases the
number Of options available to consumers, but constrains the options available to
newspaper managers.

Reductions in competition, however, have the Opposite effect. By creating
clusters newspaper managers may be fortifying their existing market positions, or
extending the geographical limits Of their markets (Scherer & Ross, 1990). Clusters also
allow managers to reduce costs by creating economies Of scale. As the choices available
to consumers are reduced, managers are less constrained when focusing circulation in
areas where cost is relatively low and advertisers want to target readers (Blankenburg,
1982). Reducing costs while increasing the newspaper’s ability to attract advertisers

increases the financial Options available to newspaper managers. They can use the

129

savings to pay the acquisition costs associated with the creation Of clusters. They also
can use these savings to finance additional acquisitions. Or they can direct the savings to
the newspapers’ owners in the form of higher profits.

However, while individual newspaper companies may benefit from clustering, the
larger effects are likely to be negative. Economic competition is regarded as an effective
way to ensure that resources are distributed efficiently at the lowest possible price . The
United States has for decades passed and enforced anti-trust laws intended to preserve
competition and protect consumers from the effects Of market power that accrues to
businesses as competition is reduced (Breit & Elzinga, 1996). There has been at least one
case where the Justice Department joined an anti-trust suit filed by a competitor which
blocked a horizontal merger between two competing newspapers in northwest Arkansas
(Shaver & Lacy, 1999, p. 741).

Federal anti-trust guidelines suggest that when mergers and market power are
evaluated, one factor that should be examined is the potential effect of a “‘small but
significant and nontransitory’ increase in price” (Breit & Elzinga, 1996, p. 413). The
question is whether such a price increase can be sustained in a market because there are
no substitutes available to consumers.

Results in Table 9 show the average advertising cpm was 44% higher in clustered
markets than non-clustered markets in 1988. The average ad cpm was 13% higher in
1998. Shaver and Lacy (1999) speculated the government assumes competition between
newspapers and other media is sufficient to keep advertising prices low. However, these
results call such assumptions into question and suggest further study to examine whether

price differences in clustered markets may be significant under the Justice Department

130

guidelines. The results support Shaver and Lacy’s (1999, p. 740) call for the Justice
Department to examine whether clustering has anticompetitive effects on advertising
prices.

The decrease in competition associated with clustering raises concern about areas
other than prices. First Amendment theory argues that competition is also important in
the marketplace of ideas. The financial commitment theory makes a specific link
between competition in this marketplace when it comes to covering the news, and
competition in economic markets (Lacy, 1992a; Litman & Bridges, 1986). Economic
competition increases the quality of news coverage (Lacy, 1988; Lacy, F ico, & Simon,
1989; Lacy & Martin, 1998; Lacy, Shaver, & St. Cyr, 1996). Competition between
newspapers also increases the diversity of news and other ideas available to readers,
which is precisely the kind of competition that First Amendment theory suggests is in
society’s interest. Two studies have examined how newspaper competition affects such
interests.

The first study (Vermeer, 1995) examined how the increased diversity Of ideas
available in a “media-rich environment” (p. 98) affects the outcome of elections. The
study concluded that as newspaper competition increased, elections were more
competitive and decided by closer margins. Vermeer suggested the decline in newspaper
competition “may therefore have effects that go beyond competition for advertising”
(p. 104).

The second study examined the relationship between newspaper competition and

public Opinion in a survey about the most important problems facing the United States

131

 

(Lasorsa, 1991). The study concluded that neWSpaper competition made a significant
contribution in explaining increases in the diversity of responses tO this question.
Therefore, clustering may reduce both the quality of news coverage and the
diversity Of ideas available to readers. If this is the case, then clustering’s effects on
society go beyond higher prices paid by advertisers — clustering may also negatively
affect democratic processes. This study does not provide any evidence that this is the
case, but it does suggest a need for additional research into the potential for such effects.
Lacy and Shaver (1999) pointed out competition will be almost impossible to
restore in markets where clusters exist. The urgent need for further study of clustering

and its consequences that they first identified is reinforced by this study.

132

APPENDIX A

VARIABLES IN THE STUDY

133

Cluster = number of papers in cluster, from Editor & Publisher International Year Book
listings and maps.

Percent = percent of cluster's circulation contributed by paper. From Editor & Publisher
International Year Book listings.

Distance = Distance, in miles, to closest paper that is member of cluster. From DeLorme
Street Atlas 7.0

Subscription price = Annual subscription price from Editor & Publisher International
Year Book. Based on multiplication Of price for longest available period if no price was
listed for a full year. Price within county or city zone used when available.

Ad price = Price of an SAU. From Editor & Publisher International Year Book.

Wires = number of wire service subscriptions from Editor & Publisher International
Year Book.

HQl = number of competing papers headquartered in counties where a paper circulates.
From Standard Rate & Data.

HQ2= number of other clustered (or common ownership) papers headquartered in
counties where a paper circulates. From Standard Rate & Data.

Compco = number of counties outside its home county where a paper circulates. From
Standard Rate & Data.

Compent = paper's total circulation penetration in counties outside its home county.
From Standard Rate & Data.

Avgpent = average penetration of paper’s circulation outside its home county. Compent

divided by Compco.

134

 

Hshlds = total households, in thousands, for all counties in a paper's market. From
Standard Rate & Data.

AHI = Average household income for all counties in a paper's market. From Standard
Rate & Data.

Radio = number of radio stations in paper’s home town. From Broadcasting & Cable
Yearbook.

TVAD’I = number of television stations in Area of Dominant Influence that includes
paper’s home county. From Broadcasting & Cable Yearbook.

Catvsub = number Of cable subscribers for company serving paper’s home town. From
Broadcasting & Cable Yearbook in 1988. From Television & Cable Factbook in 1998.
Profit = Return on sales for paper’s owner. From Compustat records.

Cost = Newspaper division costs for paper’s owner. From Compustat.

Inside = Percent of stock controlled by insiders at paper’s parent company. From Value
Line Investment Survey.

Divers = diversification index, calculated as HHI using total revenue for different
business segments. Expressed as percentage w/ 1 = 100% of revenue from newspapers.
From Compustat.

Adcpm = ad cost per thousand circulation. Calculated as ad price/paper’s circulation
multiplied by 11,000.

covcpm = subscription cost per thousand circulation. Calculated as subscription
price/paper’s circulation multiplied by 1,000.

Paper’s penetration = paper's total circulation penetration for all counties in its market.

Calculated as circulation/hshlds. From Standard Rate & Data.

135

Clustered penetration = other clustered papers’ total circulation penetration for all
counties in paper’s market. Calculated as clustered circulation/lishlds. From Standard
Rate & Data.

Daily penetration = competing daily total circulation penetration for all counties in
market. Calculated as total daily circulation/hshlds. From Standard Rate & Data.
Weekly penetration = weekly total circulation penetration for all counties in market.
Calculated as weekly circulation/hshlds. From Standard Rate & Data.

Wirepm = wire subscriptions per thousand circulation. Calculated as wires/paper’s
circulation multiplied by 1,000.

Radio penetration = radio market penetration. Calculated as radio/hshlds.

TV penetration = television market penetration. Calculated as tvadi/hshlds.

Cable penetration = cable market penetration. Calculated as cable subscribers/hshlds.
Staff = size of news staff as reported on survey questionnaire.

Percent news = average percent news as reported on survey questionnaire.
Percent ads = average'percent ads as reported on survey questionnaire.
Inches = average number Of column inches reported on survey.

Sharel = answer to survey question about sharing newsgathering resources.
ShareZ = answer to universal desk question.

Share3 = answer to question about shared administrative expenses.

Share4 = answer to question about shared production resources.

News/ad ratio = percent news/percent ads.

136

APPENDIX B

SURVEY LETTER AND QUESTIONNAIRE

137

 

Date

MANAGING EDITOR
NEWSPAPER
ADDRESS

Dear:

The enclosed confidential questionnaire is for a study of the journalistic and economic
effects of cooperation among jointly owned newspapers. Much has been said about the
advantages and disadvantages of such cooperation, but few systematic studies have
examined these questions. This study's objectives include an examination of how such
cooperation affects the distribution of resources for news coverage.

Yours is one Of xxx United States newspapers selected for this survey. Some papers were
selected because their location near jointly owned papers makes cooperation possible;
other papers were selected to provide a comparison group.

The project needs as many responses as possible tO produce reliable results. Please take
two minutes from your busy schedule to help.

The identifying number on the questionnaire is used to track responses and plan follow-
up mailings. All identifying information will be destroyed at the conclusion Of the
survey. The confidentiality of the questionnaire will be protected to the maximum extent
allowable by law.

The data collected from individual newspapers will be aggregated for analysis. The
aggregate results will be publicly available this fall, but individual newspapers will not be
identified.

Please indicate your voluntary agreement to participate in this study by completing and
returning the questionnaire in the enclosed self-addressed, stamped envelope. Enclose a
business card if you would like a brief report on the results of the study.

If you have questions, please contact me the Michigan State University School Of
Journalism at (517) 353-6530 or by e-mail at martinhu@msu.edu You may also wish to
speak to the acting director of the school, Dr. Stephen Lacy, or to e-mail him at
slacy@msu.edu You may also contact the chair Of Michigan State University's
committee supervising research involving human subjects, David E. Wright, at (517)
355-2180.

Thank you for your help,

Hugh J. Martin
Project Director

138

 

Newsroom Resources Questionnaire

 

This first section concerns the resources and space available for news coverage in your
newspaper.

1. How many full-time reporters, photographers, artists, editors and copy editors are on your
staff? Please include employees in your newsroom and, if applicable, your bureaus.

Size of news staff
2. What is the average ratio Of news to advertising in weekday editions Of your newspaper?

Average percent news/average percent advertising / Don’t know

3. What is the average size of your weekday newshole in column inches?

Average number Of column inches Don't know

 

 

 

This second section concerns resource sharing among newspapers in a group. If your
newspaper is not owned by a group please skip this section.

1. DO you share local newsgathering resources with other newspapers in your group. For
example, do reporters regularly write local stories that are published in more than one newspaper?

Yes No

2. Does a universal desk edit and prepare local copy for your paper and other newspapers in your
group?

Yes NO

3. Does your newspaper group have a central Office to handle local administrative items such as
payroll processing, expenses, or Other items for your paper and other papers in your group?

Yes NO

4. Does your newspaper share production resources, such as a press or delivery trucks, with other
newspapers in your group?

Yes NO

 

Thank you for taking the time to complete this confidential questionnaire. Please write any
comments you wish to make on the back of this questionnaire. Please return the
questionnaire in the enclosed stamped envelope.

139

 

APPENDIX C

CORRELATIONS FOR VARIABLES USED IN REGRESSIONS

140

 

ooN

ooo;

0585 €3,330:

oom

nevvm.

ooo;

ho—
{noovr
new—Vt

ooo;

(“can
Beau

oom
oommmr
ooNVMr
duo—v.

ooo;

4:0;
\2F

ooN
nnwwvr
nnwovr
now_m.
no__>.

ooo;

.Eoo
.zonz

oom

m_f

vvo.
u_v_r
mmor
who:

ooo;

ocoo

9.xeo>>

oom
a*__o.
sum—m.
.nhowr
nnommr
ounwvr
nov_.

ooo;

.an
xzeo

oom
owo.
nmo.
m__r
nnwo_r
uVm_r
omor
nuommr

ooo;

.Eoo

oom
nnoovr
newsmr
anomc.
n*_oo.
oamoo.
Omor
enooor
nnOVN-

ooo;

.Eog

oom oom
noooNr mo_r
mm_r _wor
.wo. mm_.
nmo_. o_o.
*n__~. mvo.
omou wvo.
mo_r m_o.
anoomr uncomr
*e_o_. nov_.
ooo; ovor
ooo;

8320 Mtge.— oocsmfi E85;

336% ENSEEU .S\§OQEOEOD a. refuel Ammo \ Nov 2an

com

nnomw.

nowx_.

ohor

om_r

now—i

mmo.

owo.

ooo.

sac—r

moor

*§_vw(

ooo;

0N5
uofisru

com OON
oN_r oeor
.oe_r .~e_r
moo. neon
..ee~. .ms_.
..mmm. om_.
woo; mmo.
mmo. woo.
moo. emor
emo. omor
mwor awe;
e__r seer.
mwor Neor
coo; ..mme.
coo;

Ea 3:3 Eng .95 Emu o< 8E .nzm auto c<

 

.coogofia 038 5.3 $86 52 2a 22:. .
$25-8 :32 moo 2: 3 E3555 fl :ocflotou e.
325-3 _o>o_ So 2: on 63536 3 5:28.00 .1.

oom

ou_mm.

moo.

v__r

nmor

moor

wmor

nom_.

**VB_.

«*mNNr

aivcmf

aehomr

semen.

anm_v.

eamov.

ooo;

oom

namBN.

«room.

envom.

nuoom.

snowmr

anoom.

§©m_.

oo_.

*Nh_r

wwor

mm_.

**_©N.

snowm.
omo.

nmv_.
ooo;

va.

oom

eavmw.

§*_om.

eon—r

new—mt

neowmr

o__.

*iWW—.

aeho_.

nammmr

anONNfi

ooor

{avow.

m__r

oN_u

sic—o.

anmwv.

ooo;

2.

2:85
mEocomso:
«Eon 230

.Eoo >._t

.an 2on

.an 3x83

.Eom ESQ
.an hams—U
o:OQmLoamm

8535
“cocoon
95m cams—U
Ea 3:3
pcmm.n=m
Eco o<
oocm.o:m

euro o<

1411

ooN

ooo;

0E8: _

ooN

{envo.

ooo;

€323.31 .255 03.5

mo—

«eon—V

suOmmr

ooo;

oom
tsoomr
.«vcmr
neovm.

ooo;

.an >9

oom

nowhvr

anewVE

«awmv.

§*_hh.

ooo;

.an 063—

oom

wmf

vno.

vnor

«mv_r

n__v

ooo;

.an >333

ooN

{nu—o.

aaovfl

cement

aawmmr

**mer

omm

ooo;

.Eua EEO

com
.4var
.tmxmr
..Nmn
..oMm.
..eee.
ewor
..moer

ooo. _

oom

wo_r

awe—t

§MV_.

*Icmv.

nemmv.

vxor

_m_r

mmf

ooo;

.28; @893 Ea 3:3

203$ 3:59 .wazoterotob a. 28.594 memo N ”to @3th

.cosmbocom 03.8 2:3 898 mi 2m Bush .
$6:an .68. mod 2: a Eocene e 8:288 ..
625$ on: So a: .6 Became a 5:26:60 :

oom ooN
440mm- om_r
summmr om_r
unwmm. _mm
snowm. nnv_m.
esonw. uumOm.
mmor noor
owor omor
semw_. ooo.
noose. euomo.
coo; era—w.
ooo;
Eng .96 Ego v4.

oom

§*_BN.

*ava.

nvor

4*VmV(

ouoovr

who.

*aCON.

doommr

aahhmfi

aeoomf

{*wwmf

ooo;

outg .nzm

ooN

anWVm.

«comm.

nnwomr

aeoer

unnomr

one.

newMN.

anhmmr

ancmvr

nammmf

anommr

anmnm.

ooo;

88 2

Z.

088:.
33:34.0:
.2ch oEaU

.Eon >F

.an 06am

.an >333

.Eon EEO
.an 959$
Eu 8: 3
E9. .25
Ego o<
ootu.p:m

coca o<

1442

oom

ooo. _

oEOoE

oom

egmov.

ooo. _

mEofimso:

oo—

N:.-

ooo. _

.. Eon
2an

com
.32..
41.3..

moo.

ooo. _

End
>._.

oom oom
unmovr uvv_r
steam; __m-
mmo. moo.
en_>o. hmor
ooo: _oor
ooo;

Eco; Econ

35 583

ooN
nammm.
onoov.

oo_ .-
:oomr
*nmwvr

o~o.-

ooo.—

.Eoo
baa

oom
moor
3o.
Nwor
demo _ .-
unvoNr
mmor
tow _ .-

ooo. _

Eco;
bow—:0

oom
.2;er
nrwmmr

m2.
«Low.
Elmo.

o~or
taoowr
:Cmr

ooo. _

.Eog
Poona

 

.cocabocoa oEmo 5:3 mommo oo— oS 22E. .
325-3 3...... 3o 2: a 385:? a 5:26:60 ..
325-8 .22 so on. a 38%? a 856:8 2.

com oom oom com com ooN oom oom
mwor mmo. moor uummmr .*_mmr nuwmmr snow_. n*_vM.
omor _mo. _vor enmomr sec—mr neommr u~v_. mmcts
coo: __o. owor m_o. wmo. nmo. omo. owo;
semoN. w__r cmo. ntmmv. sew—m. renew. downmr *e_omr
eumnm. moor wmo. semen. eemmm. uumom. snoomr epoomr
moor v_o. who. ovo. omo. _oor .50. moor
wm_r ovo. moor sumo—r 800:: nm.: 450—. uevw_.
snmmmr eumnmr nom_. mm_. tumo_. smm_. boo. Ami
tumw_. oo_. vwor nm_. moo. n_or .05.; t*v_mr
ooo._ ooo. moor mvo. vwor coo; No_r __m-
coo; «sown: .«mmmr nnmomr suwwmr ~m_. temwm
ooo._ renew. 4.0mm roomm. over moor
ooo; ot_mm. esooo. onwmm; nemmmr
ooo: oe_vw. omo. neo_mr
ooo; .amm- two—r
ooo: 4400—.
coo:
oNB
853:0 Eoouom 8.3.0 Ea 8:3 Ema .nzm Ego u< 3:... .25 ootq c<

Z

oEOoE
@3545:
Leon vino
Eon. >._.

.an 06am

Eon bxoo 3

Eva EEO
Eon SEED
Eon mmoamm

8.835
Eooeom
ONE SEED
Ea 3:3
Eng .96
Ego v<
ootg.n:m

ootg o<

mammal 3.48450 AOXEEEEEOU a. 23.28% moo N ”men 2an

143

oom

ooo;

EEOo:_

oom

**oov.

ooo;

mEocomso:

oo_

*©h_.

ov_r

ooo;

..Eon oBaU

ooN

**oNNr

**Nmmr

mmo.

ooo;

.Eon >rr

oom

**w_vr

**_vvr

ono.

**ovc.

ooo;

oom

**wvmi

moor

moor

vwor

mmor

ooo;

ER. 2ch Eon x383

oom

*ahov.

**wmm.

__or

*mo_r

**__mr

Nnor

ooo:

Eon EEC

oom

**womfi

**mer

__m

**o_©.

**th.

ovor

**thr

ooo;

.Eon mLonan

oom

**OVNV

**o©mr

w_f

**ocv.

**hmv.

_wo.

___r

**omm.

ooo;

En 3:3

oom

**owmr

**mvv{

om_.

**whv.

*unwv.

.mo.

moo.

**NOM.

**NNB.

ooo;

pcnn.n:m

 

.cosmbocon oEmo 5:3 momao 02 one 82:. .
.ooozmrmo 35. mod 2: E Emoc_cw_m : cousotou _._
$25.3 3.6. So 2: E .52.:ch : 532880 e...

oom

om_r

*hm_f

who.

**_©m.

**m©m.

moor

w_o.

_wo.

**mm©.

*nwmm.

ooo;

Enon<

oom

**VVM.

*noom.

om_r

**©OV(

**womr

mmo.

**VON.

**mwmr

**hmv{

**©©mr

**ONmr

ooo;

oocn.:sm

oom

**OVN.

**mom.

mm_r

**Nmmf

**wmmr

wmor

N_f

**5Nmr

**m_vr

**mmmr

**OONV

**WN©.

coo;
ootno<

. Z

oEOoE
8.0530:
.Eon 2me
.Eon >n.

.Eon 06mm

Eon >333

Eon EEO

Eon ozonmn

En 3:3

Pcnn.n:m

Eno o<

ootn.psm

ootn o<

EoQom SEED .B\.€o.:e\o.tob m. ransom mono N Nov 2me

144

REFERENCES

145

American Newspaper Markets Circulation 89-90 (1989-1990). Malibu, CA: American
Newspaper Markets.

Asher, D. (1999, December). Who owns what. Presstime, 29-30.

Bass, J. (1999, ). Newspaper monopoly. American Journalism Review, July/August, 64-
86.

Blankenburg, W. B. (1982). Newspaper ownership and control of circulation to increase
profits. Journalism Quarterly, 59(3), 390-398.

Blankenburg, W. B. (1989). Newspaper scale & newspaper expenditures. Newspaper
Research Journal, 10(2), 97-103.

Blankenburg, W. B., & Ozanich, G. W. (1993). The effects of public ownership on the
financial performance of newspaper corporations. Journalism Quarterly, 70(1),
68-75.

Breit, W., & Elzinga, K. G. (1996). The Antitrust Casebook: Milestones in Economic
Regulation ( 3rd ed.). Orlando, FL: The Dryden Press. _

Bucklin, R. E., Caves, R. E., & LO, A. W. (1989). Games of survival in the US
newspaper industry. Applied Economics, 2] , 631-649.

Dertouzos, J. N., & Trautman, W. B. (1990). Economic effects of media concentration:
Estimates from a model of the newspaper firm. The Journal of Industrial
Economics, 39, 1—14.

Doran, H. E. (1989). Applied Regression Analysis in Econometrics. New York: Marcel
Dekker, Inc.

Ferguson, J. M. (1983). Daily newspaper advertising rates, local media cross-ownership,
newspaper chains, and media competition. Journal of Law & Economics, 26, 635-
654.

146

Fu, W.-J. (2000). Three essays on the economics of networked and traditional
information services. Unpublished Dissertation, Northwestern University.

Govindarajan, V., & Fisher, J. (1990). Strategy, control systems and resource sharing:
Effects on business unit performance. Academy of Management Journal, 33(2),
259-285.

Greene, W. H. (2000). Econometric Analysis ( 4th ed.). Upper Saddle River, NJ: Prentice
Hall.

Gujarti, D. (1992). Essentials of Econometrics. New York: McGraw-Hill, Inc.

Hickey, N. (1997, November/December). Will Gates crush newspapers? Columbia
Journalism Review, 28-36.

Hill, C. W. L., Hitt, M. A., & Hoskisson, R. E. (1992). COOperative versus competitive
structures in related and unrelated diversified firms. Organization Science, 3(4),
501-5 21.

Hill, C. W. L., & Hoskisson, R. E. (1987). Strategy and structure in the multiproduct
firm. Academy of Management Review, 12, 331-341.

Jessell, H. (Ed.). (1999). Broadcasting & Cable Yearbook I 999. New Providence, NJ:
R.R. Bowker.

Lacy, S. (1984). Competition among metropolitan daily, small daily, and weekly
newspapers. Journalism Quarterly, 61(3), 640-644, 742.

Lacy, S. (1988). The impact Of intercity competition on daily newspaper content.
Journalism Quarterly, 65(2 Summer), 399-406.

Lacy, S. (1989). A model of demand for news: Impact of competition on newspaper
content. Journalism Quarterly, 66(1), 40-48, 128.

Lacy, S. (1992a). The financial commitment approach to news media competition.
Journal of Media Economics, 5(2), 5-21.

147

 

Lacy, S. (1992b). Ideas for prospering in a changing market. Newspaper Research
Journal, 13(3), 85-94.

Lacy, S., Coulson, D. C., & Cho, H. (2001). Competition for readers among US.
metropolitan daily, non-metropolitan daily and weekly newspapers. East Lansing,
MI: Michigan State University.

Lacy, S., & Dalmia, S. (1993). Daily and weekly penetration in non-metropolitan areas Of
Michigan. Newspaper Research Journal, 14(3&4 Summer-Fall), 20-33.

Lacy, S., & Davenport, L. (1994). Daily newspaper market structure, concentration, and
competition. The Journal of Media Economics, 7(3), 33-46.

Lacy, S., & Dravis, S. (1991). Pricing of advertising in weeklies: a replication.
Journalism Quarterly, 68(3), 338-344.

Lacy, S., Fico, F ., & Simon, T. (1989). Relationships among economic, newsroom, and
content variables: A path model. Journal of Media Economics, 2, 51-66.

Lacy, S., & Martin, H. J. (1998). Profits up, circulation down for Thompson papers in
80s. Newspaper Research Journal, 19(3 Summer), 63-76.

Lacy, S., Shaver, M. A., & St. Cyr, C. (1996). The effects Of public ownership and
newspaper competition on the financial performance of newspaper corporations:
A replication and extension. Journal-ism & Mass Communication Quarterly,

73(2), 332—341.

Lacy, S., & Simon, T. F. (1993). The economics and regulation of United States
newspapers. Norwood NJ: Ablex Publishing.

Lacy, S., & Simon, T. F. (1997). Intercounty group ownership Of daily newspapers and
the decline of competition for readers. Journalism & Mass Communication
Quarterly, 74(4 Winter), 814-825.

Lasorsa, D. L. (1991). Effects Of newspaper competition on public Opinion diversity.
Journalism Quarterly, 68(1/2), 38-47.

148

 

Litman, B. (1988). Microeconomic foundations. In R. G. Picard & J. P. Winter & M. E.
McCombs & S. Lacy (Eds), Press Concentration and Monopoly: New
Perspectives on Newspaper Ownership and Operation (pp. 3-34). Norwood NJ:
Ablex Publishing.

Litman, B. R., & Bridges, J. (1986). An economic analysis Of American newspapers.
Newspaper Research Journal, 7(3), 9-26.

Litman, B. R., & Sochay, S. (1994). The emerging mass media environment. In R. E.
Babe (Ed.), Information and Communication in Economics (pp. 233-269).
Boston: Kluwer Academic Publishers.

Lewis, R. (1995). Relation between newspaper subscription price and circulation, 1971-
1992. The Journal of Media Economics, 8(1), 25-41.

Maddux, D. (Ed.). (1999). Editor & Publisher International Yearbook (79 ed.). New
York: Editor & Publisher.

Markides, C. C., & Williamson, P. J. (1996). Corporate diversification and organizational
structure: A resource- based view. Academy of Management Journal, 39(2), 340-
367.

Martin, H. J. (1997, Feb. 7-8). Viewing Readers as a Resource: A Transaction Cost
Approach to Studying Newspaper Behavior. Paper presented at the Winter
Regional Conference, Association for Education in Journalism and Mass
Communication, St. Petersburg, FL.

Martin, H. J. (1998). Measuring newspaper profits: Developing a standard of comparison.
Journalism & Mass Communication Quarterly, 75(3 Autumn), 500-517.

Merrilees, W. J. (1983). Anatomy Of a price leadership challenge: An evaluation Of
pricing strategies in the Australian newspaper industry. Journal of Industrial
Economics, 31(3), 291-311.

Meyer, P. ( 1995, December). Learning to love lower profits. American Journalism
Review, 17, 40-44.

149

Morton, J. (1997, July/August). Chains swallowing other chains. American Journalism
Review, 19, 52.

Moses, L. (2001, January 22). Newspapers brace for a pending storm. Editor &
Publisher, 134, 37-38.

Nicholson, W. (1995). Microeconomic theory: Basic principles and extensions ( 6th ed.).
New York: The Dryden Press.

Niebauer Jr., W. E., Lacy, S., Bernstein, J. M., & Lau, T.-y. (1988). Central city market
structure's impact on suburban newspaper circulation. Journalism Quarterly,
65(3), 726-732.

Norusis, M. J. (1993). SPSS for Windows Base System User's Guide Release 6.0.
Chicago: SPSS Inc.

Norusis, M. J. (1994). SPSS Professional Statistics 6.]. Chicago: SPSS Inc.
Presstime. (2000, February, 2000). A decade of deals. Presstime.

Picard, R. G., & Rimmer, T. (1999). Weathering a recession: Effects Of size and
diversification on newspaper companies. The Journal of Media Economics, 12(1),
1-18.

Rosse, J. N. (1967). Daily newspapers, monopolistic competition, and economies Of
scale. The American Economic Review, 5 7(2 May), 522-533.

Scherer, F. M., & Ross, D. (1990). Industrial Market Structure and Economic
Performance. (3rd ed.). Boston: Houghton Mifflin.

Shaver, M. A. (1995). Application of pricing theory in studies of pricing behavior and
rate strategy in the newspaper industry. The Journal of Media Economics, 8(2),
49-59.

150

Shaver, M. A., & Lacy, S. (1999). The impact of interrnedia and newspaper competition
on advertising linage in daily newspapers. Journalism & Mass Communication
Quarterly, 76(4 Winter), 729-744.

SRDS Circulation 99 (1999). Des Plaines, ILL: SRDS.

Tabachnick, B. G., & Fidell, L. S. (1996). Using Multivariate Statistics. (3rd ed.). New
York: HarperCOllins College Publishers.

Taliaferro, M. C. (Ed.). (1999). Television & Cable Factbook( VOl. 67). Washington,
DC: Warren Publishing.

Thompson, R. S. (1988). Product differentiation in the newspaper industry: an hedonic
price approach. Applied Economics(20), 367-376.

Tirole, J. (1988). The Theory of Industrial Organization. Cambridge, Mass: The MIT
Press.

Velez, O. (Ed.). (1989). Editor & Publisher International Year Book. New York: Editor
& Publisher.

Vermeer, J. P. (1995). Multiple newspapers and electoral competition: A county-level
analysis. Journalism & Mass Communication Quarterly, 72(1), 98-105.

West, D. V. (Ed.). (1989). Broadcasting & Cable Yearbook '89. Washington, DC:
Broadcasting Publications, Inc.

151

 

 

 

 

 

 

 

 

 

iii...»
3...»

7.2.1....

 

new”

R1 ”91’..le Ilfi'iEG71)

'rr.

.. :_
353.
it?
:5 .

”(I‘V‘LQ’W