3 (MI- mar;
‘ -9

:IS'I’ _.~:
y- 1 Ah .. .
’0‘ “4&5 -

 
    
 
 

1..
D

I I‘ ”'0‘ “3;: s .‘

W .I A : I». III}
-‘ '-III’".‘ «:5. '= ..
I; "III???" "‘1- WI" "IL

   

I IQ; :‘ .
'I' 1| . 'I ‘
u'o " 'E'Ifi'IH I J. I ' :;mv"“

) :‘II 1'
u ' .I .
I Ji‘ I'h'.I"*g' 95‘
WINE" - I' ' . :I
.‘r’ ’3 l. l I'I I
., ' I

.' IIJII ,PII‘I" ‘ “5% 5&5‘ III? I
, . .‘,I'.'I'I“7' I'II'IIF ,(Ifmm I II ,I.

    

  

*l‘I It i I‘.I\U

1&3; I' In”; 'fdx

IIM:L'

 
 
   

I' .I w .
7‘?" '. 1"."
,. M 9%: _. 3'
‘I‘I'I‘_ ,lIII‘. a,"
‘I I ”3?. ,}I\:‘II IL

 
 

I .: M“ . *IW" ‘IIIII ‘I- ' "~33. - ..

\I‘x‘I' " r): I,"‘II I, I‘"."I" . (I... II ' I’II" iI’III :‘II lIlI-'y I. II"
If”)? {MI ’I_.7'I‘.I ‘.I “I mam "& (III #1!qu IIII< g‘wggfyfiug"gflfl. "iv-“jut. -.
II” IIPI’IIII. I I --"’ III .th“ mi". 5'..I"3‘I'II.CIFI.I..‘*3

 
   
   
  
   
 
  
   
  
 
 
 
 
  
 
   

 

     

I I .3” I W, 5 I19" I11.) .
:I'I'I‘ fl - . {II
QM}. r“ . I,“ '1' v- ’. .I II I . ,
I .- \f‘JfiI‘ III!I"I",I{III'H 'I‘I’; ’.,'II. I”) I” I I‘V'W II'I' I)“ m “If: g’II' fiyé- ..‘ “IMF”?
”5+1"? (11:. 41,": “wk: ,1.” I . Iggy": 1M fig“: I't‘JIIQYI :'m.I.I.,I I1"! '1" 5‘30 {III
IIIgII' “III "II 3301:? i? I" It; "II (rva'IuI'lfi'p'I'n ”1“?" I."‘T"IJI"..' ‘W' (\g \E I
.. I ,I I:',III I‘ I. II"IPIII‘I u fwgijiv. II \: ,c:"g';. 0’ ”,1.
.. I JI! 2:2. I I It! 40$ 'III “k.“ “MW“
‘ :‘III' ’6‘. "IT "A" ‘ I'C'X' flit” :Il' lIII "IWI‘LHI QIIIC'I'J III’IIII'II, “g!” .IIII Ilia; 1'14: 3"! 0‘3“,“ that"
I:,',II, I :I‘II .' II‘IIII,|",I,I' r "I. III." '1’, Fun,“ \ LIIIII 1‘- 2}! :4 .I"~I I'IIIIIII om)‘: III III). r”({.¢.“I‘.I'I.II,-
'Irll' I'IbAIIII, 1*, 'I" . “I: ,I'IZI‘ IIII) IKI'M'II 'I' II'VI'II:II HIIIIIIII“ ‘ ' " III} "gif'III'I' III 'IJKIkI-IIIH . ‘1'? "I." 'I
'II "’ I ‘I' II'I’I “I, .. II-' " 3‘. I I'I‘ Im' .‘1 Aw "'I \‘I.I . ~. In; IIIII."".‘"»'I(2
M. .I' IIIEIII‘II .II'IIIIsIIII II." "I II p " I . . II 3' l I.I,I.:I.Ia§7 542.4... .IIII'IIIII'IIII'I'III L. ”‘I
I ‘ 1 'I II}, 'l,‘ Ilz.‘ 9:31“ : IIKII J'I" "I III‘ III .- “‘5‘,“ 3,], 0:: .' ," . ' 'III.;:IIII " “III-'1 I .II" 5%.]; tIIIIIff'III ' II '.I‘ I", I'JQIIIfiW'
M ‘ III" Il""1 IJII IIIII"'I | “:21.”th I'II;‘r’s'3"I'III .II,"I II .It. "I I I ' '.\'.‘ ' I {I 9"“ I?!“ '4'" ”I“? III'II‘I "' I'
I III ..-I‘II.1|I fIIIIIII I. IIII.:‘I- .III,,II ..II"I“'I I '_IrI IIIII'IIIII' II.I'r\I,I'I.If.,'III.'I,' II. Iii-I I I] '8 Inf", ,,-‘III'II .I}
' .' 'Il.‘ ',III I' I -'I II. I ’I I'.‘ J-I 'I .I' ’I '. .1. .". ‘I II"?! . I I‘J‘uIllPI‘ I"
'I.i'1l"I'II‘I - ..I ~ In. , . .. . 4-. I. I.“ ~ .
.’I‘ I‘I".'I"I I. 3. ' .l 'I\'I "IIIII‘III'III'IIIIIII'I '.," "‘ I’ «I 'I'I--II"§II'-.I " . ." I‘I‘I'II'U '..
. II‘,I»..I..,I:. ...».~’ ,fiw’f I',-- . I. II-Ii'wfljw ‘ .v - . I 7'. I.[- I
l"I!I"' I" 'II'II I.I- 'w‘ [I -. , ,’I': III: .' ’.'-..' "'“I .‘I'II'I".I.‘ II'I',‘ I ""I- I.' .
.I! I I’.‘ I. -~ --'| . .I. «II‘I'I II‘III -'-I. II -: -< . I '.I'II I... I"
I I IuhII -}. I. "l I IIII,‘. II:V.fI’ II'III'I I." ,| II 1 ‘I".I. I III‘I. 51"; I‘ , \I'. fr,“ I "'J .fi'l
I" . I.'.' "II JIM" .CI-II l--I 5. .. . . Id'i’I .. I. T; ._ . II'I' I‘II‘ ..'1'.I,I'V"I'r
‘ . I“). "(.I... II‘?JII' . .'IL Ib' ”VII" I. ' III] dIII'"? IIIII II'II. If " . ' "T. "I
II .I . -" IIIHII.I".I‘ . ' I'fiz'I'I IMiIIII “I,“ 'I‘ "LIE . 'I r-I
5 I I I_,... I‘ .I\"l. 'I.I" f." I\"' ,,,I , I'I,II- III I. I IQ. M | *p
I’MIII 'II',.'.',' III 'I’.‘ II.” II '~I| IIIIIII I. 'II IIIIIII , I III I III ill; I‘III: .I'II'III I, I I -II)"III1I,fiI’. I'I'I‘lf‘ ,.
«III I... .. In' . III III-tr... ' . III-III. .;
II

I I IIIIIII.I:‘I.

"III'I‘I'lI'"?Iir'I'I«I I" l ‘ ..;r.;. III-“3.1. :‘NII’I‘II‘II
.II II,.II,I-II II' IQIIIIIIIIIIII ' -I$.I[130;.§QJI. g
IIIII N“ ”I .I. "‘3'" III III {3}“ 'IIWWI'I; I ' IIIIII I g
. 1|er I31"; ”MW-"I33“ ,IIIIII’pt‘II I IIIfI‘II’ I .'
fi‘Ihl than... fi'Il{ I'vqlnf' I','

I {III “*IIsz‘ .II .

I
H
'I-II‘“: Ir: j. §‘ ..I .,.I:~ [513's III I: III :.,I.I -. , I .J r
IIIIIh. III." (II I" I I'I,'I.? I" .‘II,"I'II“:"!' 'II'I'l‘ I VIII, .I'KIIIII' «WI! 'I‘K' II. II"}'I|I'.] Lu IIIIII'j'I I'EI’IQ‘J- '30}: I'r(I If‘-I'I‘II’}'III,‘vII' "I I“
”“II‘ E III} “191%“! II 'I (I IIIIII I 4" I-LI;I-:'I'I--II[{!:IIII I1 .

      
  
  
  

.‘tl.'

5:—

 

  
 
 

M“

 

This is to certify that the

thesis entitled

MASSMOD: A COMPUTER SIMULATION OF THE
MASS MEDIA INDUSTRY
1945-1960

presented by

Jayne Winifred Zenaty

has been accepted towards fulfillment
' of the requirements for

Ph. D. degree in MESLMEdil

/%u 75/4

Major professor

 

Dateiept. 12, 1980*

 

 

l

OVERDUE FINES:
25¢ per W per item

RETURNING LIBRARY MATERIALS:

Place in book return to remove
charge from circulation records

 

MASSMOD: A COMPUTER SIMULATION OF THE
MASS MEDIA INDUSTRY
1945-1960

by

Jayne Winifred Zenaty

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Mass Media Ph.D. Program
College of Communication Arts and Sciences

1980

w, .4"
.2 J/

’0' #5

ABSTRACT

MASSMOD: A COMPUTER SIMULATION OF THE
MASS MEDIA INDUSTRY 1945-I960

By
Jayne Winifred Zenaty

This research applied systems simulation methodology to the study of
the economic and structural behavior of the mass media industry from 1945
to l960. Using historical data from a time period marked by the intro-
duction of a new communications technology--television, the study had two
major purposes: (I) the quantification of abstract theories about media
structure and economics and the identification of integral relationships
among media components, accomplished by the construction and validation
of a computer simulation model called MASSMOD; and (2) the demonstration
of the feasibility and usefulness of systems simulation to the study of
media behavior, planning and policymaking, shown by a series of experi-
ments which manipulated two policy variables in the model and compared
alternate results to the benchmark model. Because of the shortage of
past research that has applied systems simulation to the mass media as an
industry, the research was designed to be exploratory.

MASSMOD, written in Minnesota FORTRAN, was built at a highly aggre-
gate, industry-wide level. It consisted of four sectors: broadcast (AM
and FM radio, VHF and UHF television), print (newspapers and consumer mag-
azines), advertisers and consumers. Ordinary least squares regression
routines were used to develop structural equations for the system. Post-
dictive validation techniques, using Theil's (l96l) Inequality Coeffi-

cient, was employed to validate the model's output.

Zenaty

The system represented by MASSMOD showed a remarkable stability
during 1945-1960, when television, first VHF and then UHF, was introduced
to the system. The number of magazines and newspapers published in the
period remained fairly constant; revenues grew slightly as both adver-
tising expenditures on each medium and circulation increased. The print
sector was not integrally linked to the broadcast sector in terms of
sensitivity to change. AM stations increased in number and listenership,
although earnings declined as advertising expenditures were transferred
to television. FM stations grew steadily, while revenues and expenses
declined, then took an upward swing. UHF television seemed the most
unstable subsector, with most variables decreasing in value and earnings
in the red.

The experiments performed on the system, which assumed that initial
conditions were the same as the historical situation, indicated that the
system would remain remarkably the same under different policy parameters.
Only a total FM and/or a total UHF broadcasting system would make those
media prominent and profitable, though still not as successful as histo-
rical AM radio or VHF television. UHF television had little effect on
VHF television, even when UHF frequencies were introduced in 1945, in-
stead of l952 as was the case historically.

MASSMOD presents a :simple, valid mathematical representation of the
mass media industry from 1945-1960. It lived up to its purpose as an ex-
ploratory effort into the use of systems simulation as an appropriate tool
for study of the behavior of the mass media industry. Supported by ad-
ditional data, the model needs to be expanded and improved, so that mathe-
matical relationships and theoretical speculations which are not included

in MASSMOD, especially in the FM and UHF sectors, may be investigated.

To be used as a real forecasting tool, the model needs to be updated to

Zenaty
describe industry behavior in the 1960's and 1970‘s.

Despite its shortcomings, MASSMOD has pulled together elements of a
media system too often studied in isolation, and quantified abstract
relationships. It is a simulation of a historical reality, available for
experimentation and questioning. At the minimum, this study has sug-
gested an approach that someone with access to a rich data base could use
to make accurate predictions of the consequences of policy decisions

made to influence the behavior of the mass media system.

@Copyright by
JAYNE NINIFRED ZENATY
1980

ACKNOWLEDGEMENTS

Like the proverbial tip of the iceberg, this bound dissertation is
one small, tangible sign of a much larger reality: two years of course
work in the Mass Media Ph.D. Program, and two years of dissertation
research, permeated throughout by the guidance, challenge, wisdom,
support and friendship of faculty and friends.

The guidance committee plays a significant role in the life of a
doctoral student; mine contributed encouragement, trust, and the
challenge to try new things, and included friendship as an additional
bonus. I am most grateful 'to Thomas F. Baldwin, my chairman, for his
patience, insight and belief in my abilities; to Martin P. Block for
his occasional impatience, navigation through the details and the
motivation to keep going (and get finished); to John D. Abel, for sus-
tained support, new viewpoints and a fierce dedication to research and
the rigors of the scientific method; and to John B. Eulenberg, for his
cheer, his broad vision and hope for the computer/communication future
and his camera. I also wish to express my thanks to Robert Schlater,
chairman of the Department of Telecommunication at Michigan State, who
provided assistantship support, and to the wives and families of my
committee members, who made me feel welcome and at home.

To my life-time guidance committee of two--my parents--my thanks.
No amount of formal schooling or advanced degrees can match their

wisdom, understanding, comfort, support and love.
ii

Friends provide the emotional strength and encouragement to
persevere through doctoral studies; several of mine are integrally a
part of my doctoral experience and this document: Ann, Mary and Heather,
for secretarial help and much more; Helen and my friends at Clarke
College for encouragement to pursue graduate studies; Kate, Joey, Brian,
and Eric for being there. Thanks should also go to my colleagues in
the Department of Telecommunications at Indiana University for their

patience and support during the final phases of this research.

Chapter
I.

II.

TABLE OF CONTENTS

INTRODUCTION .......................
Problem Statement ..................
The Purpose of the Study ..............
Assumptions and Research Questions .........

Assumptions ..................
Research Questions ...............

Summary .......................

LITERATURE REVIEW ....................

Media Simulations Relevant to This Study ......
A Media Allocation Model ............
A Time Allocation Model ............
An Advertising Policy Model ..........
Summary ....................
The Use of Simulation in Policymaking ........
Policy-Related Simulations ...........
Models of Industries ..............
Summary ....................
The Use of Simulation in Theory Development .....
Summary ....................
Mass Media Simulations ...............
Summary ....................

Conclusion .....................

iv

7
8

IO

11
11

25
27
28
3O
30

Chapter £29£_

III. THE MODEL ........................ 32
Methodology and Data Collection ........... 33
Problem Formulation .............. 35
Collection of Real World Data ......... 35
Formulation of Mathematical Models ....... 36
Estimation and Evaluation of Parameters . . . . 38
Formulation of a Computer Program ....... 38
Validation ................... 39
Description ..................... 41
Historical Perspective ............. 41

An Overview .................. 43

The Print Sector' ................ 45
Newspaper Subsector ............ 45

Magazine Subsector ............ 50

The Broadcast Sector .............. 55

The AM Subsector ............. 55

The FM Subsector ............. 59

The VHF Subsector ............. 63

The UHF Subsector ............. 68

The Advertiser Sector ............. 72
Probability Distributions ......... 79

The PARAM Array .............. 81
Calculation of Advertising Expenditures. . 85

The Consumer Sector .............. 86
Validation of the Benchmark Model .......... 94
IV. THE EXPERIMENTS ..................... 114
Controls on Advertising Expenditures ...... 115
Rationale ................. 115
Results--Experiment l ........... 116
Results-~Experiment 2 ........... 117

Summary .................. 118

Chagter Page

Spectrum Allocation ................. 122
Rationale ................... 122
Results--Experiment 1 ............. 127
Results--Experiment 2 ............. 128
Results--Experiment 3 ............. 133
Results--Experiment 4 ............. 136
Results--Experiment 5 ............. 146
Results--Experiment 6 ............. 147
Sumnary .................... 158

Postscript ..................... 159

V. CONCLUSIONS AND FUTURE DIRECTIONS ............ 161

Conclusions ..................... 162
MethodOlogy .................. 162
Theory ..................... 163

Future Directions .................. 168
Model Improvements ............... 169
Further Experiments .............. 171
Model Expansion ................ 173

APPENDIX A--REGRESSION EQUATIONS ............. 176
APPENDIX B--COMPUTER LISTINGS .............. 186
REFERENCES ........................ 200

vi

LIST OF TABLES

Lame Bass
1. Glossary--Newspaper Subsector .............. 47
2. Glossary--Magazine Subsector ............... 52
3. Glossary--AM Radio Subsector ............... 58
4. Glossary-~FM Radio Subsector ............... 62
5. Glossary--VHF Television Subsector ............ 66
6. Glossary--UHF Television Subsector ............ 70
7. Glossary--Advertiser Sector ............... 75
8. Parameters and Variables Used to Calculate Average

Advertiser Expenditure by Media ........... 83
9. Glossary--Consumer Sector ................ 88
10. Benchmark Validation Data--Newspaper Subsector ...... 96
11. Benchmark Validation Data--Magazine Subsector ...... 99
12. Benchmark Validation Data-~AM Radio Subsector ...... 101
13. Benchmark Validation Data--FM Radio Subsector ...... 103
14. Benchmark Validation Data--VHF Television Subsector . . . 105
15. Benchmark Validation Data-~UHF Television Subsector . . . 107
16. Benchmark Validation Data--Advertiser Sector ....... 108
17. Benchmark Validation Data-~Consumer Sector ........ 110
18. Advertising Experiment Two-—Thei1's U .......... 119
19. Frequency Allocation Experiments-~Thei1‘s U ....... 129

20. Spectrum Allocation Experiment One Output
Compared to Benchmark Model ............. 130

21. Spectrum Allocation Experiment Two Output
Compared to Benchmark Model ............. 134

vii

Table Page

22. Spectrum Allocation Experiment Three Output

Compared to Benchmark Model ............. 137
23. Spectrum Allocation Experiment Four Output

Compared to Benchmark Model ............ 141
24. Spectrum Allocation Experiment Five Output

Compared to Benchmark Model ............. 148
25. Spectrum Allocation Experiment Six Output

Compared to Benchmark Model ............. 153
A1. Regression Equations-~Newspaper Subsector ........ 176
A2. Regression Equations--Magazine Subsector ......... 177
A3. Regression Equations--AM Radio Subsector ......... 178
A4. Regression Equations--FM Radio Subsector ........ 179
A5. Regression Equations--VHF Television Subsector ...... 180
A6. Regression Equations-~UHF Television Subsector ...... 181
A7. Regression Equations-~Advertiser Sector ......... 182

A8. Regression Equations--Consumer Sector .......... 184

viii

LIST OF FIGURES

Figure Eggg_
1. Flowchart for Planning Simulation Experiments ...... 34
2. General System Block Diagram--MASSMOD .......... 44
3. Block Diagram--Newspaper Subsector ............ 46
4. Block Diagram--Magazine Subsector ............ 51
5. Block Diagram--AM Radio Subsector ............ 57
6. Block Diagram-~FM Radio Subsector ............ 61
7. Block Diagram--VHF Television Subsector ......... 65
8. Block Diagram--UHF Television Subsector ......... 69
9. Block Diagram--Advertiser Sector ............. 74

10. The Gamma Distribution .................. 80
11. Block Diagram--Consumer Sector .............. 87

ix

CHAPTER I

INTRODUCTION

Problem Statement

Mass media watchers--from amateur consumers to trained academicians--
are fluent in the "blue sky" possibilities of new communications technol-
ogies, most notably, cable television, satellites, fiber optics. But,
little has been done to actually assess, much less predict, the impact
of such technological developments on the consumer and the media industry
or to identify optimum conditions fbr the acceptance and diffusion of
that technology among the population. With no such research in exist-
ence, media policymakers, such as the Federal Communications Commission
(FCC), are forced to develop regulations which react to, rather than
anticipate, the effects of the technology and media industries must plan
somewhat blindly.

Reasons for this lack of research are twofold: the relative youth
of the mass media and the complexity of the policymaking process itself.

The mass media, as a relatively new area of academic interest, have
been studied using the techniques of many disciplines-~psychology,
sociology, economics, aesthetics, information science, marketing, manage-
ment science, economics and more. These studies, for the most part, each
have emphasized one perspective and also one medium. Thus, a fair amount
of infbrmation about the mass media and their effects does exist, such

as the consumption studies of Steiner (1963), Bower (1973), Roper (1977),
1

2
and the Newspaper Advertising Bureau (1972, 1978); the functional

approach of Bogart (1956), Mendelsohn (1964), and Wright (1975); and

the industrial economics positions of Noll, Peck and McGowan (1973),
Eoyang (1974), Owen (1975), and Stuart (1976). What does not exist is

a synthesis of this research, an integration of these disparate studies
which defines the interrelationships of mass media effects and compo-
nents, such as broadcasters, advertisers, publishers and consumers, to an
extent which makes prediction of the behavior of the mass media system
possible.

Secondly, policymaking in the public sector and in business is a
difficult process. Values and goals in a pluralistic society can be
inconsistent and conflicting, while the interrelationships of various
components of a process are defined by complex feedback loops. Ideally,
having identified various policy alternatives to a specific situation,
policy and industry decision-makers would like to test each option and
weigh possible consequences and outcomes, befbre passing a new law or
starting a new program. Such experimentation with real life, large-
scale social systems is at best unfeasible, if not impossible. Because
the real system is so complex and unavailable fer observation, it is
difficult to conduct replication studies, which maintain experimental
controls and introduce willful manipulation of variables. Once the
real system has been touched, the outcomes and consequences cannot be
reversed. Also, often when a future and hypothetical event is suggested
for introduction to the system, there is no possibility of studying it
in the recent or distant past because it is new, and has never occurred
before.

A possible solution to this two-edged problem of lack of a EEIQEL

consideration of mass media policy alternatives is the application of

3
systems science and computer simulation techniques. System simulation

provides a problem-solving methodology to deal with complex, dynamic,
interlocking multivariate relationships which describe a particular
situation or environment. Grounded in synthetic philosophy, the systems
approach "emphasizes the interrelationship among different parts of a
complex problem and the need to take a holistic look at the problems that
have been created or worsened by a piecemeal attack." (Chen, Ghausi &
Sage, 1975, p. 340) By constructing a working analogy of the structure
and interrelationships of a problem under study which replicates histor-
ical data, the systems designer can identify the complex internal inter-
actions within the system model, as well as observe the effects of
informational, organizational and environmental changes on system opera-
tion by altering the model. (Naylor, Balintfy, Burdick & Chu, 1968,

p. 8)

While systems simulation has not been applied as yet to the mass
media, it has been used successfully in analyses of industrial (Cohen,
1960; Balderston & Hoggatt, 1962) and marketing (Orcutt, Greenberger,
Korbel & Rivlin, 1961; Chorafas, 1965; Amstutz, 1967; Griggs, 1970)
behavior, economic predictions (Duesenberry, Eckstein & Fromm, 1960;
Adams & Burmeister, 1973), urban and regional studies and policymaking
(Forrester, 1969; Hamilton, Goldstone, Milliman, Pugh, Roberts & Zellner,
1969; Anundsen & Lindgren, 1972), and world fOrecasting (Forrester, 1971;
Meadows, Meadows, Randers & Behrens, 1972; Boyd, 1972; Burnett & Dionne,
1973; Behrens, 1973; Schiesser, 1976; Hobert, 1977; Zaiser & Schiesser,
1977; Naill, 1977). One of the most well-known simulations is The Club
of Rome's Limit§_39“§rgwth_model (Meadows et_al,, 1972) which was created
to identify and study the dominant elements and their interactions that

influence the long-term behavior of world systems.

4

Critics of the use of systems simulation in the social sciences are
skeptical that the complex interrelationships and behaviors of compo-
nents hihuman systems can be represented in abstract, mathematical equa-
tions and that sufficient data exist on which to base a model. Propo-
nents<rfthe methodology, such as Jay Forrester, use those very arguments
to urge the adoption of systems simulation in studying the behavior of
social systems:

Our social systems are far more complex and harder
to understand than our technological systems. Why,
then, do we not use the same approach of making
models of social systems and conducting laboratory
experiments on those models before we try new laws
and government programs in real life? The stated
answer is often that our knowledge of social systems
is insufficient fbr constructing usefu1 models.

But what justification can there be for the apparent
assumption that we do not know enough to construct
models but believe we do know enough to design new

social systems directly by passing laws and start-
ing new social programs? (Forrester. 1973. p. 6)

The Purpose fl fle My
The present research applied systems simulation methodology to

study the behavior of the mass media industry from 1945 to 1960, a peri-
od which was marked by the introduction of a new telecommunications tech-
nology--te1evision--into the existing system. The study was divided into
two parts: first, the construction and validation of the basic simula-
tion model, MASSMOD, from existing census and industry data, incorporat-
ing the economic and structural behavior of four key components--broad-
casters (AM and FM radio, VHF and UHF television), publishers (newspapers
and magazines), advertisers and consumers (detailed in Chapter 3); and
second, the experiments, which used the benchmark model as a control, to
manipulate various policy variables to observe the effects of organiza-

tional and environmental changes on the operation of the system (Chapter 4).

5
The research sought to accomplish two things:

1. By stating conscious, explicit, definite and logical relation-
ships among variables within and among the four components in the model
in mathematical form, the study hoped to contribute to the theoretical
base which describes mass media behavior and effects. The quantification
of abstract theories about media structure and economics, as well as the
integration of such theories across the boundaries of specific media,
identified integral relationships among components and variables within
the complex structure of the industry. The theoretical importance of
this work is stated best by Kuhn: "At the level of theory, a piece of
research makes a substantive contribution, as for example, by identify-
ing certain relationships among a set of variables." (paraphrased in
Krulee, 1972, p. 53)

2. By manipulating various policy variables incorporated into the
model, and comparing alternative results against the benchmark model,
the study sought to demonstrate the feasibility and usefblness of the
application of systems simulation to the study of mass media behavior,
effects, planning, and policymaking. Specifically, the policy variables
which were manipulated were those which were of interest not only during
the period under study, but are also being discussed today: regulation
of advertising expenditures and the number and type of spectrum frequency
allocations for each broadcast medium. Such experimentation using a
simulation model provided answers to many "what if" questions, without
altering the real-life system.

The creation of an historical model of the mass media industry and
its validation against existing time-series data, provided a framework
in which to identify key theoretical relationships among variables which

describe the dynamics of the system and to generate and test hypotheses

6
about the effects of certain policies and decisions. The building of a

total industry system model, linking broadcaster, publisher, advertiser
and consumer components, allowed for the synthesis and integration of
many single medium and single effect studies which have been conducted
in the field.

Television sparked a major revolution in mass communications. Its
growth and development was managed and guided by businessmen who had
little notion of its far-reaching effects, save as an economic gold mine.
It was heralded by consumers who lived in a postwar era of relative
affluence and enjoyment, and regulated by government policymakers on an
§g_hgg basis, in the best interest of the status gug. At no time did
long-range planning, or a consideration of alternative positions, seem
necessary or possible to these people.

Now in 1980, as the United States moves from an industrial to an
infbrmation economy (Robinson, 1978, p. 55), revolutionary changes in
infbrmation technology, involving computers and telecommunications, stand
ready to cause upheavals in markets, institutions, law and politics. In
order to plan and legislate effectively, decision-makers must have a
method at their disposal to explore the implications of these new tech-
nologies, and to assess the merits and drawbacks of various marketing
schemes, technical requirements, and industry structures. It is to this
end that the present research was directed--to encourage the development
of models of the mass media system, now and in the future, which are able
to predict alternative outcomes and consequences of various policy and

planning decisions.

7
Assumptions and Research Questions

Assumptions

MASSMOD, the mass media industry model developed in this study,
dealt with all model components on a highly aggregate level, as AM
broadcasters, FM broadcasters, VHF television broadcasters, magazine
publishers, advertisers, consumers, without regard to markets, type of
programming or content, demographic variables, type of product adver-
tised, or ownership or group affiliation. The choice of aggregation
level was made fer two reasons: first, data was more readily available
in industry-wide terms; and second, it was felt that the general struc-
tural interrelationships in the entire system could be more readily
identified without the additional complexity of a more detailed, more
microscopic approach.

The model assumed that broadcasters and publishers operate in such
a way as to maximize profits, which is to say to maximize the difference
between revenues and cost. Both broadcasters and publishers seek to
establish as large an audience as possible to maximize their advertising
revenue, since advertising demand is a function of audience size. Pub-
1ishers can rely on an additional source of revenue from subscriptions
to their periodicals, aware that the cost of the subscription can affect
their circulation, while the major source of revenue for the broadcaster
is advertising. Both print and broadcast managers recognize that the
amount of money invested in their publishing costs (fer talent, programs,
stories and quality of production) can also influence the size of their
audiences.

The model also made the assumption that all media in the model are
substitutable in terms of advertising placement and therefbre compete for

the same advertising dollar. Ignored were attempts to reach special

8
audiences with one specific medium, or to make use of a medium's unique

ability to convey a message, such as newspaper or magazine coupons.

Growth within a medium was assumed to depend in part on the profita-
bility of the medium in the previous year. High profits motivate more
investors to develop stations and publications. Since profits depend on
audience size, consumer use is also a factor in the growth process. The
model also assumed that cross-media ownership (ownership of more than one
medium by one party) was responsible fer the rapid growth of television
in the time period under study. Broadcast station growth was assumed to
depend on the number of channels available fer licensing, with growth
occurring until the spectrum allocations are completely used.

Gross National Product and disposable income per capita were assumed
to be major economic factors included in the model as exogenous variables.
It was also assumed that the model would not always identify and include
variables which were significant predictors of various dependent variables
and that some variables included in the model might not always be signif-
icant predictors; simple growth functions based on a yearly index were
used at these times.

Other specific assumptions about model components are detailed in

Chapter 3, the description of the model.

Research Questions

The first part of this study dealt with the possibility of construct-
ing a mathematical model of the mass media industry, from which a com-
puter simulation could be developed. Goodness of fit tests were used to
assess the validity of such a simulation against real world data.

The experimental segment of the study was concerned with the manip-

ulation of various policy-oriented variables and a comparison of results

9
against the benchmark model. These manipulations revolved around two

basic issues:

(1) Control on Advertising,Expenditures. What would be the effect
of externally imposed controls of advertising expenditures on media
growth? Since media depend at least in part on advertising revenues for
their income, and profitability affects future growth, would limiting
advertising spending ultimately curtail media growth? If such controls
were deemed necessary and/or desirable, what type of control would be
best--a tax on expenditures, a limit based on percentage of sales, or
some form of depreciation write-off (Bloom, 1976)? TWo manipulations,
one involving a tax and the other a percentage of spending limit were
perfbrmed to assess the effect of such controls on the system.

(2) Spectrum Allocation. What role did the FCC determination of
the use of segments of the radio spectrum play in the growth of broadcast
stations? Did more available channels fer a medium necessarily mean
more stations on the air? What effect did the assignment of two or more
different band segments to one type of telecommunications (radio.
television) have on revenues and growth? If television were all UHF,
or radio all FM, would the media industry be different today? The
benchmark model assumed that an unlimited number of channels were avail-
able for broadcast stations. This was the case in 1945-1960; the AM
radio band did not become saturated until the 1970's, and VHF television
continued to grow until the mid 1970's. Six manipulations were perfbrmed
to address these questions: a no-freeze situation, allowing UHF tele-
vision to begin development in 1945; a limit imposed on the number of
channels available for the four broadcast media; a VHF television, AM
radio only situation; a VHF television, FM radio only case; a UHF tele-

vision, AM radio run; and a UHF television, FM radio run.

10
Summary

This study applied the use of systems simulation to explore the
interrelationships among variables in the complex mass media industry
system from 1945 to 1960. Borrowing the technique from industrial ana-
lysts, policymakers and fbrecasters, the research hoped to gain valuable
insights into dependencies and feedback loops within the total media
complex, and, by manipulating policy variables included in the model, to
suggest the use of simulation in media policymaking and planning. The
development of the benchmark model within an historical time period was
seen as a necessary step in the eventual development of a later genera-
tion model to be used in the forecasting of the effects of new telecommu-

nications technologies on the media industry and its consumers.

CHAPTER 2

LITERATURE REVIEW

A review of the literature relating to the mass media reveals that
systems simulation methodology has not yet been applied to media problems
involving policymaking or industry structure (or at least such work has
not been released for publication) and that the use of simulation in
general has been sparse. However, simulation studies designed to aid in
policy analysis have been employed in other social science disciplines.
suggesting that the technique might be appropriate to the mass media as
well.

This chapter first presents in some detail three media-related sim-
ulations which are directly applicable to this study--Gensch's (1973)
advertising media allocation model, called AD-ME-SIM, Block's (1975)
simulation of consumer time allocation and Bloom's (1976) study of adver-
tising policy and competition. A discussion of the use of simulation
experiments in policy analysis and planning, including examples from
other social science disciplines, and in the analysis of industry struc-
tures fellows. The role of simulation in theory development is considered
briefly. The chapter concludes with a brief description of other simu-
lations which have dealt with the mass media, but which have little, if

any, relationship to the use of systems simulation.

Media Simulations Relevant to This Study

Three media-related simulations exist which have a direct bearing
11

12
on the model developed in this study. An advertising sector in the

model allocated an annual advertising budget across various media which
exist in the model. The technique used by Gensch (1969, 1973) in
AD-ME-SIM to predict a hypothetical population's reading and viewing
preferences, discussed below, was adapted to predict a hypothetical popu-
lation of advertiser's buying preferences. The prediction of these adver-
tising expenditures is a probabilistic process using appropriate proba-
bility distributions. The techniques used in Block's (1975) time alloca-
tion model, described below, were adapted fer this purpose. In addition,
Block's model offers an approach to the allocation of media time by
consumers, which, although it was not applied to the model in this study,
could be included in further, more sophisticated revisions of the model.
The present model also considered the effects of policies regulating the
amount of advertising in the broadcast and print media on the total
system. Bloom's (1976) simulation of advertising policies and competi-
tion, also discussed below, suggested possible advertising restraints

which were used in this study.

A Media Allocation Model

Gensch (1969, 1973) created a decision-making model of advertising
media selection and scheduling which defines a theory of how to send a
message that will reach a defined target population, use the most cost-
efficient advertising forms, and reach the target population the desired
number of times within a given period. The three-stage model generates
output which includes the weekly and cumulative numbers and percentages
of people in the target population reachedturthe proposed media schedule,
as well as the cost, frequency, and number of exposures, adjusted

accordingly to the subjective evaluation of the media and the value of

13
exposures to different members of the target audience. (Gensch, 1969,
pp. 209-210)

The first stage of the AD-ME-SIM model, the data generation stage,
is of particular interest in establishing the probability of behavior of
a hypothetical population. Gensch used two months of actual data from
Brand Rating Research Incorporated (BRI) to establish basic weekly prob-
abilities of viewing and reading fer a year, adjusting for trend and
seasonal influences by using a Monte Carlo system (1973, p. 117) A
similar technique. using data for the top 100 advertisers, was used in
this study to predict the media selection behavior of the advertisers.
Because Gensch deals with individual, rather than aggregate data, which
allows comparison of specific media schedules, his entire selection

process was not applicable to this study.

A Time Allocation Model

Block (1975) developed a model called TIMMOD, which simulated human
time allocations in order to assess the impact of broadband communication
network technology on consumer marketing communications. A stochastic
model programmed in FORTRAN, TIMMOD develops a series of probability
distributions which describe a single individual's decision-making
regarding the allocation of time from real-world decision processes,
which in this case are derived from the 1965 U.S. Time Use Survey (1966).
The model is limited by design to be a forecasting tool, not a general
model fer the human allocation of time.

The simulation generates the time allocations for an entire day
for a single individual and then cumulates over many individuals to
create a population. The time allocations to various daily activity

categories are based on two probability distributions: the first

14
describes the activity selection behavior of the hypothetical population,

the second describes the duration of the selected activity. The specific
probability distribution is determined from the Erlang family, with
different parameters to describe different shaped distributions for dif-
ferent activity categories. Experiments were conducted using the model
by manipulating the specification of these probability distributions,

fbr both activity categories and duration. A modified version of Block's
technique, using probability distributions, was used to create adver-
tiser buying habits in the advertiser sector of the present model.
Gensch's technique to create a hypothetical population's habits was

modified to incorporate the probability distribution approach.

An Advertising Policy Model

Bloom (1976) developed a computer simulation model of an hypo-
thetical grocery manufacturing industry to explore the competitive
effects of seven proposed controlstwiadvertising expenditures--two
advertising taxes, three limits on advertising outlays, and two depre—
ciation requirements on advertising. The model was patterned after the
nutritional submarket of the cereal industry. using data available from
the Harvard Business School “Life Cereal Case." The deterministic model,
written in FORTRAN, contains four sub-models that interact in a yearly
cycle: (1) a manufacturer's submodel, (2) a potential rival manufac-
turer's submodel, (3) a retailer's submodel, and (4) a consumer's sub-
model. The industry is assumed to have three brands belonging to three
firms at the start of each run of the model which simulates eight years
of the industry's behavior.

A group of structure and performance variables--number of firms and

brands. market concentration, market advertising expenditures, profits,

15
selling costs, prices and brand quality--were monitored in each computer

run to give an indication of how competition in the simulated industry
was affected by a control. Twentybone computer runs representing the
partial completion of an 8x2x2 factorial design were executed. The
three factors controlled were (1) type of control on advertising expend-
itures (eight types); (2) degree of sensitivity of the industry sales
response function to changes in total advertising goodwill in the indus-
try (high or moderate); and (3) types of objectives managers would be
satisfied to reach (profit streams or cash flow stream). The results
revealed that the seven controls on advertising expenditures were unable
to stimulate vigorous competition in the simulated industry, and sug-
gested the possibility that controls would tend to lessen competition in
some industries.

Bloom's model was able to investigate the effects of controls on
advertising expenditures advocated by various economists, consumer advo-
cates and policymakers. While he does not claim universal generalization
of the model, Bloom has provided necessary data which can be considered
by policymakers, such as the Federal Trade Commission, before any laws

or regulations regarding advertising expenditures are actually passed.

Summary

The Gensch, Block and Bloom models hada direct impact on the struc-
ture of the model built in this study. Gensch's approach to the crea-
tion of the habits of a hypothetical population, based on limited data,
combined with Block's use of theoretical probability distributions, was
used to predict the media buying behavior of the advertisers in the model.

Bloom's suggestions for advertising restraints were introduced into

the model during the experimental runs. The general organization of his

16
model, with interlocking components, or building blocks, was also fol-

lowed in the construction of the model.

The Use of Simulation in Policymaking

The use of systems simulation as a tool in policymaking and analy-
sis has been debated in the literature. Proponents cite the integrative
capacity of systems theory, which synthesizes bits and pieces of infor-
mation from various perspectives and sources into a comprehensive whole
(Pugh, 1977, p. vii), and its ability to discern the basic structure of
complex and dynamic societal systems (Fitz, 1979, p. 4). Critics point
to the difficulty in validating the model and the gap in philosophies
between model builders who maximize the theoretical content of models,
and policymakers, who pragmatically want to maximize the accuracy of
information estimates relevant to policy (Pugh, 1977, p. 5). Fromm,
Hamilton and Hamilton (1974), in a survey of 250 simulation models
revealed that "at least one-third and perhaps as many as two-thirds of
the models failed to achieve their avowed purposes in the form of direct
application to policy problems." (p. 3)

Nonetheless, a major impetus behind the use of computer simulation
by decision-makers and policymakers "is the possibility of testing and
evaluationg alternative decision rules, strategies and policies before
they are put into effect on actual business and economic systems."
(Naylor, Wertz & Wonnacutt, 1967, p. 703) Because of their ability to
deal with complexity, simulations can assist in policy analysis in
areas of social and administrative sciences by developing a general
understanding of the real world situation, by fbrecasting the future of
a real-world system, and by simulating the impacts of alternative

policies on the real-world system. (Pugh, 1977, p. 1) Sage (1979)

17
points out that systems engineering is an appropriate combination of the

mathematical theory of systems and behavioral theory in a useful setting
appropriate for the resolution of real-world problems. He concludes
that systems science is most effective in policy analysis when the dis-
parate perspectives of the sponsor-~who seeks information for decision-
making, the c1ient--who has real-world problems to attempt to solve, and
the researcher-~who seeks to build a workable model are brought into a
symbiotic heterogeneous complimentary relationship one with another
(10. 6).
The potential value of systems simulation as a tool in policymaking
can be seen when the elements of policymaking are examined. Dror (1968)
finds six interconnected phases in the pure rationality model of decision-
making, often presented as the universal ideal:
1. Establishing a complete set of operational
goals, with relative weights allocated to
the different degrees to which each may be
achieved.
2. Establishing a complete inventory of other
values and of resources, with relative

weights.

3. Preparing a complete set of alternative
policies open to the policymaker.

4. Preparing a complete set of valid predictions
of the costs and benefits of each alterna-
tive, including the extent to which each
will achieve the various operational goals,
consumer resource, and realize or impair
other values.

5. Calculating the net expectation for each
alternative by multiplying the probability
of each benefit and cost for each alterna-
tive by the utility of each and calculating
the net benefit (or cost) in utility units.

6. Comparing the net expectations and identi-
fying the alternative (or alternatives, if
two or more are equally good) with the

18

highest next expectation. (Dror, 1967,
p. 132)

Lasswell (1971) summarizes the policymaking process in terms of five
tasks: (1) goal clarification, (2) trend description, (3) analysis of
conditions, (4) projection of developments, and (5) invention, evalua-
tion and selection of alternatives. (Lasswell, 1971, p. 72) He suggests
that simulation models are most helpful in inventorying and evaluation

of alternative policies.

Policy-Related Simulations

Despite criticism that simulations are too theoretical or mathe-
matical, computer models have been developed which aid policymakers in
their decision process in such disciplines as urban, regional and world
analysis and forecasting; industrial dynamics; and economics. Systems
theory as a problem-solving methodology has proved particularly helpfbl
in analyzing conditions which describe a system, projecting developments
resulting from a policy option, and inventing and evaluating possible
policy alternatives. Simulation models function as a substitute for
real-world experimentation when such testing is either impossible, or
too costly, or both.

Forrester (1965) pioneered the use of systems theory in the analy-
sis of complex organization, specifically industrial management. His
initial concept of "industrial dynamics" has been applied to many other
fields which exhibit positive and negative feedback processes in the
course of their growth and regulatory action. His models are classic
examples of systems simulation.

One such application is the life cycle of an urban area described
in Urban Dynamics, which uses a stochastic, non-linear, macro-model,

written in DYNAMO, a computer simulation language. Forrester defines

19
the boundary of the city in terms of its interacting components of level

and rate variables, which are grouped into three subsystems: business,
housing and population. The model provides a means of assessing the
effects of various policies regarding housing (e.g., amount of low income
housing; ratio of industrial land to residential land), job training
(e.g., programs to train skilled workers) and business growth (e.g..
encouragement of new business) on population, unemployment and taxes in
major urban areas. Housing densities, family sizes. personnel needed in
business units, taxes levied and taxes needed are major parameters defined
at the start of the simulation. The model traces the urban life cycle
within a period of 250 years starting with empty land, growing to full
land occupancy, maturing through a realignment of internal urban balance,
and emerging into an equilibrium characterized by stagnation with its
unemployment, faltering industry, and increased taxes. It then alters
various policies governing variables to find out and predict alternate
outcomes.

Forrester (1971) applied his industrial dynamics techniques to a
model of world interactions, WORLD 2, a preliminary effort in the dis-
cussions of The Club of Rome, which later generated the Limits to Growth
model, WORLD 3. The world system is defined as man, his social systems,
his technology, and the natural environment; these interact to produce
growth, change and stress. The main objective of the Limits to Growth
model was to identify and study the dominant elements and their inter-
actions that influence the long-term behavior of world systems, in order
to suggest survival policies in areas such as population, energy, food
supply, pollution, natural resources, and capital investment. The model
is built at a high level of aggregation, such that distinctions between

developed and underdeveloped countries are not made. In determining the

20
level of aggregation the team felt that "questions of detail, of indi-

vidual nations, and of short-term pressures can be asked much more sensi-
bly when the overall limits and behavior modes are understood."
(Meadows gt_al., 1972, p. 96)

Several extensions of the Limits to Growth model have been developed,
which add sectors, disaggregate the principal components of the model,
or change model relationships based on empirical data. Boyd (1972)
included a technology sector, with the assumption that increasing tech-
nology will increase available food and reduce pollution and the deple-
tion of natural resources. The model reverses the basically pessimistic
projections of the original WORLD 2 model. Burnett and Dionne's (1973)
GLOBE 6 model breaks the principal components into developed and develop-
ing nations, elements of world trade, a population sector with five dif4
ferent age groups and greater structural detail.

Hamilton EELél: (1969) applied systems simulation to regional analy-
sis, specifically dealing with the economic growth in the Susquehanna
River Basin. The model, written in DYNAMO, interrelates and projects
demographic and employment variables simultaneously, and is designed to
render inexpensive experiments on the effects of change in parameter
values, such as increased employment opportunities, a relaxed housing
market, lay-offs of workers. to determine policies which optimize the
economic growth of the Basin. The model is data based, and therefbre
fairly disaggregated to allow for the analysis of subregions within the
Basin. The use of systems analysis and computer simulation allows the
model to be dynamic, incorporating explicit feedbacks and lagged rela-
tionships among sectors and variables.

The Center fer Environmental Study modelled a simulation of the

Grand River Basin environment in Michigan after Forrester and the

21
Susquehanna River Basin Project. (Anundsen & Lingren, 1972) The

simulation was created to assist policymakers and planners in the Grand
Rapids area to assess the effects of potential policy changes and growth
in five major areas (pollution, government, health, economics, and social
factors and technology) on the existing environment. Access to the simu-
lation is provided through various terminals located throughout the
governmental community, so that policy alternatives and ideas may be
easily explored and assessed.

In economics, simulations have been used to predict outcomes based
on various fiscal policies. Duesenberry, Eckstein and Fromm (1960)
developed an econometric model of quarterly movements of the gross
national product (GNP) to test fiscal policy measures. Their aggregation
fbllowed the national income accounts of the Department of Commerce, with
a time interval of a quarter of a year, variables adjusted fer seasonal
fluctuation, and all relationships established from time series data.

The recursive model is divided into two parts: first, the total GNP is
built up; then, the disposable income which would be generated is com-
puted. Several lags exist in the model: consumption depends on the
preceding quarter's disposable income; inventory investment depends on
past values; dividends depend on levels in the preceding quarter. The
system is actually summarized by feur equations: an inventory function,
a consumption function, a function relating personal income to GNP, and
a function tying personal income to disposable income.

Several large scale macro-econometric models have been developed for
forecasting and policy analysis such as the Wharton model, the Fed-M.I.T.-
Penn model, the Brookings model, and the Michigan model. As a class,
these models are moderately large systems of simultaneous equations

(from fifty equations to as many as several hundred), which are dynamic

22
and non-linear. (Adams & Burmeister, 1973, p. 126) The best known of

the aggregate models is the Wharton quarterly econometric model, a
living model which is modified and re-estimated when data, institutional

change'and new economic theory suggest it. (Adams & Burmeister, 1973)

Models of Industries

Since the present study attempts to investigate the interrelation-
ships present in the structure of the mass media as an industry, several
classic computer models from the economics literature are presented here
to illustrate the use of simulation in analyzing industry structure.

A classic macro-simulation of industry is Cohen's (1960) computer
model of the shoe, leather and hide sequence, which describes the aggre-
gate behavior of shoe retailers, shoe manufacturers and cattlehide
tanners between 1930 and 1940. Chief exogenous variables in the model
are the Bureau of Labor Statistics consumer's price index, disposable
income, and the stocks of hides held by hide dealers. The major endog-
enous variables are divided into four components of the industry:
retailers, manufacturers, tanners, and hide dealers.

The mathematical form of the model is a complex, non-linear system
of lagged simultaneous difference equations with one month as the basic
time period. Cohen actually built two models: one--a "one-period-change
model," resembles the classic econometric model, and is intended to ex-
plain the values of the endogenous variables far only one time period
ahead into the future. It assumed that the lagged endogenous variables
refer to their actually observed values and that the lagged endogenous
variable at one time period becomes an exogenous variable in the next.
The other model, a "process model," was designed to explain the determi-

nation of the endogenous variables for an arbitrarily large number of

23
future time periods. In this instance, the equations of the model and

the observed time paths of the exogenous variables are treated as a
closed dynamic system; the values of the predetermined endogenous vari-
ables are the values generated by the model, not the actually observed
values.

Cohen's results show a very close correspondence between the time
paths generated by the process model and the actual values on an annual
comparison basis. Although the models do not present a completely
acceptable monthly aggregate description of the industry (particularly on
three major variables: retailers' receipts of shoes, manufacturers'
production of shoes, and manufacturers' receipts of leather), Cohen
concludes (1960, p. 65) that the evidence provided by the annual compari-
sons is consistent with the hypothesis that the models incorporate some
of the mechanisms which determine the behavior of individual firms in
these industries. Improvements in perfbrmances of the process model over
the one-change model is ascribed to the selfeequilibrating, self-correct-
ing features incorporated into the behavioral mechanisms of the process
model which insure that the generated time paths never depart too far
from the observed paths in the estimation of parameters. Such features
were absent from one-period-change model.

Balderston and Hoggatt's (1962) simulation study of the lumber
industry was designed to examine the intimate dynamics of a market-~how
market infbrmation and decentralization of market decisions and institu-
tional alignments affect and are affected by economic fbrces--viewed as
a complex system of behavior in which information is limited and costly.
The authors viewed their study not as a business analysis, but as an
experiment in constructing a general system model.

Amstutz's (1967) computer simulation of competitive market response

24
moves from a macromodel to a micromodel. He believes that "while an

aggregate model may generate correct answers at a point in time, it
provides little or no insight into the reasons far these answers. The
micro-analytic simulation has the potential to provide the right answers
fer the right reasons." (Amstutz, 1967, p. 113) After construction of
a macromodel of the marketing environment, he formulates detailed behav-
ioral micromodels of the five major elements: the manufacturer, the
consumer, the retailer, the distributor, and the salesperson. Amstutz
validates three functional forms in the model: the effect of attitude

on purchase, the effect of attitude toward appeals on noting a promotion,
and the effect of display size on point-on-sale display placement. The
model allows for simulating individual behavior in the marketing decision

with various initial parameters.

Summary
This survey of the literature suggests that while the problem-solving

methodology of systems simulation has not been applied to the analysis of
policy alternatives or industry structure within the mass media field, it
has been used successfully in other social science disciplines such as
urban, regional and world planning; world and national economics; and
industrial dynamics. The mass media, when viewed as an industry system,
present a series of complex, dynamic interrelationships and feedback
loops which are not easily assessed by a straightforward linear analysis.
Hence, the use of simulation experiments should provide two major outputs:
a better theoretical understanding of the structure of the mass media
industry, including key variables and relationships; and second, a frame-
work for decision-making, which supplies the potential effects of various

policy alternatives on the existing industry structure.

25
The data bases used to build the industry models cited in this

chapter suggest that enough historical time series data are available
about the mass media industry to build a working model. The experience
of other model builders further suggests that the initial mass media
industry model should be macroscopic in nature, dealing with components
at their most highly aggregated level. Further research can disaggregate

model units and add greater structural detail.

The Use of Simulation in Theory Development

The use of simulations in the development of theory has also been
debated in the literature, most forcefully around the validity of a
simulated system as a representation of a real system. As an operating
model of a real system, a simulation can help theory development by
"raising questions, demonstrating gaps, helping discriminate between the
important and the unimportant, generating testable hypotheses and serv-
ing as vehicles for the communication or comparison of theories."
(Schultz & Sullivan, 1972, p. 6) Since so many real systems are complex
and unavailable for observation, studies dealing with factors and rela-
tionships crucial to the system are much more easily conducted in simu-
lation than in the real system.

Raser, Campbell and Chadwick (1970) suggest that the strategy of
science should be to test a theory in all of its settings of application,
and to keep it as general as possible for as long as possible (p. 185).
They cite the need fer inter-nation simulation (INS) in the area of
international relations "because of the very great difficulty we have in
observing even recent international relations as a test of theory and
because of the comparatively very great power of experimental manipula-

tions in probing the implications of theory."

26
For example, Brody's (1963) study of the impact of changing power

relationship on communication and alliance patterns in an inter-nation
simulation gradually introduced nuclear power to all of seven nations in
a simulated world. He was then able to analyze the effect of this devel-
opment on the international system, an impossibility in the recent or
distant past, since it was a future and hypothetical event.

The problem caused by the confounding of cause and effect in the
probing of theories is also one that can be overcome by simulation.
Raser and Crow (1968) investigated the effects of the development of an
invulnerable nuclear retaliatory force by one nation in an international
system. In the real world at the time of the study, the United States
was developing such a force, and major world changes were also occurring,
but the determination of which changes in the international system were
caused by the developing invulnerability and which were simply co-occurring
was impossible. Raser and Crow created a five nation world with alliance
ties, general strength levels and political characteristics designed to
simulate the Cold War system of 1962. Using repeated runs with experi-
menter-controlled research and development of an invulnerable nuclear
weapons system, measures on a cluster of independent variables fbllowing
the interventions were taken while other variables were fbrced to fluctu-
ate. This replication and experimental intervention allowed the research-
ers to rule out more alternative explanations in their attempt to estab-
lish causality than if they had simply observed the singly occurring and
uncontrolled real world.

Probing theory based on observations alone raises another problem--
that of degrees of freedom. Based on a statistical concept, the point
is that a plausible explanation (or line or correlation), which accounts

perfectly for the behavior of a single case, is possible only when the

27
number of observations is equal to the number of explanatory concepts.

If there are no observations left over, there are no degrees of freedom
to use to test the goodness of fit of the theory. Simulation, according
to Raser £3 31. "answers the degrees-of—freedom problem...through its
open-ended possibility of generating new instances, new re-runs, new
observation points greater in number than its explanatory or descriptive
contents." (1970, p. 187)

Raser gt El- (1970) also suggest that the process of building simu-
lations can aid powerfully in theory generation for two reasons: first,
as a framework or construct to guide the integration of part theories and
isolated data bits, and second, as a stimulant and goad to scholars which
results in the generation of theory. "We may find that the act of trying
to build may be the best strategy for learning more about that which we

are trying to simulate." (p. 188)

Summar

Thus, using the field of international relations as one example, the
application of simulation to probe theory, by enabling an examination of
causality and providing degrees of freedom in observation, and to create
theory, by forcing scientists to turn vague generalizations and verbal
theories into conscious, explicit, logical and definite relationships,
becomes more apparent. It is hoped that the attempt to state explicit
relationships among elements of the mass media and to probe those theo-
retical statements by experimental manipulation will yield richer theo-
retical insights into the mass media industry.

Simulation at least creates a certain amount of intellectual soul-
searching as theories are brought into the open. As Kaplan says (1964,

pp. 268-269):

28

Models have this merit, that they do not allow us

to comfbrt ourselves with the notion that we are
following up an "idea" when we are only moving from
one observation to the next in the hope that some-
thing will turn up. Too often the hypotheses with
which we work are at home only in the twilight
regions of the mind, where their wavering outlines
blend into a shadowy background. There they are safe
from sudden exposure, and are free to swoop down for
sustenance on whatever datum comes their way. Models
are at any rate conscious, explicit, and definite;
there is nothing ghostly in their appearance or
manner; they look healthy even up to the very moment
of their death.... The model saves us from a certain
self-deception. Forced into the open. our ideas

may flutter hopelessly; but at least we can see what
bloodless creatures they are. As inquiry proceeds,
theories must be brought out into the open sooner or
later; the model simply makes it sooner.

Mass Media Simulations

While there is a lack of computer simulations in the mass media
literature which employ systems science to construct dynamic, stochastic
models, several descriptive, deterministic simulations have been devel-
oped, particularly in the area of media allocation. Since this research
deals with mass media simulations, a brief description of media-related
simulations is presented below, fer the sake of completeness.

The first computer simulation used in the area of mass media was the
Simulmatics Media Selection Model, designed in 1962 to test advertising
media allocation plans. The simulation determines probabilistically an
estimate of the media audience, described by various socio-economic fea-
tures, and the reach and frequency of a proposed media schedule. (Kotler
& Schultz, 1972, pp. 517-518) Gensch's AD-ME-SIM model, described earlier
in this chapter, is an extension and improvement of this model.

The COMCOM (Communist Communication) model developed by Popkin (1965)
is an application of the Simulmatics simulation to the mass media system

in the Soviet Union. The first part of the simulation generated the

29
hypothetical descriptions of individuals in the population, based on

aggregate data tables and some basic theoretical assumptions; the second
run, the actual simulation, measured exposure to a series of media
messages, using Monte Carlo randomization methods to introduce new
variables. The output of the simulation was the exposure probabilities
of an audience fer a particular communist country over a given time
period.

Kessler and Pool (1965) developed Crisiscom, a computer simulation
of human infbrmation processing during a crisis, which represents a set
of human decision-makers who process infbrmation and the exchanges of
infermation among them. Infbrmation is generated from imbalances among
cognitive elements and relations which comprise the network, while crises
an: characterized by high rates of generation and exchange of infbrmation.
The simulation measures salience, affect and credibility of the communi-
cation processes.

Kramer (1969) simulated the United Nations public infbrmation cam-
paign conducted in Cincinnati, Ohio, in 1947 and 1948. The simulation
was designed to predict reach and frequency of message flow from known
information about the population, media habits and the placement of
messages in the mass media. The first stage of the simulation generates
the model population from basic age, sex, geographic region and literacy
data and determines probability exposure to various media. The second
stage, like the COMCOM model, calculates the actual message exposure
probabilities for the audience.

Little and Lodish (1969) developed a simulation of advertising media
allocation called MEDIAC. an on-line computer system which selects and
schedules media to maximize total market response. The user supplies

various media options, a budget, and objective and subjective data about

30
media options and the desired audience. MEDIAC selects options and

schedules over time, based on a mathematical optimization routine which
uses total market response as the optimization function.

Hanneman (1969) and Carroll developed a stochastic simulation model
of the diffbsion of innovation to cliques in any small, relatively closed
system. The simulation, called SINDI I (Simulation of Innovation
Diffusion) deals with opinion leadership, clique structure, channel
structure and the amount of knowledge available to clique members as a

new idea spreads over time through a social system.

Summary

This overview of computer simulations related to the mass media
shows that although systems methodology has not been used in conjunction
with the actual simulation, computer models have been built which deal
with populations of people, media habits and infbrmation structures.
This suggests that enough data exist to support model-building hypoth-
eses and relationships. The overview also highlights the use of simula-
tion as a data-generating tool for decision-making, which provides
output from the model under a variety of conditions and policies, such

as media schedules.

Conclusion
This review of the literature highlights the role of computer simu-
lation and systems science in the analysis of the behavior of complex
systems. While such techniques have not been applied to the mass media
industry as a system, the studies discussed here support the feasability
of such an application, both as an aid in policymaking and in theory-
building. The complexity of the Limits £9 Gggyth models and the aggre-

gate approach which the model builders used indicate that even very large

31
systems described by minimal data can be modelled and analyzed. The

structural-economic models of industry suggest a precise definition of
system boundaries, either by historical time period or identified system
components, in which to build such a model.

Thus, supported by the efforts of authors included in this chapter,
Chapter 3 describes MASSMOD, a simulation model of the mass media indus-
try in the historical period 1945-1960. The works of Gensch, Block and
Bloom were incorporated into the model in the advertising component. The
construction and analysis of the model sought to identify underlying
structures within the industry. The experiments described in Chapter 4
represented an attempt to use the model in a policymaking mode by simu-
lating the impacts of alternative policies on the real-world system

represented by MASSMOD.

CHAPTER 3

THE MODEL

This study sought to explore the effects of the introduction of
television on the AM radio, fledgling FM radio, daily newspaper, maga-
zine and advertising sectors of the mass media industry. Systems simu-
lation at the macro level was selected as the main methodological tool
to build, experiment with and attempt to validate a computer model of
the industry from 1945 to 1960 in order to make some tentative state-
ments about the interrelationships of variables among the sectors and
the effects of manipulating policy variables within the system. The
study was, therefore, a piece of exploratory research, designed to
generate some hypotheses on the effects of various policy variables on
industry sectors and consumers and to fbrmulate a richer theory about
the behavior of various system components.

Indirectly, the research hoped to show the feasibility of using
systems simulation to investigate the effects of the introduction of
any new sommunication technology on the existing mass media industry
structure under various policy alternatives. To demonstrate the appli-
cability of systems analysis to this type of problem, the decision was
made to fellow the example of Cohen (1966) and Balderston and Boggatt
(1962), who chose to model specific industries in specific historical
time periods for which historical time-series data were readily avail-

able. The starting date of the period under study here, 1945, marks the
32

33
commercial beginning of television; the closing date, 1960, the year

when color and cable television both began to have an impact on the
system. (The year 1960 was chosen as a cut-off date to avoid the possi—
ble confounding influence of color and cable television. In 1960, color
television owners and cable subscribers were less than one percent of

the population. Both began to grow significantly after 1960.)

Methodology and Data Collection
Systems simulation was judged to be an appropriate methodology
because, as Forrester states, it can be particularly useful for dis-
covering the counterintuitive nature of systems. (Forrester, 1961) It
was hoped that some relationship among system variables would be indenti-
fied that had not appeared elsewhere.
Naylor g__gl, (1968, p. 23) suggest that planning simulation ex-

periments involves a procedure consisting of the following nine elements:

1. Formulation of the problem.
2. Collection and processing of real world data.
3. Formulation of mathematical model.

4. Estimation of parameters of operating
characteristics from real world data.

5. Evaluation of the model and parameter
estimates.

6. Formulation of a computer program.
7. Validation.

8. Design of simulation experiments.
9. Analysis of simulation data.

A flowchart of this process is given in Figure l.

 

I m

FORMULATION
OE PROBLEM

 

 

 

I (2)

COLLECTION AID
PRINISSING OF DATA

 

 

 

 

I (3)

FORMULATKNI N
MATHEMATICAL moon

I (u

ESTIMATION N
PARAMETERS

 

 

 

 

 

 

 

 

 

cvuumou ‘5’ REJECT
ormmu *fifih

accm noon :5,

FORMULATION OF
COMPUTER PROGRAM

I m

VALIDATION

J (8)

DESIGN OF
EXPERIMENTS

I (9)

ANALYSIS OF
SIMULATION DATA

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Naylor Eliélu 1968, p. 24)

FIGURE 1
FLONCHART FOR PLANNING SIMULATION EXPERIMENTS

34

35
Problem Formulation

Essentially, this study had as its objective the estimation of the
effects of certain changes in policy variables, such as the number of
available AM, FM and television frequencies, cross-media ownership, and
advertising expenditures, on the endogenous variables in the system.

It sought to simulate the effects of various alternative policy decisions
on the media-consumer system in order to provide a more complete data

base for decision-making.

Collection of Real World Data

The benchmark model was constructed using real world data about
advertisers, broadcasters, newspaper and magazine publishers and con-
sumers from 1945-1960. Much of the infbrmation regarding audience
sizes; numbers of stations, magazines and newspapers; expenditures and
profits; gross national product and disposable income were contained in
Sterling and Haight's (1978) §g1g3_tg Communication Industry Trends.
Population data, such as birth and death rates and household size, were
obtained from U. 5: Census data presented in Historical Statistics and
Statistical Abstracts. Additional information about publishing costs
was found in the ngggsfl f Manufactures, while advertising expenditures
fbr the top 100 advertisers were found in annual listings in Advertising
Agg,

Most of these data were presented at the macro, or aggregate, level;
that is, data existed which described the behavior of all AM radio sta-
tions as a group, all advertisers as a group, all magazines as a group.
At times the groupings were not consistent for all variables within a
sector. For example, numbers of magazines and their circulations were

broken down in a grouping of consumer and farm magazines, while magazine

36
revenues and subscription costs were given for gll_published magazines.

It was assumed that consumer and farm magazines represented a bulk of
total non-trade, non-business magazine volume. Additionally, no source
reported the average yearly expenditure per medium for an advertiser;
hence it was assumed that the behavior of the top 100 advertisers would
be sufficient. Newsprint consumption was reported annually for all
publishers, while circulation figures and number of newspapers were
reported fOr daily and Sunday papers only. The number of FM households
in the United States for each year in the study was impossible to locate,
although scattered in-house National Association of Broadcaster publica-
tions provided some data. Hence, all data were not in perfect conformity
with each other, although the decision was made to use the existing data
mindful of their weaknesses.

All dollar figures were converted to constant 1972 dollars, using
the Gross National Product Deflator Index, to avoid the effect of
inflation. The choice of a base year within the period under study may
have been desirable; however, 1972 was chosen as a base since much of
the data was taken from Sterling and Haight (1978), who reported their
figures in 1972 dollars, and who provided yearly deflator indices for

the 1972 base.

Formulation of Mathematical Models

The formulation of a mathematical model requires the reduction of
hypotheses about the operation of a system to a level of abstraction
which specifies model components, variables and parameters, and func-
tional relationships. Naylor gt _1, (1968, p. 29) declare the process

an art and suggest that it draws upon previous studies in the field, a

bit of intuition and trial and error approaches. A modular building-

37
block approach was used in the model construction, which generated a

set of models describing the major subsectors of the system. These
subsectors were then synthesized into the general model.

Specific components and relationships among variables for each sub-
sector are discussed in detail in the second section of this chapter.
However, a word should be said about the number of variables in the
model and its complexity. Since this was a first, exploratory attempt
at building an intermedia model, an effort was made to limit both the
number of endogenous, or output, variables which would be examined in
the study and also the number of exogenous variables which determine
those outcomes. The number of variables in a model and its complexity
are directly related to programming time, computation time and validity.

In developing mathematical models of the various system components.
the potential difficulties outlined by Naylor gt_gl, (1968, p. 32) were
borne in mind: first, certain types of variables which affect the system
may be impossible to quantify or measure, such as the influence of the
FCC or the participation of parent corporations in the operations of
their subsidiaries. Second, the number of variables to be considered
in a system may exceed available computer hardware. Most computers
have a limit on memory storage; too many variables could put a program
over this limit, although the CDC 6600 computer used for this study has
a large memory available. Third, some significant exogenous variables
may not be known. Fourth, some relationships between exogenous and
endogenous variables may not be known and may be impossible to obtain.
These difficulties were encountered in describing the FM radio and UHF
television subsystems, where a simple time variable was finally used to
generate the endogenous variables in the system, when no other endogenous

or exogenous variables were significant predictors. Fifth, in some

38
cases the relationships between variables affecting the system may be so

complex that they are unable to be expressed in one or more mathematical
equations. This might be the case in the media system in attempting to

quantify cross-media ownership.

Estimation and Evaluation of Parameters

Ordinary least-squares regression routines, part of the Statistical
Package for the Social Sciences (SPSS) (Nie, Hull, Jenkins, Steinbrenner
& Bent, 1975) were used to estimate values of the parameters in the model
from the historical data collected. Assumptions made in generating in-
puts to the model were tested by subjecting the parameters and equations
to F-tests of statistical significance. The overall F-ratio for a
regression equation was examined for its significance in expressing a
relationship among variables, as was the square of the multiple regres-
sion coefficient (R-square) to determine the amount of variance
accounted for by the variables in the equation. Attempts were made to
choose equations which were significant at the .05 level and which
explained as much variance as possible. In addition, the significance
of each regression coefficient was examined to determine whether a
variable contributed to the ability to predict the behavior of the
system. Regression summaries for each equation in each subsector,
including R-square, overall F-ratios and the probabilities for signifi-

cance of individual coefficients, are presented in Appendix A.

Formulation of a Computer Program

The mathematical models describing the system were converted into
computer programs using Minnesota FORTRAN for the CDC computing system.
Subroutines which handled the generation of pseudorandom numbers and

stochastic variates, such as the Erlang family, were included. The

39
model used SOOOO octal words of core memory and required approximately

five seconds of central processing time.

Validation

The validation stage of the simulation attempted to prove the model
to be true, which implies a set of criteria to differentiate "true" from
"not true," and the ability to apply those criteria to the model. The
multi-stage verification procedure of Naylor and Finger (1967) was used,
which incorporates the methodologies of rationalism, empiricism and
positive economics, and implies that each of these approaches is neces-
sary, but not sufficient, for solving the problem of verification.

The first stage of the Naylor and Finger procedure calls for the
formulation of a set of postulates or hypotheses describing the behav-
ior of the system of interest. This set of postulates is formed from
the researcher's already acquired general knowledge of the system or
from her/his knowledge of other similar systems which have already been
successfblly simulated -- the rationalist approach. The selection of
postulates includes the specification of components and the selection of
variables, as well as the formulation of functional relationships.

The second stage of the multi-stage verification process--that of
empiricism--calls for an attempt to verify the postulates on which the
model is based subject to the limitations of existing statistical tests.
Naylor and Finger suggest that postulates which cannot be falsified
empirically be retained "tentatively" as there is no reason to assume
they are invalid just because they cannot be tested.

The third stage of the procedure consists of testing the model's
ability to predict the behavior of the system under study--the positive

economics approach. Two alternatives are available to test the degree

40
to which data generated by computer simulation models conform to observed

data-~historical or retrospective, prediction and forecasting, or prospec-
tive, prediction. Retrospective prediction procedures were employed to
validate the model in this study.

From the many statistical techniques available to assess the good-
ness of fit of the model, such as chi-square or Pearson's product moment
correlation coefficient, Theil's (1961) Inequality Coefficient was cal-
culated to determine the degree of conformity of the time paths generated
by the computer simulation to the historical data. The coefficient U
provides an index which measures the degree to which a simulation model

provides retrospective predictions of observed historical data:

“1 2
, \[5-2(P,-A,)

_l_ 2 _1_ 2
\Inz": T n”:

where P1,...,Pn are the predicted values

 

U

 

A1,...,An the corresponding actual outcomes.
U varies between 0 and 1. If U equals 0, there are perfect predictions.
If U equals 1, the predictions are no better than random.

Theil's U avoids the need to categorize data required by chi-square
analysis. It is also more sensitive to the actual degree of association
between the simulated and historical time series than Pearson's product
moment correlation, which shows only the degree of linear relationship.
Also, the Inequality Coefficient, unlike the Pearson r, is not invariant
across additive transformations. The output for each endogenous vari-
able was considered valid when U assumed a value of 0.2 or below, a

limit recommended by Theil.

41
Description of the Model

Historical Perspective

Television was dramatically introduced to the public by David
Sarnoff, president of the Radio Corporation of America (RCA), during a
broadcast from the New York World's Fair on April 30, 1939. Four years
earlier, FM radio had been presented to the public by its inventor,
Edwin Armstrong, on April 26, 1935. These two new broadcast delivery
systems were to compete head-on for the support of the public, the
broadcasting industry, and the Federal Communications Commission (FCC),
while at the same time causing an upheaval in the worlds of AM radio and
magazines.

AM radio found itself hard hit by television, which replaced it as
a major source of entertainment and national advertising dollars. To
meet television's threat, radio sacrificed network advertising dollars,
and turned local, factual and musical. (Wood, 1971, p. 304) FM offered
competition for audiences, since it boasted clearer reception, but other
events were to lessen FM's impact on its AM relative. Newspapers were
not to feel the direct impact of the invasion of television, since their
advertising was mostly local in nature. In fact, 1945-1960 were prof;
itable years for newspapers, as the loss of daily newspapers in some
cities was offset by the successful starts of others. Hence, the number
of daily newspapers in this period remained fairly constant, although
when analyzed at a more microscopic level by city size, there were sig-
nificant changes in the industry. (See Sterling & Haight, 1978, p. 23)
Circulation remained even with population expansion through 1960 (Emery,
1972, p. 620), although the amount of time devoted to newspaper reading
dropped significantly as television gained its hold on the public. Mass

magazines, once the only medium to provide advertisers with national

42
coverage, were bitterly conscious of television's presence in the media

world. To survive, they changed from a truly national mass medium to
more regional fare, to a more specialized editorial content altogether,
or to oblivion. Thirty-two of the 250 largest magazines between 1950
and 1960 either ceased publication or were merged into other magazines
(Rucker, 1968, p. 205). According to the Magazine Publishers
Association, 39 percent of its members in 1960 operated at losses and
another 25 percent had profits of less than 5 percent of revenue before
taxes. Most of this can be attributed to advertisers' shift to tele-
vision to reach the mass audience.

In itself, the structural reorganization of the mass media industry
caused by FM and television makes the period of great interest to media
historians and industrial analysts. World War II postponed the normal
growth of the industry until 1945, which marked a logical start to such
a study. A number of policy decisions by the FCC also affected the out-
come of growth in the period and are of interest to policymakers and
planners facing similar decisions today.

As FM radio and television developed, FCC commissioners reacted
to their growth in a series of policy decisions. In 1945, the Commission
issued a major spectrum frequency reallocation plan, which moved FM from
its established place in the spectrum (VHF band) to make room for tele-
vision. In 1948, as television applications flooded the Commission and
a variety of technical improvements and developments were suggested, the
FCC issued a freeze order to suspend all television licensing activity.
The freeze was lifted in 1952, by an order which added ultra-high fre-
quencies (UHF) for television use. From a policy standpoint, questions
have been raised regarding the frequency reallocation itself, the number

of channels actually designated for FM radio and television, and the

43
reactionary approach typified by the freeze order. Variables in the

historical model built in this study were manipulated to represent
modifications in the actual FCC decision-making in order to answer some

hypothetical questions asked of the period.

An Overview

MASSMOD has four principal components: (1) an advertiser sector;
(2) a print sector, composed of newspaper and magazine units; (3) a
consumer sector; and (4) a broadcast sector, with AM radio, FM radio,

VHF television and UHF television modules. (See Figure 2.) These four
sectors interact in a yearly cycle.

Key variables of interest in the system include the size of the
broadcast audiences and publication circulations; the number of broad-
cast stations and publications; the profits of broadcasters and publish-
ers; and the media expenditures of advertisers. Several key policy
variables, manipulated during the experimental runs described in Chapter
4, were important to the system: advertisers' external media expenditure
constraints and the number of broadcast frequencies available for assign-
ment. Exogenous variables which were included were the Gross National
Product (GNP), disposable income, population birth and death rates,
retail sales, number of AM receivers, and newsprint consumption.

The interaction of the sectors revolves around the functional
relationship among advertising expenditures on each medium, publication
and broadcasting profits, number of publications and broadcasting stations
and audience size. Advertiser expenditures, generated stochastically,
were based on audiences and number of media during the previous year.
These expenditures influence profitability, which in turn affects media

growth or decline, which ultimately affects audience size. This basic

72.3.. E

20:30 12>
dam 78_Umw._d0 u 0

802303 u m
umoOU 305a,;

cosmw<212<mw<3 28.8 553 .Emmzmm

N uzzw:

Woks
, mmM
- EXP.
RETAIL SALES

 

 

MEWS PRINTS

 

 

 

 

 

 

 

 

 

    
     

 

 

 

 

 

 

 

 

 

 

 

 

 

w
WM a 6
d
U 9: L a \
H©Mull§am 425m Motown ”IQ fleefiauua NOPomw 353 a V©
ammFam>o< _.L. , use): hiya amass:
flHS .mwmlfi s m m 222.536
M c m .65
D N
A w .
9 .W
mmu2u53<
dokomm dupes: w MOPuwW
m: nous 1510 0.40am , 0.3 d win £1.00
® meQW534
Nmmiaz e
ww .5,” me .95 W.
a m w ,1. Wm mmm
mm. a m mm a g m
e .

44

45
relationship varies somewhat from submedium to submedium. but essentially

describes the system and the linkages of its components. Each component

is detailed in the following sections.

The Print Sector

The print sector has two parallel components--newspapers and maga-
zines. Due to limitations on the data available, the newspaper subsector
considered daily and Sunday newspapers only; the magazine subsector
included consumer and farm publications.

Newspaper Subsector. For each year t, the newspaper subsector
determines the number of daily and Sunday newspapers (NEWSNO); the
average yearly subscription cost for newspapers (NEWCOST); the total
yearly publishing costs (NPUBCST); annual revenue (NEWSREV) and earnings
(NEWEARN). Circulation figures (NEWSCIR) from the previous year t-l are
passed to the subsector from the consumer sector, which in turn receives
the number of newspapers and newspaper subscription costs for year t
from the newspaper subsector. Total advertiser expenditures on news-
papers (TNEWBUY) for year t are supplied by the advertiser sector, which
uses the number of newspapers at t-l in its calculations. The subsector
uses two exogenous system variables in its processing: GNP and yearly
newsprint consumption (NEWSPR). A block diagram of the subsystem is
presented in Figure 3; Table 1 contains a glossary of all variables
used in the subsector.

Real world data used to derive functional relationships within the
subsector were gathered from a variety of sources. The number of daily
and Sunday newspapers is based on data taken from the Ayer Directory o_f_
Publications (Sterling & Haight, 1978, p. 22), which includes all publi-

cations serving a general circulation. The limitation of daily and

«opummmam mma<am3mzuuz§w<3 V803

m $5.2“.

 

 

._ ,
oi:

 

f

ca 2.3L
”$95322 Sou“.

 

 

 

 

 

 

 

 

”3.8.328 9r

 

F >uun2mz4

 

 

 

I
V

Nun. raged 0..

 

 

15332

 

 

 

.P WOO 3&2

_ I

 

  

 

    

 
  

11$
1 , ”39.5254 0...

w

 

NuéDnZu O...

amiamzoo £08.

I
V

46

Variable

GNP

NEARNI

NEWCOST

NEWEARN

NEWSCIR

NEWSCRI

NEWSNO

NEWSPR

NEWSREV

NPUBCST

NSUBREV

TNEWBUY

47

TABLE 1

GLOSSARY
NEWSPAPER SUBSECTOR

Description

U.S. Gross National Product

Total newspaper revenues for 1944;
millions of 1972 dollars

Annual average subscription
cost per person for newspapers

Total newspaper revenues;
millions of 1972 dollars

Total circulation of daily
newspapers (morning, evening)
in thousands

Total circulation of newspapers
for 1944

Number of daily newspapers,
morning and evening

Newsprint consumption; in
thousands of tons

Sum of advertising revenue
and subscription revenue;
millions of 1972 dollars

Newspaper publishing costs, cost
of materials; millions of 1972
dollars

Total newspaper revenues from
newspaper purchases; millions
of 1972 dollars

Total advertising expenditures
on newspapers; millions of
1972 dollars

Source

Sterling & Haight,
pp. 111-112

Sterling & Haight,
p. 157

calculated
Sterling & Haight,
p. 157

Sterling & Haight,
p. 20

Sterling & Haight,
p. 20

Sterling & Haight,
p. 20

Historical Statis-
tics R 218-223

calculated

Census of Manufac-
tures 1963. 1972

Newspaper Adver-
tising Bureau.
1972. 1978

Advertising Age

48
Sunday newspapers was imposed for two main reasons: figures were

available for numbers. circulation, costs and revenues for this cate-
gory; and, given the aggregate, national level of the model itself, the
daily/Sunday newspaper seemed the most prominent competitor of magazine,
radio and television. The newspaper audience is described in terms of
circulation figures obtained from census data, realizing that circula-
tion is a more accurate measure than individual readership, due to the
unreliability of pass-along readership reports.

Publishing costs are those reported in the ng§g§_gj_Manufactures;
it is difficult to determine from the census's categorization scheme
whether these are just for daily/Sunday papers or all newspapers, and
whether the costs include production and personnel expenditures or
merely production. Despite these uncertainties, the figures were the
only ones available. Subscription costs are based on an in-house report
by the Newspaper Advertising Bureau, estimating the annual expenditure
on weekday papers by consumers. Earnings are also based on ggg§g§_gj;

Manufactures and annual Survey gleanufactures data, and are those

 

reported as total income (sale of advertising and newspapers) from
newspaper products only. (Sterling & Haight, 1978. p. 157) Many valu-
able and more precise breakdowns of information were initiated by
government and trade organizations after 1960.

Two feedback loops extend from the circulation variable. Circu-
lation at t-l, received from the consumer subsector, determines total
subscription revenue (NSUBREV) at t. In the simpler of the two loops.
circulation at t is a function of this total subscription revenue (in
actuality, annual subscription costs which is derived from the total
subscription revenue).

In the second loop, the total subscription revenue at t, when

49
combined with advertising revenue. determines the total newspaper

revenue (NEWSREV) at t, which is related to total earning (NEWEARN) at
t. Total earnings at t in part determine the number of newspapers at
t+1 which influences newspaper circulation and begins the feedback loop
again.

Specifically, subscription revenues are a function of the circu-
lation at t-l:

NSUBREV(t)=540.2 + 0.018*NEWSCIR(t-l)
While this lagged relationship is indirect, it is required by the model
in order to make the interconnection of subsectors possible.

From this, the average annual subscription cost can be calculated,
mindfu1 of the fact that circulations are reported in thousands in the
model and revenues in millions of dollars:

NEWCOST(t)=NSUBREV(t)/NEWSCIR(t-l)*1000

Total revenues are computed as the sum of subscriptions and
advertising expenses:

NEWSREV(t)=NSUBREV(t) + TNEWEUY(t)

Publishing costs are a function of the tons of newsprint consumed
in the production process:

NPUBCST(t)= -l4021.34 + l797.32*ln(NEWSPR(t))
It has been suggested by Owen (1975, p. 36) that publication costs are
based on the circulation of the publication and that economies of scale
should exist in the production process. Since the high level of aggre-
gation of the model does not consider individual differences in circu-
lation and production costs. this relationship does not reach a level of
significance in a regression equation. Also, it would be much more
desirable to use the cost of newsprint for any given year in the rela-

tionship. Such figures for the time period under study for daily and

50
Sunday newspapers were not available.

Newspaper earnings are the difference between revenues and
publishing costs:
NEWEARN(t)=NEWSREV(t) - NPUBCST(t)
The number of newspapers for year t is a function of the GNP and
newspaper earnings at t-l:

NEWSNO(t)=1505.27 + 61.59*1n(NEWEARN(t-1))
- O.27*GNP(t)

Since newspaper earnings depend on advertising revenue and subscription
income, both of which depend on circulation, it should be noted that the
number of newspapers is determined (indirectly) by the interaction of
readers' (subscription income) and advertisers' (advertising revenue)
demands. (Owen, 1975, p. 34) Udell (1978) suggests that paper supply
is also a key determinant of number of newspapers; however, it was not
significant in the regression equation.

Magazine Subsector. The following variables are computed for year
t in the magazine subsector: number of general and farm magazines
(MAGNO), magazine publishing costs (MPUBCST), subscription revenues
(NSUBREV), total revenues (MAGREV), earnings (MAGEARN), and profits
(MAGPROF). GNP is an exogenous variable to the subsystem. Circulation
figures from the previous year t-l (MAGCIRC) are passed from the consumer
sector, which receives the number of magazines at year t as subsequent
input. The advertiser sector supplies total media buys on magazines
(TMAGBUY) in year t as input; this calculation derives in part from the
number of magazines at t-l passed from the magazine subsector. Figure
4 is a block diagram of the system; a glossary of subsector variables
is presented in Table 2.

Data from which the historical model was built were accumulated

 

 

  

 

 

 

FRom Comma
.43? e.__
' ' MAGCIRc ——Er—-)[mwecs1'

To Awea‘flseR r—

 

 

 

 

MGGCOST’

Lag
Mubkev
FRoM ADVEI'TISER / 1'
\i Trans 3 w +
To NEITHER
E— L

E f mmm FM's—— MAGRGV

To CNSUMGQ +
4

® mAGEAQN

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I
M

 

 

 

FIGURE 4
BLOCK DIAGRAM--MAGAZINE SUBSECTOR

51

Variable

GNP

MAGCIRC

MAGCIRI

MAGCOST

MAGEARN

MAGNO

MAGREV

MEARNI

MPUBCST

MSUBREV

TMAGBUY

52

TABLE 2

GLOSSARY
MAGAZINE SUBSECTOR

Description

U.S. Gross National Product

Magazine circulation, general
and farm; in thousands

Magazine circulation for 1944

Average subscription cost per
magazine per year; 1972
dollars

Total magazine revenues;
millions of 1972 dollars

Total number of general and
farm magazines

Magazine revenue, sum of adver-
tising revenue and subscription
revenue; millions of 1972
dollars

Total magazine earnings fOr 1944;
millions of 1972 dollars

Magazine publishing costs;

sum of materials, payroll, wages,
new capital; millions of 1972
dollars

Total magazine revenues from
subscriptions; millions of
1972 dollars

Total advertising expenditures
on magazines; millions of 1972
dollars

Source

Sterling & Haight,
pp. 111-112

Sterling & Haight,
pp. 342-343

Sterling & Haight,
pp. 342-343

Sterling & Haight,
p. 177
Sterling & Haight,
p. 170

Sterling & Haight,
p. 342

calculated

Sterling & Haight,
p. 170

Sterling & Haight,
p. 176

calculated

Advertising Age

53
from several sources. Magazine circulation figures and number of

periodicals are those of the Magazine Publishers' Association (MPA) for
the total per-issue circulation of those general and farm magazines
audited by the Audit Bureau of Circulations (ABC). The figures include
the larger magazines of general appeal, as well as some smaller periodi-
cals that specialize in the farm market. (Sterling & Haight, 1978.

p. 34) It should be noted that membership in the ABC is voluntary, so
that changes in the circulation or the number of periodicals could be a
function of the periodicals audited by the ABC instead of changes in the
audience's general buying behavior.

Publishing costs are taken from the ggggggugj_Manufactures and
include materials, payroll, wages and new capital expenditures. Total
revenues are also based on §_e_g§_u_s_gf_ Manufactures data and describe
earnings from periodical products only, a distinction made by Sterling
and Haight to eliminate revenues received by magazine publishers from
other businesses, such as newspapers (1978, p. 170). Subscription costs
are the average annual library subscription price gathered from appro-

priate editions of The Bowker Annual. which utilized data from the

 

American Library Association. While the cost of magazines to libraries
may not represent the actual cost to the consumer, and also includes
more than general and farm magazines, these figures were the only such
subscription data available for the years under study.

One feedback loop operates in the magazine subsector. Magazine
circulation at t-l influences publishing costs and subscription revenue
at time t; from this, magazine revenue at t is determined, which affects
the number of magazines at t+l, which in turn affects circulation.

Magazine publishing Costs are a function of the GNP and lagged

magazine circulation:

54

MPUBCST(t)=329.17 + 0.003*MAGCIRC(t-1)
+ 1.32*GNP(t)

Magazines use higher quality paper than newsprint, hence newsprint
supply is not included in the equation as it was in the newspaper sub-
sector. Paper supply and/or consumption figures for magazines are not
readily available, since the production operation of magazines is often
farmed out to independent printers who do not itemize nor report their
paper use.
The average annual subscription cost of magazines is derived from

the publishing cost:

MAGCOST(t)=5.56 + 0.001*MPUBCST(t)
From the calculation of the average subscription cost, total magazine
subscription revenue can be computed by multiplying the average cost by
the magazine circulation:

MSUBREV(t)=MAGCOST(t) * MAGCIRC(t-l)/1000
An adjustment is made to convert the result into millions of dollars.
While the circulation at time t would be a more appropriate value in
this equation, the sequence of the model requires that the preceding
year's figures be used instead.

Total magazine revenue should be the sum of subscription income

and advertising income:

MAGINC(t)=MSUBREV(t) + TMAGBUY(t)
However, this calculation did not correspond to revenue figures reported
by the industry, possibly because the revenue figures in the Mg:
Manufactures are fOr all periodical products, while the model to this
point deals with general and farm magazines. Since it was not desir-
able to add additional variables to the model, the revenue figure

reported by the industry was regressed with the total income figure

55
derived from the above equation to correct the anomaly:

MAGREV(t)=1073.93 + 0.58*(MSUBREV(t) + TMAGBUY(t))
Magazine earnings were computed as the difference between revenue and
expenses (publishing costs):
MAGEARN(t)=MAGREV(t) - NPUBCST(t)
The number of magazines at year t is a function of magazine
revenues during the previous year and the GNP at t:
MAGNO(t)=204.417 - 0.014*MAGREV(t-l) + 0.144*GNP(t)
Indirectly, the number of magazines reflects both consumer and advertiser
demands, since both subscription revenue and advertising revenue are

components of total magazine revenue.

The Broadcast Sector

The broadcast sector is represented by three modules in the model:
a radio subsector, which has parallel AM and FM components; a VHF
television subsector and a UHF television subsector. Because the model
is constructed at a highly aggregate level, the distinction between
network affiliated stations and independents, as well as the factor of
competition between stations in a market, was not considered.

The AM Subsector. For each year t, the AM radio subsector deter-
mines the number of commercial AM radio stations (AMNO); and the annual
expenses (AMEXP), revenues (AMREV) and earnings (AMEARN) for the
stations. The number of households with AM radios (AMHH) from the
previous year t-l are passed to the subsector from the consumer sector;
this sector receives the number of AM stations at year t as input.
Advertising expenditures on AM radio (AMBUY) fur year t are delivered
from the advertising sector, which uses the number of stations at t-l in

its calculations. GNP is the only exogenous variable in the module;

56
one policy variable--channels available for AM radio use (AMCHAN)--was

manipulated during experimental runs described in Chapter 4. A block
diagram for this subsector is presented in Figure 5; Table 3 contains a
glossary of variables used in the module.

The model concerns itself only with commercial (i.e., those which
sell advertising time) AM radio stations; the number of stations for
each year is that reported by the Federal Communications Commission.
Expenses, revenues and pre-tax earnings are also FCC figures as given
in annual financial reports on the radio industry. The figures include
not only AM stations, but AM-FM combinations--those FM stations which
are owned by AM stations. The FCC did not require separate reporting
of AM and FM revenues for these jointly owned stations until 1969
(Sterling & Haight, 1978. p. 204), which made the analysis of FM
stations particularly difficult in this model.

In detail, the equations in the submodel are as follows:

The number of AM radio stations in year t is a function of the GNP
for that year and the earnings of the stations during the previous year
t-l:

AMNO(t)=2727.41 - 3.13*GNP(t) - 1.78*AMEARN(t-1)
+ 213.46*RNYR

RNYR can be considered a growth variable; it represents the number of

the year and makes more of a contribution as the model progresses through
time, and the industry becomes more well-established. The use of RNYR
could also indicate that a significant variable has not been identified
and included in the model. Since earnings are derived from expenses,
which are affected by audience size, the number of stations is indirectly
determined by audience. (Blau, Johnson & Ksobiech, 1976, p. 199)

The number of stations in year t is a factor in the total expenses

«chummmzm OBS. z<iz§o<3 V603
m maze:

020

 

 

 

 

 

 

 

 

 

 

% >312 _ f % >324 a 02:34 W
Noe—534 some L . , - _ Max:628 B

- _ FIE t
. «955254 Our
_ i E T!
a55< 115
r a

MmiamZou Sam...

 

 

 

 

 

 

 

 

 

 

 

Variable

AINCOME

AMBUY

AMCHAN

AMEARN

AMEXP

AMHH

AMNO

AMREV

GNP

RNYR

58

TABLE 3

- GLOSSARY
AM RADIO SUBSECTOR

Description

Difference between AM station
revenues and expenses;
millions of 1972 dollars

Total advertising expenditures
on AM radio; millions of 1972
dollars

Number of channels available
fOr AM commercial use

Total pre-tax earnings for all
AM radio stations (includes AM
and FM combinations); millions
of 1972 dollars

Total expenses for all AM radio
stations (includes AM and FM
combinations); millions of

1972 dollars

U.S. households with AM radio
receivers; in thousands

Number of commercial AM radio
stations

Total revenues of AM radio
stations (includes AM and FM
combinations); millions of
1972 dollars

U.S. Gross National Product

Time index: NYR-l

Source

calculated

AdvertisingpAge,
calculated

FCC Reports,

 

manipulated

Sterling & Haight,
p. 203

Sterling & Haight,
p. 203

Sterling & Haight,
p. 367

Sterling & Haight,
p. 43

Sterling & Haight,
p. 203

Sterling & Haight,
pp. 111-112

calculated

59
of stations in that year. The number of AM households at t-l is also

a determinant:

AMEXP(t)=513.04 + 0.00002*AMHH(t-l) + 0.07*AMNO(t)
Since broadcasters attempt to maximize audience size in order to attract
maximum advertising dollars and hence profits, this equation is quite
expected.

AM station revenues are determined mainly by advertiser expendi-
tures, although these do not totally account for reported AM station
income. The growth variable RNYR is also included in the equation.

Part of this anomaly could be due to the inclusion of the FM revenues
of AM-FM station combinations.

AMREV(t)=118.423 + O.760*AMBUY(t)
+ O.689*(RNYR**2)

AM station income is computed as the difference between revenue and
expenses:
AINCOME=AMREV(t) - AMEXP(t)
This computation is not equivalent to the pre-tax earnings reported fbr
AM stations in the FCC Annual Reports, although the components of the
regression equations for AM revenues and AM expenses are significant
and the predicted values validate with the historical data. AM earn-
ings are therefore calculated as follows in the model:
AMEARN(t)=59.059 + 0.49*AINCOME
The FM Subsector.~ The FM radio subsector determines for each year
t the number of FM stations in operation (FMNO), and their annual
revenues (FMREV), expenses (FMEXP) and pre-tax earnings (FMEARN).
Advertising expenditures on FM stations (FMBUY) for the year under
consideration are passed to the subsector by the advertising sector; this

sector uses the number of FM stations in the previous year t-l to arrive

60
at the expenditures for year t. GNP is the only exogenous variable in

the subsystem, although calculations also depend on the growth variable
RNYR. One policy variable, representing the number of FM channels
available for commercial use (FMCHAN), was included for experimentation
in Chapter 4. A block diagram of the FM subsector is given in Figure 6;
Table 4 contains a glossary of variables.

This subsector was, by far, the most difficult to model due to lack
of data, or lack of data available in the desired categories. Data on
the number of FM stations were derived from annual FCC Reports and refer
to commercial stations actually on the air, regardless of license status.
Figures for expenses, revenues and earnings are also taken from FCC
Reports, but include only independent FM stations (stations not owned by
an AM station company). As previously mentioned, the FCC did not require
AM-FM station combinations to provide a detailed breakdown of their
financial affairs until 1969. Hence, the monetary figures used in this
subsector most likely underestimate actual expenses, revenues and prof;
its, but there is no data available to quantify this underestimation.

The number of households with FM radio receivers does not enter
into the calculation of any variables in the subsector. This could be
due, at least in part, to the scarcity of data used in the regression
estimating equations. Neither the FCC nor the National Association of
Broadcasters (NAB) nor the National Association of FM Broadcasters nor
the Electronics Industries Association nor audience measurement services
such as Hooper or A. C. Nielsen maintained any separate data on FM
households or the number of radio receivers capable of FM reception
during the time period involved in this study. Four scattered data
points were gleaned from in-house publications in the NAB library pro-

duced by NAB's Broadcast Measurement Bureau (1949) and FM-ghasis (1959).

achummmnm o~o<¢ Zunnz<mw<~o xUOAm

o mmzwmu

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

72?“.
732.. _ Beau 02st
«3.5»; Sud _1 9.... v
«3.532 0%

 

 

 

6

M
C:
M

61

Variable

FMBUY
FMCHAN

FMEARN

FMEXP

FMHH

FMNO

FMREV

GNP

GROWTH

RNYR

62

TABLE 4

GLOSSARY
FM RADIO SUBSECTOR

Description
Total advertising expenditures

Number of channels available
for FM commercial use

Total pre-tax earnings for
commercial FM radio stations;
millions of 1972 dollars

Total expenses for commercial
FM radio stations; millions
of 1972 dollars

U.S. households with FM radio
receivers; in thousands

Number of commercial FM radio
stations

Total revenues of commercial FM

radio stations (excludes com-
binations); millions of 1972
dollars

U.S. Gross National Product

Constant to describe growth of
FM radio stations

Time index: NYR-l

Source

Advertising Age

FCC Reports.

 

manipulated
Sterling & Haight.
p. 205

Sterling & Haight,
p. 205

NAB. 1959; NARTB.
1952

Sterling & Haight,
p. 205

Sterling & Haight,
p. 205

Sterling & Haight,
p. 157

calculated

calculated

63
and the National Association of Radio and Television Broadcasters

(1952, 1954).

The number of FM stations for year t is determined by an exponential
relationship with FM station earnings during the previous year t-l and
the GNP:

FMNO(t)=exp(7.208 - 0.096*FMEARN(t-l)
-0.006*GNP(t))

Station earnings are calculated as the difference between revenues
and expenses:
FMEARN(t)=FMREV(t) - FMEXP(t)
where revenues are a function of the advertisers' expenditures on FM
commercial time and the cube of the growth factor RNYR:

FMREV(t)= -16.518 + 0.086*FMBUY(t)
+ 0.0023*(RNYR**3)

FM station expenses are calculated using only the time variable RNYR in
its cubed and squared form. As with most of the other equations ulti-
mately used in the broadcasting submodels, many other variables suggested
by theory, such as number of FM households, number of FM stations, earn-
ings, GNP, were included in regression equations, only to discover that
they were not significant predictors of the variable under consideration.
Therefore, the only significant equation derived for FM expenses was as
follows:

FMEXP(t)=ll.42 + 0.019*(RNYR**3) - 0.28*(RNYR**2)

The VHF Subsector. This subsector determines the number of com-
mercial VHF television stations (VHFNO) at year t. It also computes the
revenues (VHFREV), expenses (VHFEXP) and earnings (VHFEARN) for those
stations during the same year. The consumer subsector provides this
module with the number of households with VHF television receivers

(VHFHH) during the previous year t-l. The advertising sector delivers

64
advertiser expenditures on VHF television (VHFBUY) for year t, and uses

number of VHF television households at t-l in its computations. As in
the other broadcast subsectors, GNP is the only exogenous variable; one
policy variable--number of channels in the VHF bandrrfthe spectrum
allocated for commercial television use (VHFCHAN)--were manipulated
during experimental runs. Figure 7 is a block diagram of the VHF
subsystem; a glossary of variables is given in Table 5.

The number of commercial VHF stations for each year is that
reported by the FCC (Sterling & Haight. 1978, p. 49). Regular tele-
vision station operation was approved by the FCC in 1941, but was inter-
rupted by World War II. After the war. growth was still slow, due to
high station construction and operation costs, the limited number of
VHF-only channels, and allocation problems. The FCC froze the proces-
sing of station applications in 1948; the freeze was lifted in 1952.
Increase in station numbers during this period is the result of
stations which had approved applications prior to 1948 actually going
on the air. Revenues, expenses and earnings for VHF commercial stations
are drawn from official FCC data on television station economics
(Sterling & Haight, 1978. p. 206).

One feedback loop is involved in the model, with the number of VHF
stations as the keystone. VHF expenses are a function of VHF households,
the number of VHF stations and advertiser expenditures for VHF commercial
time. VHF revenue is based on advertiser expenditures, which depend in
part on the number of VHF stations during the previous year. The dif-
ference between revenues and expenses is pre-tax earnings for year t,
which after a two-year delay, affects the number of VHF stations at t.
Hence, number of stations influences earnings which in turn determines

number of stations.

mowummmam 233543 uz>uuz<mw<5 xuobm

s “#53“.

 

 

Saute.)

 

 

f

 

5

V
NOM.Pflfl>6( OP 6

 

 

 

>muuz>

a? 738:?
«3.532 :3...

 

 

 

   

 

013
..ai I
deDmoGu 20.3....

_ , .

V
Unwrrdmw’ea 0L.

axmu...)

 

 

 

 

Variable

GNP

RNYR

VHFBUY

VHFCHAN

VHFEARN

VHFEXP

VHFHH

VHFNO

VHFREV

66

TABLE 5

GLOSSARY
VHF TELEVISION SUBSECTOR

Description

U.S. Gross National Product

Time index; NYR-l

Total advertising expenditures on
VHF television; millions of 1972
dollars

Number of channels available for
commercial VHF stations

Earnings before taxes for com-
mercial VHF television stations;
millions of 1972 dollars

Expenses for commercial VHF
television stations; millions of
1972 dollars

Number of U.S. households with VHF
television receivers; in thousands

Number of commercial VHF television
stations

Revenue of VHF commercial television
stations; millions of 1972 dollars

Source

Sterling & Haight,

pp. 111-112
calculated
Advertising Age

 

FCC Reports,
manipulated

Sterling & Haight,
pp. 208-209

Sterling & Haight,
pp. 208-209

Sterling & Haight,
p. 372

Sterling & Haight,
p. 49

Sterling & Haight,
pp. 208-209

67
Specifically, VHF station revenue is related to advertising expend-

itures as follows:
VHFREV(t)= -24.63 + 0.48*VHFBUY(t)

It might be expected that VHF station revenue would be accounted for in'
full by advertising revenue. 'This, however, is not the case, at least
according to the data used in the regression equations. Stations do
have other sources of income, such as promotional fees, use of talent,
use of facilities. However, advertiser expenditures are not broken down
according to VHF and UHF stations, and the assumptions made as to the
proportional breakdown may be incorrect.

Station expenses are computed according to the following
relationship:

VHFEXP(t)= -32.78 + O.489*VHFBUY(t)
- 0.00533*VHFHH(t-1)

Stations base their expenditures on the number of households to be
served, as well as the revenue they derive from advertisers. This
result is consistent with real life data on a market by market basis.
Stations in larger markets, determined by number of households, tend to
invest more money in programming and personnel expenses than stations
in smaller markets. Of course, their revenue is also increased due to
larger advertiser buys, so the increased expense is offset by increased
revenues.

Earnings are the difference between revenue and expenses:

VHFEARN(t)=VHFREV(t) - VHFEXP(t)

The submodel acknowledges the slow growth of television stations
and the subsequent freeze in the granting of applications by using two
separate equations:

VHFNO(t)= -149.156 + 0.343*GNP(t)

68
The second equation takes into account earnings after a two year delay

and adds the growth variable, RNYR, not unexpected due to the freeze:
VHFNO(t)= -l77.29 + 0.303*GNP(t)
+ 0.6*VHFEARN(t-2)
'+ 20.42*RNYR

Since no FCC financial data were available for the years 1945-1947,
the subsector reports values of O for expenses, revenues and earnings
for these three years.

The UHF Subsector. This subsector begins operation in the model in
1953, the first year the FCC authorized the use of frequencies in the
ultra high band for commercial television. It determines the number of
UHF television stations (UHFNO) at year t as well as the revenues
(UHFREV), expenses (UHFEXP) and earnings (UHFEARN) for those stations.
The subsector receives the number of households with UHF television
receivers (UHFHH) at t-l from the consumer sector and the revenue from
advertising on UHF stations (UHFBUY) from the advertiser sector. Number
of UHF stations at t-l is used by the advertiser sector to determine all
buys. UNYR is a growth variable comparable to RNYR in the AM, FM and
VHF subsectors; in 1953, UNYR equals 0. The number of channels avail-
able for commercial UHF stations (UHFCHAN) was a policy variable manip-
ulated in experiments in Chapter 4. A block diagram of the subsystem
is shown in Figure 8; Table 6 is a glossary of variables in the sub-
sector.

Figures fOr number of commercial UHF stations are taken from FCC
Annual Reports. The number of households with UHF receivers is a
derived value based on UHF penetration percentages reported by the
Advertising Research Foundation and the U.S. Census Bureau and the
number of homes with television receivers, gathered from NBC Corporate

Planning data (Sterling & Haight, 1978, p. 373). Revenues, expenses

moeummmzm 2335.3: mzzuuzéosn— 50.5
m $53...

6

 

 

 

 

 

 

 

02....23

 

 

 

 

 

 

 

 

 

 

 

 

 

LF 53::
«09.533. 6a....

 

 

 

 

 

 

 

 

9.
a
3‘
t—K—J
Hi,-
be
r—h—

 

5.3.93

 

64.. w
1....» r5522 0...

 

 

Iiuzb T

 

 

2T

 

amiomzou tom“.

69

Variable

UHFBUY

UHFCHAN

UHFEARN

UHFEXP

UHFHH

UHFNO

UHFREV

UINCOME

UNYR

70

TABLE 6

GLOSSARY
UHF TELEVISION SUBSECTOR

Description

Total advertising expenditures
on UHF television stations;
millions of 1972 dollars

Number of channels available for
UHF commercial use

Total earning befOre taxes for
UHF television stations; millions
of 1972 dollars

Total expenses for UHF television
stations; millions of 1972 dollars

Number of households in U.S. with
UHF receivers; in thousands

Number of commercial UHF television
stations

Revenue of commercial UHF television
stations; millions of 1972 dollars

Difference between UHF station
revenue and expenses; millions of
1972 dollars

Time index; NYR-9; 1953=O

 

Source
Advertising_Age,
calculated

FCC Reports,
manipulated

Sterling & Haight,
pp. 208-209
Sterling & Haight,
pp. 208-209

Sterling & Haight,
p. 372

Sterling & Haight,
p. 49

Sterling & Haight,
pp. 208-209

calculated

calculated

71
and pre-tax earnings for commercial UHF stations are those reported by

the National Association of Broadcasters in their Television Financial

 

Reports (Sterling & Haight, 1978, p. 209).

Feedback in this subsector revolves around number of UHF stations,
determined by station earnings after a one-year delay, but which in turn
contributes to the determination of both expenses and revenues, which
generate earnings. The number of UHF stations at t-l is used by the
advertiser sector to calculate UHF advertising expenditures, which also
contribute to the prediction of UHF revenue, thus adding to the com-
plexity of the feedback loop.

Specifically, the number of UHF stations is determined by the
following equation:

UHFNO(t)=128.21 - O.55*UHFEARN(t-1)
+ O.87*(UNYR**2) - 13.86*UNYR

UNYR functions as a growth variable, negative in this case, since the
number of UHF stations decreases from 1953 to 1960. This is partially
due to the fact that television sets were not required to receive UHF
signals until the 1962 All-Channel Receiver Act was passed. Hence,
without a guarantee that audiences would be capable of receiving their
signals, station profitability was in jeopardy. Number of UHF receivers
was tested in this equation, but was not a significant variable. Num-
ber of UHF households reflects indirectly in the equation, as it affects
expenses which in turn affect earnings.

The equation for expenses is:

UHFEXP(t)=26.68 - 0.001*UHFHH(t-l) + 0.26*UHFNO(t)

Revenue is based an advertiser expenditures and the number of

stations:

UHFREV(t)=34.32 - 0.02*UHFBUY(t) + O.27*UHFNO(t)

72
Station income is calculated as the difference between revenue and

expenses:
UINCOME=UHFREV(t) - UHFEXP(t)
However, this income figure, like that in the AM subsector, does not
equal the reported pre-tax earnings for UHF stations, even though the
predicted values for both revenue and expenses validate with historical
data. Therefore, another regression equation was inserted in the model
to accomodate the discrepancy:
UHFEARN(t)= -5.380 - 2.395*(UINCOME**2)
- 6.030*UINCOME

The Advertising Sector

This sector computes the total advertising expenditures by adver-
tisers for each media subsector in the model at t (NEWSBUY, MAGBUY,
AMBUY, FMBUY, VHFBUY, UHFBUY) based on input from the media subsectors
regarding the medium's circulation or audience, and the number of pub-
lications or stations in operation during the previous year t-l. Oper-
ating as a media planning-buying function, it first calculates the
amount of money spent on local, national spot and network advertising
before allocating expenditures to the particular media. It also moni-
tors retail profits (RETPROF) as the difference between retail sales
(SALES), supplied to the model as an exogenous variable, and advertising
expenditures. For reasons to be discussed below, this sector is sto-
chastic--i.e., the results are based on probabilistic rather than com-
pletely determined outcomes. Minimum and maximum values for one adver-
tiser's expenditures on a particular medium are provided to the sector
as exogenous variables, as is a parameter for each medium used by the
random number generator (the PARAM array). A policy variable which

controls the amount of advertising expenditures (ADEXPOL) was included

73
in the model for manipulation during the experiments described in

Chapter 4. A basic diagram of the sector is presented in Figure 9; a
glossary of variables is given in Table 7.

Generally, annual advertiser expenditures are reported by medium,
without regard for VHF/UHF or AM/FM distinctions, and by the type of
advertising--network, national spot and local. Figures used in the
model, reported by Robert J. Coen of McCann-Erickson, Inc. (Sterling &
Haight, 1978, p. 124), are estimates, which vary greatly in validity
depending on the medium. Broadcast figures are taken from FCC annual
financial reports; newspaper data are for daily and Sunday papers only.
Retail sales figures for each year were taken from appropriate volumes
of Statistical Abstracts. In order to locate minimum and maximum ex-
penditures by advertisers by medium, it was assumed that the advertising
buys of the top 100 advertisers represented a major portion of all buys
in a particular medium; hence, annual Advertising Age reports of the top
100 Advertisers by medium were used to provide the necessary minima and
maxima, and also to determine estimates of the average annual advertiser
expenditure by medium.

The selection of a media schedule by an advertiser is a multi-
faceted problem, which includes at the least such variables as the
distribution system of the product, the target population, the firm's
budget size. and the characteristics of the product (Gensch, 1973.

p. 75). Other considerations include the availability of time or space
in the media, the quality of the advertising copy and the viewing or
reading habits of the population. While a number of mathematical methods
exist, such as the optimizing approach of linear, non-linear and dynamic
programming, and the non-optimizing approach of iteration or marginal

analysis, heuristic programming and simulation, Gensch maintains that

 

 

 

I

 

CI RC.

 

 

 

 

 

 

 

E,
FROM meme AV 1
4 M6 1| H H MED: A
5035561123 5 1 B” Y

No . ____J(.—_‘

_.L_

 

 

 

 

 

fl

 

4L

ERLANG
GENERATOR
100 Times

News ;
mac.
wet RAD

5961' RAD
LKAL. RAD
NGT ‘rv
SPOT TV
Locust. TV

 

 

 

 

 

 

TOTA
ADVPoL MEDIAL—

Boy

 

TO MEDIA

 

\

fl
suesec-roes

FIGURE 9
BLOCK DIAGRAM--ADVERTISER SECTOR

74

Variable

AMBUY

AMHH

AMNO

AVLRD

AVLTV

AVMAG

AVNEW

AVNRD

AVNTV

AVSRD

AVSTV

75

TABLE 7

GLOSSARY
ADVERTISER SECTOR

Description

Total advertising expenditures
on AM radio; millions of 1972
dollars

U.S. households with AM radio
receivers; in thousands

Number of commercial AM radio
stations

Average expenditure per adver-
tiser on local radio spots;
millions of 1972 dollars

Average expenditure per adver-
tiser on local television spots;
millions of 1972 dollars

Average expenditure per adver-
tiser on magazines; millions
of 1972 dollars

Average expenditure per adver-
tiser on newspapers; millions of
1972 dollars

Average expenditure per adver-
tiser on network radio spots;
millions of 1972 dollars

Average expenditure per adver-
tiser on network television
spots; millions of 1972 dollars

Average expenditure per adver-
tiser on spot radio; millions
of 1972 dollars

Average expenditure per adver-
tiser on spot television; millions
of 1972 dollars

Source

AdvertisingAge,
calculated

Sterling & Haight,
p. 367

Sterling & Haight,
p. 43

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight.
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

BUDGET

FMBUY

ISEED

MAGCIRC

MAGNO

NEWSCIR

NEWSNO

NYRLRD

NYRLTV

NYRMAG

NYRNEW

NYRNRD

NYRNTV

NYRSRD

NYRSTV

76

TABLE 7 (cont'd)

Advertising budget fOr all media;
millions of 1972 dollars

Total advertising expenditures
on FM radio; millions of 1972
dollars

Externally determined seed for
random number generator

Magazine circulation for general
and farm magazines; in thousands

Total number of general and farm
magazines

Total circulation of daily news-
papers (morning, evening); in
thousands

Number of daily newspapers,
morning and evening

Index for local radio ad expendi-
tures in PARAM array

Index for local television ad
expenditures in PARAM array

Index for magazine ad expendi-
tures in PARAM array

Index for newspaper ad expendi-
tures in PARAM array

Index for network radio ad expend-

itures in PARAM array

Index for network television ad
expenditures in PARAM array

Index for spot radio ad expend-
itures in PARAM array

Index for spot television ad
expenditures in PARAM array

Sterling & Haight,
pp. 122-129

AdvertisinggAgp,
calculated

Sterling & Haight.
pp. 342-343

Sterling a Haight,
p. 342

Sterling & Haight,
p. 20

Sterling & Haight,
p. 20

OBDGT

PARAM

RADNO

RADTOT

RETPROF

SALES

TBUY

TLRDBUY

TLTVBUY

TMAGB UY

TNEWBUY

77

TABLE 7 (cont'd)

Over-budget indicator used to
monitor difference between ad-
vertising budget and actual ex-
penditures

Array of average advertising
expenditures for each medium,
minimum and maximum values, and
k values

Total number of commercial radio
stations (sum of AMNO and FMNO)

Total advertising expenditures
on radio (both AM and FM); (sum
of local, spot and network radio
expenditures); millions of 1972
dollars

Total retail profits; difference
between retail sales and adver-
tising expenditures; millions of
1972 dollars

Total annual retail sales;
millions of 1972 dollars

Total advertising expenditures
across all media; millions of
1972 dollars

Total advertising expenditures
on local radio; millions of 1972
dollars

Total advertising expenditures
on local television; millions
of 1972 dollars

Total advertising expenditures
on magazines; millions of 1972
dollars

Total advertising expenditures
on newspapers; millions of 1972
dollars

calculated

Advertising,Age

calculated

calculated

calculated

Census of Manufac-

tures

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

TNRDBUY

TNTVBUY

TSRDBUY

TSTVBUY

TVNO

UHFBUY

UHFNO

VHFBUY

VHFHH

VHFNO

78

TABLE 7 (cont'd)

Total advertising expenditures on
network radio; millions of 1972
dollars

Total advertising expenditures on
network television; millions of
1972 dollars

Total advertising expenditures on
spot radio; millions of 1972
dollars

Total advertising expenditures on
spot television; millions of 1972
dollars

Total number of television stations
(sum of VHFNO and UHFNO)

Total advertising expenditures on
UHF television; millions of 1972
dollars

Number of commercial UHF television
stations

Total advertising expenditures on
VHF television; millions of 1972
dollars

Number of U.S. households with VHF
television receivers; in thousands

Number of commercial VHF television
stations

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

Sterling & Haight,
pp. 122-129

calculated

AdvertisingyAge.
calculated

Sterling & Haight,
p. 49

Advertising Age,
calculated
Sterling & Haight,
p. 372

Sterling & Haight,
p. 49

79
"...there is no single acceptable. unified theory of making media selec-

tion decisions." (1973. p. 17)

The lack of a unified theory of media selection is compounded in
this model by the highly aggregate nature of the variables. No time/
space availabilities are offered, nor is it possible to define selected
target audiences from the population, which is assumed to be homogeneous.
The concept of market segmentation as a tool in planning advertising ex-
penditures was to come later. (Kotler, 1972. pp. 165-166) In addition,
no single product decisions are involved, only total expenditures.
Therefore it was decided to make the selection of media by advertisers
stochastic, selecting annual advertiser expenditures by medium from
probability distributions instead of completely determined equations.

Probability Distributions. A stochastic simulation involves the
replacement of an actual statistical universe of elements (such as
advertiser media selection behavior) by its theoretical counterpart,
described by some probability distribution, and then sampling by means
of some type of random number generation from the theoretical population,
according to Naylor _p _l. (1968. p. 69). The task becomes one of
selecting an appropriate probability distribution to replace the actual
statistical universe.

A probability distribution represents the distribution of the
relative frequency of the occurrence of all possible events under con-
sideration. The shape of such a distribution is determined by the
behavior of events within the system. Distributions can be divided
into two general types, theoretical and empirically derived. An empir-
ically derived probability distribution assumes that data are available
to generate the distribution, while the theoretical distribution is

represented by a mathematical function or rule. Theoretical distribu-

80
tions are preferred for use in computer simulations.

The gamma family of probability distributions was selected for use
in the advertising sector of the model. As Zehna points out, the gamma
family is so extensive that it "is a fairly safe assumption to make as
a model for an experiment described by almost any nonnegative random
variable." (1970, p. 148) The gamma distribution has two parameters,

k which is the number of successes per interval or unit space, and a
which is the reciprocal of the average number of successes per interval.
One of the most powerful properties of the distribution is its ability
to change shape from an extremely skewed exponential distribution to a
near normal distribution by changing only the k parameter. Figure 10

shows the effect of changes in k on the shape of the distribution.

 

 

 

FIGURE 10

THE GAMMA DISTRIBUTION

The computer function for generating deviates from a gamma distribu-
tion developed by Pritsker and Kiviat (1969, p. 99) was employed in the
model. Actually, because the cumulative distribution function fOr a
gamma distribution cannot be formulated explicitly (Naylor g_ugl., 1973,

p. 88), the k parameter is limited only to integer values and an Erlang

81
distribution results. The equation for the Erlang distribution used by

Pritsker and Kiviat is:

1 -
1.x”) 3 Wm)" ‘ exp(-px) x>0
= O . otherwise
The function itself, called ERLNG, with its FORTRAN source code. is con-
tained in Appendix B. It employs the inverse-transformation method to
calculate the random deviates and the CDC Minnesota FORTRAN pseudorandom
number generator function RANF.

The PARAM Array. The ERLNG function of Pritsker and Kiviat requires
the establishment of a two-dimensional array PARAM which defines parame-
ter values for each probability distribution to be generated. FOr the
MASSMOD simulation, this required 128 rows of information to accomodate
eight different media buying behaviors (newspapers, magazines, network
radio, spot radio, local radio, network television, spot television,
local television) for 16 different years. For each medium for each year,
a row contained the mean of the distribution, i.e., the average media buy
per advertiser among the top 100 advertisers, divided by k; the minimum
value of the distribution; the maximum value of the distribution; and the
desired value of k.

The need to examine advertising expenditures at three different
levels for broadcasters was suggested by Stuart (1976) who found that
television's effect on radio was most profound at the network sales
level, since network radio was the principal medium for which adver-
tisers substituted television. Spot and local sales continued to rise,
but at growth rates lower than pre-television rates. (Stuart, 1976,

p. 136) Newspaper advertising was considered at the local level only,

since 85 percent of newspaper advertising is lbcal or regional. (Udell,

82
1978, p. 28) The Federal Communications Commission's policy of localism,

manifested both in its assignment of UHF frequencies for television
stations (Noll, Peck & McGowan, 1973, p. 103) and its limitation on the
power of FM radio stations (FCC Docket #6051, 1945) suggests that these
media would be used to a greater degree for local advertising than for
national spot or network advertising.

Minimum and maximum values, as well as the mean of the distribution.
were obtained from appropriate annual Advertising Age top 100 listings.
Estimates for the value of k for each medium were made by examining out-
put from the runs of the function using various k values, and comparing
them to actual historical data. Values of k used in the model for each
medium are given in Table 8.

Means for each distribution for each medium were estimated with
regression equations using appropriate variables from the print, broad-
cast and consumer sectors of the model.

Average advertising expenditures on newspapers at time t (AVNEW)
are a function of newspaper circulation and the number of newspapers in
the previous year (t-l):

AVNEW=24.285 + 0.0026*NEWSCIR(t-1)
- 0.067*NEWSNO(t-1)

Average expenditures on magazines (AVMAG) also depend on circula-
tion and number of publications but are responsive to the number of
households with VHF television receivers during the previous year. This
is not unexpected since general circulation magazines competed with
television for the mass audience advertising dollar:

AVMAG=9.194 + 0.000018*MAGCIRC(t-l)
+ 0.000070*VHFHH(t-1)
- 0.011*MAGNO(t-l)

Advertising behavior for radio buys is broken down by the industry

Media Buy
AVNEW

AVMAG

AVNRD
AVSRD
AVLRD
AVNTV
AVSTV
AVLTV

83

TABLE 8

PARAMETERS AND VARIABLES USED TO CALCULATE
AVERAGE ADVERTISER EXPENDITURE BY MEDIA

Variables
NEWSCIR(t-l), NEWSNO(t-l)

MAGCIRC(t-l), VHFHH(t-1),
MAGNO(t-1)

AMHH(t-l). VHFHH(t-l)
AMHH(t-l).VHFHH(t-l), RNYR**2
SQRT(RADNO(t-1))

VHFHH(t-1), RNYR**2
VHFHH(t-l), RNYR**2
TVNO(t-1), RNYR**2

|7r

N

0105000101

84
into network buys (which decreased in size as television took over as

the mass advertising medium), spot buys (made by national and regional
advertisers on a station by station basis) and local buys (made by local
advertisers and merchants).

Average network radio advertising expenditures (AVNRD) at time t
are determined by the number of households with AM radio receivers at
t-l and also by the number of households with VHF television receivers:

AVNRD=5.39 - 0.000043*AMHH(t-1)
- 0.000068*VHFHH(t-1)

Average spot radio expenditures (AVSRD) also are determined by the
number of AM and VHF households during the previous year. In addition,
the square of the growth variable RNYR appears in the significant
equation:

AVSRD=2.45 - 0.0000077*AMHH(t-1)
- 0.0000035*VHFHH(t-1)
+ 0.012* (RNYR**2)

Average local radio expenditures (AVLRD) at t are a function only

of the total number of radio stations, both AM and FM (RADNO), at t-l:
AVLRD=2.881 + 0.046*SQRT(RADNO(t-l))

Average expenditures for television advertising are similar to those
of radio, both in their categories and their computations. Network
(AVNTV) and spot (AVSTV) purchases in year t are functions of the number
of VHF households during the previous year and the square of the growth
variable RNYR. When UHF households was added as an additional variable
in the equations, the regression was not significant.

AVNTV=2.95 + 0.0044*VHFHH(t-l) - 0.044*(RNYR**2)
AVSTV=1.20 + 0.00011*VHFHH(t-1) + 0.0095*(RNYR**2)
Average local television advertising expenditures (AVLTV) are

determined by the total number of television stations, VHF and UHF,

85
(TVNO) at t-l and the squared growth variable:

AVLTV=1.56 + 0.009*TVNO(t-l) - 0.0080*(RNYR**2)

Calculation of Advertising Expenditures. Once values are stored
in the PARAM array. the actual calculation of advertising expenditures
can occur. Since all assumptions in this sector are based on the behav-
ior of the top 100 advertisers, the total expenditures per medium for
each year are computed by summing the results of a call to the ERLNG
function 100 times. The total available budget for advertising expend-
itures on newspaper, magazine, radio and television space and time is
computed as a function of retail profits (RETPROF) during the previous
year and the year index RNYR:

BUDGET=4474.8948 - 0.00l3*RETPROF(t-l)
+ 413.43*RNYR

Expenditures are checked after each call to ERLNG, and if the expendi-
tures exceed the budget, the buying process stops before 100. Once the
total buys are calculated, network, spot, and local categories are com-
bined for radio and television to produce a dollar total for all adver-
tising, regardless of type, on the particular medium:

RADTOT=TNRDBUY + TSRDBUY + TLRDBUY

TVTOT=TNTVBUY + TSTVBUY + TLTVBUY

Since no industry figures exist which split the advertising revenues

between AM and FM radio stations or VHF and UHF television stations, two
assumptions were made. AM radio was assigned 80 percent of all radio
advertising expenditures; FM radio was assigned 20 percent. All tele-
vision advertising expenditures were assigned to VHF television until
1952; after 1952, 75 percent of all buys were attributed to VHF and 25
percent to UHF. Since determining this division of expenditures was not

a main purpose of the model, these constant assumptions were used across

86
all years of the model. Obviously, a more responsive non-linear function

could be determined and included in future models to make this estimation

more precise.

The Consumer Sector

This sector generates the percentage of U.S. households having AM
(AMHHPC) and FM (FMHHPC) radio receivers and VHF (VHFHHPC) and UHF
(UHFHHPC) television receivers for year t, as well as the percentage of
the adult population subscribing to newspapers (NEWSPC) and magazines
(MAGPC). Percentages are converted to households and persons using
calculations of the number of U.S. households (POPHH) and the total
population (POP) for each year and output to the media subsectors and
the advertising sector. Annual birth (BIRTHR) and death (DEATHR) rates
and the average annual household size (HHSIZE) are exogenous variables
to the subsystem. Audience percentage calculations are based on informa-
tion input to the sector by the media subsectors, such as number of
magazines, newspaper cost or number of AM radio stations, or on the
exogenous variables disposable income (DISINC) or number of AM sets
available (AMSETS). No policy variables operate in the sector. A block
diagram of the subsystem is given in Figure 11. Table 9 presents a
glossary of variables used in the sector.

Population data was obtained from Historical Statistics, which
provided annual estimates of the population, annual estimates by age
(adult was defined as 18 and over), households by number of persons, and
the number of births and deaths per thousand population. (Both birth
rates and death rates for a given year were based on the population in
that same year; data were modified to use the previous year's population

as a base for each of calculations in the model.) Newspaper circulations

NmmFumaod PF
F

«Scum mmznmzouulzgwfin goofi—

 

 

I
mafia 303 .5 an. 8

dueraméd o...
N

/
mdohumwaaw ._M<JO(0~_D 0.?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

moCPZMUWMQ .
, «In—WE

 

 

 

 

 

 

 

 

 

 

 

 

2 mm :2...
.2235 .583»:
.65 «222.535
ELENA.
073
.c. mearzmuamd
(Bus.
A. 0.3.12”:
nmuzmdst
Fundamen—
53935
h. 8522 .F t. «:28
I :68 _u... r don. w
+

 

 

 

 

 

Lb urbem

87

Variable

ADULTPC

AMHH

AMHHPC

AMNO

AMSETS

BIRTHR

DEATHR

DISINC

FMHH

FMHHPC

FMNO

HHSIZE

MAGCIRC

NENCOST

88

TABLE 9

GLOSSARY
CONSUMER SECTOR

Description

Percentage of U.S. population l8
years and older

U.S. households with AM radio
receivers; in thousands

Percent of all U.S. households
with AM radio receivers

Number of commercial AM radio
stations

Number of AM receivers produced
annually; in thousands

Annual U.S. birth rate; total live
births per l000 population

Annual U.S. death rate; total num-
ber of deaths per 1000 population

Disposable personal income, per
capita

U.S. households with FM radio
receivers; in thousands

Percent of all U.S. households
with FM radio receivers

Number of commercial FM radio
stations

Average size of U.S. households
Magazine circulation for general
and farm magazines; in thousands

Annual average subscription costs
per person for newspapers

Source

calculated
Sterling & Haight.
p. 367

Sterling & Haight,
p. 367

Sterling & Haight,
p. 43

Electrical Mer-
chandising

Historical Statis-
tics 85-10

HistoricalStatis-
tics Bl67-l80

Sterling & Haight,
p. ll4

calculated
Sterling & Haight,

p. 205

Statistical Ab-

 

stracts No. 35

Sterling & Haight,
pp. 342-343

calculated

NEWSNO

NEWSPC

POP

POPADLT

POPHH

UHFHH

UHFHHPC

UHFNO

VHFHH

VHFHHPC

VHFNO

89

TABLE 9 (cont‘d)

Number of daily newspapers,
morning and evening

Percent of adult population
purchasing daily newspapers

Annual population of the U.S.;
in thousands

Population of persons l8 years
and older in U.S.; in thousands

Number of households in U.S.;
in thousands

Number of households in U.S.
with UHF receivers; in thousands

Percent of U.S. households with
UHF receivers

Number of commercial UHF tele-
vision stations

Number of U.S. households with
VHF television receivers; in
thousands

Percent of U.S. households with
VHF receivers

Number of commercial VHF tele-
vision stations

Sterling & Haight,
p. 20

calculated
Historical Statis-
tics A23-28

Historical Statis-
tics A29-42

Historical Statis-
tics A335-349

Sterling & Haight,
p. 372

Sterling & Haight,
p. 372

Sterling & Haight,
p. 49

Sterling & Haight,
p. 372
Sterling & Haight,
p. 372

Sterling & Haight,
p. 49

90
are those listed in Historical Statistics fOr all daily newspapers

(Sterling & Haight, l978, p. 333); magazine circulation figures are those
for the total number of copies sold of general and farm magazines, re-
ported by the Magazine Publishers Association (Sterling & Haight, l978,
p. 342). Percentages fbr both magazines and newspapers were calculated
as the number of single copies purchased per 100 adults.

The number of households with AM radio receivers are drawn from
U.S. Bureau of the Census figures, reissued by the Radio Advertising
Bureau (Sterling & Haight, l978, p. 368). No such figures were avail-
able from similar sources for FM receiver households. Scattered in-
house publications from the National Association of Broadcasters library
provided FM figures for 1948, 1952, 1954, and l959. Percentage figures
for VHF receivers are those of NBC Corporate Planning as reprinted
annually in Television Factbook (Sterling & Haight, l978, p. 372). UHF
percentage figures are those reported by the Advertising Research
Foundation (Sterling & Haight, 1978, p. 372).

Disposable personal income statistics are based on Bureau of
Economic Analysis calculations appearing in Historical Statistics.
Disposable income is personal income that remains after all tax and other
obligatory payments to government (Sterling & Haight, l978, p. ll4).
Number of AM sets produced annually is data drawn from annual tables in

Electrical Merchandising.

 

Population calculations for the sector are as fbllows, with all
figures in thousands of persons or households. Total U.S. population
fbr each year is the population of the previous year minus the deaths
in that year, plus the live births in that year:

POP(t)=POP(t-l) + (BIRTHR (t)*POP(t-l)
- DEATHR(t)*POP(t-l))/1000.

9l
Birth and death rates are given as number of births or deaths per

thousand persons. Since population figures in the model are already in
thousands, a simple multiplication produces numbers of persons born and
dead. These numbers must then be divided by 1000 to maintain the re-
porting base.

The number of U.S. households fbr any year is calculated by dividing
the total population by the average household size for that year:

POPHH(t)=POP(t)/HHSIZE(t)

To derive the number of adults in the population from the total popu-
lation, without employing a sophisticated time delay function with popu-
lation maturation rates fbr various age groups, which was judged too
sophisticated and computer-time and space consuming for this model, adult
population percentage data was regressed with the year of reporting
(l945 equals 0; l960 equals l6) to yield the following relationship:

ADULTPC(t)=0.709 - 0.0045*RNYR
This percentage was then used to compute the number of adults:
POPADLT(t)=ADULTPC(t)*POP(t)

Household percentages for the various broadcast media are based on

the following equations:

AMHHPC(t)=l.039 - 0.000013*AMNO(t)
- 0.0000086*AMSETS(t)

Hence, the percentage of households with AM radio receivers is determined
by the number of AM stations in operation and the number of AM receivers
produced during that year. Number of households with AM receivers is a
simple calculation:
AMHH(t)=AMHHPC(t)*POPHH(t)
Percentage of households with FM receivers was a more difficult

relationship to express, since so little data about FM households were

92
available. The best regression equation used disposable income and the

natural log of the year index RNYR as predictor variables:

FMHHPC(t)=0.043 + 0.093*ln(RNYR)
- 0.000018*DISINC(t)

Number of households with FM receivers was reached as:
FMHH(t)=FMHHPC(t)*POP(t)

Percentage of households with VHF receivers was best predicted by
disposable income and the cube and square of the growth index RNYR.
Other variables, such as number of VHF stations, number of television
receivers produced and cost of receivers were entered in the equation,
but were not significant or produced a singular matrix, from which no
unique solution could be derived.

VHFFHPC(t)= -l.058 - 0.004*(RNYR**3)
+ 0.00045*DISINC(t)
+ 0.0097*(RNYR**2)
Number of VHF households follows:
VHFHH(t)=VHFHHPC(t)*POP(t)
UHF household percentages begin to be calculated in 1954. The
growth curve for UHF receivers could not be fit by one regression equa-
tion. A growth factor, based on the UHF year index, is added to the
UHFHHPC equation from 1957 onward:
GROWTH= -0.039 - 0.0014*UNYR + 0.00144*(UNYR**2)

The complete equation for UHF television household percentages is:
UHFHHPC(t)=0.0081 + 0.042*ln(UNYR) - GROWTH

UHF households:
UHFHH(t)=UHFHHPC(t)*POP(t)

The percentages of newspapers and magazines sold when compared to
the total U.S. population are based on disposable income, number of

publications, and in the case of newspapers, cost.

93
NEWSPC(t)=0.463 + 0.000050*NEHSNO(t)

- 0.0033*NENCDST(t)
+ 0.000018*DISINC(t)
Circulation for daily and Sunday newspapers is calculated as
follows:
NEHSCIR(t)=NEHSPC(t)*POPADLT(t)

Magazine percentages were derived from this equation:

MAGPC(t)= -0.536 + 0.0058*MAGNO(t)
+ 0.00022*DISINC(t)

The circulation for magazines follows:
MAGCIRC(t)=MAGPC(t)*POPADLT(t)

Calculations in the sector are based on historical data in all
cases and hence are deterministic. Initially, a stochastic sector deal-
ing with the allocation of consumers' discretionary time on media and
the media expenses of consumers was envisioned. Robinson and Converse
(1972, p. 198) fOund that areas of greater television set ownership
showed greater amounts of primary activity time devoted to television,
and greater amounts of secondary viewing as well. However, they sug-
gest no quantitative relationships, nor the interrelationships among
other media and their time allocation. Block (1975) developed a sto-
chastic computer simulation called TIMMDD which allocated discretionary
time among a variety of activities, including media use. Two key prob-
lems kept a consumer time allocation subsystem from inclusion in the
model. First, the highly aggregate level of the model deals with the
total U. S. population as if it were one entity and does not allow for
a categorization of media users into heavy or light, in the manner of
TGI, or high and low, reader/non-reader. To accomplish this type of
approach, habits of a hypothetical population would have to be created,

in a manner similar to that of Gensch (1973) or Block (1975). This

94
would add greatly to the complexity of the model.

However, even if such a simulation like TIMMOD were added to MASSMOD,
the other key problem still remains: the conversion of amount of minutes
allocated to media to number of persons and number of households for ad-
vertiser and media financial calculations. The complexity of this con-
version is created in part by the aggregation level. Perhaps in fUture
models which are run at a more microscopic level, such considerations
can be taken into account. The decision was made to use a completely
deterministic calculation of population media behavior in the initial

version of MASSMOD.

Validation of the Benchmark Model

As mentioned previously in this chapter, the final step in creating
a simulation model is the validation. Since this was a model created
from historical data to describe a specific time period, postdictive
validation procedures were used. A goodness of fit test, Theil's U
(discussed earlier in this chapter in the Methodology and Data Collection
section), was used to determine how well the data generated by the
equations and relationships contained in the model corresponded to
actual historical data. A Theil's U value of 0 indicates perfect agree-
ment between predicted and actual data; a value of 1 indicates the
results are no better than chance. A cut-off value of 0.2, suggested by
Theil as adequate for predictive ability, was used to evaluate the
results.

The creation and validation of a simulation model is a cyclical
process. Regression equations are derived from historical data to pre-
dict values for both the time period under study and future periods for

each subsystem. Variables are selected as predictor variables for the

95
endogenous variables under consideration and regressed. The overall

correlation coefficient (R-square) and the F-ratio for the equation are
examined for significance, as are the individual coefficients. When the
results are not acceptable, variables are removed, added, or presented
in another non-linear fbrm until a satisfactory equation is developed.
The values generated for endogenous variables by those equations is then
compared to the historical data and subjected to a goodness of fit test.
If the results are not acceptable, a new regression equation is sought
which meets the tolerance level of the validation test. Once each sub-
system has been validated individually, they are combined into one model,
and all data generated by each subsystem are passed as input and output
to the various sectors. Again, the values generated by the model fbr
various variables under study are matched with their historical actual-
ities to determine validity. If a variable fails to validate, further
modifications are made on the prediction equations, and the validation
procedure begins again.

Hence, the results presented in this section are the products of a
fair amount of trial and error and refinement, seeking equations and
parameters which not only satisfy explicit and intuitive theory, but
also the quantitative rigor of the goodness of fit test. The MASSMOD
benchmark model validates on all endogenous variables, and is thus
assumed to be a valid description of the mass media industry from 1945-
1960. It is this benchmark model that was manipulated in experimental
runs described in Chapter 4.

Tables 10 through 17 display the validation data fOr key variables
on a subsector by subsector basis. Table 10 contains results for the
newspaper subsector. The model's prediction of number of newspapers

compared to the historical data yields a U of 0.006; annual newspaper

com.ocnm
oow.~mmm
oom.mmmm
cow.mmmm
och.mamm
oom.~cmm
occ.~om¢
oo~.mom¢
oom.~ome
ocm.~omv
com.~mme
oom.m¢~v
ooo.oooo
oom.ommm
ooo.oooc
ooo.oooo

_mspa<

>mmm3mz

momooomo.

om~.mN~m
«mm.vo~c
Nam.mmmm
¢m~.m¢mm
«mm.¢pmm
pmm.mmmm
me.wo¢m
owm.~mpm
¢o~.m~om
mmm.omme
mom.mm~¢
mam.~mm¢
mm¢.m—¢¢
meo.mom¢
Nam.owo¢
mom.mmmm

uapa_uaea

omm.m~
co~.m~
opm.-
coo.c~
c-.-
cpm.m~
o-.om
om¢.a~
omm.m~
oco.mm
o~m.-
oo~.m~
omm.m~
omm.-
oom.mm
omo.mm

a

u

hmouzmz

mammoomo.

om~.-
omw.n~
oom.-
omo.m~
oup.mu
oom.m~
ooe.m~
c-m.m~
opm.mm
o—n.m~
oem.m~
o~m.m~
oeo.m~
om~.m~
oom.m~
o-.om

mau_vogm

«Chuummam mmm<mm3mz

<h<o zouh<oH4<> x¢<zxozmm

op ugm<p

ommp
emmp
mmwp
camp
«mm-
pew—
oom—
msmp
swap
mnwp
omm_
emmp
smmp
msmp
Nnmp
omwp

_mzuo<

umpa_caea

ozmzmz

mvopmmoo.

www—
www—
mmwp
enm—
pemg
amm—
oemp
mem—
comp
ommp
pomp
pump
mmwp
mmmp
mow—
pmwp

2 m.4~mzh

comp
amm—
mmmp
nmmp
mmmp
mmmp
«map
mmmp
Nmmp
Pmmp
ommp
mem—
mem—
sump
oemp
memp

 

emm>

96

Au_u=ouv o_ u4m<~

mmm.eon_
om~.wmmp
w~¢.omo_
e~m.oosp
oom.~e~—
op—.Nmm—
www.mwmp
mvo.ommp
mo—.mme—
mpo.mmmp
mmm.oomp
wem.m~mp
ooo.oooo
Fem.—vop
ooo.ocoo
ooo.cooo

pmspu<

ovme—n—o.

mm¢.mmm—
m-.m~m—
coo.~mo_
wo—.~o~p
www.moup
Nom.momp
mem.~mm_
pwm.o~m—
mm¢.omv—
omm.m~¢—
mmm.¢oep
m—_.~mm_
www.momp
om—.moop
use.mmm

www.mom

umpu_umea
pmumzaz

: m.4~m:h

comp
mmmp
mmmp
sma—
0mm—
mmm—
«map
mmm—
Nmmp
_mmp
ommp
mem—
memp
swo—
oemp
mvmp

cam»

97

98
subscription costs a U of 0.03; newspaper revenue 0.05 and newspaper

publication costs 0.02.

In the magazine subsector (Table 11), comparisons of model pre-
dictions with actual time-series data fbr number of magazines produce a
U of 0.02; for annual subscription costs 0.01; and fbr magazine revenue
and publishing costs, 0.02 and 0.02, respectively.

Broadcast subsector validations are presented in Tables 12-15. AM
radio data (Table 12) yields a U of 0.03 fbr number of AM radio stations;
a U of 0.02 for revenue, 0.02 fbr expenses and 0.13 for earnings. FM
model data (Table 13) produces the fbllowing values of U: fbr number of
FM stations, 0.11; for FM station revenue, 0.12; for FM expenses, 0.06;
and for earnings 0.14. In the VHF television subsystem (Table 14),
Theil's U calculations generated a 0.06 for number of stations, 0.11 for
revenue, 0.11 for expenses and 0.13 for earnings. Table 15 contains data
for the UHF sybsystem, which began predicting values for variables in
1953. Number of UHF stations validates with a U of 0.01; revenue, 0.09;
expenses 0.02; and earnings 0.13.

Table 16 contains data fbr total buys by the top 100 advertisers by
medium as generated by the stochastic advertiser sector and as reported
by Advertisinngge. Total adviitising buys fbr newspapers predicted by
the model produced a U of 0.08 when compared to actual data. For maga-
zine buys, U was 0.04. For the broadcast media, AM buys yielded a U of
0.03; FM buys 0.03; VHF buys 0.12; and UHF buys 0.09.

Consumer sector data is presented in Table 17 for both population
data and various media audiences. The Theil's U fOr total population
figures is 0.01; for household populations, 0.005; and for the adult
population, 0.01. For newspaper circulation, U is 0.01, while for maga-

zines it is 0.02. Number of households with AM radio receivers produced

oe~.~mom
oom.mmmm
om¢.mmvm
ooo.oooo
ooo.oooo
oco.oooo
ooo.~mmm
ooo.oooo
ooo.oooo
ooo.oooo
o-.mo-
ooo.oooo
ooo.ooco
chm.omo~
ooc.oooo
omm.mmmp

_msau<

>m¢¢<z

omemmmpc.

mow.mpo~
mm~.mmo~
mom.¢w¢~
nnm.m¢m~
cwo.mom~
~cm.m¢¢~
¢m¢.~—¢N
cem.ome~
mam.om-
Non.ommm
mmm.¢mpm
~¢m.¢mpm
m~_.mmom
cow.mo-
mm~.—~—~
¢o~.a¢cm

capu_uae¢

ov~.~
coo.“
cm¢.~
om~.n
cm~.~
om~.~
onm.m
oo~.~
oo—.~
ocm.o
om~.~
omw.m
omm.m
cum.m
ooo.o
ooo.o

_asaa<

wmmmowpo.

om¢.~
om¢.~
oam.~
omm.~
omm.~
opm.~
ooN.~
omm.m
cp~.~
oo—.m
owo.~
omo.~
ooo.~
omm.o
o—o.m
oeo.~

napaccaca

hmoum<=

«chommmzm mz-<w<=
<H<o zo~h<o~4<> xm<z=ozmm

_— m4m<h

mum
emu
CNN
mum
mmm
mmm
mmm
mmw
NmN
new
omw
new
mmm
new
mmm
mpm

penuu<

mmmmommo. = m.4~mzh

«mm
«RN
New
mew
new
mom
mmm
mom
mmm
mmm
pmm
new
«cm
New
mcm
mmm

capu_caea

oom—
amm—
mmmp
«mop
amm—
mmmp
«mop
mmmp
Nmmp
me—
ommp
mvmp
mvmp
Nem—
camp
memp

 

cam»

99

 

100

upomommo. 2 m.4~mzh

o~m.omm~ pmc.mmw— oom—
omo.¢¢m— eno.mpwp ama—
oom.~m~— mm~.omm_ mmmp
onm.mo~p eec.mmm~ swap
omo.amm~ N—m.mmmp mmap
oce.emmp mn¢.mwo— mmm—
oom.ocop mmm.~mo— emo—
om—.¢Fo_ mom.m~op mmmp
ooo.mom— wmo.ommp Nmmp
o¢m.—~m— mpo.mvm— —mmp
omm.mme— opm.vm¢— amm—
opn.mmep msm.movp mem—
omm.mmmp o~e.mmm— memp
05¢.Nmm— mup.—~mp Kemp
ooo.oooo mm~.mmmp ovmp
om¢.¢opp mmm.mm¢— mem—
szuu< UwHUPUmLm me>

Hmumzmz

Au.u=oav _P msm<c

 

101

NNNPBNNO. mm-~_mo. somomemo. = m.sHm=h
oo~.mom Now.mo~ ooo.o~m “mm.c~m omen o~¢m comp
oom.mc~ omw.em~ coo.a~m mac.eom cmmm mmmm mmmp
co..mm~ www.meu ooe.~a~ ~¢m.m¢~ ompm oo~m mmm_
com.~_k mom._m~ oom.m¢~ Nom.¢¢~ moon «_om “mm.
oos.mme Nam.m_~ co_.¢c~ cm@.mm~ eNmN “Nam 8mm,
oom.~mm mwc.mo~ oom.m¢~ 5mm.¢o~ seem “mew mmm.
oom.~mo omm.~mm com.umo -_.om~ _NmN oemm qmmp
coo.m.~ smo.o~o ooo.~cm www.mhm Foam mmmm mmm.
oo~.eoN mpc.mmm oom.mom oum.m_m .mmm ~m_~ «mm_
oo~.mmo mom.¢mm ooo.mm~ m.~.~m~ NmNN oe¢_ _mm_
oo_.No~ m_m.omo oom.m~m _Nm.m~m omom Pam. amm—
oou.aho “mo.~¢o oc~.mm~ osm.Fm~ _Nm_ ~w~_ mem—
oom.mem _mm.mmo oom.mo~ cum.._m _~o_ moo, mam.
oom.emm mmo._~o oom.om~ mm~.¢uu woo. mm¢_ Resp
oom.omm mmm.amm ooa.¢m~ mn~.m_~ new om__ mem_
oou.m~w cam.mmm ooo.omh _mw.wm~ spa omm mew.
.Hmmmmm .mmmmmmwum .flmmmmm mmmmmmmum .flmmmmm uapu_uaeg emm>

axmz< >m¢=< ozz<

«chummmam o~o<m z<
<H<a zomh<om4<> xm<zzuzmm

Np m4m<h

 

102

—mm~¢mmp. a m.4~m:h
oom.mo mmm.mm cem—
oow.mm mop.mm mmmp
oom.cm Noo.~m amm—
ooo.em www.mo mmmp
oo~.m~ o-.—w ommp
ooe.mu mem.mw mmmp
ooo.om mmm.mm cmmp
o0¢.mm mmo.oop mmmp
coo.mo_ mmm.pmp Nmm—
oom.oo— Nmm.—N— —mm—
oo~.n~— mmm.¢v_ ommp
ooo.~op Fem.—mp mews
oo~.omp mmo.m¢— wemp
oom.¢¢— wN—.em~ Kemp
oom.¢m— www.cp— oem~
oom.-m em~.m¢p memp
posuu< cmuupvmgm me>

zm<uz<

Av.u:ouv Np mam<h

 

103

msmvmmmo. mmoomm—P. —o~p¢wop. = m.a~m=b
oom.pp opo.~— oo¢.m mmu.@ wmm mum oom—
ocn.m ~P~.w oo¢.c -¢.m mmm mmm amm—
oom.¢ o¢¢.m com.m opo.m mmm owe amm—
oom.m mom.m ocF.m —om.~ omm ncm sma—
oom.~ mem.~ oo~.~ mm¢.~ oem mum mmmp
oom.~ om—.~ oom._ mo—.N Nmm cmm mmmp
oom.~ moe.~ oom.p pmm.— com «mm «mop
oom.~ vwo.m ooe.p cum.p 0mm —om mmmp
oom.m mo—.¢ oom.p cmo.m mmo mom Nmmp
oom.m mom.m oop.~ o~o.~ mum com Pmm—
oom.n N¢N.m oco.~ smN.m mmm ems ommp
oom.m emp.m coo.m mmm.~ com was memp
oom.m smm.m oo~.~ omm.~ mme mmc mem—
ooo.o m¢¢.op ooo.o m~m.— oe— —pm Kemp
ooo.o Ncp.~_ ooo.o mom.o ac mop cemp
ooo.o m~¢.~p ooo.o mam.p we pm mcm—
pmauo< nouowumgm —m:uu< kuuwvmgm _mauu< uquwvmgm Lom>

mxmzm >umzm ozzm

mahummmzm oHo<m Zu
<h<c zo~h<omd<> xm<zzuzmm

mp m4m<h

Aw.p:ouv mp u4m<h

oom.mn
oom.Nu
coo.pu
con.on
oom.on
oo~.cu
coo.pn
oom._u
con.—u
oop.m..
oom.¢n
oom.ou
cow.mu
ooo.o

ooo.o

ooo.o

pmzuu<

uapu_umea

zm<mzm

omempmm—.

mum.mu
om~.~u
mm¢.~u
emu.—i
opfio:
«no.0:
mmu.ou
mmp.—u
Nuo.Pu
www.mu
mm¢.mn
mon.mu
nme.on
Nm¢.mu

mmm.o—u

www.mu

= m.4~m:h

comp
mmmp
wmmp
mmmp
ommp
mmmp
emmp
mmm—
Nmmp
_mm—
ommp
mcmp
mem—
sem—
mem—
mqm—

 

me>

104

 

105

nmpemmop. mmewmoop. emmmmwmo. a m.4~u:k
oom.~mm «pm.mm~ oop.mom omo.¢mm oec ewe comp
oom.mwm mmm.cmo oom.m~m c—N.emm mme m¢¢ mmmp
oop.pmm mom.omo ocu.mmm mmm.eom P_¢ 0mm wmmp
oo¢.n¢m mmm.mmo com.omm o—p.wwm pmm omm mmmp
ooe.~om mam.moo oc—.—~o mm¢.wm~ ¢vm Pmm ommp
oom.m—¢ pN~.m~m ooo.—om mme.meo mom me mmmp
oo~.omm m—m.m~¢ coo.nm¢ mo¢.~mm mmm mmm «map
ooo.m- ~N~.mmm oco.m~m mco.o¢¢ amp mom mmmp
cam.mme omw.—mm ooo.mmm ¢mo.mmm mo— 00— Nmmp
oo~.mmm mmP.m~¢ com.—p¢ www.mme mop amp —mm~
oom.¢p~ mmm.~m~ oom.~m~ mmo.mom mm pm ommp
oom.mpp mom.~¢~ co~.mo e—m.mm~ pm mm memp
coe.e¢ mm~.opp oov.op mmm.m__ up mm wemp
ooo.o ooo.o coo.o ooo.o N— pp hemp
oco.o ooo.o ooo.o ooo.o 0 mp oemp
ooo.o ooo.o ooo.o ooo.o w o mvmp
—m:uu< coau_uwgm —c=uu< twpowvmgm pmzpu< umuuwumga gum»

mxumz> >mmmz> ozmr>

«Chommmzm zo~m~>m4mh mz>
<h<o zouh<oua<> xm<z=ozmm

e— mqm<h

Au.u=oav ¢_ usm<h

oom.opm
oom.mmp
oo—.mvp
oom.~v—
oo~.mo_
oo—.~¢_
oom.mo—
coo.mm
oom.mm
oom.~m
oc~.npu
oo—.mvu
ooo.m~:
ooo.oc
occ.oo
ooo.oo

_a=»u<

zx<muz>

 

mmmmmmmp. = m.4~mzh
Nom.Om— coa—
mmo.m~m mmmp
ome.opm mmmp
mp—.omp swap
www.mop mmmp
mpm.om— mmm—
~m¢.mp~ vmmp
www.mm mmmp
mom.~o Nmmp
m~o.ce pmmp
N-.om 0mm—
NP¢.N_ memp
oo—.m mem—
ooo.o sem—
ooo.o mea—
ooo.o memp
vaUFumgm me>

106

 

 

 

107

cepmoomp. : m.4~m:~
oo¢.o —mm.mu comp
oo~.ou cmo.¢n mmmp
oo¢.mn mpo.mu mmmp
ooe.mu —mo.~u NmmF
coo.mu —om.~u mmmp
oo¢.~u own.mu mmm—
com.opu mm~.w—u ema—
oon.opu Nom.opu mmmp
Pm=HU< bmvuwflwgm gmm>

zz<umzn
ccempmmo. mummmmmo. Nomoemoo. a m.4~m:p
ooe.¢¢ omm.—¢ ccm.¢¢ moo.o¢ mm as comp
oom.~¢ omm.me oom._¢ pao.N¢ um ¢n amm—
ooa.me mmc.m¢ oom.mm mmu.~e em mm wmmp
com.o¢ mn¢.oe oo—.p¢ pom.¢¢ om mm “mop
oo~.¢m mmm.m¢ con.pm mwp.m¢ mm mm ommp
oo—.¢m mom.mm oo~.m¢ www.mm ep— epp mmmp
ocm.mm —¢m.mm oom.N¢ moo.mm pmp FN— «mm—
oo¢.m~ ou~.m~ co~.~— pom.wm c o mmmp
—m:uu< vaUFumgm pmaau< vouuvumgm —w=go< umuupuwga Lcm>

mxmuza >mmuzn czar:

achoum zomm~>MAMH m3:
<H<d zo~h<o~4<> x¢<zzuzum

m— u4m<h

 

108

moomommo. o—mco—eo. nommmmuo. : m.4~m:»
~.now o.cm~ .mmmp .owmp .mmmm .emoe comp
m.m~m c.nmn .mm~— .omep .emmm .pmme mmmp
o.~mm o.v~o .NF—p .mm—_ .mpmw .moee mmmp
m.omm o.mmo .mmpp .mmmp .wmom .cwme Kma—
m.o~m o.mmm .momp .mmmp .empm .mcme mmmp
«.mpn o.mm~ .mm—p .mm—F .¢¢om .ovme mmm—
m.m¢~ o.oon .vmop .ompp .umec .mmwm cmmp
m.m~m o.mom .moop .ommp .moee .mmom mmmp
c.0om o.mmm .pmm .vmm .memv .epmm «map
e.o¢w o.—¢m .emm .wmm .mmmm .Nmmm pmmp
N.mom c.mom .Nmm .mmm .Nomm .momm emmp
m.mmm o.—mm .pmm .mem .Pmom .nmom memp
c.0em o.¢om .mmm .m—m .cmmm .vemw wemp
o.m~m o.mmm .emm .mxm .oomm .mewm Kemp
o.m~m o.~wm .mwm .mnm .pmmm .mmmw ovmp
o.oom o.m¢m .mpm .nom .eevw .epom memp
—m:uo< tapewomgm Pasuo< umuuvumgm szuu< voguwcmga me>

>3mz< >=mw<zh >=mzmzh

mohumm zmmahmm>a<
<h<o zouh<oua<> xx<£zuzmm

mp m¢m<h

mN.Nmm
om.ncm
om.m~m
cm.mcm
mn.mm¢
mm.¢~¢
oo.mmm
oo.mmm
om.mm_

paspu<

>3muxa

m-mmomo.

oo.—on
oo.mmc
oo.m¢m
oo.mmm
oo.—mm
oo.vm¢
oo.omm
oo.m~m
oo.¢m~

nauu_uaea

om~.m~np
oom.mom~
oom.w~mp
ocm.opm~
om~.oo¢p
omm.ms~—
ooo.~—op
coo.pmm
oom.cwm
ooc.mnm
oom.o—m
oco.o——
coo.ooo
ooo.ooo
ooo.ooo
oco.ooo

_a=»u<

>3mmz>

momm—oNp.

coo.mop~
ooo.¢pmp
coo.p¢m~
oco.¢om—
ooo.~mo—
ooo.~mmp
ooo.~n_p
ooo.omm
ooo.mmm
ooo.moo~
ooo.mmo
ooo.mmm
ooo.~mm
ooo.ooo
ooo.ooo
ooo.ooo

umpu_umea

Au.peouv o. usm<~

m.—o~
N.em_
o.mmp
N.omp
~.omp
w.mmp
N.mm—
¢.~o~
o.mP~
m.~—N
w.mmm
N.NP~
m._pm
«.mom
¢.~om
o.m-

pozpu<

Nome-awe.

o._m—
o.mw—
o.mop
o.em—
o.mmp
o.om—
o.—m—
o.pom
o.m~N
o.opm
c.om~
o.m_~
o.m-
o.mF~
o.mm—
c.—p~

emuopuaea

: m.4Hm:H

oom—
mmmp
wmmp
mmmp
cmmp
mmm—
wmm—
mmm—
Nmmp
me—
ommp
memp
mvm—
nem—
ovo—
mvmp

 

Lcm>

109

 

110

emecoooo. -¢~meoo. usmeoooo. a m.s_mzh
om5¢__ m¢m~P_ o_m~m ouemm omm~N_ m_mm~_ comp
mmms.. meow.. Nompm e-_m mush“. ween“. mmm.
mm_mp_ «N_P__ Nmeom mmmme _epenp come“. mmm_
mo_~_. ~o~c.. memae “meme ¢-P~_ mmmmo_ “ma.
wm¢o__ Newmop mmums osmme _Nmmo_ magma. omm_
momma. o-mop amuse _mmue mummo. Roomm_ mmm.
mnemo_ N¢-o_ mamas mNmoc _mmmo_ «omoo_ «mm_
msmno_ -~8o_ mwmme among mommm_ m_m~m. mmm.
mmeoo_ mmmmop eoeme ma_m¢ «mmcm_ _memm_ ~mm.
mummOP names, eomqq “wees ~m~¢m_ moomm_ _mm_
emcee. Nmemop mocme oeeme «mo_m_ QMNOm_ omm.
megmo. Neo~o_ ooooo Ncsms mmpme_ mamas. asap
Booms. ~m~.o_ ooooo mo__¢ _mooe_ evnce. mem—
emkoo_ _omoop ooooo memmm m~_¢¢. me.e¢_ Rem.
_ommm exams cocoa mommm mmm_¢_ om~_e_ osm—
Nnmmm magma ooooo mo_Pe m~¢mm_ mokmm_ mes.
.Hmmmmm .mmmmflmmum .flmmmmm .mmmmflmmum .Hmmmma umpu_umea ema>

pso<aoa acazx aoa

«ehumm mmzamzou
<h<o zo~p<o~4<> x¢<zzozmm

up u4m<h

 

111

mmoemmpo. mmwpmewo. Pommmmoo. a m.4Hm:h
mmpom msmme «meomp Noemm— Nmmmm mmumm comp
Omvme mowwe mmmmm— m—mmmp oomwm oemsm mmmp
oomme mumme mummmp mwmosp mpemm mowcm wmmp
oomme mmmm¢ cpepm— ommmup mcmum ommom mmmp
oomoe Nmmuv panmmp nmemmp Nopnm awnmm 0mm—
oomme c—mmv mommnp mmmmmp nepom popmm mmm—
oo—me momme nmaem— muepmp msomm e—eem «map
ocmve mone¢ mmomcp pmomop uneem mommm mmm_
comme popee acmmmp upmmmp _mmmm mommm Nmop
com—e oemme com—mp em—mmp mpowm ommmm pmmp
oomoe omope ammuep —mom¢p mwmmm Nemmm emmp
oommm ommoe m—smep moouep memmm Nowpm memp
mmcnm Noemm mamoep mmeoep mmmwm Nompm meow
ocmmm swoon sommmp powmmp muopm m—mom mcmp
mommm commm mmmomp mnmmmp mmmom ovmom camp
oo—mm momnm ecu—N. pmm¢ep ewmme mmemq memp
szuu< vmvuwvwgm PQ=HU< kuuwtwgm Pmauu< umuuwumsm me>

==z< um~uu<z m~um3mz

Au.a=ouv ~_ msm<~

 

112

omoomeoo. mmmmommo. ooomoooo. o m.4_mzp

momm onom omnme omomm ooooo ospmp ooop
o—mm o—pe omome eomov ooomp —m¢~— amm—
oomm Nome mwope ooome oooo oeopp ommp
onmm Nome oooom pnepe oooo NmNP— “mop
oooo pmom ooovm oommm oooo emoop ommp
oooo mon— oomom moopm oooo uppop mmop
oooo sum oooom mpvmw oouo —¢¢o emmp
oocow coopm oooo omom mmo—

oommp eomop oomo ompo Nmop

owmop mmowp ooo noes —mmp

mmom oooo ooo ppmo omop

oco men ooo momm ocop

«up —po_ pun —omv oco—

ep opm ooo moom meo—

w o ooo mom oqop

o o ooo —_¢ meo—

pmsuu< omuupomgm pmauo< vmuupomgm —w:uo< umHUPomLa com»

:xmzo Izmz> :zzu

Au.p=ouo N_ usm<h

113
a U of 0.02; for households with FM receivers, 0.10. Households with

VHF receivers had a U of 0.05, while those with UHF receivers yielded a
U of 0.09.

CHAPTER 4

THE EXPERIMENTS

One of the major benefits of constructing a simulation model of a
real-life system is the ability to alter conditions within the model in
order to evaluate various alternative policy options or different courses
of action. As previously stated, this opportunity to experiment with the
system is useful both in theory-building, to evaluate various hypotheses
about the system and to provide more observation points, and in decision-
making, to offer probable outcomes of several alternative policy param-
nters. This chapter describes eight experimental manipulations per-
formed with MASSMDD, which sought to explore the results of different
operating parameters on the mass media industry. The experiments were
groups around two policy issues: external controls on advertising ex-
penditures and allocation of radio spectrum space for different types of
broadcasting. Other potential experiments are discussed in suggestions
for further research in Chapter 5.

The endogenous variables generated in each of the four subsystems
in MASSMOD were viewed as structure and performance criterion variables
and were monitored in each experiment to study the effect of the policy
manipulation on the simulated mass media industry. These variables
included number of boradcast stations and number of publications;
audience size and circulation; revenues, expenses and earnings for each

medium; advertising expenditures and retail profits. The following
114

115

discussion of the experimental runs includes a rationale for each exper-
iment and an analysis of output from the experimental run compared to

the benchmark model output using Theil's goodness of fit test.

Controls on Advertising Expenditures
Rationale

Proponents of external controls on advertising expenditures fall
into two main groups: those who argue that unlimited advertising can
reduce competition in a multi-brand, multi-firm industry (Kaldor, 1949;
Bain, 1956; Comanor & Wilson, 1967; Greer, 1971), and those who main-
tain that too many commercial interruptions in broadcast programs destroy
the programming itself. (Doan, 1975; Dominick, 1976) While the first
group specifically proposes a control on the amount of money a firm is
able to spend on advertising, the second group deals with a reduction in
the number of commercial minutes available to advertisers seeking to buy
broadcast time. Nonetheless, this reduction could result in a decrease
in advertising revenues received by broadcast stations, which constitute
a major share of station incomes.

In a study addressing the first concern dealing with advertising
and competition, Bloom (1976) applied constraints which represented
proposed controls on advertising expenditures posed by critics, in his
study of advertising and competition:

(1) a five percent straight tax on current-year advertising

subtracted from after-tax profits;

(2) a progressive tax on advertising of 2 percent per

million computed as follows:

rate = .02 x (advertising expenditures/1,000,000)

(3)

This
(1)

This

different

5 percent

116

a 10 percent limit on the advertising/sales ratio,

accomplished by adjusting the profit equation to

subtract 100 percent of advertising spending that

exceeds 10 percent of sales;

a 20 percent limit on the advertising/sales ratio;

a depreciation requirement, with a fast write-off

of advertising outlays:

current advertising expense = .6A(t) + .15A(t-1) +
.15A(t-2) + .1A(t-3)

a depreciation requirement, with a slower write-off

of advertising outlays:

current advertising expense = .3A(t) + .3A(t—l) +
.2A(t-2) + .2A(t-3)

study used two of Bloom's suggestions:

a straight tax of 5, 10 or 20 percent on current

year advertising, subtracted from after-tax profits;

a 10 percent limit on the advertising/sales ratio,

accomplished by adjusting the profit equation to

subtract 100 percent of advertising spending that

exceeded 10 percent of sales.

Results--Experiment 1

experiment actually consisted of three runs, each with a

percentage value for the tax on advertising expenditures (TBUY).

The tax was deducted from retail profits in the following manner, using

as an example. Taxes of 10 and 20 percent were imposed similar-

RETPROF(t) = SALES(t) - TBUY(t) - (.05*TBUY(t))

117

Since the budget for advertising expenditures for a given year is a
function of the retail profits of the previous year, it was assumed that
this reduction in profits should have an effect on subsequent media buys,
which in turn would affect the revenues of the media in the system.
However, all criterion variables in the experimental run were identical
to those in the Inenchmark model, except, obviously, retail profits.
Bloom's (1976) experimental runs with such a tax tended to make the
advertising expenditures of all brands lower (p. 81), which eventually
diminished competition in the industry. It is possible that the highly
aggregate nature of MASSMOD, several levels above consideration of a
brand or firm, masked this effect, since a decrease in retail profits
of one million dollars brought about by the tax was still only an overall
reduction of 0.3 percent. Moreover, the equation in the model which
generated the annual advertising budget displayed a negative relation-
ship between retail profits and the budget, such that a decrease in the
total retail profits resulted in an increase in the advertising spending
budget. Given ‘this structural relationship in the model, the tax on
advertising expenditures indirectly causes an increase in the advertising
budget for the next year, rather than a disincentive as was the case in
the Bloom experiments. This problem is addressed in more detail in the

summary section of the advertising control experiments.

Results-~Experiment 2

This experiment levied a 100 percent tax on all advertising expend-
itures in excess of 10 percent of sales revenues. This was accomplished
using the following equation, applied when advertising expenditures

(TBUY) exceeded the 10 percent limit (LIMIT):

118
RETPROF(t) = SALES(t) - TBUY(t) - (TBUY(t) - LIMIT))

where LIMIT = .10*SALES(t).
When expenditures did not exceed the limit, retail profits were computed
as:

RETPROF(t) = SALES(t) - TBUY(t)

Table 18 presents values of Theil's U for criterion variables in
the experimental run when compared to those variables in the benchmark
run. In this (:ase, a U with the value of 0 indicated perfect agreement
between the runs (i.e., no change), whereas a value approaching l indi-
cated a more radical change caused by the experimental manipulation.

The largest U value (only .0393) is for retail profits, an expected
result since it was the variable being manipulated. This small change
from the benchmark supports the lack of changed output in Experiment 1
in this section, as the limit which was deducted from profits here was
greater than the straight tax imposed there. Some broadcasting revenues
changed slightly, reflecting a very small increase in advertising buys

for each medium.

Summary

Manipulations of external advertising controls did not produce
substantive changes in the criterion variables when compared to the
benchmark model, eliminating a discussion of the relative merits and
demerits of each policy alternative. The manipulations did, however,
highlight two structural relationships in the model. The first is the
inverse relation between retail profits and advertising budgets, which
means that a decrease in profits actually produced an increase in the
budget for advertising expenditures. Secondly, the components of the

model are insensitive to changes introduced in the advertising sector.

TABLE 18
ADVERTISING EXPERIMENT TWO

119

THEIL'S U
Variable Theil's U
NEWSNO .0002
NEWSCIR .00002
NEWCOST 0
NEWSREV 0
NEWEARN .0038
NPUBCST 0
MAGND 0
MAGCOST 0
MAGCIRC .0001
MAGREV .0011
MPUBCST .00004
MAGEARN .0035
AMND .0007
AMHH 0
AMREV .0038
AMEXP .0002
AMEARN .0123
FMNO .0054
FMHH 0
FMREV .0237
FMEXP 0
FMEARN .0182
VHFND 0
VHFHH 0
VHFREV .0033
VHFEXP .0042
VHFEARN .0003
UHFNO 0
UH FHH 0
UHFREV 0
UHFEXP .0006
UHFEARN .0314
TBUY .0032
TNEWBUY .0038
TMAGBUY .0041
AMBUY .0048
FMBUY .0049
VHFBUY .0032
UHFBUY .0031
RETPROF .0393

120
While the former relationship seems intuitively correct and can be

supported by actual industry behavior, the insensitivity of the model's
components to changes in advertiser buying behavior seems incorrect,
since advertising revenues are a major source of income for both print
and broadcast media, and revenues affect the behavior of the entire
media subsystems.

While the original benchmark model validates on all criterion
variables using ‘Theil's goodness-of-fit test, it is at this point in
the simulation modelling process that Naylor and Finger's (1967) multi-
stage verification procedure is best employed. The two-edged question
faced by the researcher is whether the relationships expressed in the
model are in fact correct, and the media sectors arg_insensitive to
changes in advertiser behavior, or whether some statement of relation-
ship and/or selection of relevant variables is not correctly stated or
used. If, upon examination, all structural relationships seem to be
rational and correspond with reality, then the model's results should be
accepted, and interpreted accordingly. 0n the other hand, if some
underlying assumptions or equations seem less than obvious or realistic,
a reexamination of those components for possible structural misstatements
and data problems is in order, before accepting the model's output as
correct.

In the case at hand, a false assumption resulting in a misstatement
of the relationship between profits and advertising budgets, could
explain the model's insensitivity to changes in the advertising budget.
What the model computes as retail profits are not retail profits at all,
but the difference between retail sales and advertising expenditures.
The equation fails to include production or personnel costs, as well as

any other marketing costs not considered advertising. Another exogenous

121
variable, retail production costs or retail expenses, should be included

in the equation to make it accurate.

The difference between retail sales and advertising expenditures was
the figure used in determining the regression equation which predicted
the annual advertising budget. It is this equation which shows an in-
verse relationship between budget and "retail profits," (i.e., as profits
decrease, the advertising budget increases.) While this relationship in
itself can be easily rationalized, and in fact, is probably correct,
since a drop in sales might precipitate a more vigorous advertising
campaign, the parameters and coefficients do not reflect the actual
mathematical relationship since an incorrect assumption, and therefore
incorrect data, were used to derive the equation. If the direction of
this relationship is correct, however, the placement of the tax, whether
a direct tax on all expenditures as in Experiment 1, or a tax on excess
expenditures as in Experiment 2, is incorrect, as the tax should serve
as a deterrent to advertising expenditures, not an incentive as is the
case when a decrease in profits results in an increase in budget. The
advertising budget 5 hould be the first value calculated for each run
of the subsector each year. Then, on the basis of that proposed budget,
the tax should be imposed, so that the actual amount of advertising
expenditures for the year have been reduced by taxation.

Additionally, data used to derive the relationships for this sector
may have been mismatched. Retail sales figures supplied to the model as
exogenous were used for all sales made at the retail level across the
country. The behavior of advertisers in each buying category (newspapers,
local radio, network television, etc.) was based on the transactions of

the top 100 advertisers in that category, which obviously does not

122

include all retailers. However, data describing the retail sales, pro-
duction costs and other expenses which were suggested for inclusion
earlier for the top 100 advertisers are not readily available and data
on the behavior of all retail advertisers, whatever their size, are also
not available.

Hence, the experiments conducted to investigate the effect of
external advertising control policies on the model system did not yield
output which could lae analyzed and compared to determine the optimum
policy option and its impact on the media system. Instead, the attempted
experimental runs with their lack of differences in output, highlighted
a structural flaw in the model itself, which might have gone undetected
if the experiments had not been undertaken. This discovery highlights
the role of experimentation in simulation not only for investigation of
various experimental states imposed on the system, but also for the

refinement and correction of the basic model specifications themselves.

Spectrum Allocation

 

Rationale

These experiments dealt with the impact of the FCC's determination
of how various ssegments of the electromagnetic spectrum are used on the
growth of the broadcast industry. They manipulated only the amount of .
the spectrum assigned to a particular medium, and assumed that the width
of a channel within that assignment was fixed, according to FCC stan-
dards. In other words, the number of available channels was increased
or decreased by increasing or decreasing the size of the assigned band
and not by using a given band more or less efficiently in terms of

channel size.

123

This spectrum allocation policy variable was crucial to the model
in representing the regulated nature of the broadcast sector. Deter-
mining the uses to which portions of the electromagnetic spectrum may be
put, as well as licensing the use of the frequencies, constitutes the
essential regulatory power given to the FCC by the Communications Act of
1934. (N011, Peck & McGowan, 1973, p. 97)

At first, AM radio stations simply applied to the FCC for a license
in whatever community and on whatever channel they chose, as long as
they could prove non-interference with other stations in the area. This
same policy has remained in effect for AM radio, although the FCC deter-
mines at what power and during what hours stations may broadcast in a
three-tiered structure of powerful, regional clear-channel stations;
strong metro stations; and local stations.

The same type of allocation process was used initially for FM
radio stations and VHF television stations, which chose a community and
a channel on which they would provide service. Facing both a monstrous
number of applications and potential interference problems, in 1945 the
FCC issued an allocation plan which determined in advance where stations
could be built and which was based, not on a definite demand for
stations, but rather on control of interference. In addition, the FCC
shifted FM radio from its 1940 assignment in the VHF band to a new
location in the UHF portion of the spectrum. This policy was to have a
profound effect on the development of the industries. (Webbink, 1969,
p. 537)

The 1945 allocation plan moved FM radio from its assignment in the
VHF band at 42-50 MHz to the UHF band at 88-108 MHz to allow for 13
VHF television channels to be used in the top 140 markets of the country.

(FCC, 1945) The television assignment was criticized at the time as too

124

meager; members of the Radio and Television Planning Board (RTPB) com-
missioned by the FCC from the industry had recommended 25 or 30 channels.
(Sterling & Kitross, 1978, p. 233) While the assignment gave FM more
channels, it also required the manufacture and consumer purchase of all-
new FM transmitters and receivers to pick up the signals at the new
frequencies, setting the development of FM radio back at least a decade.

Inundated by the postwar flood of applications for television
licenses, the FCC issued a "freeze order" in 1948 to suspend all licens-
ing activity until a study had been made of the channel-allocation prob-
lem. (13 Federal Reg. 5860 (1948)) Realizing that 121 television chan-
nels were not adequate for a nationwide system, in 1952 the Commission
issued its Sixth Report and Order (FCC, 1953) which lifted the freeze
and provided for the use both of the VHF channels and 70 UHF channels
(470-890 MHz) set aside in 1945 for television experimentation and
future broadcasting. The Qrggg provided for the intermixture of VHF
and UHF stations within the same markets, a policy which critics said
would create for UHF a serious competitive disadvantage with the VHF
stations. (CBS, 1956, p. 792; DuMont Laboratories, 1949, p. X-87)
They suggested a policy of deintermixture, or reassignment of channels
by rulemaking to make a given community either all VHF or all UHF.

The deintermixture debate, along with a debate on the concept of
VHF drop-ins and a change in AM channel width from 10 to 9 kHz, con-
tinues today. A more radical proposal calls for a complete deregulation
of broadcasting (Levin, 1970, p. 209; Webbink, 1969, p. 537) which would

allow the economics of the market place to determine the value of

 

1Channel 1 (44-50 MHz) had been turned over to safety and special

services in 1948 in return for the exclusive assignment of the remain-
ing 12 stations to television. (Sterling & Kitross, 1978, p. 298)

125

broadcast licenses and earnings and the division of spectrum. (Levin,
1964, p. 153; Owen, 1975, p. 92) Hence, the manipulations of the
spectrum allocation policy variable have implications not only within
the historical time period of the model, but also on policy debates
today. 9
The benchmark model assumed that an unlimited number of channels
were available for broadcast stations. This was effectively the case
during the period. Six manipulations were run to address questions
raised by FCC actions.
(1) Earlier development of VHF television. To investigate
the effect of the 1948-1952 freeze on VHF television
license applications and the effect of the late intro-
duction of UHF use for television, the model was run
under the assumption that both VHF and UHF channels
were available for use in 1945.
(2) Limit on number of channels available for commercial
use. As stated before, the benchmark model was run
assuming an unlimited number of channels were avail-
able for use by AM, FM, VHF television and UHF tele-
vision. This experiment set a limit on the number
of channels which was reflected in the number of
licensed stations.

The 1952 Report and Order made available 2,053
frequency assignments for television in 1,291 com-
munities; over two-thirds of the assignments were
UHF channels; 252 frequencies were reserved for
educational use. The previous VHF—only assignment

table had made available 405 assignments in 140 metro

126
areas. (Webbink, 1969) On the radio side, the 1945

spectrum reallocation established 100 available
channels for FM, 20 of which were for educational
use, which created 3,623 frequency assignments across
the country. Prior to the reassignment, only 40
channels were available. AM spectrum use was never
governed by a table of assignments; the 107 channels
set aside for AM radio were used in 1977 by a total
of 4,497 stations. (Sterling & Haight, 1978, p. 44)

This experiment placed the following limits on
frequency assignments for each medium:

AM radio: 50 channels, allowing 2,100 stations;

FM radio: 40 channels, as allocated prior to
1945, permitting 1,450 stations;

VHF television: 12 channels, with 400 frequency
assignments;

UHF television: 30 channels, with 600 frequency
assignments.

Thus, in this experiment FM radio and VHF tele-
vision retained their assignments before the 1945

reallocation and the 1952 Report and Order respec-

 

tively. AM radio frequencies and UHF television
assignments were cut in half.

A VHF television, AM radio broadcasting system. All
television stations were assumed to transmit on VHF
frequencies, which were unlimited in number. All
commercial radio stations used amplitude modulation.

No UHF or FM sectors were included.

127
(4) A VHF television, FM radio broadcasting system.

In this experiment, no AM radio broadcasting
was included.

(5) A UHF television, AM radio broadcasting system.
All television was of the UHF variety, begin-
ning in 1945, with all radio using amplitude
modulation.

(6) A UHF television, FM radio broadcasting system.
As before, this experimental run operated only

with UHF television and FM radio components.

Results--Experiment 1

 

In this experiment, the calculation of UHF variables was begun in
1945, instead of 1953, when UHF commercial television began historically.
Two additional modifications were made in the model. First, the GROWTH
variable was removed from the new calculation of UHF household percent-
ages in the consumer sector:

GROWTH = -0.0394 - 0.00144*UNYR + 0.00144*(UNYR**2)
UHFHHPC(t) = 0.0081 + 0.042*1n(UNYR) - GROWTH
GROWTH was included in the original equation to adjust for the curvi-
linear function representing the rapid growth in the number of UHF
households. Its presence under the experimental adjustments made the
number of households nonexistent.

Secondly, UHF earnings, which in the benchmark model were calculated
as:

UHFEARN(t) = -5.380 - 2.395*(UINCOME**2) - 6.030*INCOME
where INCOME = UHFREV(t) - UHFEXP(t)

were allowed to be the difference between revenues and expenses (in

128
other words, equal to UINCOME above). In the equation in the benchmark

model, when UINCOME was historically negative, the negative coefficient
of INCOME served to make the final term of the equation positive, which
increased the value of UHFEARN (i.e., made it less negative than it
might otherwise have been.) However, in this experimental run, the nega-
tive coefficient caused an overall decrease in earnings when income was
positive, a result which does not correspond with reality. Positive in-
come should produce positive earnings.

Table 19 contains values of Theil's U for all model variables when
values from the experimental runs were compared to the benchmark. No
variables outside the UHF subsector seemed to be affected greatly by the
early addition of UHF television. Media revenues were most sensitive to
the entry, mainly because of a slight increase in advertiser expenditures
for each medium, but the U's are small (.0127 to .1132). The additional
years of UHF service resulted in an overall increase in the number of
UHF stations and UHF households in 1960. Revenues and expenses were
about 25 percent greater, while UHF station earnings were positive,
instead of negative, but showed a downward trend. Table 20 presents
UHF subsector output for the experimental run compared to the basic
model.

The results of this experiment indicated that the earlier addition
of UHF television into the media system had little effect on the more
established media, making fears of a competing medium drawing advertising

revenue from other media unfounded in this system.

Results--Experiment 2

 

In this experimental run, a maximum number of stations for AM, FM,

UHF and VHF was established using the CHAN variable. In the radio sector

Variable

NEWSNO

NEWSCIR
NEWCOST
NEWSREV
NPUBCST
NEWEARN

MAGNO
MAGCIRC
MAGCOST
MAGREV
MPUBCST
MAGEARN

AMNO
AMHH
AMREV
AMEXP
AMEARN

FMNO
FMHH
FMREV
FMEXP
FMEARN

VHFNO
VHFHH
VHFREV
VHFEXP
VHFEARN

UHFNO
UHFHH
UHFREV
UHFEXP
UHFEARN

TBUY
TNEWBUY
TMA GB UY
AMBUY
FMBUY
VHFBUY
UHFBUY
RETPROF

TABLE 19

FREQUENCY ALLOCATION EXPERIMENTS
THEIL'S U

.0004
0
0
.0127
0
.0176

.0022
.0020
0
.0166
0
.0522

.0044

.0001

.0157
.0011
.0547

.0338
0
.1132
0
.0654

.0013
0
.0372
.0485
.0031

.4254
.4333
.4620

.4217

.8166

.0219
.0178
.0633

.0202

.0204
.0360
.1189
.0003

000000

000000

.1511
.0048
.0089
.0352
.0704

.0180
.0664
.0370
.0481
.0033

.0043
.0003

.0009
.0109
.0020

.0113
.0114
.0032
.0034
.0001

Exp 1 Exp 2 Exp 3

000000 000000

.0251
.0008
.0772
.0063
.2349

.0320
.1275

.1599
.0125

.0054

.0966
.1238
.0123

129

Exp 4

.0076
.0066
.0005
.1510

0

.2207

.0064
.0064
.0005
.0547
.0050
.1912

.9805
.9914

.0033
.0107
.1446
.1959
.0108

.0235
.2247
.2426

.8996
.1393

.0662

Exp 5

.0005
.0001
.0001
.1032

.0184

.0035
.0034
.0002
.0287
.0048
.0937

.0547
.0863
.1608
.0134
.4241

.4326
.4333
.4394
.4253
.8682

.0984
.0186
.1159
.1983

.3680
.0049

.9830
.9922

.4314
.4476
.4249
.8918

.0222
.1872
.2801
.9126

.3969
.0006

omp.~m
o¢~.o¢
n—o.o¢
P-.ee
ooo.¢¢
pmm.m¢
mvm.e¢
mmo.m¢
~o~.o¢
mum.ov
m—m.pm
_mm.mm
voo.~m
mo~.—o
mpn.¢o
poo.mm

m mxw

>umm=o

moo.o¢
poo.me
oom.~¢
pon.¢¢
mop.o¢
omm.mm
ooo.om
poo.om

OOOOOOOO

oozmm<=

Amooz xm<z=ozmo oh ou¢<mzoo
pomhoo uzo hzmzmmumxm zouh<oo44< zomhommm

om m4m<h

osom
op—e
some
Nom¢
pmow
moup
mum

OOOOOOOOO

 

mu
on
mm
mm
om
«_—
pm-

COOOOOOOO

ooop
omop
omop
“mop
emo—
mmop
emo—
mmo—
«mop
pmop
omo—
ovop
ovo—
Kemp
ooo—
moo—

 

cam»

130

zm<mmzo

—mm.~
omo.¢
opm.m
poo.m
pom.m
omn.m

mm~.o—a
moo.o—n

OOOOOOOO

oozmm<z

Au.peouo ON m4m<e

Noo.oe
uwm.n¢
ooo.m¢
mm_.m¢
mmo.—¢
mmo.o¢
mmm.o¢
5mm.o¢
m—o.o¢
vmo.p¢
mmm.me
www.me
omm.mc
pom.~m
omn.¢m
om~.o~

_ mxu

mxmmzo

omm.pv
omm.~¢
mmo.me
om¢.oe
mmo.o¢
mom.mm
poo.mm
o-.o~

OOOOOOOO

oozmm<z

oom—
amo—
omo—
mmop
ommp
mmop
«mop
mmo—
Noo—
pmm—
omo—
oemp
mem—
hemp
oeop
mcop

Law»

131

132
if the limit on the number of AM stations was reached, the overflow of

new stations was added to the FM total, under the assumption that a
broadcaster seeking an AM station would prefer an FM station to no
station at all. In the television subsectors, if the limit on stations
was reached, no further stations were allowed for that year. Unlike the
radio case where AM and FM stations were assumed to be substitutable, the
interchangeability of VHF and UHF stations did not seem to be supported
by historical data. A broadcaster might well have preferred no station
at all to owning an UHF television station.

Theil's U values for variables in this run are given in Table 19.
No changes at all occurred in the print sector; differences were minimal
in the advertiser, VHF and UHF sectors. AM stations reached their limit
in 1952 (U = .1511), which caused the number of FM stations to increase
somewhat (U = .0180). The increase in FM stations was not as great as
the absolute difference between the number of AM stations in the bench-
mark model and the limit, however, since the number of AM stations for a
year is part of a feedback 100p which determines earnings and number of
stations.

When the limit of 2,100 stations was reached in the AM subsector,
AM expenses declined, which brought about an increase in AM earnings.
This increase in earnings was inversely related to number of stations;
thus, as expenses declined and earnings increased each year, the number
of AM stations in the following year decreased. At first, this seems
contrary to industry behavior; one would expect that as earnings/profits
in an industry increased, investors would be attracted to the industry,
and more corporations (in this case, stations) would result. On the
other hand, what seems to be occurring is that revenues tend to be

distributed over a smaller number of stations which record high earnings.

133
Holding the number of AM stations available constant at 2,100 highlights

this effect--stations which can earn high revenues exist, while those
that don't are forced to go off the air. This relationship has implica-
tions for those who propose to decrease the AM channel width from 10 kHz
to 9 kHz, thus providing more channels and ultimately more stations.
Either total AM earnings will decrease with the increase in stations,
causing earnings per station to decline, or the newly established sta-
tions will quickly go off the air, because the earnings will remain con-
centrated in the hands of the most profitable stations. More analysis
at the micro (station) level is needed to pursue these questions.

In the FM subsector, the decrease in total number of radio stations
(AM plus FM) caused a decrease in local radio advertising expenditures,
which were based on number of radio stations on the air. This ultimately
resulted in a decrease in FM revenue and earnings. These results are'

presented in Table 21.

Results--Experiment 3

 

This Experiment set the maximum number of FM and UHF stations at 0
throughout the run of the model to study the results of an all AM, all
VHF broadcasting ssystem. All other conditions in the model remained
constant.

There was no change in the print sector. Since all radio and tele-
vision advertising expenditures were assigned to AM and VHF respectively,
these variables showed an expected increase (U values of .0966 and
.1238). Number of AM stations decreased (.0251), as revenue (.0772) and
earnings (.2343) increased, due to the inverse relationship discussed in
the summary of Experiment 2. Number of AM households showed little

change, an anticipated result, even with the deletion of FM broadcasting,

o~¢.oNp
—o—.mpp
moo.oo

mom.oo

mom.oo

moo.po—
oNN.po—
N_—.—pp
ene.Nm—
Nmo.pN—
mmo.v¢~
Peo.~m_
umo.o¢p
oN~.¢mp
www.cpp

eon.m¢p.

N mxm

zm<uz<

mmm.oo
oop.mo
Noo.nm
NmN.mo
:N. 8
meo.~o
oom.so
omo.oo—
omm._m—
Nmo.—Np
mmo.¢¢p
peo.—m~
mno.oep
oN—.¢mp
ooo.¢pp
co~.me~

oozmm<t

moo.moo
moo.moo
ppo.moo
—oo.moo
¢o~.moo
omn.moo
omm.moo
mN~.moo
ooo.moo
mom.¢mo
opm.omo
Nmo.Neo
~mo.oNo
mmo.~No
omo.mom
coo.mom

N mxu

axwz<

pN m4m<H

omo.om~
omo.oms
moo.ovm
moo.pm~
Nmm.o_~
mNo.oo~
on.~oo
eoo.oho
o—o.ooo
www.cmo
opm.omo
umo.N¢o
pmo.oNo
mmo._No
omo.oom
coo.mom

oome<z

ouoo: xm<zzozmm oh omm<mzoo
bamboo oZH hzmzmmmaxm zo-<oo44< zomhomam

oopN
oopN
oo—N
oopN
oopN
oo-N
oo—N
oopN
oo—N
meo—
Poo—
Nom—
moop
movp
oop—

N mxm

omen
ommm
oon
«Pom
mNoN
nmoN
oemN
omNN
NmpN
oqmp
poop
Nom—
moop
moc—
oop—
oNo

oozmm<z

ooo—
omop
omop
smo—
omop
mmmp
emop
mmo—
Nmmp
FmoF
ommp
oeop
ovmp
Kemp
comp
meop

 

emm>

134

zm<u2m

NNN.m
ooN.N
va.N
emN.p
o——.o
voo.o
mom.o
om~.~
Nmo.p
ooN.m
mmv.m
oom.m
NNv.o
va.o

mmo.o—u

mNo.o

oozmm<z

>mmzu

mm~.o
«Ne.m
opo.m
pom.N
omv.N
oop.N
pNo.~
oNo.—
omo.m
omo.N

moN.m.

mmm.N
ono.N
o~o.~
Nom.o
omm.p

oozmm<z

Av.a:oov FN m4m<h

omm
moo
NNN
onm
omm
mvm
oNo
Pom
mom
oom
eNm
men
mNo
ppm
mop

N mxm

ozzu

mum
“mo
ooo
Nvm
NNm
omm
eNo
pom
mom
mom
eNm
oes
mNo
ppm
mop
Po

oome<z

ooop
omop
omop
Romp
omop
mmm—
emo—
mmmp
Nmo—
—mop
ommp
ovop
ovo—
Neo—
ovop
mvmp

Law»

135

136

since AM and FM audiences tend to be duplicated.

The number (If VHF television stations declined slightly (U = .0320)
due to a small decrease in earnings (.0125) brought about by increases
in expenses (.1599) and revenues (.1275), which are both directly
related to advertising revenues in an inverse relationship. Television
stations rely almost exclusively on advertising expenditures for their
revenue, so an increase in expenditures translates into an increase in
revenue. Increased revenue from advertising, generated by a large tele-
vision audience, encourages stations to invest more money in programming
and personnel expenses, knowing that the additional expenditures will
result in larger audiences and higher revenues. As the model is con-
structed, this is the stated direction of the relationship due to time
lags; i.e., advertising buys for the year under consideration are
calculated first, using audience data from the previous year. It could
be that the sequence begins with a larger investment in programming
expenses, which attracts a larger audience, which in turn attracts more
advertising dollars.

Table 22 contains summary results for this Experiment.

Results--Experiment 4

 

This Experiment maintained the all-VHF television broadcasting con-
dition from Experiment 3, but shifted radio to an all-FM state. Hence,
the UHF and AM radio subsectors were eliminated from the run.

One modification was made in a model equation, which, in the end,
did not prove satisfactory. The number of FM stations in the benchmark
model had been calculated using this equation:

FMNO(t) = exp(7.2081 - 0.09624*FMEARN(t-l) + 1.3663*ln(RNYR))

The equation worked well when FMEARN was negative or minimal, the

MNm.o¢p
Nom.N¢p
N—p.m_p
opo.mN—
oo—.m¢—
oms.om—
Nom.Nm~
omm.Nmp
mo—._oN
NNP.oo—
mNo.o~N
~o¢.o—N
—oo.NNN
moo.—pN
moo.—o—
opN.NNN

m mxm

zz<mz<

mmm.oo
oop.mo
Noo.mm
NmN.mo
omN.po
meo.uo
omn.~m
omo.oop
omm.~m—
—mo._Np
moo.¢ep
peo.pmp
mno.oep
oN_.em—
ooo.¢——
con.m¢—

oome<=

mo~.mmo
moN.oNo
o—m.omo
pmN.omo
moo.—oo
N—o.¢oo
oom.ono
oop.ooo
www.mmm
moo.oom
on.mom
ooo.mNo
omm.mmo
Nom.NNo
poo.omo
ooo.opm

m mxm

>mmz<

NN m4m<h

Nmo.oNo
moo.ooo
N¢o.oem
Nom.¢¢n
ooo.mo~
«mm.¢oh
Nmp.omu
omo.NN~
on.mpo
mpN.NoN
pmm.mNo
onm.~o~
on.p_o
omN.¢Nm
mmm.m—N
pmo.omn

oozmm<z

ouoo: xm<z=ozmo oh omm<mzoo
bamboo ummzh hzmzmzmaxm zomh<oo44< zomhoumm

ooop
omop
omop
smo—
omop
mmmp
emop
mmo—
Nmop
poop
omop
meop
oeop
Neop
oeop
memp

 

cwm>

137

eN¢.¢oNF
com.vo—p
mmn.om——
mo¢.Nm_P
om¢.poo
Nmm.epo
moo.—NN
ooo.Noo
NNm.oem
ooN.om¢
mno.oom
opo.omN
omo.N——
o

o
o
m mxm

>ummz>

omo.¢oo
e—N.¢oo
ooo.ooo
opp.moo
mmv.oom
mm¢.m¢o
moe.umm
ooo.o¢¢
coo.mmm
ooN.om¢
mmo.oom
o—o.mmN
omm.mpp

o

o

o

oozmm<z

Noo.omop

oeo.mem
m_¢.mmm
oom.~oo
mom.NNo
NNN.~oo
moo.——o
ovm.spm
con.mne
oo—.m—¢
mmm.noN
mom.~vN
om~.opp
o

o
o
m mxu

axuuz>

¢~N.mmm
oom.o~o
oom.ooo
moo.ooo
oom.moo
pr.m_m
mpo.MN¢
NNN.Nmm
oNo.pmm
om—.mpe

mmo.NoN.

mom.N¢N
oom.op~
o
o
o

 

oome<z

Ae.oeooo NN m4m<e

Foe

vow
wee
mom
can
me
poN
mmN
moN
oop
oN—

oom—
mmop
omop
smo—
omop
mmop
emo—
mmop
Nmop
mep
omop
oeop
ovmp
Noo—
oqop
mvop

 

ccm>

138

Av.ucouv NN wom<h

Ncm.m¢N
mmm.o—N
com.moN
mu—.mo—
pno.om—
oop.mN~
—mo.op_
ovm.mo
o—o.¢o
oso.o¢
NNN.oN
N_¢.Np
oop.m
o

o
o
m mxm

zm<muz>

Noo.omN
mNo.mNN
oo¢.o—N
upp.oo—
coo.Nop
mpm.omp
Nm¢.m_p
NoN.oo
eoN.No
mmo.o¢
NNN.oN
N—¢.Np
oo_.~

o

o

o

oozmm<z

ooo—
omop
omop
smo—
omop
mmop
emo—
mmop
Nmo—
_mo—
omo—
meop
oeop
Neop
oqop
meop

 

cmm>

139

140
historical case. However, in this experiment, FM earnings increased

substantially, since all radio advertising expenditures were channelled
into this subsector. Before a change in the equation, this resulted in
no FM radio stations and forced the program to find the exponential of

a negative number. The sign of the FMEARN coefficient was changed from
negative to positive. However, after the modification, the argument of
the exponential became so large that the number of FM stations had to be
reported in scientific notation (maximum value: .32E+19). No attempt
was made to make further modifications in the equation to bring it into
sensible range. It seems safe to say, however, that under the conditions
which existed between 1945-1960, the growth of FM radio was contained,
and that given a free rein, it would have developed faster, though not
as fast as the benchmark equation would suggest.

The enormous values for number of FM stations affected the entire
run of this Experiment. Local radio advertising expenditures depended
on the total number of radio stations the previous year. This resulted
in a major share (1f each year's advertising budget being allocated to
FM radio, with smaller allocations to the remaining media. Thus, news-
paper advertising buys (U = .2247), revenue (.1510) and earnings (.2207)
decreased as did magazine advertising buys (.2426), revenues (.0547) and
earnings (.1912). Other variables in the print sector had U's below .01.
In the VHF subsector, advertiser expenditures also decreased (.1393),
affecting revenue (.1446), expenses (.1959) and earnings (.0108). FM
revenues (.9805) and earnings (.9914) were radically affected. Table 23

contains summary data for these variables.

eoo.m~oN
omm.no~N
omo.NoNN
cum.mooN
ooN.noeN
oem.moNN
voo.vaN
omm.NN—N
omm.¢mo_
Noe.oom—
oNo.~Nmp
omo.opo~
mon.mooN
omm.pmoN
mpm.~m_m
«mo.NooN

v mxm

zm<mzuz

mpo.mom¢
m—_.mNm¢
on.pmm¢
oN~.m¢Nv
ooo.oo—e
m—o.mpN¢
mem.onom
om—.—mom
non.N¢mm
oem.mmmm
oom.oon
omn.oeNm
noN.Non
oo¢.o¢Nm
mpm.nm—m
nmo.NooN

oozmm<=

oom.m-e
oop.nome
NNP.¢mm¢
m¢~.mom¢
mm—.moN¢
mpo.Noom
meo.nosm
m~o.mmom
«No.mmvm
N_~.moem
omn.No~m
ooo.mmpm
omo.Nnom
op_.uoom
Nos.oNov
moo.momm

e mxu

>u¢m3mz

mN mom<h

omN.sNNo
vmo.eo—o
Nmo.momm
eoN.m¢om
cmm.¢pom
pom.mmom
Foo.oo¢m
on.Nm~m
eoN.mNom
mom.omoe
mom.mmne
ooo.nmme
mo¢.mpe¢
meo.mome
Nom.oNo¢
moo.nomm

oozmm<z

Amooz xm<zzozmm oh omm<azoo
bamboo moo; Hzmzmmmmxm zomh<oo44< tomhoumm

mN—m
nooN
moNN
mNoN
omoN
NNeN
—oNN
om—N
omo—
oom—
oom—
omop
NoeN
mva
mnmN
epoN

e mxu

oome<z

>om3mzp

«moo
—va
oo¢¢
oome
NoNe
ovme
omom
mmom
opmm
ummm
mon
Room
ceoN
NeoN
mumN
cpoN

oom—
omop
ooo—
smo—
omop
mmop
cmop
mmop
Nmop
poo—
ommp
meo—
oeop
Kemp
oeop
mcop

 

Law»

141

voo.¢oe
oo¢.o~e
moo.o—m
NoN.mme
«No.mme
meN.Nmm
moN.epm
¢N_.NN¢
mpo.¢m¢
m¢o.mm¢
———.oom
oNo.on
oo~.¢mn
mop.oss
moo.N~N
eoo.MNo

e mxu

z¢<mo<z

mpp.~mm
—Nm.Nvo
oom.vNN
Nmo.oom
N—¢.ohm
NNN.ooN
oom.o-
NMN.~¢o
om¢.omo
m¢—.oNN
oo¢.omo
o~m.—o~
ooN.o—m
ooo.Nmo
moo.N~m
woo.mNo

oozmm<z

moo.oNMN
eem.momN
NNo.moNN
NoN.¢_NN
pom.~NNN
omm.oNNN
m—o.mm—N
poo.mooN
on.oNoN
mNo.¢omp
N—o.onop
—o¢.pmmp
oNN.—N—N
pom.N¢—N
omN.pnpN
eoN.m¢oN

e mxm

>mmo<z

ooo.o—oN
mo~.omoN
mmm.¢o¢N
mmo.mva
eNo.oomN
Nom.o¢¢N
¢o¢.N—¢N
oem.on¢N
omm.ooNN
No~.o~NN
mom.¢m—N
Nmo.¢opN
mmp.umoN
coo.moNN
oMN.-PN
eoN.o¢oN

oozmm<z

Au.p:ouv MN mom<~

mum
«on
mom
pno
oNN
Nos
moo
eem
o—m
owe
New
Foe
wmo
mum
mum
Now

e mxm

oo2mm<z

>= 85,..—

ooN—
oNep
mm——
omN—
mNNp
mops
omp—
oNNp
woo
woo
oNo
mco
mpo
mum
ono
moo

ooo—
mmop
omop
“mop
omop
mmop
emo—
mmo—
Nmop
_mop
omop
oeop
o¢op
Kemp
eco—
mem—

 

cmm>

142

ooo.ooo
ooo.m—m
oom.m¢m
Nmm.m¢¢
smo.mom
o~o.o_m
mou.ooN
vo~.~mN
eoN.NoN
eo~.oo_
oom.¢mp
ooo.pp—
www.mo
o

o
o
e mxm

axmmz>

¢—~.mmm
oom.omo
mom.omo
moo.ooo
oom.moo
~N_.m—m
mpo.mN¢
NNN.nmm
oNo.pmm
oo—.mpv
mmo.~oN
mom.~¢N
oo~.o—p

o

o

o

oozmm<z

oNN.moo
mum.—¢N
—oo.oms
em—.omo
meN.Nmm
oN~.¢e¢
moo.¢o¢
mmN.on
cmo.—NN
Nmm.o—N
pen.mmp
coo.oN~
nno.oop
o

o
o

e mxm

>mxmz>

ouo.¢om
epN.¢oo
ooo.ooo
o—~.ooo
mm¢.oom
mm¢.m¢o
mo¢.~mm
ooo.o¢e
voo.mom
ooN.om¢
muo.oom
o—m.omN
omo.mpp

o

o

o

oozmm<z

Av.u=oov MN m4m<b

omop mo—N
comp epop
oNo— pco—
pomp com—
moNp Nmop
mum Nomp
coo FNPP
mmm ooo
opo mom
mom ooop
oom moo
mpm mom
poN mmN
o o

o o

o o

v mxm oozmm<z

>omuz>

oom—
omo—
omm—
smo—
omop
mmop
«mop
mmop
Nmop
pmop
omo—
ovop
oeop
Neo—
cem—
mem—

 

Law»

143

a_+mNm.
op+mNm.
o_+mmp.
-+umm.
N~+uoo.
~p+mem.
Np+mmn.
Np+me_.
m_+mo_.

m_+meN..

¢p+uom.
mo+mmm.

omnoeomp

oNoNo
omNNN
mN

v mxm

ozzm

mum
nmo
ooo
New
MNm
omm
¢No
pom
mom
com
eNN
we“
mNo
ppm
mop
Po

oozmm<z

mooe oop
opoe No—
osme oo—
eo—v map
prw mop
mopv oop
more —o_
Nmpe _oN
Foam o—N
moon oPN
omen NNN
«Nem opN
oemp oNN
pNo— mpN
oNo— cop
mmop ——N
e mxm oozmm<z
>=m2m

Au.u:ouV MN u4m<p

oMN.mmN
No¢.oNN
voN.mpN
NNo.eo~
ooo.oo—
omo.¢m—
ooo.m—_
oov.oo
omo.oo
msm.om
peo.mN
oom.¢—
ooe.n

o

o

o

e mxu

zm<muz>

Noo.omN
mNo.mNN
oo¢.opN
n_~.omp
coo.No~
mpm.om~
Noe.m—p
NNN.oN
eoN.No
mno.o¢
NNN.oN
N—¢.Np
oop.m

o

o

o

oozmm<z

ooo—
ommp
omo—
mmop
emo—
mmop
«mop
moo—
Nmo—
_mop
omop
mem—
oeop
meo—
oom—
meop

 

cmm>

144

poo.mmm
oNo.onm
Noo.omm
mum.e¢m
o¢¢.omm
o—o.omm
coo.mmm
opo.omm
eoM.MNm
mme.ooN
mm¢.moN
ooo.ooN
eeo.p¢_
o¢¢.NPP
oNo.oo

¢¢o.No

c mxm

zx<mzm

NNN.m u
ooN.N u
NN¢.N u
cmN.p a
opp.o u
coo.o u
mon.o u
om—.— u
Nmo.— u
ooN.m n
mm¢.m a
oom.m u
NN¢.o u
ch.o u
mmo.o~u
mNo.m u

oome<z

Ae.u=ouo mm m4m<e

poo.som
NeN.oom
Nmm.eom
m~_.oem
oom.o¢m
Noo.omm
oom.omm
Non.F¢m
Nom.NNm
poo.pom
un—.oom
o—p.oN—
eNo.Pmp
voo.NNp
oom.-

moo.¢m

e mxu

>m¢2m

mmm.o
NN¢.m
o—o.m
pom.N
om¢.N
oop.N
pNo.P
oNo.—
omo.N
omo.N
noN.m
mmm.N
omo.N
ono.~
Nom.o
mom.~

oome<z

oomp
amm—
omop
“mop
ooo—
mmo—
«mop
mmop
Nmop
_mop
omo—
ovop
oeop
Neop
ovo—
meop

me>

145

146
Results--Experiment 5

This Experiment imposed an all AM, all UHF system on the broadcasting
industry for the time period under study. The UHF subsector was adjusted
in a manner similar to Experiment 1, so that it could operate from 1945,
instead of beginning in 1953.

In the model, magazine advertising expenditures were determined in
part by the number of VHF television households. Therefore, eliminating
the VHF sector which set the number of VHF households to 0, caused a
drop in magazine advertising expenditures (U = .1159). Since earnings
were dependent on advertising buys, this contributed to a decrease in
overall magazine earnings (.0937). The UHF television audience generated
in this experiment should most probably have been substituted for the VHF
audience in the equation. Including both UHF and VHF households in the
equation did not prove to be significant--possibly because they are
duplicated audiences. Other variables in the print sector were not
substantially changed, with U values of .01 or less.

The elimination of VHF households led to an increase in radio
advertising expenditures, all of which were assigned to AM radio (.1983).
This increase resulted in increased AM revenues (.1608) and earnings
(.4241). Since earnings and number of AM stations are inversely related,
the number of 11M stations decreased (.0547). This relationship was
considered earlier in this chapter. The decrease in stations caused an
increase in AM households (.0863), due to the inverse relationship
between those two.

The macro level at which the model is constructed does not allow
for an analysis of station preference types by AM households. However,
the increase in AM households which occurred when there were fewer AM

stations might be explained by the higher profitability of those stations

147
(recalling that a decrease in stations brought about an increase in

earnings), which might suggest better quality programming. On the

other hand, the directionality of the relationships in the feedback
loop could be reversed. A larger audience will attract more advertising
dollars, which in turn increases revenues and earnings, which are dis-
tributed to those AM stations which remain in existence to share the
profits. Higher earnings due to larger audiences tend to reduce the
number of AM stations on the air.

The assignment of all television advertising expenditures to UHF
television brought about a substantial change in values for that vari-
able (.3680). Number of stations increased (.4326), mainly as a result
of the growth function built into the equation, but this increase helped
both expenses (.4253) and revenues (.4394) to increase also, with a
resulting increase in earnings (.8682). UHF households (.4333) in-
creased, also as a result of a growth function. Results for this

Experiment are found in Table 24.

Results--Experiment 6

 

This Experiment maintained an FM radio, UHF television broadcasting
system, with results similar to those of Experiment 4, with its FM
subsector, and Experiment 5, with its UHF system. Again, the exponential
relationship which determined number of FM stations resulted in very
large values for this variable, which in turn increased advertising
expenditures for FM (.9126), while decreasing advertising buys for the
rest of the media. This increase in advertising expenditures for FM in
this Experiment was even greater than that in Experiment 4 (compare
Tables 23 and 25), since the variable of VHF households, which kept

radio ad expenditures in check, had been removed.

oemp oNn

ommp NNN
_oN— «no
«NN— moo
sopp mos
—mNp mom
—m—p ooh
oepp moo
ooo— mum
ooop peo
oopp ooo
omo— pno
moop ¢om
voo— moo
c—op Non
MNoF new
o mxo oozmo<z
>ooz<

oop.omo
_pn.—¢o
o¢o.omo
~N_.moo
mop.¢N~
ovo.ooo
Noo.ooo
o—o.mNo
ooN.mNo
oN—.¢¢o
mmo.omm
o~¢.—on
e~m.mom
upo.oNN
moN.oNo
Noo.¢oo

o mxo

z¢<oo<z

eN o4o<h

o~—.no~
FN~.Neo
oom.¢Nn
Nmo.oom
Np¢.mmm
moN.oo~
ooo.o-
nmN.~¢o
ome.ooo
mep.oNN
oo¢.o~o
opm.pmm
oom.opn
ooo.Nmo
moo.Nm~
coo.MNo

ooZom<z

omooz ¥m<zzozoo :hmz oom<mzoo
bamboo m>~m hzozfimooxo zomh<oo44< zomhommo

omo
ooo
pno
oNo
oN__
mpo
moo
ooo
oeo
moo
mom
moo
moo
oou
ono
o—m

m 9m

ooN—
oNep
oop—
ooN—
oNN—
oopp
oop—
oNNF
coo
ooo
oNo
oeo
m—o
omo
Nmmo
moo

oozom<z

ooop
moop
oom—
Noo—
oomp
ooop
comp
moop
Noop
poo—
ooop
ovo—
ovo—
“wo—
oeop
mem—

 

S.
CU
G.)
>-

148

mom.ooN—

NNm.pNNP
Noo.oop—
ooN.oo_p

FN¢.ooop
e¢o.NN_—
oNN.mmo—
ueo.oNo—
oo¢.ooo
moN.moo
mo~.-o
oom.opo
Noo.n¢o
ooo.eoo
oom.ooo
oNo.ooo

o mxu

>omz<

noo.oNo
moo.ooo
Nvo.o¢~
Noo.¢¢m
ooo.mon
nom.¢om
NN—.oos
ooo.NN~
oNo.o_o
mpn.No~
Pno.oNo
omm.pms
on.~po
ooN.eN~
onm.mpm
poo.oo~

oozoo<z

Izz<

onooe
Noooe
o~m~¢
ommme
Noo~¢
opooe
ooooe
momcv
popee
oeoNe
omopv
ooooe
Noqom
“Noon
eomom
moNnm

oome<z

Ae.p=ouv «N uom<c

ozz<

onem
comm
oon
«pom
NNoN
NooN
ocoN
ooNN
NmpN
oeop
poo—
Non—
moop
ooep
ompp
oNo

oozmm<z

oomp
moo—
ooo—
moo-
oomp
oomp
com—
moop
Nom—
Poo—
oomp
oeop
ovo—
hemp
ovo—
oemp

 

cmm>

149

ozuzo

oozmm<z

omm
eoe
oom
Nun
Nnm
Nom
mum
onm
oev
coo
ooo
moo
mom
o

o

o

o mxo

>oouzo

pom
omo
Nvo
omo
—oo
¢o¢
omm
NNm
emN

OOOOOOO

ooZmo<z

Ae.peou V cm uom<h

ooo.¢mm
ooo.on
¢o~.moN
oo¢.NoN
ooo.NoN
op¢.NNN
ooo.NmN
poo.oMN
nop.oNN
ooN.o—N
poo.oNN
FNo.om—
omo.oPN
coo.mm_
one.ooN
osm.opN

omm.oo
oo~.mo
Noo.mo
NoN.oo
oNN.po
meo.mo
mos.no
omo.oo_
omm._m_
Noo.pNP
omo.¢¢~
—¢m.—m—
mno.o¢p
oN—.vm~
ooo.epp
co~.m¢~

ooZom<z

oomp
oomp
oom—
Romp
ooo—
ooo—
coo—
moop
Noo—
poop
ooop
oeop
ooo—
Kemp
ovo—
oqop

 

com»

150

No—.o¢
coo.oe
emm.¢¢
NNo.N¢
.vp¢.p¢
omn.o¢
omm.ov
omn.o¢
oNo.Fv
oNo.N¢
Npo.¢¢
pom.o¢
ooo.ne
poo.~o
omm.vo
oNN.oN

o mxm

mxouzo

ooo.p¢
omm.Ne
Noo.m¢
on¢.oe
mNo.o¢
moo.oo
—oo.No
oNN.oN

OOOOOOOO

oozoo<z

omo.oo
oom.mo
mpo.~o
pom.o¢
moo.~¢
o-.o¢
oom.o¢
NNN.oe
oNs.¢¢
«MN.¢¢
Noo.o¢
ooo.¢¢
oop.No
omN._o
mpn.¢o
—oo.om

o mxo

>omm==

ooo.o¢
poo.Ne
ooN.N¢
pom.e¢
mo—.o¢
moo.oo
ooo.oo
poo.oN

OOOOOOOO

oozmo<z

Ae.peooo ¢~ mom<e

omom
oppe
NoNe
Nome
—ooN
oomp

N
(N
:6

000000000

ooo—
ooo—
ooop
Noop
oom—
ooop
coo—
moop
Noo—
—omp
ooo—
ovop
ovo—
Neo_
ovo—
meo—

 

cmm>

151

Ae.peoov cw usm<e

 

zx<omzo

Qw-
mm
on
Q'N
I I

ooZom<z

ooop
ooo—
oomp
“mop
ooop
ooop
comp
moop
Noop
pomp
ooop
o¢op
oemp
mem—
o¢m~
meo—

 

cam»

152

ooN.¢on
omo.oepm
oom.mooN
eoF.Noom
soo.oooN
vNN.oooN
com.mmoN
o—o.ooNN
moo.om—N
puo.noop
poo.¢oo—
oom.¢oo_
omo.~moN
on.ooom
Poo.Nm_m
o_o.oooN

o mxo

opo.nom¢
opp.on¢
oNo.—ome
oN_.m¢N¢
ooo.oop¢
o—o.mpN¢
mem.onom
2:. Son
non.N¢om
oeo.oomm
ooo.oon
on~.o¢Nm
soN.non
oo¢.o¢Nm
mpm.nmpm
Nmo.NooN

oozom<=

oon.MNpo
oo~.oNo¢
ooo.oNoe
mNm.¢on¢
Noo.oom¢
oou.omme
m—n.ooo¢
_o—.oNom
oom.ppom
po—.euvm
oNo.o¢mm
omo.p¢pm
MNo.moom
ooo.em—¢
oo¢.o—oe
oeo.momm

o mxo

>mmm3mz

oN u4o<h

ooN.~NNo
ooo.eopo
Noo.momo
eoN.o¢oo
¢oo.epoo
pom.omoo
poo.oo¢o
oNo.no—o
coN.mNoo
ooo.omoe
ooo.mone
ooo.-o¢
oo¢.m~ee
oco.oom¢
No~.oNoe
ooo.momm

oozom<z

ooooz xm<==ozoo oh ooa<mzoo
bamboo xmm hzozomomxo zouh<oo44< zomhoomm

emoe
—Noe
ooev
oomc
NoNe
ovmq
omom
omom
epom
mmmm
oon
Room
eeoN
NvoN
mmoN
epoN

oozmm<z

oom—
oomp
ooop
hoop
oom—
ooo—
coop
moop
Noo—
poop
ooop
oeop
oemp
Kemp
ovop
oeop

com»

153

emo.mo¢
ooo.¢N¢
oom.ooe
opn.ome
who.op¢
oo¢.~oe
Noe.eme
ooo.~N¢
oem.o_¢
«pp.oo¢
ooo.~oe
Noo.opo
ooo.¢oo
mno.opm
moo.¢No
Noo.¢oo

o mxm

zm<mo<z

opp.~o~
FN~.N¢o
oom.¢NN
Nmo.oo~
Npe.onn
moN.ooN
ooo.omn
NMN.N¢o
oo¢.ooo
m¢_.on
ooe.omo
o—m. SN
oo~.o—N
moo.Nmo
moo.N~N
coo.mNo

oozoo<z

woo.onN
PoN.ooNN
ooo.ooNN
moo.moNN
NMN._o_N
oom.popN
onm.¢NoN
No~.mooN
om—.¢poN
poo.emoN
poo.NNmp
op—.MNop
_oN.omoN
q—p.pooN
omp.vNNN
poo.oNoN

o mxo

>omo<z

ooo.oFoN
oom.oooN
omo.¢oeN
nso.oeoN
eNo.oooN
Non.oe¢N
¢o¢.NpeN
oeo.o~¢N
ooo.ooNN
No~.oNNN
ooo.¢mpN
Noo.¢oPN
onp.nmoN
coo.moNN
mmN.PNFN
eoN.o¢oN

oozmo<z

Ao.u=ouv oN o4m<h

mun
_mo
oou
meo
moo
ooo
oNo
moo

ooNP
ova
om—F
ooNp
oNNP
oo——
ompp
oNNP
coo

ooo

oNo

ooop
ooop
ooop
Noop
ooo?
ooop
coop
moop
Noop
poo—
ooop
oqop
ocop
Reap
oeop
oeop

 

me>

154

ooo.Nu¢
mso.oo¢
oNo.o¢¢
mo¢.oN¢
oo~.mpe
N—N.oom
ooN.on
«NN.oom
oom.oom
o—¢.oom
oeo.ooN
moo.NNN
ooo.—o—
moo.oop
—No.oo

oeo.oo

o mxm

zm<mzm

NNN.o u
oon.N a
NN¢.N n
emN.— a
op—.o u
eoo.o u
oom.o u
oop.— a
Nuo.p u
ooN.m n
oo¢.m .
mom.o s
NN¢.o n
N~¢.o n
ooo.o~u
oNo.o u

oozoo<z

ooo.¢o¢
ooo.oo¢
ono.oo¢
oNo.oNe
mom.o~¢
Noe.umm
Nou.mnm
oom.—om
o—o.¢om
Non.mpm
Noo.ooN
nop.ooN
ooo.pop
PoN.oNP
mMN.mN

¢N¢.-

o mxo

>mmzm

mmm.o
NNe.o
opo.m
—om.N
om¢.N
oop.N
PNo.p
oNo.—
omo.m
omo.N
NoN.m
oom.N
ono.N
ono.~
Nom.o
ooo.p

oozoo<z

Ae.ueoao mm mom<~

mp+mNm.
op+mNo.
op+uop.
~_+moh.
up+moo.
Np+m¢o.
-+mmm.
np+oe—.
o_+oo_.
op+ovs.
¢P+ooo.
mo+omm.

onmoeomp

oNoNo
onNNN
MN

muo
Koo
ooo
moo
oNo
omo
«No
Poo
ooo
ooo
eNm
oem
oNo

oom—
ooop
ooop
moo,
ooop
ooo—
coop
moop
Nomp
Poop
ooop
oeoF
ovop
mem—
oeop
oeop

 

emm>

155

cNF.oo
omo.oo
Noo.No
omN.po
meo.ov
oo_.o¢
omo.o¢
ooo.o¢
ooo.o¢
oNo.oe
ooo.me
poN.oo
NNm.mo
ooN.po
m—N.¢o
poo.om

o mxo

>ommzo

ooo.o¢
—oo.N¢
oo~.N¢
—om.¢¢
mo—.o¢
ooo.oo
ooo.oo
poo.oN

OOOOOOOO

oozoo<z

ouom
op—e
NoNv
Nomc
pooN
ooup

R
[N
m

OOOOOOOOO

Ae.oeouo mm “Jeep

0540
[\N

r—QmmN
Nr-O‘mm
PP

OOOOOOOOQ

ooo—
ooo—
ooop
Noo—
ooop
ooop
coo—
moop
Noop
poo—
ooo—
ovo—
ovop
Kemp
oeop
ovop

 

com»

156

z¢<umzo

Pom.N
emo.¢
opo.m
—oo.N
—oo.N
oNn.o

omN.opu
Noo.opu

OOOOOOOO

oozoo<z

Ae.aeouo m~ u4m<e

m¢o.o¢
oo¢.o¢
voo.ve
oom.Ne
eoo.~¢
eoN.ov
oNo.om
ooo.o¢
moo.o¢
noo.N¢
NN~.m¢
oo~.o¢
ooo.~¢
—oo.po
omm.¢o
oNN.oN

o mxo

oozoo<z

oxouzo

ooo.p¢
omm.N¢
Noo.m¢
ou¢.o¢
mNo.o¢
moo.oo
poo.no
oNN.oN

OOOOOOOO

ooo—
ooo—
ooop
Romp
ooop
ooo—
coop
moo—
Noo—
—oop
ooo—
oeop
ovmp
neo—
ovop
oom—

 

me>

157

158
The decrease in ad buys forced newspaper revenues (.1270) and

earnings (.1842) down, as well as magazine revenues (.0618) and earnings
(.2191). The increased revenues from advertising increased FM revenues
(.9830) and earnings (.9922). UHF subsystem variables increased as they
did in Experiment 5; number of stations (.4314) and households (.4334)
on the basis of yearly growth functions, revenue (.4476) and expenses
(.4249) as a result of increased number of stations, and earnings

(.8918) as the resultant difference between the latter two. Speculating
on reasons for system behavior in this experiment is difficult, since so
many of the structural equations for FM and UHF use only time indices as

predictor variables.

Summar

Manipulations of the use of the electromagnetic spectrum by various
broadcast media in various combinations resulted in little disturbance
of the print media. Since most changes in the broadcast sector were
relatively small in themselves, this result is not unexpected. When
variables in the print sector were affected, it tended to be the result
of changes in the allocation of advertising dollars which caused the
difference. Several relationships within the broadcast sector were
highlighted by these experiments: the inverse relationships between AM
earnings and number of stations, which suggested that as the AM industry
became more profitable, high-earning stations kept other stations from
sharing in the industry profits; and between number of AM stations and
number of AM households, another link in the earning/number chain, since
number of households is a predictor of earnings.

The very large values for number of FM stations in Experiments 4 and

6 caused changes within the entire subsystem, but still only yielded

159
values for Theil's U around 0.3. This either suggests a problem of

sensitivity within the model structure, or the remarkable stability of
the industry system in the midst of an uncontrolled addition of FM radio.
Historically, the AM radio industry was stable, and prepared to make
substantial investments in VHF television. The indifference of radio
broadcasters to the FM system, in addition to the FCC's reallocation of
spectrum space to FM forcing the manufacture of new receivers, could well
have made the (affect of FM on the system so minimal. Even with differ-
ent channel availability parameters, the model functioned from initial
conditions from the historical period under study.

Another structural problem was highlighted when the UHF subsector
was manipulated. Changes in values of the criterion variables occurred
not because of changes in variable values within the subsystem, but
solely because of the growth function built in by the presence of UNYR.
When UNYR equals 0 in 1945 instead of 1953, values change dramatically,

but only because of a time index.

Postscript

 

Thus, a basic simulation model such as MASSMOO can be manipulated
to reflect changes in the conditions under which the model operates to
accomplish two things: the evaluation of various policy options and
their effect on the system, such as comparing results across all six
spectrum allocation manipulations; and the identification, or at least
the emphasis, of structural relationship within the model which run
counter to assumed behavior: for example, the inverse relationships
between number of AM stations and AM earnings, and between number of AM
stations and AM households, suggest that the subsector stations may

tend to operate as (an oligopoly, instead of as competitors.

160
One other comment should be made about the Experiments. Every

manipulation made here assumed that the initial conditions in the system
in 1945 were exactly the same as they were in the benchmark. Also, the
modified models for the most part retained equations for criterion
variables that were established using the historical data. Thus, for
example, in Experiments 5 and 6, the pattern for number of UHF stations
across all 16 years is parabolic, reflecting the initial function in the
benchmark model.

Chapter 5 contains a summary of conclusions to be drawn from exam-
ination of both the benchmark model and the experimental manipulations.
It also ciiscusses the weaknesses of the present MASSMOD version and

makes suggestions for further research and future directions.

CHAPTER 5

CONCLUSIONS AND FUTURE DIRECTIONS

This research applied systems simulation methodology to the study
of the economic and structural behavior of the mass media industry from
1945 to 1960. Using historical data from a time period marked by the
introduction of a new communications technology--television, the study
had two major purposes: (1) the quantification of abstract theories
about media structure and economics and the identification of integral
relationships among media components, accomplished by the construction
and validation of a: computer simulation model; and (2) the demonstra-
tion of the feasibility and usefulness of systems simulation to the
study of media behavior, planning and policymaking, shown by a series of
experiments manipulating two policy variables in the model.

Because of a shortage of past research that has applied systems
simulation to 'the mass media as an industry, or to decision-making and
planning within that industry, the research was designed to be explora-
tory. This chapter contains conclusions about the application of systems
methodology to the study of the media, as well as some necessarily tenta-
tive conclusions about the structure and performance of the media industry
during the introduction of television. The chapter concludes with some
suggestions of areas where future research could be conducted that would

build upon the work of this study.

161

162

Conclusions

 

Methodology

The creation of MASSMOD, a simulation model of the industry which
validates against historical data for the time period, suggests that
enough information about the behavior and economics of the mass media
industry is available to construct a mathematical model of the system.
Further, the data acquisition problems discussed in Chapter 3 should
only lessen as the model is expanded and modified to deal with decades
beyond the time frame of the original MASSMOD. While the data are still
distributed among multiple sources, industry has realized the importance
of quantitative information reported in a variety of ways to its develop-
ment, and is therefore collecting more information than it did in 1945-
1960.

Secondly, the manipulation of conditions in MASSMOD to conform to
various hypothetical situations with alternative outcomes indicates that
a simulation model can be of use in systematically testing the impact of
proposed policy actions on the mass media system without actually imple-
menting such actions in the real world. The spectrum allocation experi-
ments performed as part of this study altered the availability of chan-
nels for four different broadcasting systems to investigate the effect
of alternative policy choices. The model showed remarkable stability
across the experiments, and suggested that even unfettered FM and UHF
systems would not have been financial successes on the scale of AM radio
and VHF television. Policymakers looking back to history in the midst
of current decisions could use such a simulation model to answer ”what

if" questions that gareviously relied only on speculation for answers.

163

The search of the literature dealing with systems analysis of the
structure and economics of other industries detailed in Chapter 2 implied
that the methodology might be appropriate to the analysis of the mass
media as a multi-component industry. The construction, validation and
manipulations with MASSMOD confirmed this, if only on a first-attempt,

exploratory basis.

Theory

Since MASSMOD is an exploratory effort in the construction of a
valid simulation model of the mass media industry, conclusions drawn from
the model are necessarily tentative. Nonetheless, it is felt that the
model has sufficient validity to make these conclusions of interest.

First, the Inedia system was able to be represented in quantitative
mathematical form by just four components: print, broadcast, advertising
and consumer, and at a high aggregation level, chosen to identify overall
structure with the least complexity. Initial attempts to include a
manufacturing sector for broadcasting equipment which would provide
receiver supply and costs did not yield significant equations, and was
therefore dismissed as unnecessary to the development of broadcast
audiences. The consumer's choice to be a viewer or listener seems to be
independent of the supply and/or price of the receiver, or can be ex-
plained as well by available disposable income. This may have relevance
to the developing field of video recording technology and its acceptance
by the public.

At the same time, some model subcomponents, such as FM radio and
UHF television, were more dependent on simple time variables as predictors
than on structural variables determined within the system. When system

variables were used as predictors for UHF and FM variables such as number

164
of stations, households, revenues, expenses and earnings, the regression

equations were not significant. Without searching beyond the system for
other predictor variables, a simple time variable, which indexed the year
in the series, was used. Hence, even when values for structural varia-
bles in the system changed, the UHF and FM sectors for the most part
remained unchanged. Incomplete data and/or a real lack of structural
equations to predict the early years of these two media could be respon-
sible for this deficiency, which affected the overall sensitivity of the
model.

Secondly, what seemed on first inspection to be a potentially com-
plex system was able to be modelled with relatively few endogenous varia-
bles with simple linkages among and between them. In the model develop-
ment stage, complex relationships were proposed to predict most variables,
only to discover that some of the predictor variables did not make a
significant contribution to the equation. These variables were therefore
eliminated. For example, variables from other subsectors were included
as predictors vvithin a subsector to acknowledge a possible overlap
among subsectors--television affecting magazines and newspapers, FM
radio affecting AM radio--but most of these cross-sector variables were
not significant in the regression equations. Variables from the news—
paper subsector, such as circulation and cost, were included in early
equations for the magazine subsector, and vice versa, in an attempt to
determine how much effect the publishers have on each other. The results
were not significant, and therefore the variables were not included in
the final equations.

In the time period of this study--l945 to 1960, the print sector

was largely insensitive to changes in the broadcast sector, even though

165

magazine circulation was a function of television households. Only when
a shift in advertising expenditures occurred, as it did when an all UHF
sector was imposed on the system, did revenues change slightly. News-
papers and consumer publications do not compete with the broadcast media
for the exclusive claim to each consumer; rather, the consumer reads
such publications in addition to his/her radio listening and television
watching. Newspapers and magazines were not radically affected by the
other's behavior either, as the lack of linkages between the two sub-
systems suggested. The often expressed fear that a new medium will
siphon off advertiser dollars from existing media was not supported in
MASSMOD; rather, newspaper and magazine advertising expenditures in-
creased during the period.

The demise of mass circulation magazines in the 1960's, attributed
to television, did not occur in this time period. Further, since
television won over advertisers from magazines on the basis of audience
demographics and not mere audience size, the model may not have reflected
such an effect, since it deals with audiences at a high level of aggrega- '
tion which ignores demographic differences. A more microscopic approach
would be required to detect such an effect.

In the broadcast sector, each medium was isolated from the other,
again indicated by the lack of broadcast variables passed from subsector
to subsector. Attempts were made to include, for example, AM variables
in FM and television equations, but the results were not significant.
Several relationships existed in the AM subsector, which were contrary to
expected behavior. It was assumed that lagged AM earnings would influence
the number of AM stations in operation directly; that is, as earnings

increased more investors would be willing to establish AM stations and

166

the number of such stations would increase. However, as stations in-
creased in number, the total AM earnings decreased, and vice versa, as
number of stations decreased, earnings increased. This suggests that AM
stations may function, not as competitors, but in an oligopoly, where
strong, financially sound stations can ward off the attempts of newer
stations to enter the broadcasting market and generate higher earnings
for themselves.

Additionally, it was assumed that a greater number of stations
would attract a larger audience, but the opposite held true--an increase
in the number of stations caused a decrease in audience. The audience
variable is linked to advertising buys, which influence both AM revenues
and expenses. Thus a larger audience could result in higher advertising
expenditures which would cause higher earnings, and therefore fewer
stations. The result is unexpected in the sense that the FCC and other
access advocates have argued that more stations mean more audience poten-
tial.

The lack of structural equations using endogenous variables from
the system to predict FM variables, which instead relied on time param-
eters for estimates, indicated a failure to identify key variables for
the subsystem. On the other hand, the simplicity of this subsystem
could be attributed to FM's step-child role in a system dominated by
AM—FM combinations, with profits reported jointly, and AM interest in
the television industry.

The introduction of UHF television into the media system in 1945,
as if no freeze had occurred and the decision was made to use both VHF
and UHF bands for commercial television had little effect on the VHF
subsector, indicating its strength and ability to withstand competition.

Hence, under the conditions of the model, UHF stations might always have

167

been less profitable than VHF stations, an argument not accepted by

those UHF proponents who maintain the ranking occurred only because

VHF had an eight-year head start. The subsector equations showed that
advertising expenditures on VHF were predictors of both VHF revenue and
expenses, a logical relationship since a broadcaster would spend more on
programming to attract a large audience which in turn would attract more
advertising dollars. No variables except disposable income and time were
found to be significant predictors of the size of the VHF television
audience. While such variables as receiver supply, receiver costs,
number of AM households, circulation of newspapers and magazines were
included in regression equations in the model development stage, none

of the variables produced significant coefficients. The consumer decis-
ion to watch television, or in actuality to own a television set, must be
based on an entirely different set of variables from those included in
the model.

Lack of predictor variables for the UHF television audience followed
from the experience of the FM subsector. However, like the FM subsector,
the UHF subsector equations contained few endogenous system variables
and relied mostly on time variables. When the subsector functions early,
(Experiment 1) or alone (Experiments 5 and 6), changes in variables were
due primarily to ‘the greater range of UNYR (0 to 15, instead of O to 8).
Also, earnings were not equal to the difference between revenues and
expenses in this subsector. The actual computation of UHF earnings used
this difference, but with a negative coefficient. The relationship held
only for negative income for UHF; results were dramatically different
when UHF incomes turned positive.

As discussed in Chapter 4, advertising buys in the model were

remarkably insensitive to changes in retail profits and hence budget.

168
Also, retail profits and the advertising budget showed an inverse rela-

tionship, with the advertising budget increasing as retail profits
declined. One could argue that a decline in sales, and hence profits,
would lead a businessperson to advertise more in the hope of stimulating
more sales.

At the macroscopic level, overall, the system represented by MASSMOO
showed a remarkable stability during 1945-1960, when television, first
VHF and then UHF, was introduced to the system. The number of magazines
and newspapers published in the period remained fairly constant; revenues
grew slightly as both advertising expenditures on each medium and circu-
lation increased. The print sector was not integrally linked to the
broadcast sector in terms of sensitivity to change. AM stations in-
creased in number and listenership, although earnings declined and
advertising expenditures were transferred to television. FM stations
grew steadily, while revenues and expenses declined, and then took an
upward swing. UHF television seemed the most unstable subsector, with
most variables decreasing in value over the years and earnings in the
red. The experiments performed on the system, which assumed that initial
conditions were the same as the historical situation, indicated that the
system would remain remarkably the same under different policy parameters.
Only a total FM and/or a total UHF broadcasting system would make those
media prominent and profitable, though still not as successful as his-

torical AM radio or VHF television.

Future Directions
The construction of MASSMOD and the execution of experimental manip-
ulations suggest future directions and further research dealing with

systems simulation as a methodology to study the mass media industry and

169

to aid in policymaking. Some suggestions deal with direct modifications
of the model, others with additional experiments which could be per-
formed, and still others with possibilities for MASSMOD beyond its 1960
cutoff date. '

 

Model Improvements

Most of the weaknesses of the model have been discussed in some
detail elsewhere in this document. They are summarized here.

First, better data sources, if they exist, need to be found for the
UHF and FM subsectors to aid in the search for more meaningful prediction
equations using variables of importance to the model. The heavy reliance
on a time variable is not desirable.

The calculation of number of FM stations is an especially poor link
in the FM subsector. Either a different function must replace the ex-
ponential relationship that is used, or other variables of significance
need to be located.

Another data {aroblem exists in the advertiser sector. Compatible
data describing the advertising expenditures and retail profits of the
same group of industries and businesses need to be found to make the
allocation of advertising expenditures more sensitive to changes in
profits. This (also involves the inclusion of expenses of those same
industries so that a real calculation of profits can be made.

The allocation of advertising moneys to radio and television is made
on a two-step basis. First, network, spot and local purchases on all
radio and television stations are computed and totalled. Then, on a
straight percentage basis, 20 percent of the total is assigned to FM and
80 percent to AM. Likewise, 25 percent of the television total is

assigned to UHF, and 75 percent to VHF. A functional relationship,

170

relating percentage and time, should be introduced to put this allocation
on a sliding scale. It is not certain that data exist to describe this
function, however.

Each subsector in the system is relatively autonomous. The search
for variables and relationships that interweave system components must be
continued to make components more sensitive to each other. It is diffi-
cult to imagine that UHF television introduced in 1945 did not have more
of an effect on VHF television. It is also difficult to reconcile the
enormous changes to the FM sector in the spectrum allocation experiments
caused by the faulty FMNO calculation with the relatively small changes
in Theil's 0 when experiments and the benchmark were compared.

UHF television households did not add significance to regression
equations in the benchmark version of the model, when VHF television
households were also included as predictors. This stems from the fact
that UHF audiences duplicate VHF audiences, rather than comprise their
own unique audience. UHF-only television sets did not, and still do
not, exist; However, when the model operates in a UHF television-only
mode, the values for UHF television households need to be substituted
for VHF households, since it is the only television competitor against
print and radio. Allowing the model to run with a value of O for VHF
households in the UHF-only mode is not correct.

The allocation of advertising expenditures among the media should be
made more complex, though still remain stochastic, to introduce an ele-
ment of choice and substitutability among media options. At a more
microscopic level, this would begin to involve optimization of buys in
terms of cost, reach and frequency. A way of expressing this at a macro
level, without target audiences of product specificity, needs to be

found.

171

Finally, the consumer sector should be modified from its determinis-
tic state to a more stochastic model, integrating some of the previous
research on consumer allocations of time to the media with the prediction
of media audiences. This would expand the model from its structural,
economic representation to a behavioral system as well. The task here
is to relate consumer time allocations to the head count of persons and/
or households required by audience and circulation measures.

Also, media costs need to be considered in the subsector, in terms
of subscription costs and receiver expenditures. A cursory attempt was
made to incorporate such cost into audience regression equations, but
the results were not significant. More work needs to be done on this

problem.

 

Further Experiments

Once a basic simulation model has been constructed, the number of
experiments which can be designed for execution is almost limitless.
Included here are further experiments either initially suggested as
part of the current study, or which emerged from this research.

In the area of spectrum use, experiments in this study dealt only
with reduced or increased spectrum assignment to a medium, but did not
consider the width of the broadcast channel. In the light of the debate
over changing AM channels from 10 to 9 kHz, this might make an interest-
ing exploration.

Another area of discussion about the spectrum deals with abandonment
of the traditional FCC licensing procedure in favor of marketplace
economics. Introducing a fee or auction bid for a frequency assignment
into the system would help predict the effect of such a plan, part of a
larger deregulation package circulating among House and Senate Communica-

tion Subcommittees. (U.S. Congress, H.R. 13015, 1978)

172
Two experiments in this study imposed external controls on adver-

tising expenditures. Also mentioned in the rationale for those experi-
ments was the possibility of limiting the amount of advertising time
and space available for purchase. This would involve the advertiser
sector, and focus on supply, demand and advertising costs as they affect
media revenues and growth.

Cross-media ownership (i.e., ownership of a broadcasting entity by
a newspaper, or ownership of two or more different types of broadcasting
media) is a reality assumed to exist and operate in MASSMOD, but it is
never specifically quantified or manipulated. Revenue from cross-owner-
ship interests could explain why earnings are not the simple difference
between revenue and expenses in several subsectors. Given the FCC's
concern and actions regarding broadcast interests owned by newspapers,
it would be informative to quantify investments by one medium to another.
Also, experimentation with the subsidization of one broadcasting medium
by another (VHF television's early support from AM broadcasters, for
example) would generate data of interest to those concerned with newer
communications technologies and their possible subsidization. Manipula-
tion of such a policy variable, once quantified in the model, could
supply output which might address questions such as rate of acceptance
by the public, cost of the technology, effect on existing communications
systems with and without cross-subsidization. The difficulty of executing
such experiments is that, to the researcher's knowledge, no data exist at
the level of aggregation of this model which quantify the extent of cross-
ownership, or deal with its effects. It seems that data from individual
media owners, and/or microanalytical cross-ownership studies, may have to

be located first, and then aggregated for use.

173
The history of the growth of the mass media, both print and broad-

cast, suggests that technological developments and improvements are keys
to industry expansion and improvement. It is clear that until technolog-
ical advances bring the cost of receiving equipment down to consumer-
affordable levels, the medium does not advance. Not until printing
presses were powered by steam and type set by linotype machines instead
of by hand did newspaper circulations begin to reflect acceptance by the
masses who were able to afford a penny per copy.

Competition in the marketplace is a prime motivator of such develop-
ment: at the same time, claims about the high profit ratios of broad-
casters and some publishers have been made. It seems that some form of
control or requirement on research and development investments of broad-
casters and publishers might provide a means of reinvestment for high
profits and speed the growth of the communications technologies, which
need capital to develop and sell. Finding historical research and
development figures and building these into the model might be difficult,

if not impossible.

Model Expansion

 

Perhaps even more important than the additional experiments which
can be run with MASSMOD in the 1945-1960 time period, is the expansion
of the model itself to reflect two more decades of growth and development
within the mass media industry. The selection of the 15-year time
period to study the mass media industry was designed to provide a con-
tained system in which to establish initial structures and data bases.
Now that a model has successfully been constructed and data sources
located, MASSMOD needs to be updated and diversified to include develop-

ments such as color television, cable television, video discs and

174

videocassette recorders and pay television. Such development may be
easier in a two-step, one decade at a time approach, but more thorough
record-keeping on the part of industry trade groups can only make the
task easier than the present research. Once the model is updated to
1980, it can then be used to extrapolate beyond the data, to explore
effects of even newer technologies, such as direct broadcast satellites,
and the growth rate of video recorders and discs.

Another direction for improvement would be movement from a highly
aggregate macroscopic investigation of the system to a less aggregate
microscopic system, where the differences between individual stations,
publications, advertisers and consumers can have an effect on system
behavior. Such disaggregation could make the system more stochastic,
and certainly more complex, but it would also lead to a more detailed
understanding of the structure of the media system.

The suggestions for further research presented in this chapter
reflect the fact that MASSMOD presents a simple, though valid, mathe-
matical representation of the mass media industry from 1945-1960. As
such it has lived up to its purpose as an exploratory effort into the
use of systems simulation as an appropriate tool for study of the
behavior of 'the mass media industry. Supported by additional data, the
model needs to be expanded, and improved, so that mathematical relation-
ships and theoretical speculations which are not included in MASSMOD,
especially in the FM and UHF sectors, may be investigated. To be used
as a real forecasting tool, the model needs to be updated to describe
industry behavior in the 1960's and 1970's.

Despite its shortcomings, MASSMOD has pulled together elements of
a media system too often studied in isolation, and quantified abstract

relationships. It is a simulation of a historical reality, available

175
for experimentation and questioning. Even if one is unwilling to
accept the validity of the model built here, one should agree that this
study has shown that it is possible to build a simulation model of an
industry that can yield interesting, valuable, and sometimes counter-
intuitive insights. Thus, as the minimum, this study has suggested an
approach that someone with access to a rich data base could use to make
accurate predictions of the consequences of policy decisions made to

influence the behavior of the mass media system.

APPENDIX A

TABLE A1

REGRESSION EQUATIONS
NEWSPAPER SUBSECTOR

NSUBREV = 540.1973 + 0.01837 * NEWSCIR(t-1) F2 = 4.414
(.281) (.065) R = .329

P = .065

NEWSNO(t) = 1505.27 + 61.59 * 1n(NEWEARN(t-1)) F2 = 8.438
(.0000) (.162) R = .565

- 0.267 * GNP(t) P = .004

(.007)

NPUBCST(t) = -14021.335 + 1797.32 * 1n(NENSPR(t)) F2 = 152.83
(.0000) (.0000) R = .982
P = .0000

176

TABLE A2

REGRESSION EQUATIONS
MAGAZINE SUBSECTOR

MAGNO(t) = 204.41733 - 0.0138 * MAGREV(t-l)
(.013) (.666)
+ 0.144 * GNP(t)
(.066)

MPUBCST(t) = 329.17166 + 0.003043 * MAGCIRC(t-1)
(.000) (.022)
+ 1.3181 * GNP(t)
(.000)

MAGCOST(t) = 5.5580 + 0.00104 * MPUBCST(t)
(.000) (.004)

MAGREV(t) = 1073.93 + 0.5794 * (MSUBREV +
(.005) (.002) TMAGBUY(t))

177

132)-n wax-n 'uzo-n

‘oJU-n

130.59

441

12

51

.989
.001

.19
.988
.000

.401
.508
.004

.56
.927
.002

AMNO(t)

AMEXP(t)

AMREV(t)

AMEARN(t)

TABLE A3

REGRESSION EQUATIONS
AM RADIO SUBSECTOR

= 2727.4112 - 3.1284 * GNP(t)

(.017) (.113)
- 1.7826 * AMEARN(t-l)
(.152)
+ 213.4629 * RNYR
(.000)

= 513.0369 + 0.0000199 * AMHH(t-l)

.0000) (.0000)
+ 0.07119 * AMNO(t)
(.000)

= 118.423 + 0.760 * AMBUY(t)
1

. 94) (.000)
+ 0.689 * (RNYR**2)
(.000)

= 59.059 + 0.49 * AINCOME

(.194) (.001)

178

'UFD'TI '030‘" 112311

'UFU‘TI

144

39.
.8559
.000

35

16

.805
.973
.000

738

.662
.856
.000

.048
.534
.001

TABLE A4

REGRESSION EQUATIONS
FM RADIO SUBSECTOR

FMNO(t) = exp(7.2081 - 0.09624 * FMEARN(t-1)
(.000) (.008)
+ 1.3663 * 1n(RNYR)
(.0000)
- 0.00633 * GNP(t))
(.000)
FMREV(t) = - 16.518 + 0.086 * FMBUY(t)

(.000) (.000)
+ 0.00023 * (RNYR**3)
(.000)

FMEXP(t) = 11.42477 + 0.01899 * (RNYR**3)
(.000) (.000)

- 0.28225 * (RNYR**2)
(.000)

179

'UJU'VI

88.93
.9557
.000

296.63
.983
.000

86.478
.945
.000

TABLE A5

REGRESSION EQUATIONS
VHF TELEVISION SUBSECTOR

VHFNO(t) = -177.1866 + 0.3032 * GNP(t) F2 = 82.177
(.299) (.388) R = .965
+ 0.62554 * VHFEARN(t-Z) P = .000
(.060)
+ 20.4162 * RNYR
(.014)
VHFREV(t) = -24.63 + 0.48 * VHFBUY(t) F2 = 120.11
(.649) (.000) R = .923
P = .000
VHFEXP(t) = -32.78 + 0.489 * VHFBUY(t) F2 = 31.552
(.780) (.072) R = .875
- 0.0053 * VHFHH(t-1) P = .000

(.519)

180

TABLE A6

REGRESSION EQUATIONS
UHF TELEVISION SUBSECTOR

UHFNO(t) = 128.21031 - 0.55432 * UHFEARN(t-l)
(.000) (.083)
+ 0.87311 * (UNYR**2)
(.016)
- 13.8565 * UNYR
(.0030)

UHFEXP(t) = 26.6837 - 0.001047 * UHFHH(t-1)
(.000) (.114)
+ 0.25604 * UHFNO(t)
(.000)

UHFREV(t) = 34.317 - 0.0216 * UHFBUY(t)
(.000) (.238)
+ 0.274 * UHFNO(t)
(.021)

UHFEARN(t) = -5.380 - 2.395 * (UINCOME**2)
(.009) (.048)
- 6.030 * UINCOME
(.004)

181

'UJU‘I'I 'UJU'I‘I 'UZJ'H

“um-n

238

45

10

12

.57
.996
.000

.47
.948
.001

.93
.814
.015

.96
.838
.011

AVNEW

AVMAG

AVNRD

AVSRD

AVLRD

AVNTV

TABLE A7

REGRESSION EQUATIONS
ADVERTISER SECTOR

= 24.285 + 0.00263*NEWSCIR(t-1)
(.0000) (.000)

- 0.0667*NEWSNO(t-1)
(.000)

= 9.194 + 0.0000183*MAGCIRC(t-1)
(.000) (.610)

+ 0.0000696*VHFHH(t-1)
(.016)

- 0.0111*MAGNO(t-1)
(.593)

= 5.3917 - 0.00004251*AMHH(t-1)
(.000) (.000)

- 0.00006839*VHFHH(t-1)
(.000)

= 2.4464 - 0.000007685*AMHH(t-1)
(.0000) (.059)

- 0.00003470*VHFHH(t-1)
(.009)

+ 0.01238*(RNYR**2)
(.000)

= 2.881 + 0.0456*SQRT(RADNO(t-1))
(.0000) (.000)

= 2.94781 + 0.0004359*VHFHH(t-1)
(.0000) (.0000)

- 0.044057*(RNYR**2)
(.000)

182

'o:D'n

vac-n

'Dm'l'l

'UW'TI

.074

.912
.000

.788

.927
.000

N
N
(A)

.052

.974
.000

.728

.887
.000

0’}
01

.245

.824
.000

1796.

46

.998
.000

AVSTV

AVLTV

BUDGET

TABLE A7 (cont'd)

= 1.20097 + 0.0001101*VHFHH(t-1)
(.000) (.004)
+ 0.009543*(RNYR**2)
(.188)

= 1.562 + 0.0090*TVNO(t-1)
(.000) (.017)

- 0.0080*(RNYR**2)
(.368)

= 4474.89 - 0.00l3*RETPROF(t-l)
(o) (.375)
+ 413.43*RNYR
(0)

183

1330-31 '07011

vac-n

434.671
.990
.000

35.443
.887
.000

358.446
.982
.000

TABLE A8

REGRESSION EQUATIONS
CONSUMER SECTOR

ADULTPC = 0.709 - 0.0045*RNYR F2 = 439.31
(.0000) (.000) R = .969
P = .000
AMHHPC = 1.039 - 0.0000127*AMNO(t) F2 = 34.073
(.0000) (.277) R = .883
- 0.00000862*AMSETS(t) P = .000
(.004)
FMHHPC = 0.043 + 0.093*1n(RNYR) F2 = 3758.51
(.097) (.0000) R = .998
- 0.000018*DISINC(t) P = .000
(.121)
VHFHHPC = -1.058 - 0.0004*(RNYR**3) F2 = 176.149
(.219) (.004) R = .985
+ 0.00045*DISINC(t) P = .000
(.232)
+ 0.0097*(RNYR**2)
(.002)
UHFHHPC = 0.0081 + 0.042*1n(UNYR) F2 = 37.616
(.399) (.001) R = .862
- GROWTH P = .001
NEWSPC = 0.463 + 0.0000504*NEWSNO(t) F2 = 13.975
(.006) (.448) R = .792
- 0.00326*NEWCOST(t) P = .000
(.000)
+ 0.0000177*DISINC(t)
(.054)

184

TABLE A8 (cont'd)

MAGPC = -0.536 + 0.00582*MAGNO(t) F2 = 158.067
(.001) (.000) R = .963
+ 0.000217*DISINC(t) P = .000

(.120)

185

APPENDIX B

MASSMOD--Main Program

PROGRAH HASSHOD(INPUTyOUTPUTvTAPEb'OUTPUT)

INPLICIT REAL(A-Z)

INTEGER NITvNYR

COHHON/EXOB/GNP(16)

COHHON/NEUDATA/NEUCOST(16)vNEHSNO(16)vNPUBCST(16)vNEHSREV<16)v
NEUEARN<16)vNEUPROF<16lvNEHSCIR(16)vTNEUBUY(16)

COHHON/HAODATA/HAGNO(16)oHPUBCST<16)vHAGCOST(16)9HAGREU(16)v
MAGEARN(16)vHAGPROF(16)oHABCIRC(16)vTHAGBUY(16)

COHHON/AHDATA/AHNO(16>vAHREV(16)9AHEXP(16)vAHEARN(16)vAHPRDF(16)r
AMHH(16)vAHBUY(16)

COHHON/FHDATA/FHNO(16)yFHREV<16>9FHEXP(16)9FHEARN(16)vFHPROF(16)v
FHHH<16)9FHBUY(16)

COHNON/VHFDATA/VHFNO(16)vVHFREV(16)vUHFEXP(16)90HFEARN(16)9
UHFPROF(16)oUHFHH<16)vUHFBUY(16)

COHHON/UHFDATA/UHFNO(16)vUHFREU(16)vUHFEXP(16)9UHFEARN(16)v
UHFPROF<16)9UHFHH(16)vUHFBUY<16)

COHHON/ADVDATA/TBUY(16)oRETPROF(16)

CDHHON/POPDATA/POP(16)vPOPHH(16)vPOPADLT(16)

++++++

COHHON/OUNDATA/CROUN .
DATA GNP/564.1D477.6946893v487o7949007953305957695'59805'621.99
+ 613.79654o89668099680099679.57720o4v73608/
DATA CROUN/Oo/
C
C SET SIMULATION RUN LENGTH
C
RLNGTH-169
”7.1 o
NIT'RLNBTH/DT + 0.00001
C
C EXECUTION
C

no 30 NYRIlvNIT
CALL ADSEC(NYR)
CALL PR8EC<NYR)
CALL RADSEC<NYR)
CALL CONSEC<NYR)
CALL UHFSEC(NYR)
CALL UHFSEC(NYR)
3O CONTINUE
CALL REPORT
END

186

Advertiser Sector-~Subprogram

SUDROUTINE ADSEC(NYR)

INPLICIT REAL(A-Z)

CONHON/X/PARAN(12874)vISEED

COHHON/NEUDATA/NEUCOST(16)9NEUSNO(16)rNPUDCST<16)vNEHSREV(16)v

+ NEUEARN(16)9NEHPROF(16)9NEUSCIR<16)9TNEUBUY(16)

COHHON/HAGDATA/HAGNO(16)vNPUDCST(16)vHAGCOST(16)vHAGREU(16)r
NAOEARN(16)9HAGPROF(16)9NAGCIRC<16>vTHAGDUY<16)

COHNON/AHDATA/AHNO(16)vAHREU(16)vAHEXP(16)vAHEARN(16)vAHPROF<16)r
AHHH(16)9ANDUY(16)

COHHON/FHDATA/FHNO(16)IFNREU(16)vFHEXP(16)IFHEARN(16)9FHPROF(16)9
FHHH(16)7FHDUY(16)

CONNON/VHFDATA/VHFNO(16)vUHFREV(16)vUHFEXP(16)OVHFEARN(16)v
VHFPROF(16)rUHFHH(16)'UHFDUY(16)

COHNON/UHFDATA/UHFNO(16)vUHFREV(16)9UHFEXP(16)vUHFEARN(16)9
UHFPROF(16)vUHFHH(16)vUHFDUY(16)

CONNON/ADUDATA/TDUY(16)IRETPROF(16)

INTEGER ISEEDvNYRvNITvKKK9NYRNEU9NYRHAGONYRNTVONYRSTVyNYRLTUo

+ NYRNRDrNYRSRDvNYRLRDvIvJ
DIMENSION SALES(16)9TVNO(16)vRADNO(16)

DATA SALES/207537.v238729.9246290o9251637.1254340.!274651o'

+ 273208.9279919ov237087.9283308o9301395.9301636.,

+ 307695.9303565or319130o7319547o/
DATA((PARAN(I9J)vI'I!128)9J.1!2)/256I0o/
DATA((PARAH(I9J)71.11128)vJ'3v4)/26060128o59v32.19134.84v38o02!

+40.I941o01943o97945o84v46o06951o64952o94951.66950.954.07754o599

+273459168o47o19.92945o20v43o73944o78949.42919.62128o34931o66v
+39o20938o60953o85933o50944o44v55o31939.68!109.66933o50132o02'
+31o65929o85964o61v27o99v74o58I20o65v16.39!12.71713o08v9o09r
+7.4197.28926o60925o06924.14924.48028o52929o85927o92v31.039
+32026'26080'2602303108073804603903904.04"46058939089’45056'
+5003'560595700395907'61008,68097'67091I65033'65057965018964062!
+65o15966o67r65o593t0o'1.795o7715.86937.03124o48953o92183o859154.82

+9690°897407991350159110043'67054'3‘00 0005110999033913009'17924!

+16.98926o8932.79939.75939o86960o61166.73058o227330o71.093.8911o19v

+13o96917o24723o77930o15936.89940o22937.54937o89940o940.760
+1639.916*3.11632.932t5o'1633.932*6o/

DATA ISEED/1234/

DATA NEHSCRIIHAGCIRIrAHHHIIVHFHHI/45955o9115967.r32500ovool
DATA NEUSNOI9HAGNOIrRADNOIvTVNOI/1857.7217.0954.vao/

+++++

C
C LOCATE SPECIFIC MEDIA IN PARAM ARRAY

C

NYRNEU-NYR
NYRNAG'NYR+16
NYRNRD-NYR+32
NYRSRD-NYR+48
NYRLRD-NYR+64
NYRNTUINYR+80
NYRSTVINYR+96
NYRLTVINYR+112
RNYR-NYR-l

IF(NYR .NE. 1) GO TO 1

187

C
C
C

Advertiser Sector--Subprogram (cont'd)

INITIALIZE FIRST YEAR --LAGBED VARIABLES

PARAN(NYRNEUQI)-(24o285 + 0.00263INEHSCRI - 0.0667‘NEUSNOI)/
PARAH(NYRNE994)

+PARAH(NYRHA891)I(9.194 + 0. 00001838HAOCIRI + 0. 000069610HFHHI
-0o 0111*HAGNOI)/PARAH(NYRHAGv4)

+PARAM<NYRNRDv1)'(5 3917 - 0.000042511AHHHI - 0. 00006839!
UHFHHI)/PARAH(NYRNRDr4)

PARAH(NYRSRDOI)O(2.4464 - 0.0000076853AHHHI * 0.000034708
+

UHFHHIllPARAH(NYRSRDvA)
PARAH(NYRLRD91)I(2.881 + 0.0456tSORT(RADNOI))IPARAH(NYRLRDv4)
PARAM(NYRNTVII)'(2.94781 + 0.0004359SUHFHHI)/PARAH(NYRNT094)
PARAH(NYRSTV!1)I(1o20097 + OoOOOIIOIIUHFHHI)/PARAH(NYRST094)
PARAH(NYRLT091)I(1o562 + 0.009OITVNOI)IPARAH(NYRLTUv4)
GO TO 2
RADNO<NYR-1)IAHNO(NYR-1) + FMNO(NYR-l)
TVNO(NYR-1)-UHFNO(NYR-1) + UHFNO(NYR-I)
nvwew-24.2as + 0.002633NEUSCIR(NYR-1) - 0.0667INEUSNO(NYR-1)
PARAH(NYRNEUv1)OAUNEU/PARAH(NYRNE994)
AVMAG-9.194 + 0.0000183!HAGCIRC(NYR-1) + 0.000069680HFHH(NYR-1)

-0.0111$HAGNO(NYR-1)
PARAH(NYRHA891)IAUHAG/PARAH(NYRNA894)
AVNRD-5.3917 -0.00004251‘AHHH(NYR-1) - 0.0000683930HFHH(NYR-1)
PARAH(NYRNRDII)IAUNRD/PARAH(NYRNRD94)
AVSRD-2.4464 - 0.0000076851AHHH(NYR-1) - 0.0000347OIVHFHH(NYR-1)
+ 0.0123BI(RNYR882)

PARAH<NYRSRD91)-AUSRD/BARAH(NYRSRD94)
AVLRD-2.881 + 0.0456iSORT(RADNO(NYR-1))
PARAH(NYRLRDv1)'AVLRD(PARAH(NYRLRD94)
AVNTU'2o94781 + 0.0004359IUHFHH(NYR-1) - 0.0440573(RNYR*82)
PARAH<NYRNTVr1)-AUNTU/PARAH(NYRNT094)
AUST081.20097 + 0.000110110HFHH(NYR-1) + 0.009543¥(RNYR$12)
PARAH(NYRST991)IAVSTU/PARAH(NYRST094)
AVLTVI1.562 + 0o0090iTUNO(NYR-1) - 0.0080*(RNYR332)
PARAH(NYRLT091)-AULTV/PARAH(NYRLTV94)

2 BUDGETIO.007317ISALES(NYR) + 649.046881RNYR + 4894.7131

C
C CALCULATE TOTAL AD BUYS FOR IOOADUERTISERS

C

THAOBUY(NYR)IOo

TNEUBUY<NYR)-O.

TNRDBUY-0.

TSRDBUY-0.

TLRDBUYIO.

TNTVBUY-0.

TSTVBUY-0.

TLTVBUY'O.

TBUY(NYR)-Oo

DO 20 KKK-19100
TNEUBUY(NYR)-TNEUBUY(NYR) +ERLNG(NYRNEU)
TMAGBUY(NYR)-THAGBUY(NYR) + ERLNG(NYRHAG)
TNRDBUY-TNRDBUY +ERLNG<NYRNRD)
TSRDBUY-TSRDBUY +ERLNG(NYRSRD)
TLRDBUY-TLRDBUY +ERLNG(NYRLRD)

188

20

31

32
33

Advertiser Sector--Subprogram (cont'd)

IF(NYR oLT. 4) GO TO 20

TNTVBUY-TNTVBUY + ERLNG(NYRNTU)

TSTVBUY-TSTVBUY +ERLNG<NYRSTV>

TLTVBUY-TLTVBUY + ERLNO(NYRLTV)

CONTINUE

TDUY(NYR)ITNEUBUY(NYR) + TMAGBUY(NYR) + TNRDBUY + TSRDBUY+
TLRDBUY + TNTVBUY + TSTVBUY + TLTVBUY

GHOST-BUDGET - TBUY(NYR)

RADTOTITNRDBUY + TSRDBUY + TLRDBUY

TUTOT-TNTUBUY + TSTVBUY + TLTVBUY

AHEUY(NYR)IOoOODRADTOT

FHBUY(NYR)IO.20#RADTOT

IF(NYR - 8) 31932932

VHFBUY(NYR)ITVTOT

UHFBUY(NYR)IO.

GO TO 33

UHFBUY(NYR)‘0.75#TVTOT

UHFBUY(NYR)‘O.25*TUTOT

RETPROF(NYR)-SALES(NYR) - TBUY(NYR)

RETURN

\lbflb

Random Number Generator Functions

FUNCTION DRAND(I3EED)
DRAND-RANF(ISEED)

RETURN

END

FUNCTION ERLNG(J)
COHNON/X/PARAH(12894)9ISEED
K-PARAH(J94)

IFCK-I) 19292

PRINT 109J

FORMAT(/16H K30 FOR ERLNG 9 I7)
RETURN

R-I

DO 3 I-19K

RNUN'DRAND‘ISEED)

R-R 1 RNUH

ERLNG. -PARAH(J91)IALOO(R)
IF(ERLNG - PARAH(J92)) 49596

ERLNG-PARAH(J92)

RETURN .

IF(ERLNG - PARAH(J93)) 59597
ERLNG'PARAH(J93)

RETURN

END

190

+

+

Print Sector--Subprogram

SUBROUTINE PRSEC(NYR)

INPLICIT REAL(A-Z)

INTEGER NYR

COHHON/NEUDATA/NEUCOST<16)9NEUSNO<16>9NPUBCST<16)9NEUSREU(16)9
NEUEARN<16)9NEUPROF(16)9NEHSCIR(16)9TNEUBUY(16)

COHNON/NAGDATA/HAGNO(16)9NPUBCST<16)9HABCOST(16)9HAGREV(16)9
HAGEARN<16>9HA8PROF(16)9HAOCIRC(16)9THAOBUY<16>

COHNON/OUNDATA/CROUN

COHHON/EXOO/GNP(16)

DIMENSION NEUSPR<16)

DATA NEUSPR/3237993995o94420994781995142.9552199555799556999

57139 957329 961739 963209 963009 960599 965789 968°09/
DATA NEUSCRI9NEARNI9NABCIRI9HEARNI/45955.93148.9115967.92050./

C
C INITIALIZATION FOR YEAR ONE LAOOED UARIABLES

C

C

IF(NYR 9NE9 1) GO TO 10

NSUBREU-540o1973 + 0.018379NEUSCRI
NEUCOST(1)-NSUBREV/NEUSCRI81000o

NEUSNO(1)-1505o27 + 61.59!ALOG(NEARNI) - 0.267XGNP(1)
NPUBCST(1)-329.1717 + 0.003043IHA8CIRI + 1.3IS1XGNP(1)
HAGCOST(1)I5.5580 + 0.001043HPUBCST(1)
NAONO(1)-204o41733 -0.013SSNEARNI + 0.14486NP(1)
HSUDREU-(HAOCOST(1)RHAGCIRI)/1000.

GO TO 20

C EXECUTION PHASE

C

10

20

N8UDREV‘540.1973 + 0.01837¥NEUSCIR(NYR-1)
NEUCOST(NYR)INSUDREU/NEHSCIR(NYR-1)31000.
NEUSNO(NYR)-1505.27 + 61.599ALOO(NEUEARN(NYR-1)) - 0.267

+ IONP(NYR)

NAONO(NYR)-204o41733 - 0.01381NAGREU(NYR-1) + O.144!GNP(NYR)
NPUBCST(NYR)-329o17166,+ 0o003043¥flAOCIRC<NYR-1) + 1.318138NP(NYR)
NAGCOST(NYR)IS.5580 + 0.001049HPUDCST(NYR)
HSUBREU-(HABCOST(NYR)*HAOCIRC(NYR-1))/1000o

NPUBCST(NYR)O -14021.335 + 1797.32!ALOO(NEUSPR(NYR))
NEUSREU(NYR)INSUBREU + TNEHBUY(NYR)
NEHEARN(NYR)INEUSREU(NYR) - NPUBCST(NYR)
NEUPROF(NYR)INEHEARN(NYR) +CROUN

HAGREV<NYR)IIO73.93 + 0.5794¥(HSUDREU + TNAGDUY(NYR))
NAGEARN<NYR)IHAGREU(NYR) - NPUBCST(NYR)
HAOPROF(NYR)IHAGEARN(NYR) +CROHN

RETURN

191

GOO

000

Radio Subsector--Subprogram

SUDROUTINE RADSEC(NYR)

CONHON/AHDATA/ANNO(16)9AHREU(16)9AHEXP(16)9AHEARN(16)9AHPROF(16)9
+ AHHH(16)9ANBUY(16)

CONNON/FHDATA/FNNO(16)9FNREV(16)9FHEXP(16)9FHEARN(16)9FHPROF(16)9
+ FNNH(16)9FHDUY(16)

CONHON/EXOB/GNP(16)

COHHON/OUNDATA/CROUN

DATA ANEARNI9AHHH19FNEARNI/200o932500.9-5.0/

INITIALIZATION

RNYR-NYR-l

IF(NYR .NE. 1) GO TO 10

AHNO<1)82727.4112 - 3.1284I8NP(1) - 1.78263AHEARNI + 313.4629
AHEXP(1)'513.0369 + 0.000199IAHHHI + 0.07229IANNO(1)
FHNOLN-7o2081 - 0.096243FHEARNI - 0.00633918NP(1)
FHNO(1)'EXP(FHNOLN)

GO TO 20

EXECUTION PHASE

10 AHNO(NYR)32727o4112 - 3.12848ONP(NYR) - 1.7826lAHEARN(NYR-1)
+ + 213.46293RNYR
AHEXP(NYR)-513.0369 + 0.00001993AHHH(NYR-1) + 0.07229IAHNO(NYR)
FNNOLN-7o2081 - 0.096248FNEARN(NYR-1) + 1.3663*ALOO(RNYR)
+ - 0.00633IONP(NYR)
FNNO(NYR)-EXP(FHNOLN)
20 AHREV(NYR)-118.423 + 0o760¥AHDUY(NYR) + 0.6893(RNYR332)
AINCOHE!AHREU(NYR) - AHEXP<NYR)
ANEARN(NYR)859.059 + 0.49IAINCOHE
ANPROF<NYR)IAHEARN(NYR) +CROUN
FHREU(NYR)I -16o518 + 0.0863FHBUYCNYR) + 0o0023¥<RNYR383)
FHEXP(NYR)III.42477 + 0.018993(RNYR¥¥3) - 0.282251(RNYR*¥2)
FNEARN(NYR)-FHREV(NYR) - FMEXP(NYR)
FHPROF<NYR)IFNEARN(NYR) + CROUN
RETURN

192

 

VHF Television Subsector--Subprogram

SUDROUTINE UNFSEC(NYR)

COHHON/EXOG/ONP(16)
COHHON/VHFDATA/UHFNO(I6)9VHFREV(16)9VHFEXP(16)9VHFEARN(16)9
+ UHFRROF‘16)9VHFNH(16)9VHFDUY(16)
COHHON/OUNDATA/CROUN

RNYR-NYR‘I

IF(NYR oEOo 1) GO TO 4

IF(NYR 9LT. 4)OO TO 3

UHFNO(NYR). '17792866 9 09303238NP(NYR) + 096254‘VHFEARN(NYR-2)
+ + 2094162’RNYR

VHFREV(NYR)‘ ’24963 + 094O'UHFDUT(NYR)

VHFEXP(NYR). -32978 + 0.489’UHFDUY(NYR) ‘ 090053‘9HFNH‘NYR-1)
VHFEARN(NYR)‘VHFREU(NYR) - UHFEXP(NYR)
VHFPROF(NYR)-VHFEARN(NYR) + CROUN

RETURN

VHFNO(NYR)' -1499156 + 0940A3‘VHFERNI + 09343’ONP(NYR)

GO TO 5

UHFNO(NYR)-09

VHFEXP(NYR)-09

VHFREV(NYR)-09

VHFEARN(NYR)-Oo

VHFPROF(NYR)'09

RETURN

END

193

UHF Television Subsector--Subprogram

SUDROUTINE UHFSEC(NYR)
COHNON/UHFDATA/UHFNO(16)9UHFREU<16)9UHFEXP<16)9UHFEARN(16)9
+ UHFPROF(16)9UHFHH(16)9UHFBUY<16)
CONNON/OUNDATA/CROUN

UNYR-NYR-9

IF<NYR .LT. 9) 80 TO 1

UHFNO(NYR)-128921031 - O.55432¥UHFEARN<NYR-1) + 0.873113(UNYR¥X2)
+ - 13.85653UNYR

IF (NYR .EOo 9) UHFNO(NYR)-6o

UHFEXP(NYR)826.6837 - 0.001047IUHFHH(NYR-1) + Oo25604£UHFNO<NYR)
UHFREV<NYR)I34.317 -0.0216*UNFDUY(NYR) + 0.2748UHFNO(NYR)
UINCONEIUHFREU(NYR) - UHFEXP(NYR)

UHFEARN(NYR)I -5.3BO -2.39SS(UINCOHE382) -6.030*UINCONE
UHFPROF(NYR)-UNFEARN(NYR) +CROUN

RETURN

UHFNO(NYR)-0o

UHFEXP(NYR)-Oo

UHFREV(NYR)-0.

UHFEARN(NYR)'0.

UHFPROF(NYR)-Oo

RETURN

END

194

Consumer Sector--Subprogram

SUDROUTINE CONSEC(NYR)

.INPLICIT REAL(A-Z)

INTEGER NYR

COHHON/NEUDATA/NEHCOSTC16)9NEHSNO(16)9NPUBSCT(16)9NEUSREV(16)9

NEHEARN(16)9NEUPROF<16)9NEUSCIR(16)9TNEHDUY(16)

COHHON/HAODATA/HAONO(16)9HPUBCST(16)9HAGCOST(16)9HAGREU(16)9

NAOEARNC16)9HAOPROF(16)9NAGCIRC(16)9THAODUY(16)
CONNON/ANDATA/AHNO(16)9AHREV(16)9AHEXP(16)9ANEARN(16)9AHPROF<16)9
AHHH(16)9ANDUY(16)
COHNON/FHDATA/FNNO(16)9FHREV(16)9FNEXP(16)9FNEARN(16)9FNPROF(16)9
FNHH(16)9FHDUY(16)
CONNON/UHFDATA/VHFNO‘16)9VHFREU(16)9UFEXP(16)9UHFEARN‘16)9
UHFPROF<16)9UHFHH(16)9VHFDUY(16)
COHHON/UHFDATA/UHFNO(16)9UHFREU(16)9UHFEXP(16)9UHFEARN<16)9
UHFPROF(16)9UHFHH(16)9UHFDUY(I6)

COHNON/POPDATA/POP(16)9POPHH(16)9POPADLT(16)

DIHENSION DIRTHR(16)9DEATHR(16)9HHSIZE(16)9ANHHPC(16)9FHHHPC(16)9
UHFHHPC(16)9UHFHHPC<16)9ANSETS(16)9DISINC(16)9
ADULTPC(16)9NAOPC(16)9NEUSPC(16)

+DATA BIRTHR/20o 63924935927. 12925. 33924. 93924. 50925. 33925. 539

259 42925. 75925. 44925. 65925. 76924. 91924. 40923 80/
+DATA DEATHR/10. 72910. 10910. 30910. 0799 8799. 7699. B799. 7799 7699. 369
9 4799 5799 7799 6699.5699 54/

DATA HHSIZE/3.409396193.6393o569395093.4793o4493o4493o4393.459
3.409393993o3993o4493o3593929/

DATA AHSET3/14021909144B4o591046595996309395961o298174o695974o39
4043.1944039093067.69339398935019094100o095100909
61009 097200.0/

+DATA DISINC/23009 09 2342.9 2203.92279o92244o92377.92399.92425o9
2482.92467.92567o92634o92639.9262 59926859926979/

++++++++

++++

C
C INITIALIZATION
C
RNYRONYR-1
UNYR‘NYR-9
GROUTHI0.0
IF<NYR .NE. 1) GO TO 10
POP44-13B397.
POP(1)-POP44 + (BIRTHR(1)$POP44 - DEATHR(1)¥POP44)/1000.
GO TO 1
1O POP(NYR)IPOP(NYR-1) + (BIRTHR(NYR)*POP(NYR-1) - DEATHR(NYR)#
+ POP(NYR-1))/1000.
1 POPHH(NYR)-POP(NYR)/HHSIZE(NYR)
ADULTPC(NYR)-Oo709 - 0.00458RNYR
POPADLT(NYR)IADULTPC(NYR)3POP(NYR)
C
C CALCULATE HOUSEHOLD PERCENTAGES FOR BROADCAST MEDIA
C

AHHHPC(NYR)-1.039 - 0.0000127XAHNO(NYR) - 0.00000862XAHSETS(NYR)
IF (NYR .GT. 2) GO TO 2
FHHHPC<NYR)-0o01
IF(NYR .GE. 3) GO TO 3
VHFHHPC(NYR)IO.OOO2
GO TO 5
2 FHHHPC(NYR)-0.043 + 0.093!ALOG(RNYR) - 0.0000181DISINCINYR)

195

Consumer Sector--Subprogram (cont'd)

3 UHFHHPC(NYR)I -1.058 - 0.00043‘RNYR883) + 0.00045IDISINC(NYR)
+ + 0.00973(RNYR3‘2)
IF(UHFHHPC(NYR))49595
4 UHFHHPC(NYR)-0.009
5 AHHH(NYR)8AHHHPC(NYR)!POPHH(NYR)
FHHH(NYR)-FHHHPC(NYR)*POPHH(NYR)
VHFHH(NYR)-UHFHHPC(NYR)1POPHH(NYR)
IF(NYR .LT. 10) GO TO 6
IF (UNYR .OE. 4) GROHTH- -0.0394 - 0.00144IUNYR + 0.001443
+ (UNYR132)
UHFHHPC(NYR)'0.0081 + 0.042IALOO(UNYR) - BROHTH
UHFHH(NYR)-UHFHHPC(NYR) t POPHH(NYR)
GO TO 7
6 UHFHH(NYR)-0.
UHFHHPC(NYR)-0.

C
C CALCULATE ADULT PERCENTABES FOR PRINT MEDIA
C

7 NEHSPC(NYR)-O.463 + 0.00005043NEHSNO(NYR) - 0.003269NEUCOST(NYR)
+ + 0.00001778DISINC(NYR)
HAGPC(NYR)I -0.536 + 0.005823HA6NO(NYR) + 0.0002173DISINC(NYR)
NEHSCIR(NYR)-NEHSPC(NYR)8POPADLT(NYR)
MAGCIRC(NYR)IHAOPC(NYR)!POPADLT(NYR)
RETURN
END

196

Report Subprogram

SUDROUTINE REPORT
COMMON/NEUDATA/NEHCOST(16)9NEUSNO(16)9NPUBCST(16)9NEUSREV(16)9
NEUEARN(16)9NEUPROF<16)9NEHSCIR<16)9TNEUBUY(16)
COMMON/MAGDATA/MABNO(16)9MPUBCST(16)9MAGCOST(16)9MAOREV(16)9
MAGEARN(16)9MAOPROF(16)9MAGCIRC(16)9TMAGDUY(16)
COMMON/AMDATA/AMNO(16)9AMREU<16)9AMEXP(16)9AMEARN<16)9AMPROF(16)9
AMHH(16)9AMBUY(16)
COMMON/FMDATA/FMNO(16)9FMREV(16)9FMEXP(16)9FMEARN(16)9FMPROF(I6)9
FMHH(16)9FMDUY(16)
COMMON/UHFDATA/UHFNO(16)9UHFREV(16)9UHFEXP(16)9VHFEARN(16)9
UHFPROF<16)9VHFHH(16)9VHFDUY(16)
COMMON/UHFDATA/UHFNO(16)9UHFREV(16)9UHFEXP(16)9UHFEARN(16)9
UHFPROF(16)9UHFHH(16)9UHFDUY(16)
COMMON/ADVDATA/TDUY(16)9RETPROF(16)
COMMON/POPDATA/POP(16)9POPHH(16)9POPADLT(16)
INTEGER NYEAR9NYR
HRITE<69100)
100 FORMAT(1H1930X93A SIMULATION OF THE MASS MEDIA INDUSTRY 1945-1960
+t9/l)
URITE(69110)
110 FORMAT<1X98PRINT SUBSECTOR OUTPUT¥9/)
URITE(69200)
200 FORMAT<1X9IYEARI 3X93NEUSNO$ 4X93NEHSCIR1 2X99NEUCOST3 6X9INEUSREV
+9 6X9INPUDCSTI 6X91NEHEARN$ 6X9tNEHPROFtvl)
DO 10 NYR-1916
NYEAR-NYR+1944
URITE(69210)NYEAR9NEUSNO(NYR)9NEUSCIR(NYR)9NEUCOST(NYR)9
+ NEHSREU(NYR)9NPUDCST<NYR)9NEUEARN(NYR)9NEUPROF(NYR)
210 FORMAT(1X9I494X9F5. 094X9F7. 094X9F5. 294(4X9F9. 3))
10 CONTINUE

++++++

URITE(69220)
220 FORHAT(//91X9SYEARI 4X9IMAONOI 4X9 IMAOCIRCS 2X9 DMAOCOSTI 7X9
+ IMAGREUI 6X9¥MPUDCST¥ 6X9IMAGEARNS 6X9IMAOPROFI9I)
DO 20 NYR'1916
NYEAR-NYR+1944
URITE(69210)NYEAR9MAONO(NYR)9MAOCIRC(NYR)9MAOCOST(NYR)9
+ MAGREV(NYR)9NPUDCST(NYR)9MABEARN(NYR)9MAOPROFCNYR)
20 CONTINUE
URITE(69120)
120 FORMAT(///91X9$RADIO SUBSECTOR OUTPUTS9I)
URITE(69300)

300 FORMAT(1X93YEAR* 5X9$AMNOS 6X93AMHH3 BX9SAMREU! BX9SAHEXP¥ 7X9
+ IAMEARNS 7X9¥AMPROEi9/)
DO 30 NYR‘1916
NYEAR-NYR+1944
URITE(69310)NYEAR9AHNO(NYR)9AMHH(NYR)9AMREU(NYR)9AMEXP<NYR)9
+ AMEARN(NYR)9AMPROF(NYR)

310 FORHAT(1X9I494X9F5.094X9F6.094(4X9F9.3))

30 CONTINUE

URITE(69320)

320 FORMAT(//91X9¥YEARI 5X93FMNOX 6X93FMHH¥ 8X9 *FMREUI 8X9 *FMEXP* 7X9
+ *FMEARNI 7X9 lFMPROF¥9/)

197

Report Subprogram (cont'd)

DO 40 NYR-1916
NYEAR-NYR+1944
URITE(69310)NYEAR9FMNO(NYR)9FMHH(NYR)9FMREV(NYR)9FMEXP(NYR)9
+ FMEARN(NYR)9FMPROF(NYR)
40 CONTINUE
HRITE<69130)
130 FORMAT(///91X9¥VHF SUBSECTOR OUTPUTI9I)
URITE(69400)
400 FORMAT<1X9¥YEARI 4X930HFNO$ 5X9tVHFHHI 7X930HFREU$ 7X9XUHFEXPI
+ 6X9OVHFEARN3 6X930HFPROFivl)
DO 50 NYR'1916
NYEAR-NYR+1944
URITE<69310)NYEAR9UHFNO(NYR)9VHFHH(NYR)9VHFREV(NYR)9UHFEXP<NYR)9
+ VHFEARN(NYR)9UHFPROF(NYR)
50 CONTINUE
URITE(69140)
140 FORMAT(///91X98UHF SUBSECTOR OUTPUTI9I)
URITE<69500)
500 FORMAT(1X9IYEAR1 4X9IUHFNO¥ 5X9IUHFHHS 7X93UHFREU$7X9¥UMFEXP$
+ 6X9¥UHFEARNS 6X9tUHFPROFI9I)
DO 60 NYR-1916
NYEAR-NYR+1944
URITE(69310)NYEAR9UHFNO(NYR)9UHFHH(NYR)9UHFREU(NYR)9UHFEXP<NYR)9
+ UHFEARN(NYR)9UHFPROF(NYR)
60 CONTINUE
HRITE<69150)
150 FORMAT(///91X9¥ADVERTISER SUBSECTOR OUTPUT¥9/)
HRITE(69600)
600 FORMAT(1X9IYEAR# 10X99TDUY¥ 7X93TNEHBUYI 7X9ITMAGBUY! 9X9
+ lAMDUYt 9X98FMDUY3 8X930HFDUY¥ 8X93UHFDUY! 7X93RETPROF¥9I>
DO 70 NYR'1916
NYEAR-NYR+1944
URITE(69610) NYEAR9TDUY(NYR)9TNEUDUY(NYR)9TMAODUY(NYR)9AMDUY(NYR)9
+ FMBUY(NYR)9UHFDUY(NYR)9UHFDUY(NYR)9RETPROF(NYR)
610 FORMAT(1X9I49BF14.0)
70 CONTINUE
HRITE(69160)
160 FORMAT(///91X9ICONSUMER SUBSECTOR OUTPUT¥9/)
URITE<69700)
700 FORMAT(1X9¥YEAR¥ 1X9ITOTAL POPS 4X93HH POPS 1X9tADULT POPt9l)
DO 80 NYR-1916
NYEAR-NYR+1944
URITE(69710)NYEAR9POP(NYR)9POPHH(NYR)9POPADLT(NYR)
710 FORMAT(1X9I493F10.0) '
80 CONTINUE
RETURN
END

198

Theil's U Validation Program

PROGRAM THEILU(INPUT9OUTPUT9TAPE5'INPUT9TAPE6'OUTPUT)
DIMENSION PREDICT‘16)9ACTUAL(16)
READ(5933) EXP
33 FORMAT<A10)
READ(5934) NUM
34 FORMAT<I2)
DO 300 KKK'19NUM
PSUM'O.
ASUH'O 9
DSU"-° 9
N-O
READ(5920) IUAR
20 FORMAT(A10)
READ(59100) (PREDICT(I)9I'1916)
100 FORMAT(4F14.3)
READ(59200) (ACTUAL(J)9J-1916)
200 FORMAT<4F14.3)
URITE(692000) IVAR
2000 FORMATC/91X91A7)
URITE<691000)
1000 FORMAT(6X9SPREDICTEDS9OX93ACTUALI)
DO 10 K31916
IF(ACTUAL(K))O9998
8 PSUM'PSUM + PREDICT(K)#¥2
ASUM-ASUH + ACTUAL(K)X¥2
DSUM'DSUM + (PREDICT(K) - ACTUAL(K))##2
NIN+1
9 URITE(691100) PREDICT(K)9ACTUAL(K)
1100 FORMAT(1X92F14.3)
10 CONTINUE
TOP-SORT(1.0/N x psum) .
DOT-SORT(1.0/N X PSUH) + SORT(1.0/N 3 ASUM)
U'TOP/DOT
URITE<691200) U
1200 FORMAT(/91X9$THEILS U'I9F14.8)
300 CONTINUE
END

199

REFERENCES

REFERENCES

ADAMS, F. G. & BURMEISTER. E. Economic models. TEEE Transactions on
Systems, Man, Cybernetics, 1973, SMC-3, 19-27.

AMSTUTZ. A. E. Computer simulation of competitive market response.
Cambridge: M.I.T. Press, 1967.

ANUNDSEN, K. & LINDGREN, N. Can you put a simulation model to work in
a real community? Innovation, 1972, 29, 43-56.

BAIN, J. S. Barriers to new competition. Cambridge: Harvard University
Press, 1956.

BALDERSTDN, F. E. & HOGGATT, A. C. Simulation of magket processes.
Berkeley: Institute of Business and Economic Research, 1962.

BEHRENS, W. W. The dynamics of natural resource utilization. In
D. L. Meadows and D. H. Meadows (Eds.), Toward global_equi1ibrium:
collected papers. Cambridge: Wright-Allen Press, 1973.

BLAU, R. T.. JOHNSON, R. C. & KSOBIECH. K. J. Determinants of TV
stations economic value. Journal of Broadcasting, 1976, 20,
197-207.

BLOCK, M. P. The potential impact of broadband communication network
technology on consumer marketing communication: a computer simu-
lation experiment. Unpublished doctoral dissertation, Mass Media
Ph.D. Program, College of Communication Arts and Sciences,
Michigan State University, 1975.

BLOOM, P. N. Advertising competition, and public policy:_ a simulation
study. Cambridge: Ballinger Publishing Company, 1976.

BOGART, L. The age of television. New York: Frederick Ungar
Publishing Co.. 1956.

BOWER, R. T. Television and the public. New York: Holt, Rinehart and
Winston. 1973.

BOYD, R. World dynamics--a note. Science. 1972, 177, 516-519.

BROADCAST MEASUREMENT BUREAU. Who owns radio? Washington: National
Association of Broadcasters, 1949, 39.

200

201

BRODY, R. A. Some systematic effects of nuclear weapons technology: a
study through simulation of a multi-nuclear future. Journal of
Conflict Resolution, 1963, 663-753.

BURNETT, R. A. & DIONNE. P. J. GLOBE6: a multiregion interactive
world simulation. Simulation, 1973, 20, 192-197.

CHEN, K., GHAUSI, M. S. & SAGE, A. P. Social systems engineering: an
introduction. Proceedings of the IEEE. 1975. 63. 340-343.

CHORAFAS, D. N. Systems and simulation. New York: Academic Press,
1965.

COHEN, K. J. Computer models of the shoe,g1eather,ghide sequence.
Englewood Cliffs: Prentice-Hall, Inc., 1960.

COLUMBIA BROADCASTING SYSTEM. Statement re FCC report 49-948. In
Hearings on the UHF-VHF allogations problem befbre the Senate
Committee on Interstate and FOreign Commerce. 84th Congress, 2nd
session, part 2, 1956.

COMANOR, W. S. & WILSON. T. A. Advertising, market structure and
perfbrmance. Review of Economics and Statistics, 1967, 49,
423-440.

 

DOAN, R. K. We pause briefly fbr seven or eight commercials. In T.
Smythe & G. Mastroianni (Eds.), Issues in broadcasting. Palo Alto:
Mayfield Publishing Co., 1975. 205-209.

DOMINICK, J. R. The effects of commercial clutter on radio news.
Journal of Broadcasting, 1976, 20. 169-176.

DROR, Y. Public policy-making reexamined. Scranton: Chandler
Publishing, 1968.

DUESENBERRY. J. S., ECKSTEIN, 0., & FRDMM. G. A simulation of the U.S.
economy in recession. Econometrica, 1960, 28, 749-809.

ALLEN B. DUMONT LABORATORIES. A national television allocation plan--
volume VI. In Federal Communications Commission Docket No. 8736
gt 31.. revised October 23, l949-February 6, 1950.

EMERY, E. The press and America. 3rd Ed. Englewood Cliffs: Prentice-
Hall, Inc., 1972.

 

EOYANG, T. An economic study of the radio industry in the United States.

 

New York: Praeger Publishers, 1974.

FEDERAL COMMUNICATIONS COMMISSION. In the matter of allocation of
frequencies to the various classes of non-government services in
the radio spectrum from ten kilocycles to 30 million kilocycles.
FCC Report Docket #6051, May 25, 1945.

202

FEDERAL COMMUNICATIONS COMMISSION. Sixth report and order. 1 Radio
Egg 91:601. para 63 (1952).

FITZ, R. System methods in the public sector: a dialogue. Systems,
Man and Cybernetics Review, 1978. 7, 4-5.

FORRESTER, J. W. Industrial dynamics. Cambridge: M.I.T. Press, 1965.
FORRESTER, J. W. Urban dynamics. Cambridge: M.I.T. Press, 1969.
FORRESTER, J. W. World dynamics. Cambridge: M.I.T. Press, 1971.

FORRESTER, J. W. Counterintuitive behavior of social systems. In
D. L. Meadows and D. H. Meadows (Eds.), Toward global equilibrium:
collected papers. Cambridge: Wright-Allen Press, Inc., 1973,
3-30.

FROMM, G., HAMILTON, W. L. & HAMILTON, D. E. Federally supported
mathematical models: survey and analysis. Volume 1: Analysis
and conclusions. Boston: Data Resources, Inc., and Abt
Associates, Inc., 1974.

GENSCH, D. H. A computer simulation model for selecting advertising
schedules. Journal of Marketing Research, 1969, 6, 203-214.

GENSCH, D. H. Advertisingyplanning--mathematical models in advertising
media planning. New York: Elsevier Scientific Publishing Co., 1973.

GREER, D. F. Product differentiation and concentration in the brewing
industry. Journal of Industrial Economics, 1971, 19, 201-219.

GRIGGS, J. B. Evaluating marketing change: an application of systems
theory. East Lansing: Institute for International Business and
Economic Development Studies, 1970.

HAMILTON. H. R.. GOLDSTONE. S. E.. MILLIMAN. J. W.. PUGH. A. L..
ROBERTS, E. B., ZELLNER, A. Systems simulation for regional
analysis, an application to river basin planning. Cambridge:
M.I.T. Press, 1969.

HANNEMAN, G. J. A computer simulation of information diffusion in a
peasant community. Unpublished master's thesis, Department of
Communication, Michigan State University, 1969.

HARWOOD, K. Broadcasting and the theory of the firm. Law and Con-
temporary Problems, 1969, 34. 485-504.

HOBERT, W. T. User's manual for an introductory world energy model.
Bethlehem, Penn.: Lehigh University, 1977.

KALDOR, N. The economic aspects of advertising. Review of Economic
Studies, 1949-50, 18, 1-27.

KAPLAN, A. The conduct of inqyiry. San Francisco: Chandler, 1964.

203

KESSLER, A. R. & POOL. I. Crisiscom: a computer simulation of human
information processing during a crisis. IEEEIransactions on
Systems Science and Cybernetics. 1965, SSC-l, 52.

KOTLER, P. Marketing_magagement--analysis, plagning and control.
Englewood Cliffs: Prentice-Hall, Inc., 1972.

KOTLER, P. & SCHULTZ, R. L. Marketing system simulations. In H.
Guetzkow, P. Kotler, and R. Schultz (Eds.), Simulation in social
ggdadministrative science. Englewood Cliffs: Prentice-Hall,
1972, 481-549.

KRAMER. J. F. A Computer simulation of audience exposure in a mass
media system: the United Nations infbrmation campaign in
Cincinnati 1947-48. Unpublished doctoral dissertation,
Massachusetts Institute of Technology. 1969.

KRULEE, G. K. Individual behavior simulations--cognitive processes.
In H. Guetzkow. P. Kotler, and R. Schultz (Eds.), Simulgtion in
social and administrative science. Englewood Cliffs: Prentice-
Hall, Inc., 1972. 51-102.

LASSWELL, H. D. A pre-viewgf policy sciences. New York: American
Elsevier Publishing, 1971.

LEVIN, H. J. Economic effects of broadcasting licensing. Journal of
Political Economy, 1964, 72, 151-162.

 

 

LEVIN, H. J. Spectrum allocation without market. American Economic
Review, 1970. 60. 209-218.

 

LITTLE. J. D. & LODISH, L. M. A media planning calculus. Operations
Research, 1969, 17, 1-33.

 

MEADOWS. D. H.. MEADOWS. D. L.. RANDERS. J., & BEHRENS. W. W. Th;
limits to growth. New York: Universe Books, 1972.

MENDELSOHN, H. Listening to radio. In L. A. Dexter and D. M. White
(Eds.), People, society and mass communications. New York: The
Free Press, 1964.

 

NAILL. R. F. Managing the energy transistion. Cambridge: Ballinger

4—

Publishing Company, 19 7.

NATIONAL ASSOCIATION OF BROADCASTERS. FM-phasis. Washington: National
Association of Broadcasters, 1959, 1.

NATIONAL ASSOCIATION OF RADIO AND TELEVISION BROADCASTERS & RADIO-
TELEVISION MANUFACTURERS ASSOCIATION. How to sell more radio sets
with FM-phasis. Washington: NARTB, 1952.

NAYLOR, T. H.. BALINTFY, J. F., BURDICK, D. S., & CHU. K. Computer
simulation techniques. New York: John Wiley and Sons, Inc., 1968.

204

NAYLOR. T. H. & FINGER, J. M. Verification of computer simulation
models. Management Science, 1968, 14, 893-895.

NAYLOR, T. H.. WERTZ, K., & WONNACUTT, T. Some methods for analyzing
data generated by computer simulation experiments. Communications
of the ACM, 1967. 10. 703-710.

NEWSPAPER ADVERTISING BUREAU. A national survey of the content and
readership of the American newspaper. 1972.

NEWSPAPER ADVERTISING BUREAU. Basic facts about newspapers. July,
1978.

NIE, N. H.. HULL, C. H.. JENKINS. J. G., STEINBRENNER, K., BENT, D. H.
Statistical package fbr the social sciences. 2nd Ed. New York:
McGraw-Hill, 1975.

NOLL, R. G., PECK, M. J., & McGOWAN, J. J. Economic aspects of telg;
vision regulation. Washington: The Brookings Institution, 1973.

ORCUTT. G. H.. GREENBERGER. M.. KORBEL. J.. & RIVLIN. A. M. Micro-
analysis of socioeconomic systems: a simulation study. New York:
Harper and Brothers, 1961.

OWEN, B. M. Economics and freedom of expression-:media structure and
the First Amendment. Cambridge: Ballinger Publishing Company.

975.

ALFRED POLITZ RESEARCH INC. National survey of radio and TV sets
associated with U.S. households. Washington: Research Department.
National Association of Radio and Television Broadcasters, 1954.

POPKIN, S. L. A model of a communication system. American Behavioral
Scientist, 1965. 8, 8-11.

PRITSKER. A. B. & KIVIAT, P. J. Simulation with GASP II. Englewood
Cliffs: Prentice-Hall, Inc., 1969.

PUGH. R. E. Evaluation of policy simulation models: a conceptual
approach and case study. Washington: Information Resources Press,
1977.

RASER, J. R.. CAMPBELL, D. T., & CHADWICK, R. W. Gaming and simulation
for developing theory relevant to international relations. General
Systems, 1970, 183-204.

RASER, J. R. & CROW, W. J. A simulation study of deterrence theories.
In L. Kriesberg (Ed.). Social processes in international relations.
New York: Wiley, 1968, 372-389.

ROBINSON, G. 0. (Ed.) Communications fOr tomorrow--policy perspectives
for the 19805. New York: Praeger Publishers, 1978.

205

ROBINSON, J. P. & CONVERSE, P. E. The impact of television on mass
media usages: a cross-national comparison. In A. Szalai (Ed.),
The use of time. The Hague: Mouton, 1972, 197-212.

ROPER. B. W. Changing public attitudes tpward television and other
mass media: 1959-1976. New York: Television Infbrmation Office.
1977.

RUCKER, B. W. ‘Ihefirst freedom. Carbondale: Southern Illinois
University Press, 1968.

SAGE, A. P. Systems engineering education. Systems, Man, and
Cybernetics Review, 1978, 7, 6-9.

SCHIESSER. W. E. An introductory world fbod model. Bethlehem, Penn:
Lehigh University, 1976. (mimeographed)

SCHULTZ, R. L. & SULLIVAN, E. M. Developments in simulation in social
and administrative science. In H. Guetzkow, P. Kotler & R. L.
Schultz (Eds.), Simulation in social and apminisirative science.
Englewood Cliffs: Prentice-Hall, Inc., 1972, 3-50.

STEINER, G. A. Theypeople look at television. New York: Alfred A.
Knopf, 1973.

STERLING, C. H. & HAIGHT, T. R. The mass media: As en Institute guide
ip_communication industry trends. New York: raeger Publishers,
978.

STERLING, C. H. & KITROSS, J. M. Stay tuned-~a concise history of
Aggrican broadcasting. Belmont. Calif.: Wadsworth PubTishing Co.,
1978.

STUART, F. The effects of television on the motion picture and radio
industries. New York: Arno Press, 1976.

 

THEIL. H. Economic fbrecasts and policy. Amsterdam: North-Holland
Publishing Company, 1961.

UDELL, J. G. The economics of the American newspaper. New York:
Hasting House Publishers, 1978.

UNITED STATES BUREAU OF THE CENSUS. Histprical statistics of the
United States: Colonial times to 1970. Washington: U.S. Depart-
ment of Commerce, Bureau of the Census, 1972.

UNITED STATES BUREAU OF THE CENSUS. Statistical abstract of the
United States. Washington: U.S. Government Printing Office, 1960.

UNITED STATES BUREAU OF THE CENSUS. United States Census of Manufactures.

 

Washington: U.S. Government Printing Office, 1954.

UNITED STATES CONGRESS, HOUSE. A bill to establish certain requirements
relating to interstate and foreign telecommunications. and for
other purposes (Communications Act of 1978). H.R. 13015, Sub-
committee on Communications, 95th Cong., 2ndsess., 1978.

206

WEBBINK, D. W. The impact of UHF promotion: the all-channel television
receiver law. Law and Contemporary Problems, 1969, 34, 535-561.

WOOD, J. P. Magazings in the United States. 3rd Ed. New York: Ronald
Press. Inc., 1971.

‘WRIGHT, C. R. Mass communication: a_§ociological perspective, 2nd
5!; New York: Random House, 1975.

ZAISER, J. N. 8 SCHIESSER. W. E. An introductory world energy model.
Proceedings of the IEEE. 1975, 63, 390-396.

ZEHNA, P. W. Probability distpibutions and statistics. Boston:
Allyn and Bacon, Inc., 1970.