. . . . -

.figrfi.» ca». . mu - a . . .

... .n.hww. .....w.) . i-u:. if. .1.

' t! 3an“ ' .43 laud-ti cl “3"“; ‘ I. .
.guww.~wh$fithlwwm~

a.
.. .1. rev... on». g 316?. 3 1.73. .
v be I. n 0 It 0'0 v3.2.3?!“ - $.nfi0
. .mfiflm." .... A1. "...”. rd. .
Vi OI ' ‘vfi Mflq~VciVWo .

‘4

- . . . z
t t... n. n: .. u .n . t\ t
y . . . . ... ‘ ... i.. - .1 .. a . _.
. . . - . . . ...ntrmfisw 0.. .t
- .. . n 5.1....» . n: ..
' ‘. . . OI.
.

a.
.. m

1
"In... H . 1.. If
.I. n. . I v. A
.-. ....t...a $1,}.
. a 1.... l- .. uflUthturq r .
..-- .11..ulut 4.. “:3 5.. ..
. t' £15.. a
' I ‘4
ll . Im‘ H, . I». 3.: :7:
Ir . . 5| t. . fi ‘
I. ;

J‘q" . ’

Wammut- - wtm;
tun?
14:11 r ,

..

 

r
t I A ‘ ¢ r
. . t a ...”..l n 1,». t a ,r . . .7 ..rs. .
.1»; . I alll'w‘th 0n. 0 ‘1 ..o . V “.... '1‘... . v Jvu. AhQWo
. m I I .. Illa yl ' I .t. . . a. o. I. u! t I ”1.... n .
N, ....tt... «LINt wtwwfumlt I. tt.rt,tt.W\mhrt4%. .4 4 . . . _ .. p .. - in... m; .. -. . .
.. - 9. ttl» .gtt ‘ t - . . l . . czawtxatflht. .zt.Jxapr..i .. .
. Di. fii.“|l"tl‘0\i t o . ! \v . let. ‘1. ..I “”1001! .1 .

\L‘flttw, . t
.

 

 

0‘ .
- t .vt {ht . u 7

I I5 I I .7 I vb c it I .. I] 0 It'O‘t‘ la . .

t u I: I cut! uu t u 0-; c ..~ ~ o to t- 1 $1144.... _ . hr). at L‘ u t

-c t - n- t . . . u - t.-- t .u I . - - t r If -ltt) gfu....dfldfi\nftlvlltfiflwuflt§ t

u - c t t Int J I o . up . \t t tthal o n - I. ‘ ,.o|V{ 4L .I'tvtzwfitd }(A‘.4}‘%)... .
I: t . .. July . ttatv I m t t It I \ ttt ”.1 II. tt, I 1H, p\\ .{t’ft d}lvtiv.llxr\vbta\tl.fi.l‘l

t. I u 0.. t . ‘ . . a, t .- . . t- -1li?1:».

t . t I L t t . x t t .3 than .t

Putt}. u T

..<I -1

l\\\t\\\\\\\\t\\\\\t\t\tttttttttt‘ttt‘ttttltmmgm I

3 1293 10

 

 

THESJB l; ‘ 4 fl
Mi: $1 5 z; e sf: 5 7.6%. ts
‘ University
‘ I"

 

 

This is to certify that the

thesis entitled
Multiple regression and Discriminant
Analysis for Prediction of Audience
Share and Classification of Format
Opportunitypresented by

Phillip Charles Pettelie

has been accepted towards fulfillment
of the requirements for

Masters degreein Telecommunications

 

 

 

Dan August 1h, 1981

 

0-7639

 

OVERDUE FINES:
25¢ per day per item

RETURNING LIBRARY MATERIALS:
\

Place in book return to mauve
charge from circulation records

 

 

 

 

 

@ Copyright by
PHILLIP CHARLES PETTELLE
1981

MULTIPLE REGRESSION AND DISCRIMINANT ANALYSIS
FOR PREDICTION OF AUDIENCE SHARE AND CLASSIFICATION
OF FORMAT OPPORTUNITY

By

Phillip Charles Pettelle

A THESIS

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

MASTER OF ARTS

Department of Telecommunication

1981

5'9?

q-u ~

6H

ABSTRACT

MULTIPLE REGRESSION AND DISCRIMINANT ANALYSIS
FOR PREDICTION OF AUDIENCE SHARE AND CLASSIFICATION
OF FORMAT OPPORTUNITY

By

Phillip Charles Pettelle

This thesis takes demographic information from 58 markets nationwise,
and derives statistical models for use in prediction of audience share for
radio formats. The formats selected are the nine most popular based on
total listenership.

The data used was refined to exclude variables whose variance explan-
ation was minimal. This was achieved using Pearson product-moment
correlation and factor analysis. T-tests were used to find variables
significant among groups of markets. Discriminant analysis was used to
recognize the similarities between markets, thus classifying them as
favorable or unfavorable for some formats. Finally, multiple regression
enabled prediction of the approximate audience share that a format would
achieve within a market. I

It was found that discriminant analysis successfully classified
markets, however, the degree of success for a format within a market cannot
be determined by this method. Conversely, multiple regression achieved
predictive results superior to others in this study. Examples are displayed

in Chapter V.

Accepted by the faculty of the Department of Telecommunication,
College of Communication Arts and Sciences, Michigan State University,

in partial fulfillment of the requirements for the Master of Arts

WW

V7
Director of Thesis

degree.

ACKNOWLEDGEMENTS

It is here that I wish to make known my most special appreciation
and sincere gratitude to those who patiently withstood the rantings
and ravings of the author during this period of my life. In the
Department of Telecommunication I thank John Abel, Gary Reid, Ann
Spalding and, of course, Bob Yadon, who may have discovered a new
level of tolerance towards graduate students because of me. I'd
like to also thank my parents who shared my anxiety in completion of
this project. My gratitude and love to you both. Thanks also to all
of my friends nationwide who understood my ordeal and were glad when I
would talk about something else. Finally, a very special thank you
to Nancy who was closest to me during this time. Without you, Nancy,
completing this would have been a long and tedious task. Thanks for

your compassion and friendship.

TABLE OF CONTENTS

LIST OF TABLES .............................................. vi
CHAPTER Page

I INTRODUCTION ...................................... l

Programming as Target Marketing ................... 2

Historical Trends in Programming .................. 3

Scope of Study .................................... 6

Significance of Study ............................. 7

Notes ............................................. 9

II RELATED RESEARCH .................................. ll

Notes ............................................. 22

III METHODOLOGY ....................................... 24

General Discussion ................................ 24

Hypotheses ........................................ 28

Pearson Product-Moment Correlation ................ 3O

Factor Analysis ................................... 31

Multiple Regression ............................... 32

Discriminant Function Analysis .................... 34

Notes ............................................. 37

IV DATA ANALYSIS ..................................... 38

Pearson's r ...............................; ....... 38

T-tests of Groups Means .............; ............. 45

Tests of Hypothses ................................. 63

Discriminant Analysis Results ..................... 73

Multiple Regression Results ...................... 193

Notes..............................., ............. 109

iv

CHAPTER Page

V SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS ............. llO
Conclusions and Recommendations ....................... l20

The Passage of Time ................................... 123

APPENDIX A: MARKETS INCLUDED IN THE COMPUTATION ................ l24
APPENDIX B: FORMATS TOTAL SHARE OF AUDIENCE BY MARKET .I ........ l25
APPENDIX C: VARIABLE LIST ...................................... l28
APPENDIX D: FINAL CORRELATION MATRIX ............ , .............. l31
LIST OF REFERENCES .............................................. l49

LIST OF TABLES

TABLE Page_
1. National Share of Major Radio Formats ....................... 6
2. High Intercorrelations ...................................... 39
3. Factor Loadings of Linear Variables ......................... 42
4. Significant t-values for "Rock/AOR/Contemporary" ............ 47
5. Significant t-values for "Beautiful Music" .................. 49
6. Significant t-values for "NOR/Adult Contemporary" ........... Sl
7. Significant t-values for "Country" .......................... 52
8. Significant t-values for "News/Talk" ........................ 54
9. Significant t-values for "Black/Disco" ...................... 56

10. Significant t-values for "Spanish" .......................... 58
ll. Significant t-values for "Religious/Gospel" ................. 60
l2. Significant t—values for "Classical" ........................ 6l
13. Tests of Hypothesis Variables ............................... 64
l4. Discriminant Function Components for "Rock/AOR/Contemporary"

Market Classifications ................................. 76

15. Classifications of Ungrouped Markets for the "Rock/AOR/
Contemporary" Format ................................... 78

16. Discriminant Function Components for "Beautiful Music"
Market Classification .................................. 82

17. Classification of Ungrouped Markets for the "Beautiful
Music" Format .......................................... 83

l8. Discriminant Function Components for "MGR/Adult
Contemporary" Market Classification .................... 86

vi

TABLE
I9.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.
35.

Page

Classification of Ungrouped Markets for the "MGR/Adult

Contemporary" Format .................................. 87
Discriminant Function Components for "Country" Market

Classification ........................................ 90
Classification of Ungrouped Markets for the "Country"

Format ................................................ 91
Multiple Regression with Rock/AOR/Contemporary Audience

Share as the Dependent Variable ....................... 95
Multiple Regression with Beautiful Music Audience Share

as the Dependent Variable ............................. 96
Multiple Regression with MOR/Adult Contemporary Audience

Share as the Dependent Variable ....................... 98
Multiple Regression with Country Audience Share as the

Dependent Variable .................................... 99
Multiple Regression with News/Talk Audience Share as the

Dependent Variable .................................... lOO
Multiple Regression with Black/Disco Audience Share as

the Dependent Variable ................................ l02
Multiple Regression with Spanish Audience Share as the

Dependent Variable .................................... lO3
Multiple Regression with Religious/Gospel Audience

Share as the Dependent Variable ...................... lO5
Multiple Regression with Classical Audience Share as

the Dependent Variable ................................ 108
Example Prediction Values for "ROck/AOR/Contemporary"

Audience Shares ........................................ ll2
Example Prediction Values for "Beautiful Music" Audience

Shares ................................................. ll3
Example Prediction Values for "NOR/Adult Contemporary"

Audience Shares ....................................... 114

Example Prediction Values for "Country" Audience Shares.... ll5

Example Prediction Values for "News/Talk" Audience Shares.. llfi

vii

TABLE Page

36. Example Prediction Values for "Black/Disco" Audience
Shares ............................................... 117

37. Example Prediction Values for "Spanish" Audience Shares... ll8

38. Example Prediction Values for "Religious/Gospel Audience"
Shares ................................................ ll8

39. Example Prediction Values for "Classical" Audience
Shares ................................................ ll9

viii

CHAPTER I

INTRODUCTION

The purpose of this research is to test a methodological sequence
wherein specific market variables will be identified and treated as
criteria in the selection of a radio format that would be the most
economically sensible among the major format choices.

In the chapters that follow, the sequence of events leading to
format audience share prediction and classification will be detailed.
After the initial market data has been gathered statistical treatments
will be used to delete the unimportant variables from later tests. The
refined data will then be used in multiple regression analysis to develop
linear combinations of variables that are effective in the prediction of
the audience share of a format within a market. Then, the same refined
data set will be treated with discriminant function analysis so that
market favorability towards a format may be statistically classified.

It is hoped that this effort will increase the programmer's know-
ledge of the market factors that contribute to the success of formats
within markets. The assumption at the outset is that there are more
demographic factors involved in the success of formats within markets
than sex and age. The methods used will find the combination of variables

that are effective in determining format success.

Programming as Target Marketing
A radio station's potential audience is comprised of the market's

population, which can be further defined as an aggregate of many sub-
categorical, deScriptive characteristics of people (e.g., demographics).
The broadcast media overall, knowingly or not, will utilize similar if
not identical target marketing strategies of other profit oriented
businesses. For example, television as it exists today, and radio, as
it existed before the advent of television, both could be identified
as "textbook" mass appeal marketers. This undifferentiated marketing
approach viewed the market as an agglomeration of people. Programmers
of the medium characteristically would,

"choose not to recognize the different market segments

making up the market. (They) treat the market as an

aggregate, focusing on what is common in the needs of

people rather than on what is different. (They) try

to design a product and a marketing program that appeal

to the broadest number(s) ..."l

Radio station programmers in recent years have recognized the value of

abandoning the classic "mass appeal" approach. This was not always the
case. With radio and television employing essentially the identical pro-
gramming strategies for a finite audience, "...the result is hypercompet-
ition for the largest (audience) segments and undersatisfaction in the

smaller ones."2

Radio programmers, stifled by a stale audience approach
and increasing popularity for the new medium, sought to achieve success in
marketing to smaller, more loyal audiences. In effect, radio had resigned
the mass appeal approach to television and it's new visual movelty. The
success of radio's new endeavor was a matter of calculated survival. The
target marketing of radio allowed stations to "achieve a strong market
position in the particular segment it serves, owing to it's greater know-

ledge of the segments' needs and the special reputation it acquires."3

So, radio currently employs target marketing strategies and concepts
common to all who have a "product," and a "market." The success factor in
linking the constructs "product" and "market" is "knowledge." This factor
can make or break a target market campaign. Specifically, knowledge of
various market segment needs is critical to selection of the market segment
where financial success is most probable. This, then, is the essence of
itarget marketing, and, "...if (that) segment of the market is well chosen,
the firm can earn high rates of return on its investment."4

As discussed later, previous studies dealing in format research attempt
to define target market groups, or to monitor the success of formats (i.e.
target market strategy). Generally, studies in radio target marketing
attempt to formulate market opportunity by recognizing patterns that develop
in radio audiences as a consequence of the music programming format.

Target marketing can be said to be the programming of a radio station
with the objective that a particular group of persons sharing common demo-
graphic characteristics are drawn as the station's listenership.

The means by which target groups are obtained are varied. Most
accomplish this goal with a musical format, but some avoid music in order
to attract a different type of listener. Herein lies the challenge to the
programmer, that is, to find the most appropriate mode by which the target
listener may be attracted. In doing this, a radio station can virtually
guarantee a segment of the population to advertisers with similar target
objectives who seek contact with those listeners in their role as consumers

of products.

Historical Trends in Programming

 

As mentioned in the previous discussion, radio was once programmed

to be "all things to all people." Of course, this cannot truly be possible.

as was explained earlier. Nevertheless, a select few audience approaches
could be identified during radio's reign as the primary entertainment
medium. At the time, they were successful in obtaining certain audience
sub-groups. While the "soaps" captured the female audience and held them
for years, play-by-play descriptions of major league baseball games
attracted the male listener.5 Although these programs typified radio broad-
casusprior to television, musical formats indeed existed with profitable
results. A typical format would include, "uninterrupted sweeps of music
followed by brief news and weathercasts and by clustered conmercials and
public service announcements."6 Although musical formats were present,
the variety of formats was minimal and was usually confined to the pro-
grammer's personal conceptions of "good music." The substance of radio's
success was it's reliance on the dramatic series, sports coverage, and
utilization of known celebrities through network broadcasts.7

Television offered a new dimension to entertainment, and quickly
sent radio scrambling for ideas in order to maintain it's former prowess
as an entertainment medium. Performers once fixtures of radio programs,
"quickly made the transition from microphone to microphone and camera or

they got out of the business."8

The musical format needed a new approach,
as it could not be transferred to television with success. The most popular
current music seemed the obvious choice, and when audience rating services
such as Hooper and Trendex began publishing results of audience surveys,
competition became keen and the Top 40 format became a commercial
success.9

Successful ideas are seldom ignored, and with station programmers
willing to try anything to gain listeners, the new Top 40 format spread

from coast to coast. The late 50's and early 60's were dominated by

Top 40 radio stations doing anything and everything to capture audiences.
Meanwhile, as Top 40 dominated the AM band, a few stations began

to acquire licenses for the FM dial. However, the innovative drive

inspired by inter-market competition did not exist as it did on the AM

dial. FM had an inherent problem. That being that manufacturing companies

of radio receivers did not include FM bands along with AM bands. Since AM

dominated in popularity, FM receivers simply did not sell. The few FM

stations were almost solely playing "good music," or classical recordings,

more or less cultural offerings with little or nu audience following. As

a consequence of these factors, 1966 saw FM set saturation at only 48 percent

for homes nationwide.10
with new AM stations doing on-the-air at a surprising rate, the

spectrum became so crowded as to force the FCC in l968 to impose a freeze

on all new applications for AM stations, regardless of class, in order to

study alternatives to this situation.11

In l969, the Commission instituted
a rule to revise AM assignments. Among the restrictions of the new ruling
was the following: "Applications for new stations would not be accepted if
there is an available FM channel which the applicant could use and provide
the same substantial service benefit."12

The Commission's ruling changed the face of radio programming. FM,
once the domain of so-called "good music" was now wide open to all formats.
By l975, FM set saturation had reached 93 percent in homes, about 30 percent
in automobiles, and FM listenership represented 36.4 percent of all radio
listening.13 Overall, FM listening increased 8l percent between 1970 and
1975.M

The 1970's showed the dynamic growth of FM as a viable entertainment
medium. It had expanded from a unidimensional audience status to the

15

dominant listener share between AM and FM by 1980. During this time,

formats and format development varied strongly. AM formats, once

reliant on Top 40 and MOR, expanded to accommodate non-musical formats

in order to differentiate themselves from the superior quality of FM.
Major formats control the overall share of the total audience. With
esoteric formats locally successful in certain markets.16 By l980,

the following formats have the largest shares nationally. Here, they are

listedalong with the share of total listenership.17

Table 1. National Share of Major Radio Formats

 

format Listenership %
Rock/Contemporary 33.04
Beautiful Music l5.58
MOR/Adult Contemporary l6.ll
News/Information/Talk 9.68
Country l0.26
Black/Disco 9.76
Spanish 2.02
Classical l.27
Religious/Gospel l.24

Scope of Study

 

There were some limitations encountered in the production of this
paper which should be outlined at this point. At the time that this
study was done, the author was unable to utilize a publication whose
value would be unquestionably infinite in this situation, the population
and economic reports published every ten years by the Department of

Commerce, Bureau of the Census. Optimally, as stated above, the use of

the most recent and accurate data is necessary. The Census Bureau

data would have been a source for such information, if available

census data were relevant to current population demography. However,

at the time of completion of this paper, the l980 census figures

were not yet due to be released for at least l2 months. In addition, the
reports from the most recent available census in 1970 would obviously

be obsolete material at this date.

In order to attain the most credibility possible in current data
sources, this researcher has sought publications that contain the most
recent "projections" from the 1970 Census. The M.S.I. publication,
Survey of Buying Power and the City and Countnyata Book, two sources
used heavily in this study, contain projections based on the l970 Census.
At this point in time, these data sources are the most accurate and
reliable available, and are used with confidence in this research.

The fact that the calculations contained in the study are not based
on original census data or compiled raw figures may cause increased
error because of the problems involved in making projections. Also,
because of the limitations in data that is consistant across all of the
markets included in this study, certain variables that may be significant
as far as format descriptiveness could not be used in this research.

An example of such a variable would be the level of education of the

market's population.

Significance of the Study

 

This research may be another step in the evolution of programming
sophistication. Station programmers need increased knowledge with regard

to their respective markets and potential format successes within them.

In this case, knowledge is the disection of markets into their demographic
composition, and identifying underserviced sub-audience groups. This
research will further define these radio sub-audiences according to
demographic similarity. These "demographic descriptions" can be used to
describe markets in terms of format opportunity, or as a more reliable
justification of a station's current programming.

Most stations currently choose to refer to this listenership as
being within a particular age category. Certainly age is not the sole
determining variable for formats, and this research will reveal other
variables that add to the complete demographic description of formats,
to be characterized as a conglomeration of interrelated demographic
variables.

Mgltiple regression has been an accurate predictor of social behavior,

 

used extensively for a huge variety of purposes. A format's share of
audience within a market can be construed as a result of social behavior,
the measured listenership of a market's population. Discriminant analysis
will show how markets are similar with other markets whose audience share
for a format is the highest in the country.

This potential knowledge is of great importance to programmers in the
light of the heightened intensity of local competition. The necessity
for this type of information by programmers occurs during the act of
making a competent decision as to the likely changes in the sound format
and their subsequent effects of the market situation. This study has the

capacity to generate this sort of helpful information.

CHAPTER I--NOTES

1P. Kotler, Marketing Mangement: Analysis, Planning and Control

(New Jersey: Prentice-Hall Publishers, 1976), pp. lSl-156.

2Ibid.

31bid.

41bid.

5W.L. Quaal and J.A. Brown, Broadcast Management (New York: Hastings

House Publishers, l976), p. l79.

6E. Routt, J. McGrath, and F. Weiss, The Radio Format Conundrum

 

(New York: Hastings House Publishers, l978), pp. l-l0.

7Quaal and Brown, pp. l79-l80.

8Routt, McGrath and Weiss, p. 5.

91bid.

IOIbid, p. 8

nW. B. Emery, Broadcasting and Government: Responsibilities and

Regulations (East Lansing: Michigan State University Press, l97l), p. l22.
l2

 

Ibid, p. 123.

13Routt, McGrath and Weiss, pp. 8, 9.

”Ibid, p. 9.

15d. Duncan, American Radio, Vol. V, lst ed. (Kalamazoo: Gilmore

 

Advertising, Spring l980), p. A-22.

16

17

Ibid.

Ibid, p. 25

10

CHAPTER II

RELATED RESEARCH

Previous studies in this research area attempt to theorize a
common denominator as a basis for successful recognition of programming
opportunity. Although there are no exact models or particular studies
from which this research originates, the following summaries and
critiques are indicative of past efforts to identify a realistic pro-
gramming possibility within a market.

Neilsen and Thibodeau formulated two basic ways to generate format
opportunities; within market, and cross-market opportunity. For within
market comparisons, a relationship was identified between, "...a formats'
market share and the average share of stations with that format within a
particular market with the market shares and average market shares of
stations with different formats within a market, region, or market size."1
This can be interpreted as a basic comparison between the success of a
certain format as compared to the successes of other formats within the
market. Thus, one would look for formats where the market shares and the
average station shares were substantially greater then the average within
a particular market, region, or market size. For cross market comparisons,
Neilsen and Thibodeau related market shares of stations with a particular
format within a market, region, or market size with the market size of

similar formats in different markets, regions, or market sizes.2

ll

12

The within market comparison indicates the relative success of a
format in relation to other formats on a share basis, however, this
correlation does not provide the explanation for success of certain
formats within a market while other formats within the same market may
not be doing as well. This could be misleading for a programmer. Although
this analysis may show a higher average share for format "A" over format
"8," for example, it does not positively indicate an opportunity in the
market for an additional station programmed with format "A." Thus, there
is little evidence of possible success should a station change formats
from "B" to "A" merely on the suggestion that the station would also
experience the higher share held by the other stations in format ”A."

Neilsen and Thibodeau use the Hesbacher_et_gl. study, explained
later in this review, as a model for their study. An understanding of
Hesbacher's model should help explain the major difficiencies of Neilsen
and Thibodeau. In Hesbacher and in Neilsen, the presence of the format
pre-exists the presence of the audience within the market. In both cases,
there is little effort to understand or define the demographic character-
istics which can be the sole justification for the presence or the exclusion
of a format within a market. This situation is most evident in Neilsen's
method for cross-market comparisons, as explained above. The cross-market
comparison makes the fatal assumption that there is a negligible difference
in the demographic composition between markets. Comparisons of format
successes in different markets may be misleading because of the mutual
exclusitivity of any twomarket's demographic composition.

In both of these comparative methods, it is left to the judgment of
the programmer as to the validity of such comparisons. The lack of sign-

ificant statistical differentiation among comparative "judgments“ leaves

13

the programmer alone with subjectivity. In other words, the programmer
may not be any further informed of the market situation as a result of
the use of this analysis. Neilsen and Thibodeau, although providing a
means of monitoring format progress within and across many markets, do not
give the station programmer an accurate tool to discover whether or not
opportunities exist for formats within a chosen market.

Moomey and Skolnik tried to find major format categories by surveying
222 radio station programmers nationwide. They used factor analysis for
the identification of possible patterns in programming. The objective
was to discover the types of people perceived by the programmers to be
their audience. The results of the survey described five very broad
categories which are summarized as follows:

1. Programmers of the "General Adult" category would, "...
welcome a broad base of men and women from all social
strata, and presumably tailor their programming accordingly.
There will be wide differences of format in these types of
stations, hence the confusion usually surrounding the

term, 'middle-of-the-road'..."3

2. Programmers of the "Young Adult" category, would "...
(seek) their base audience among listeners in the 20-30
year-old category...They reason 'that these listeners will
respond to the same appeals as the youth4market if presented
in a somewhat more sophisticated style."

3. Programmers of the "Focused Adult" category would,"...

focus upon upper income professional adults whose social
background mirrors what advertising agencies deem desirable

from a marketing viewpoint...Programming found on these

stations could range from the non-personality, non-offensive music
formats to the sophisticated classical music format."5

4. Programmers of the "Focused Youth" category clearly have,
"...aimed directly at a youth audience comprising sub—teens,
teens, college students and young adults. Also included are
Negro and lower income listeners, suggesting that these
characteristics do not by themselves distinguish a separate
approach to programming."6

Finally, for the category called:

14

5. "Department Store," programmers would likely display

"...a concern with program content categories rather than

audience social characteristics. Programmers scoring

highly in this category view their target audience through

the kinds of programming they think will appeal to the most

listeners. Like a department store it takes years to build

a programming inventory and image."

The study seems to be a very valid representation of how station
programmers perceive their audiences. What was unique in their attitudes,
and confirmed in the factor analysis, was that their listeners possessed
a variety of interrelated characteristics. It could be interpreted,
then, that most programmers at the time of the survey were only beginning
to differentiate the perceptions of who their listenership was on some
demographic factors other than age. For example, the "Focused Youth"
category, whose age group was stated as "sub-teens, teens, and young
adults," is further supported by other characteristics such as ”lower
income," "Negro," and "college crowd." "Focused Adult," whose age
category was described as the 40-50 year-old age group by the factor
analysis, is supported by the demographic characteristics of "suburbanite,'I
"upper income," and "professional people." Such added descriptive terms
that are statistically confirmed, as in this case by factor analysis,
enable a programmer to understand the composition of a format's listener—
ship in other than the unidimensional term; age. Among other benefits of
this knowledge, a programmer may use this in promotional efforts for his
station by attaching positive imagery in the minds of the target listener.

Moomey and Skolnik provide a view of station programmers perceptions
of their audiences. They also begin to reveal the importance of additional
descriptive characteristics in sub-audiences. This study illustrates the

demographics that may be present within a market if a particular format

is determined to be a success. This study does not show how to confirm

15

the fact that audience sub-groups exist within a market. Although the
objective was simply to discover programmers perceptions, this is
irrelevant for programming strategies. It does not matter what
the station programmer believes to exist, unless substantiated by
statistical facts. What is more important is the determination of what
sub-groups actually exist within a market.

Lull, Johnson and Sweeney constructed a demographic profile of
audience types by radio format. They illustrated the relative demographic
uniqueness of audiences and relationships among the sub-groups. "The amount
of demographic distinctness or uniqueness obtained by radio formats has
implications for persuaders and informers who use radio to reach their
intended audiences."8 The most important confirmation of the study was
that the highly distinct audience sub-groups.in this case they were found
to be All News, Live Progressive Rock, and Top Forty, are more efficient
in the dissemination of information to specific "target" audiences.9
"Advertisers and the creators of public service messages can use this
knowledge when they select among the various formats for transmission of
their messages. Of course, when the appeal is a general one, exclusive
use of a highly distinct format should perhaps be avoided."10

Another result was the confirmation of certain variables as being
the most discriminating among those used for the knowledge of certain
formats. For example, the most discriminating variable for knowledge
of an "All News" or "Beautiful Music" format was age. Likewise, the most
discriminating variable for knowledge of "Rock“ and "Contemporary" formats
was education.n

While the Lull et al. study does little to actually assist in format

selection for markets, like the Moomey and Skolnik research, it offers

16

examples of some of the discriminating demographic characteristics of
formats. The Lull study, by use of discriminant function analysis,
found two significant discriminant functions. As mentioned above, age
was one that was perhaps the most meaningful, with residence ownership,
geographic stability, and marital status also contributing to this first

12 In the second function,

function, but to a lessor degree than age.
education was the major discriminating variable, as was stated earlier,
with geographic stability, dwelling type, sex, and age of lessor importance

13

in the separation of the groups in this function. These two functions,

headed by the variables age and eduction, accounted for 97’percent of the
inter-group variability.14

Lull, Johnson and Sweeney, like Moomey and Skolnik, have attempted
different methodologies in order to identify some demographic categories
that describe the audiences for radio formats. These studies constitute
a beginning as to possible outcomes of future research that have similar
objectives in format research.

Massey utilized discriminant analysis to obtain coefficients for

five Boston area FM stations. The study evaluates audience similarities

 

by correlations among variables being either strongly positive, and thus
being characteristic of that audience, or strongly negative and thus

15 The crux of

inferred to be a characteristic of some other audience.
the paper was the technique of avoiding misclassification of descriptive
variables. For example, if "higher occupational status" is character-
istic of two formats, but one to a greater degree than the other, it
would be important to avoid misclassification into the format group where

"higher occupational status" is actually less significant. Unfortunately,

for the example given, Massey does not choose to identify the stations

17

with formats but instead refers to them as simply stations "A," "8,"

etc. Although one could conjecture as to what station A's format might
be based on the strongly positive correlated variables, the connection

of "A" with a format could have aided station managers in relating formats
with audience characteristics.

Discriminant analysis, as Massey has illustrated, is a reliable
tool for analysis of how variables contribute to format listenership.

The problem of misclassification, which Massey addresses as essentially
the main topic of the paper, should be a lessor consideration with this
study. First, this would be because this study undertakes a two-way
analysis with clearly distinctive groups at opposite ends of a geometric
continuum, as opposed to the n-way analysis with multiple groups as
described by Massey. His focus was on the direction of the analysis
capabilities of the method, while this study will utilize the classifica-
tion useage of the method. In summary of Massey's illustration of the
use of discriminant analysis, the example given does illustrate some
sharp differences between the audiences of various unnamed formats based
on some relatively generalized categorical variables.

Hesbacher monitored the radio market in Philadelphia to discover how
sound formats relate to audience size.16 Generally, the facts presented
reflect the common knowledge held today by most programmers.‘ At the time
of publication, however, Hesbacher's observation of an apparent gradual
decline in popular music listenership and the growth exhibited by the
newer "Gold" and "Beautiful Music" formats may have been significant. The
study showed the effects of format changes on the size of the station's

audience. During the course of the study, one year or four rating sweeps,

18

two stations underwent format changes. One station was in the midst of
heavy competition in the MOR format with the leaders of the market, two
AM giants. The subsequent change of this FM station to the new "personality

17 The other station

Gold" format increased audience size by over 50 percent.
undergoing format change formerly sat at the bottom of the Philadelphia
radio market employing a "Jazz" format. This station changed to an "All
Talk" format and by the next rating period had lost half of their already
small audience.18 Hesbacher's conclusion from these observations was the
obvious, "...a change in format does not guarantee an increase in audience
size."19
A more detailed demographic analysis of the Philadelphia market may have
added understanding to both of these incidents. Without it, a manager or
researcher must use deductive reasoning to explain the occurrance given
only the very basics of the situation. For example, it seems obvious from
Hesbacher's data that the audience receptive for another non-musicial format
was not likely to exist. Similar non-musical formats existed on two other
highly rated stations within the market and drew some 28 percent of the
total audience.20 This would lead a manager to believe that there may
not be enough "room" in the Philadelphia market for a third station offering
a similar type of entertainment. At best the decision to program another
non-musical format would be risky, because the goal of the third station
would be to capture some of the audience held by the existing non-musical
stations, and this strategy means attempting to compete with two experienced
and successful operations as reflected in the shares of the total audience.
As the final result showed, the decision made proved to be unsuccessful,

probably because some important considerations were ignored. The "Gold"

format success can be attributed to the quick, visible conclusion that one

19

9,000 watt "Gold" format could not possibly service all of the Philadelphia
radio market.

The example put forth by Hesbacher indicate that there is no heuristic
reason for format success in these instances beyond awareness of obvious
situations. Hesbacher does not advance a prototype for format success,
and infers that format success depends on the proportion of the total
audience held by the station. The study can be summed up as such;
"Programming astuteness has one general principle; program for your own
audience and against other stations in the same market."21

In his dissertation concerning radio broadcasting, Haring used regres-
sion techniques to define a relationship between a format's market entry
and the number of formats existing within a market. Basically, it was
an attempt to estimate the number of signals that had to be present
within a market before a particular format could exist. This was done
for the more common formats except for, "Popular-Standard," "Contemporary,"
and "Country and Western." The exceptions were due to the "non-esoteric"
nature of these formats; that these are the first formats to normally
enter a market.22

For example, by regressing the number of "classical" formats available
in a market on the number of competing signals within a market, one could
estimate the number of signals that would be present prior to the gener-
ation of a classical format. The regression took the form: F; = aj + bjsi’

where F? is the number of formats of type j in the ith market, and Si is

the number of competing signals in the i
23

th market for which format data

is available. By Haring's analysis in this example, 22 radio signals

would have to be present in a market before a classical format could be

successfully programmed in a market.24

20

Again, as with other research reported, Haring used the presence of
a station to confirm the fact that an audience for that station's format
exists. This is as if to say that a format preceeded the audience within
the market. The heuristic results generated by Haring are too general
and inconsistent to be of value to programmers and managers. For example,
though Haring states that 22 signals are necessary within a market before
a classical format can enable an appearance, a classical format does exist
in Louisville where only l4 signals can be found. The Haring formula does
not consider the possibility of a demographic pecularity within indiv-
idual markets. The result is a format strategy formula based on existing
markets. Inconsistencies are common to Haring's reported results. For
another example to further underline the point, Haring's regression method
provides that 26 signals must be present within a market before an "All
News" format will be generated.25 However, in Davenport, where only 14
signals are licensed an "All News" format does indeed exist. Again, by
not considering that a special demographic segment may be present within
the Davenport market that could allow for the existence and the success
of an "All News" format, Haring has left himself open to inconsistencies
in his attempt to develop a predictive method for this problem.

Using regression to predict format entry appears to be inappropriate.
The number of format outlets and the relative success of the format are
likely not to be highly correlated. It is common, for example, for a
single radio station with a particular format to have the audience share
that three other stations with the same format have combined. Thus, the
factors such as "good" management is difficult to measure, or statistically

designate. It is believed that the combined efforts of stations utilizing

21

the same format within a market can be estimated using variables found

to be linear combinations. Using the same method for prediction of format
outlets ignores the intangible effects upon the potential audience. So,
it is the terms such as "good management," "station loyalty," "promotion
effectiveness," "signal quality," and so on, which make consistently
accurate prediction of format outlets difficult for anything less than

a complete, exhaustive and probably expansive consideration of these
factors upon the market's radio audience.

The generalities and inconsistencies described in some of the above
analyses indicate that much effort may be misdirected in the area of
format research. By logic and by the results of previous research, the
examination of the existing situation within markets concerning formats
already present is not a consistent method for programmers and researchers
to discover success in a format strategy. The important factor is not
the number of stations in a market formatted in a certain manner, or the
relative success two stations have enjoyed with an identical format. The
occurrance of an esoteric format, or the success of a number of popular
formats in a market, is most likely generated by the presence of one, or
a combination of diScriminating variables that are demograhically unique

to that market.

22

CHAPTER II--NOTES

1R.P. Neilsen and T. Thibodeau, "Applying Market Research Methods

to Formating Decisions for Radio," Feedback (Spring l979), p. 19.

21bid.

3R. Moomey and R. Skolnik, "Typologies of Radio Station Target

Audiences,” Journal of Broadcasting(Fall l970), p. 470.

4Ibid.

51bid., p. 469.

61bid., p. 469.

71bid., pp. 469-470.

8J.T. Lull, L.M. Johnson and C.E. Sweeney, "Audiences for Contemporary

Radio Formats," Journal of Broadcasting(Fall l978), p. 45l.

91bid.

1OIbid.

11Ibid., p. 447.

121616., p. 446.

131616.

141616.

15W.F. Massey, "Discriminant Analysis of Audience Characteristics,"

Journal of Advertising Research (l965), p. 39.

16P. Hesbacher, "Radio Format Strategies," Journal of Communication

 

(Winter 1976), p. ll0.

23

171616.

18161a.

19161a.

ZOIbid.

2'Itn'd.

22J.R. Haring, "Competition, Regulation, and Performance in the

Radio Broadcast Industry" (Ph.D. Dissertation, Yale University, 1975),
p. l05.

23Ibid., p. 106.

24Ibid.

251616.

26Ibid., p. 100.

CHAPTER III

METHODOLOGY

This chapter addresses the methodological techniques used in this
analysis. First, a general discussion of the overall objective will
be reiterated. Then, a short summary of data sources will ensue.
Since this research will involve the incidental testing of basic
heuristics in program formats, these heuristics will be structured as
hypotheses and will be defined as such in this chapter. Finally, the
procedure for the treatment of data will be explained. This will
involve the employment of four statistical methods: Pearson product—
moment correlation, factor analysis, multiple regression analysis and
discriminant analysis. Ancillary to these major statistical treatments,
t-tests will be performed as a preview of those variables that may be
signficant in a comparison among market group means. Each of these
methods will be explained as to their general purpose as a statistic,

the theory behind it's use, and it's application to this particular study.

General Discussion

 

The objective of this study is to ascertain opportunities for
programmers of radio stations. These opportunities will consist of

vacancies in radio formats within markets. For example, if a particular

24

25

station within a market were aspiring to achieve greater success by
gaining a larger audience share, management would need to consider all
possible avenues to reach this organizational goal. For radio, those
althernatives are somewhat limited, however, the first and most obvious
strategy for this objective would be a change in the programming format
aired by a station. If a consistent method were developed that could
associate certain social characteristics within a market with the success
of particular formats in that same market, a radio station programmer
could identify format possibilities by demographic composition. It is
also likely that such a statistical method would reveal some relatively
inconspicious opportunities for programming alternatives.

In order to develop this statistical method, a knowledge of markets
where format success has been outstanding must be acquired. Across the
country, there are markets where certain formats enjoy abnormal success
in relation to the national average for these formats. This study will
identify these markets, attempt to identify the demographic elements
that are predictive of the success of formats, then relate these elements
of format success to actual audience shares for formats within the
market in order to reveal a possible format opportunity that would assist
the station in achieving its respective organizational goals.

For this analysis, 58 markets listed in Appendix A were selected

from the James Duncan bi-annual publication, American Radio. The criteria

 

for the selection of these markets was an adequate sample of cases for
each format, and availability of information for each market. For
example, certain markets in the northeast were deleted from this

analysis due to inconsistent data for most demographic categories. Rather

than SMSA's, geographic areas in this part of the country are territories

26

that are as much as 300 years old. Government SMSA's only loosely
coincide with these New England areas, and as such are not as consistent
as are those in other parts of the country. For these reasons, some
northeast markets could not be included, though their presence in this
study would have provided more geographic balance to the sample of
markets.

American Radio is a single source for the progress of Arbitron

 

audience estimates across 161 United States Markets. Among the many

informative ratings analyses in American Radio is a report stating the

 

best and worst markets for various formats. This consists of a listing
of the top 15 or bottom l5 markets for the following nine formats:1

l) Rock/AOR/Contemporary

2) Beautiful Music

3) MOR/Adult Contemporary

4) Country

5) News/Talk

6) Black/Disco

7) Spanish

8) Religious/Gospel

9) Classical

The markets selected for this study are all unique in that they are
known as either a very good or very poor audience for one of the formats
listed. These, then, are the markets whose successes in certain formats
are abnormal in comparison to the national average. For a comparison

of format successes in the chosen markets based on the national average,

Appendix B lists data reported in the Spring l980 edition of American

27

Radio.2

Also, to statistically illustrate some demographic differences
between the very good markets for formats and the remaining markets,
t-tests will be performed. The results of this and other analyses will
be reported in Chapter IV.

The data gathered for the initial statistical analysis was acquired
from a surprisingly small number of sources. Again, as stated earlier,

the chief criteria for variable selection was consistency and availability.

One of the major sources was the l977 City and County Data Book, which is

 

the most recent edition of this publication. From this was obtained SMSA-
based statistics on a number of variables across the 58 selected markets.

Another major source of data was Sales and Marketing Management's "l980

 

Survey of Buying Power." Here is found the most recent population
estimates (projections based on the 1970 Census), along with buying income
averages and market economic characteristics. Also, the publication offers
the latest retail sales figures on a per market basis for a number of
store group categories. Another source of data which has already been

referred to, American Radio, is a critical information reference for the

 

existing format configurations within markets. This source tells the
number of stations per format and it's share of the total audience within
markets. The information within this publication will serve as the basis
for comparison between the results this research will generate and the
actual situation within markets. Ancillary to these sources are: Sggt_

Radio Rates and Data, where descriptions of formats and format configuration

 

can be found; and Broadcasting Yearbook, where vital information about

 

station power, formats, management, rates, and various general information

about broadcast facilities within markets can be found.

28

Hypotheses

 

During the course of this study, certain variables will be found
to be significantly related to particular formats. This would be an
opportunity to test some generalizations that people may have about the
character of a format's audience. For example, it may be held by some
that the "Classical" music format attracts the upper income level as
it's listenership, almost exclusively. Similarly, the "Rock/AORI
Contemporary" format may be thought by some to attract soley the 18-24
age group. Audience research organizations that survey radio listeners
have encouraged this minimal variety of audience description by categor-
izing survey results by two demographics; sex and age.3 Other formats
may be the victim of stereotype as they may be referenced by only a single
variable. These stereotyped references will be the subject of the tested
hypotheses.

I The research hypotheses and null hypotheses are listed as follows:

H]: Markets that show an abnormally high share for the "Rock/AORZW
Contemporary? format will also be significantly greater than the
remaining markets for the variable representing the 18-24 age group.

H2: Markets that show an abnormally high share for the "Beautiful _
[Mgsic" format will also be significantly greater than the remaining
markets for the variable representing percentage of women within the
market.

H3: Markets that show an abnormally high share for the "MQRKE
Adult Contemporary" format will also be significantly greater than the
remaining markets for the variable representing the 50+ age group.

H4: Markets that show an abnormally high share for the "Country"

format will also be significantly less than the remaining markets

29

for the variable representing income per capita.

H5: Markets that show an abnormally high share for the "News/Talk"

 

format will also be significantly greater than the remaining markets for
the variable representing the 25-49 age group.

H

 

6: Markets that show an abnormally high share for the "Black/
Disco" format will also be significantly greater than the remaining ”1
markets for the variable representing the 8,000-9,999 household effective
buying income group.

H7: Markets that show an abnormally high share for the "Spanish"
format will also be significantly greater than the remaining markets for
the variable representing percentage below poverty level.

H8: Markets that show an abnormally high share for the "Religious/

Gospel" format will also be significantly greater than the remaining
markets for the variable representing percentage of women within the
market.

H0: There is no difference in how the specified variable is
distributed among the two market groups.

To confirm or reject these hypotheses, t-tests will be performed on
each format. For each t-test, the group means will be examined for
significant differences. The groups themselves will consist of the
following: Group l will be those markets which have the best share
of a market's audience for a particular format; Group 2 will be the
remaining markets in the analysis. A further discription of this

process will be provided in the section covering t-test results in Chapter

IV.

30

Pearson Product-Moment Correlation

 

Prior to treatment of data by higher order statistical methods,
variables found to be highly correlated with other variables included in
this study must either be normalized or deleted from further statistical
interpretation. To accomplish this task, Pearson product-moment correlation
coefficients of variables will be computed. The Pearson correlation

coefficient is a zero-order correlation and is used to measure the strength
4

\

of the relationship between two interval level variables. The strength

of the relationship would indicate both the goodness of fit of a linear

regression line to the data, Fnd‘WBenwthe coefficient is squared, the

proportion of variance in one variable shared by the other. For this

 

 

situation, computation of Pearson product-moment correlations is a very
convenient way to compute a large number of coefficients in the search
of highly intercorrelated variables.

The problem of high intercorrelation introduced above is called
multicolinearity. The presence of multicolinearity among the data has the
potential to cause misinterpretation of later results in statistical
treatments of the data.

Without deletion or normalization of variables, the data would be

somewhat ambiguous in that two or more variables would represent the

..-—1*-
__‘___._fi_._ -...-.”-

same trend in the character of the population. ES? Exampie. total marketni’g
population and total households across the 58 markets have a Pearson's

r of .997. XThis is because as populations increase across the 58 markets,

so do the total households within those markets. These variables are

almost perfectly correlated, and could be interchangeable in most

situations as representations of a market's population. In other words,

this is an almost perfect linear relationship. For this analysis, it is

31

necessary to have only a single variable represent total population,
so other variables that are highly correlated with total population
as total households is shown to be above, must be deleted to avoid

ambiguity, or normalized as a representation of that variable. The
criteria for extreme multicolinearity will be from the .8 to l.0 range.6

It is in these cases that normalization or deletion will be the alter-

natives to correct this situation.

Factor Analysis

 

It is likely that a large number of variables will exhibit the
same strong relationship indicated between total population and total
households. In order to objectively select a single variable to represent
the block of linear relationships, those variables highly intercorrelated
will be subject to a factor analysis. This treatment is perhaps the
most objective means for the selection of a single variable representative
of the linear relationship among those variables listed in Table 2. The
goal here is to reduce subjective selection of the variable list as much
as possible. The resultant factor matrix will show all the linear variables
in Table 2. The goal here is to reduce subjective selection of the
variable list as much as possible. The resultant factor matrix will
show all the linear variables in Table 2 loading on a single factor. The
highest variable loading would be the variable most descriptive, and that
variable will be thus selected as the appropriate representative variable
of the block of linear relationships in Table l.

The remaining variables will also be treated with the factor analysis
routine. The rationale for doing this is deletion of the variables whose
basic meaning is relatively trivial. For example, if a factor analysis

were performed with 30 factors and approximately 60 variables, variable

32

loadings on distant factors would be likely candidates for deletion.

In this hypothetical situation, should a particular variable load only

on the 29th factor, it could be deduced that this variable's relative
meaning was insignificant enough that it's deletion from further analysis
would not be unreasonable. It is proposed that this method would be a
statistically proper means for deletion of variables.

Multicolinearity can be dealt with to the degree that it is to be
reduced as much as is reasonably possible. One cannot carelessly delete
variables that add significantly to the explanation of relationships.

In this research, this goal of reasonable deletion of variables has been
accomplished with the grounds for deleting being: a) extreme inter-

correlation, and b) no relative importance in general for format description.

Multiple Regression Analysis

 

After the previous mentioned statistical treatments are employed
as a data reduction sequence, multiple regression will be used to attempt
to define a linear prediction equation for a format's audience share.
Multiple regression will be the first technique used here to explore the
various combinations of predictor variables that comprise a linear rela-
tionship model for a format's resultant audience share.

Multiple regression's fundamental purposes are to predict dependent

variables and to test research hypotheses.7

However, since an ancillary
purpose of this research is to discover variables that contribute to the
success of formats, no specific hypotheses have been developed here. For
this study, the identification of variables contributing to the linear
relationship with the dependent variables precludes hypothesis formation.

Thus, as stated above as a major purpose of the method, multiple regression

33

used here attempts to construct a predictive association between a
criterion variable and a set of predictor variables.

The basic principles of regression analysis used in a bivariate
case can be expanded to include a larger number of independent variables.
The form of the equation is as follows:

Y'=A+B]S]+BX+BX3+ ................ +BX

2 2 3 k k
where Y' represents the estimated value of the dependent variable, Y,;&
is the Y intercept, X1 k are thewdependent variable values, and

81 k are the corresponding regression coefficients.8

The meaning of
a regression coefficient in the above equation, 81 for example, stands for

the change in Y with a change in one unit of 4’ when X2
9

...... k are h91d
constant, partialled out, or controlled for.

In the construction of the models that will illustrate the relation-
ships between the format audience share and the predictor variables,
step-wise multiple regression will be used in which the variable contri-
buting the most significant explained variance will be entered first, the
variable explaining the next most entered second, and so on. This proced-
ure will cease due to predetermined statistical criteria so that when no
additional variables meet the criteria, the processwfill stop. The reason
for such criteria is so that only the best "explainers" of variance are
allowed into the prediction equation. Also, such criteria arelikely to
limit the sheer number of variables into a smaller, more understandable
amount.

This step-wise method of regression was chosen because alternative
techniques deny the researcher the opportunity to examine the equation

variable by variable, and thus showing the interactions of contributing

variables within the predictive equation. The step-wise method allows

34

an ordered, logical understanding of each component ()f the regression
equation, thereby better relating the contribution of each variable

towards it's relationship with the dependent variable.

Discriminant Function Analysis

 

Discriminant analysis is used when groups of cases are defined
a priori and the purpose of the analysis is to distinguish the groups from

10 In the situation of

one another on the basis of their score profiles.
this study, the a griori groups are to be distinguished by success of
formats within markets, as measured by audience shares. One market

group would have the best shares for a format, the other would have the
worst shares. From these grouped markets, an unknown number of discrim-
inating variables that measure market characteristics will be selected.
The mathematical objective is to then take these discriminating variables
and weight and linearly combine them so that the two groups are forced to

be as statistically distinct as possible.11

By taking these variables and
combining them mathematically, it is hoped to attain a single dimension
with "favorable" markets for a format at one end of the continuum, and
"unfavorable" markets at the other.
The linear combinations, or "discriminant functions" take the form:
D

= d1 Z + d. Z + ..... +d. Z

1 1 1 12 2 1p p

Where Di is the score on discriminant function i, the d's are the weighting
coefficients and the 2's are the standardized values of the p discrim-

inating variables used in the analysis.12

\.
In order to determine the effectiveness of the derived function in
actual discrimination and separation of groups, statistical tests will be

included along with the results at the end of the next chapter.

35

These statistics include: Wilk's Lambda, which is a measure of the
relative discriminating power within the variables; and, canonical corre-
lation coefficients, which is an indication of the actual separation

13 Wilk's Lambda is also the criterion for selection

of the two groups.
of the discriminating variables during the step-wise selection procedure.
The minimization of this statistic is the determination of the best
combination of variables that separate the respective groups.14 Also,
since actual membership of group cases is known prior to computation,
observing the proportion of correct classifications is a means to check
the functions ability to distinguish between groups.15

Once the functions are derived, it is then possible to achieve the
objective of this method for this study, the classification of markets
based on favorability towards formats. The status in this regard for
those markets which formed the functions is already known, so reporting
this would only be redundant. However, a market not belonging to either
group will be classified according to which of the first two market groups
it's characteristics are most similar. To designate the degree of
similarity, a standard score is given that market, then placed along the
function continuum. The actual classification into a particular group
is dependent on it's proximity to the respective group controids, or
means, of the first two defined market groups.

It is not beyond possibility, then to classify markets entirely
outside the boundaries of this study by calculating a discriminating
score, and again comparing it to the group centroids. However, as
mentioned in the opening chapter, caution should be maintained as to the

credible application of this study's results because the reliability of

such computed results is fully dependent on how the gathered data

36

approximates actuality within markets. Nevertheless, classification
of new markets would involve an equation similar to the earlier figure
used for selection of discriminating variables. Thus, the basic

classification equation would appear as follows:
CL: Cilv-l + cizvz + ccccc + Cipvp + cio

where Ci is the score for group i, the cij's are the coefficients with

CiD being the constant and the V's are the raw scores of the discriminating

variables.16

37

CHAPTER III--NOTES

1J. Duncan. American Radio, Vol. V, lst ed. (Kalamazoo:

Gilmore Advertising, Spring 1980), pp. 26-27.

21bid.

3Arbitron Radio Audience estimates for the Lansing-East Lansing
Market (New York: Arbitron Radio, October/November 1980), p. 3.

4N.H. Nie, C.H. Hull, J.G. Jenkins, K. Steinbrenner, and D.H. Bent.

Statistical Package for the Social Sciences, 2nd ed., (New York: McGraw-
Hill, 1975), p. 280.

51bid.

61bid., p. 340.

7F.N. Kerlinger and E.J. Pedhazur. Multiple Regression In Behavioral

Research, (New York: Holt, Rinehart and Winston, 1973), p. 361.

8Nie, op. cit., p. 328.

91bid., p. 330.

101616., p. 435.

”1616.

121bid.

131616., p. 440.

”1616.. p. 447

151bid., p. 445

161616.

CHAPTER IV

DATA ANALYSIS

This chapter reports the results of statistical treatments applied
to the data from the 58 markets listed in Appendix A. The first section
shows the process of variable deletion by Pearson product-moment
correlation and factor analysis. Then, t-tests will be performed to
discover variables that are distinctive among those markets that do
exceptionally well with a particular format. This section also answers
certain hypotheses suggesting possible relationships in this regard.
After that, the use of discriminant function analysis as a means for
formats will be reported. And finally, linear models constructed with
multiple regression analysis will show the capacity of this method for

prediction of audience shares of formats within markets.

Pearson's r

 

Appendix C contains a listing of the original set of variables that
was subject to treatment with Pearson product-moment analysis. Notice
that age, income, format, and market retailing categories are represented
in this analysis by both the original projected number of persons within
that category, and the normalized value of the original figure. For
example, during the Pearson analysis, the 18-24 age category in Lafayette

is represented by the original projected number of person within that

38

39

TABLE 2. High Intercorrelations

 

Variable

Correlation With Total
Retail Sales

 

NNNNNNNNN—l—J—l—lu—l—l—J—J—J—J
mummwa-aouaoowosmwa—Ioso

CDNO‘UT-§WN-J

Total Retail Sales

Total Population

18-24 Age Group

25-34 Age Group

36-49 Age Group

50+ Age Group

8,000-9,999 EBI Group

10,000-14,999 EBI Group
15,000-24,999 EBI Group

25,000+ EBI Group

Total Households

Public School Enrollment

Total Bank Deposits

Total Housing Units

Total Manufacturing Establishments
New Capital Expenditures

Total Retail Establishments

Total Wholesale Establishments

Total Wholesale Sales

Total Monthly Social Security Payments
Total Food Sales

Total Eating and Drinking Place Sales
Total General Merchandise Sales
Total Furniture Outlet Sales

Total Automobile Dealer Sales

Market Graduated Buying Power Index
Market Total Effective Buying Income
Market National Advertising Dollars

1.0000
.9817
.9818
.9851
.9818
.9515
.9228
.9367
.9761
.9871
.9715
.8650
.8593
.9519
.8816
.9412
.9662
.8895
.8881
.8923
.9870
.9939
.9943
.9957
.9512
.9934
.9816
.9650

 

1'

 

4O

category (22,184), and also by an index which represented 22,184 as a
percentage of Lafayette's total population (16.3%). This was done for
age, effective buying income(EBI), and format and economic categories in
anticipation of high intercorrelation between these demographic items.

Recall that the objective of treatment with Pearson product-moment
correlation is the discovery of areas of high intercolinearity. It is
anticipated that groups of variables will share extensive multicolinarity.
This is a problem when a large number of variables are intercorrelated
between .80 and 1.00.1 When high intercorrelation, or multicolinearity,
occurs between variables, it means that those variable items are in a
linear relationship with one another. Consequently, retaining more than
one of these variables is useless since the meanings that these variables
portray are vitually identical. Pearson's r locates highly inter-
correlated variable groups that require the special attention of the
researcher who in turn may alleviate the problem in order that ambigious
interpretation of later results may be avoided.

A Pearson's correlation matrix was computed using the 88 variables'
listed in Appendix C. Table 2 shows those variables which were inter-
correlated at .8 or higher. The general meaning of this block of
variables is logical in that as market sizes vary, the values of the
variable categories also change at a near linear rate. Table 2 also
shows those variables that are correlated with total retail sales.

These relationships may be more important than others because of this
figures importance to the potential income sources for local radio sales

efforts. As stated previously, retaining only one of this list of

41

variables is necessary to represent the linear trend shared by the
items listed in Table 1. This task, along with the identification and
deletion of trivial variables listed in Appendix C, is accomplished by
the application of factor analysis.

Factor analysis is a data reduction technique whose coefficients,
when squared, indicate the amount of variance explained by the relation-

2 With this principal in mind,

ship between the variable and the factor.
factor analysis will enable the most objective means possible for the
selection of a single variable from Table 1 for retention. It is most
likely in a factor analysis that those variables listed in Table 1 will
all load on the same factor. Squaring the coefficients will show the
variable which explains the most variance, and thus will be the variable
retained as the best representation of the meaning of those variables
listed in Table 1.

Table 3 shows the factor loadings for those linear variables listed
in Table 2. In order that computer calculations could be made possible,
however, some of the linear variables were deleted prior to the

computation of factors:

public school enrollment

total bank deposits

new capital expenditures

total wholesale establishments

total wholesale sales

market national advertising dollars
It was felt that these variables would not be indicative of the most

variance explained, which is one of the goals of employment of this

42

 

 

 

TABLE 3. Factor Loadings of Linear Variables
Variables Factor Loading

1. Total Population .99232
2. 18-24 Age Group .98991
3. 25-34 Age Group .99168
4. 35-49 Age Group .99106
5. 50+ Age Group .98257
6. 8,000-9,999 EBI Group .96927
7. l0,000-l4,999 EBI Group .97875
8. 15,000-24,999 EBI Group .98894
9. 25,000+ EBI Group .98301
10. Total Market Households .99250
11. Total Housing Units .98890
12. Total Manufacturing Establishments .94393
13. Total Retail Establishments .99026
14. Total Monthly Social Security Payments .93688
15. Total Retail Sales .97063
16. Total Food Sales .98385
17. Total Eating and Drinking Place Sales .97677
18. Total General Merchandise Sales .96553
19. Total Furniture Outlet Sales .96716
20. Total Automobile Dealer Sales .86096
21. Graduated Buying Power Index .98799
22. Market Total Effective Buying Income .99258

 

 

43

method. Reference to these correlations in Table 2 show that they are
not among the highest intercorrelations.

The factor loadings shown in Table 3 all were aligned on the first
factor, as predicted earlier. Squaring these coefficients reveals that
the variable "total effective buying income" explains 98.52 percent of
the variance between this variable and the factor. Other high factor
loadings include: "total market households," (98.5%); "total population,"
(98.4%); "25-34 age group," (98.34%); "35-49 age group," (98.22%); and
"total retail establishments," (98.06%). Because "total effective buying
income" statistically explains the majority of variance it will be
retained, while the remaining linear variables listed in Table 2 will
be deleted.

Deletions of trivial variables, determined by weak factor loadings
inside of the fifteenth factor were also considered concurrently. It
was decided by this researcher that due to the degree of subjectivity
involved in the interpretations of factor loadings, very liberal
criteria would be the rule as to variable deletion by this method. The
author strongly feels that more objective dealings with variables can
occur only by the use of methods more objectively interpretable, such as
the forthcoming multiple regression. Therefore, only those variables that
contribute little to overall variance explanation, that is, no better
than a .3 factor loading by the fifteenth factor, are appropriately
deleted from further analysis.

In the factor analysis, all 88 variables were included. Those
variables found to be relatively insignificant to variance explanation

based on factor analysis results are listed below:

44

- manufacturers with 100+ employees

- hotel, motel, trailer park and camp receipts

- crime rate per 100,000

- general merchandise establishments

- eating and drinking establishments

- normalized number of "Rock/AOR/Contemporary" formats
- normalized number of "Beautiful Music" formats

- normalized number of "MOR/Adult Contemporary" formats
- normalized number of "Country" formats

- normalized number of "News/Talk" formats

- normalized number of "Black/Disco" formats

- normalized number of "Spanish" formats

- normalized number of "Religious/Gospel" formats

- normalized number of "Classical” formats

The first five variables listed above were consistently below set criteria
outlined earlier and had factor loadings explaining less than ten percent
of variance between the variable and the factor. Inclusion of these
variables would be of little advantage in further interpretation.

The deletion of the normalized format presence variables is the
result of weighing differences between this group and the non-normalized
group. For example, the non-normalized group showed generally higher
intercorrelations with the "total effective buying income" variable.

Also, the non-normalized groups highest factor loadings appeared mostly
within the first factor, the factor explaining the most variance among the
variables and factors. The formats that loaded on other than the first
factor showed extremely high loadings. This indicates greater interaction
by this group, non-normalized format presence variables, with the

factor explaining the greatest amount of total variance. Therefore,

so as not to include redundant items, the normalized variables

45

for format presence within a market will be deleted from further
analysis.

With the variable reduction completed the revised variable list
now consists of 44 items. approximately one-half of the original
number of variables. Appendix D is a Pearson correlation matrix
showing the bivariate relationships between those variables remaining

at this point in the analysis.

T-tests ofggrogp means

 

For this section, t-tests were performed on the data in order to
discover whether or not market data were differentiated on the basis
of format success. Also within this section are the results of the
hypotheses stated in Chapter III.

The t-test was employed for two basic reasons. First, as
mentioned above, it can identify variables which were significantly
unique to markets that exhibited outstanding success for certain formats.
Appendix B shows these markets as they are grouped according to format
success. The second purpose of performing the t-test is confirmation
of hypotheses. The primal structure of each of the hypotheses in
ChapterIIIwas merely the conjecture of a statistical uniqueness shared
by a format and a variable. Because of the basic nature of t-tests,
this method enables a viable testing of the hypotheses.

T-tests are a statistical method for the purpose of testing whether

3 For this

or not the difference between two sample means is significant.
study, the two groups, one with high shares for a format and the other

remaining cases, pre-exist this analysis. Thus, the objective is to

46

find the effects which significantly differentiate the first group

from the second.4 The goal of "significance" is ‘tO be

"indicative of" or "signifying" a true difference between the two groups.5
In effect, what is being tested is the null hypothesis (H0), or the
hypothesis of "no difference." Should the probability computed by the
t-test be smaller than the significance level chosen, H0 is rejected.
Conversely, should the probability computed by the t-test be larger than

the chosen significance level, H is not rejected.6 Nevertheless, this

0
does not imply that the null hypothesis is, in reality, true. Thus are
the risks of inferrential statistics.

A t-test was performed between groups defined as the following:
Groupfll was those markets included in Appendix B which showed an abnormally
high share for the "Rock/AOR/Contemporary" format: Group 2 was the
remaining cases in the analysis. Table 4 shows the means, standard
deviations, standard error, t-value, and two-tailed probability estimates
for those variables shown to be statistically significant at the .05
alpha level.

As can be seen from examination of Table 4, most variables that
have significant probabilities have negative t-values. What is then
deduced from Table 4 is not what variables are uniquely associated with
the success of the "Rock" format, but which variables are significantly
ngt_associated with the success of the "Rock" format in these markets.

For example, the "median age" variable means show that markets that
do well with "Rock" formats have a generally younger population. This

is to be expected to be, in fact, true when considering possible

characteristics of this format's audience. Also to be expected is the

47

 

 

 

. .,- omN. ooN._ NooN.P N

NNo om om o NoN. coo. mooo. F apostoa xoopo co oooEoz
. . - ooo. ooo.o mNoN.m N

mpo mm mm N Fmo. mom. memo. F managed mzmz mo mgmgm
. . - oNP. NoN.F oomN.F N

moo om FF o Poo. Noo. oooo. F mooseoa 5362 co gonzo:
. . - ooo._ omo.PF ooNo.o_ N

oNo om oN N ooo.N moo.o omoo.o F moosaoo No: to oeoom
. . - ooo. oom.N oooo.o N .

ooo om _N N Poo. oNo.F oooo.F P mooeaoo No: to Loosoz
. . - ooN. .Po.o NNNo.oF N NoELoL utmoz
moo oo oo N NFo.F Noo.o NoPN.PP P Foc_ooooo co ozoom
ooo. om No.o ooo. on.o opNo.oo N oosooa sooooosooooo

oNo.P ooo.o mooN.oo P \No<\xooo co oooom
. . - oNP. moo. Nomo._o N
Foo oo oo o Noo. NFo.N ooop.om F ooeoz co mooooooamo
. . - oNo. Foo.o oNNo.oN N
_oo oo oo o ooN.. Noo.o oooo.oN F oooeo oo< +oo
. . - ooo. ooN.N Nooo.o_ N -
opo om mm N _oo. oo_.o NNNo.o_ P ooooo oo< oN o_
. . - ooo._ NNF.FP oNoo.oo N mooa< oooco oo
o_o oo oo N oNo.N oom.o PooN.PN P oop>to moooooocoo
. . - oNo. ooo.N mooN.oF N oo<.co meow»
ooo om NN N ooo. ooo.o mooN.N _ mo e6>o mooooooeoo
. . - Foo. mop.o oopo.oo N
ooo oo oo N NoN. oNo.N oooo.NN P oo< oooooz
auwprnmnoca soummgo mzpo>uu Locum cowpow>mo com: manage mpnowcm>
—wou-ozh mo mwemmo egoccoum ugoucmum

 

 

oxgoeoasmucouwmo<\xuom= goo moo—o>-a ucoowmwcmwm

.e m4m<h

48

high positive t-value for "share of the Rock/AOR/Contemporary format."
Another t-test was performed for groups defined as follows:
Group 1 contained the markets that showed the highest "Share for the
Beautiful Music Format; and Group 2 was the remaining cases in the
analysis. The variables that had significant probabilities at the .05
level are listed in Table 5 along with the accompanying statistics.
Dissimilar to the pattern found for the "Rock" format, "Beautiful
Music" exhibited predominantly positive t-values.
Notice the contrasts between the variables that are duplicated
in Tables 4 and 5. The median age for "Beautiful Music" is indicative
of a generally older audience for this format. This speculation is
reinforced by the positive values of the "50+ age group" and
"65 and over age group" variables and the negative values for both the
"18-24 age group"and "25-34 age group." The strong significance of
the "hotel,motel receipts" variables may be due to the fact that a
majority of the markets that were included in Group 1, and thus the
best in share for "Beautiful Music," are in Florida, a state known for
large touristtwade. This variable's interpretation should then be
with caution, however, the strength of older age groups in this analysis
is a very important and obvious consideration for programmers of this
format.
A t-test was performed for markets grouped as follows: Group 1
contained those markets which possessed the highest shares for the

"MOR/Adult Contemporary" format; Group 2 was the remaining cases. The

49

 

 

 

Noo. om oF.N “mm” wmm m mum“ m m mooeLoo No:
09. e 6: mm; wfi wwwmnm m 9...: 3:95.. mam...”
mop. om oo.~ WWW” WWMHF wmwwuw m muoseod coma: Fowwuoomm
so. a 9.. mm”. m3 wwmwm No 9.5 .2 a.
so. e .2. mm mm: ngnfi N. 9.5 .2 one
ooo. om No.N- “mm” WWW m mmwm w” m ooooo moo oN-o_
Poo. om om.o mm” o “mm WP mmww WP m mooooooo Pooo: .Poooz
ooo. om PN.o mmm N mmm w wwmm M, m meoo> mo Looo oowmmowmoo
.. om”. om mom ..
xuppwaonoeo Eocmmcu mapo>lu Lace“ compom>mo com: manage m—nomco>
Poouuoz» we mmemwo aconcoum ugoucopm

 

 

=oomoz Poooooooo= coo moooo>-o ooooocoooom .o oooop

50

The significant variables at the .05 level are listed in Table 6.

It is immediately apparent that only a very few variables are
unique to this format. This may confirm a common belief that many
hold about the "MOR/Adult Contemporary" format. That is, that the
"MOR/Adult Contemporary" format is the "all purpose" format, being
inoffensive to anybody. Also, as being the generalized format seeking
a mass audience as opposed to a segmented audience.

Perhaps the only variables worthy of notice in this t-test are
"percentage below poverty level," and "15,000-24,999 household EBI
group " This could be inferred to mean that the "MOR/Adult Contemporary"
format succeeds where there exists a slightly higher than average level
of income, or standard of living. However, practical determination of
"slightly higher" could be difficult to interpret from market to market.

For the "Country" format, a separate t-test was performed on
markets grouped as the following: Group 1 contained those markets which
possessed the highest share for the "Country" format; Group 2 was the
remaining markets in the analysis. Table 7 shows the significant t-values
at the .05 level.

The negative t-values for "income per capita" and "15,000+ family
income" infers that those markets successful in the "Country" format
have generally lower income levels throughout the market. Also of
negative association are the "number of Rock/AOR/Contemporary formats,"
"number of Beautiful Music format," and "number of Black/Disco formats."
This would suggest that part of the success of the "Country" format may
mean the lessor incidence of these above formats within the same market.

Again, interpretation of "lessor" is left to the judgment of the observer.

51

 

 

 

. . m—F.F owm.n omom.Fp N

000 mm mN w me.p own.m mNm¢.mm P mumELom mo: mo memcm
. . P—m. me.N oo¢m.N N

coo om No m mew. omm.N ooom.m _ mpmsgom mo:
. . mow. m—m.m oomm.mp N mpcwssmwpnmumm

oFo oo oo N ooN._ ooo.o oooo.FN _ oooxooeo ooo oooooo
. . one. mmm.m o¢mo.om N

ooo oo oo N ooo. ooN.P ooNo.oo F ooo ooo.oN-ooo.o_
. . u rum. ooF.o owpm.pp N augm>oa Zome mcw>w4

Noo oo oo N ooo. ooo.? oNoo.N _ oooooo moo ooooooozoo

Nuwpwnonogo soummcm mopo>lu Locgm cowaow>mo com: moooeo mpnopgo>
poopuozh to mwemmo vcoucopm cooncoum

 

 

=NgooooEooooo o_oo<\ooz. coo moooo>-o oooooooooom .o ooooo

52

 

 

Poo. om oo.N- WWW” mm” P wwww F m moosooo Noopo
ooo. om oo.o ”Mm M ”mm m mmww mm m poseoo Ngooooo co oooom
.. NoH NNNN NNNNN N ......
ooo. om No.o- oNN.o _oo.m wwmm m m mooseoo oomoz Pocoooooo
_No. om No.N- WWW” WWW w WWW“ w m moosooo Nooo
.. NNNH NNNN NNNNNN N .....NNNNNNN.
.. NNNH. NNNN NNNNNN N ...,NNNNNNNNNNNN.
ooo. mm mN.v “mm” wwmnfl wwwwum m mucmssmwpooumu ouo<
ooo. om o_.N WWW ” WWW MP mmmm mm m moooooooom oooo moo
.. NNNN NNNN NNNNNN N
.. NNNNN NNNNNN NNNNNNN N ......
.. NNNN NNNN. NNNNN N ...NNNNN...

 

ouwpwnonocm Eoummcu
Pwoulozh we mmcmmo

mopo>lu Loccm
oeoooooo

cowuow>wo com:
oeooooom

moooew mpnowgo>

 

 

=Ngooooo. Loo moo_o>-o ooooocooooo .N Noo<o

53

For the "News/Talk" format, a t-test was performed on variables
grouped as the following: Group 1 contained those markets that
showed the highest share for the "News/Talk" format; Group 2 contained
the remaining cases in this analysis. Table 8 shows the significant
t-values at the .05 level.

Apparent here is a higher level of income likely within large
metropolitan areas as exhibited by the variables; "income per capita,"
"15,000+ family income," and "25,000+ household EBI group."

The variables that show positive or negative t-values confirm
common speculation that this particular format thrives best within the
largest markets as opposed to more moderately sized markets. However,
one cannot infer that this format will succeed exclusively within
such a market. This is shown by the fact that "News/Talk" acquires a
substantial share of audience in the Youngstown (15.8 share), and

7 These two markets, as an example,

Monterey (14.4 share) markets.
while larger than many who sustain a good share for the "News/Talk"
format, cannot be considered as being of a similar character as the
metropolitan markets such as Philadelphia, St. Louis, or New York.

The t-test for the Black/Disco format contained markets grouped
as the following: Group 1 contained those markets that held the
highest market share for the "Black/Disco" format; Group 2 was the
remaining markets in this analysis. Table 9 displays the t-values
significant at the .05 level.

Positive t-values that will be indicative of the demographic des-

cription of this format aside from the obvious "share of Black/Disco

format" are fewer than most formats in this analysis. The variable,

54

 

Noo. om oN.o- Mum” WWW w wwwm m” m o_ooomoo: ooquHmommmop
.. NoH NNN. NNNNN N ..NNNNNNN.
ooo. om N_.o “MW” WWW _ mwmm w” m oooco moo oo-oo
ooo. om NF.N WNW” mm“ ” owmm w” m ooozo moo oo-mN
”Fe. om mm.N- WNW” mmmmnm wmmmnwp m mpouoz .mpmpoz
.8. e .2- mm”. a; mwmwnm N mug... wmmnwmg
ooo. om _P.m- WWW m WWW my mmwm m“ m moeoooooom oooo moo
ooo. om om.o WMMHN “mmmm wwww MM m moo_o5oo oeoooo +ooo.m_
.. NNNNN. NNNNNN NNNNNNNN N
ooo. om em.m mmwnw wmmnmp wmenwm m cone: mmoucmoemo
.. NNNNNN NNNNNNN NNNNNNNNN N
N_o. om om.N WNW m menm” WWW” mm m Nooom ooooeoo oooooooeoo
xuwpwaonogm Eocmmcm mopo>-u Losgm cowuow>mo cow: mozoew mpnowco>
Fwoulozp No mmcmmo ucoucoum ucoucopm

 

 

=N_oo\mzoz= Loo moopo>-o ooNUNNNoooo

.w m4m<h

55

 

 

 

ooo. om oo.o WWW” mmm _ wmmmn_ m Pooommopo wmsmmMom
ooo. om oo.o WWW” WWW”, wwmm o m moowooom Pouommo_o
ooo. om oo.o wwwn WWW ” mmmm m m mooseod Noopo
ooo. om oo.NF WWW”, “mm w wmwm W, m Npoo\mzozmwmsummom
ooo. om oo.o ”mm” Mwmn mmmeN m mooseoo Npoo\mzoz
o_o. om No.N- mm“ M mmm MP wwwm MN m mooeooo Ngooooo to oeoem
ooo. om N_.N “mm” ”mm W wwwm w m ooooooomoooo mwn%flwmoz
mpo. om m¢.N mmmn MMMHM wwwmnm m oposeom owmoz Fz$wuoomm
Ppo. om No.N- Mm“ ” WWW WP mmww mm m ooscoo Nooo co oooom
ooo. om wN.e WWW” mwmnm wwwmnm m mmeLom xoom
_oo. oo oo.m- WNW” WWW M mwmm m” m mooosomopooomo oooooom moo
ooo. om mo.N- mam” mww W mmmm m m mooosomooooomo oooo
.oo. oo No.o WWW ” WWW M ”mm” MN m opooommumowommMoN
zuwpmnmaoem Eoummgo mz—o>-p eocem comoow>mo com: monogo mpnowgo>
Fwou-ozp 4o mmgmwo vconcopm ncoucopm

 

owsthzou .w m4m<h

56

 

 

 

. . oNN. Noo., oNoN., N oesooo
oNo om oN N ooN._ ooo.o Nooo.o F FoomooNNooooopoo co oaoom
. . FoN. _oo.m ooop.o N
ooo om No o ooo._ Poo.o ooNN.NN _ ooeaod Noo_o Lo ocoom
. . NNN. ooo.p FooN.P N
Foo om No N ooo. ooP.N oooo.o _ mooseoa Noopo
Noo. om oN.N- ooo.? ooN.F_ oNoo.o_ N ooeooo Neoooosooooo
Foo._ oNo.o oooo.o F opoo<\ooz co oeoom
. . - ooo. ooo.N oooo.o N moooaoo
oNo om oN N oNo. ooo.._ ooNo._ F Nooeoosooooo oPoo<NNoz
. . - oNo. __o.o NNNN.oF N mooooomopooomo
opo oo oo N Poo. ooN.N mo_o.o_ F oooxoooo ooo oowooo
. . oNo. Noo. Nooo.N N mooosomt_ooomo
ooo om No N No_. ooo. oNoo.o _ omooooooeoz Poeoooo
. . Noo. oNF.N oNNo.NF N
Noo oo oo o ooo. ooN._ ooFN.o_ P mucosomtoooomo oooo
. . ooo. oNo.o No_N.o_ N Poooo
N_o oo oo N oN_.F _No.N FNmo.o_ P ooeo>oo zopoo oooooooeoo
. . oo_._ ooo.N Nopo.o N
ooo oo oo o NoN.N ooo.N oo_N.oN F Noopo oooooooeoo
xuopmnmnoea genome; mapm>-p LoLLm cowpmw>mo cow: monocw mpaowco>
poohuozh mo mmemma veoucopm ugoucopm

 

 

.oomNonoopo. ooo mooPo>-o oooowooooom

.m m4m<~

57

"percentage black" is likely to be exclusive of this format. Also,
of importance for this format audience is a seemingly lower income
level as inferred by the "percent below poverty level" variable.
"Food establishments" and "general merchandise establishments" have
highly significant probabilities for this format.

Unique in this t-test treatment for this format is the t-value for
"share of Religious/Gospel format." The Black/Disco t-test is the only
instance aside from "Religious/Gospel" itself where this incident occurs.

The remaining format representing minority listeners, "Spanish"
was subject to a separate t-test where markets were grouped as the
following: Group 1 contained those markets with the highest share
for the "Spanish" format; Group 2 was the remaining formats in the
analysis. Table 10 lists the variables and their t-values that were
significant at the .05 level.

Again, as with "Black/Disco," this format shows indications of a
combination of urban lower income areas, as can be seen by the t-values
for "percentage urban population," "percentage below poverty level,"
"crime rate per 100,000 persons," and "15,000-24,999 household EBI
group."

Notice that while some of the positive t-values in conjunction with
the format variables infer the presence of this format within the larger
markets ("percentage urban," "crime rate per 100,000 persons"), others
contradict this suspicion. For example, the negative t-value for
"number of News/Talk formats" infers that a format typical of the larger
metropolitan areas is not present in the same markets as "Spanish."

It is likely that geography is a consideration for this format.

58

 

 

 

. . ooo. Foo. oooN. N poseoo
ooo oo oo N ooo.o opo.op ooNo.o_ _ omoooom Lo ocoom
. . Noo. ooo. ooo_. N
ooo oo op NP ooo. Noo.? oooN.o _ mooscoo omwooom
. . ooo. oop.P oooo. N o o o m3.
oNo oo on N Poo. ooN.F ooNo.P _ up at o NP o\ z
F_o. om No.N- NNo._ oop.__ oNFo.oP N Noseoo ococoooooooo
ooo.F Noo.o oNoo.o F opoo<NNoz co ozoom
. . - oNo. oNo.N ooN_.Po N ooozo ooo
ooo oo oo o ooF.F Poo.o oNoo.oN P opooomoo: ooo.oN-ooo.oF
ooo. No N_.N ooN.ooN Foo.ooo_ ooo.oNoo N moomcoa ooo.oop
ooo.ooo ooN.oooN Noo.oNoo _ ooo mooo motto
. . _om. _oN.N oooo.o_ N moomooo ooo.F
ooo oo oo N ooN.P moo.o oooo.N_ F coo oooo ooeoo
. . - Noo. ooN.o oooo.PF N mooxopoeN +oo_
ooo om oN N ooo. ooo.N ooNo.N _ .oooz to oooooooeoo
. . - oNo. ooo.o oopo.oo N 6656: ootooooo
ooo om No o mNp.o ooo.F_ oNoo.oo F ooozo oooooooeoo
. . oNo. NNo.o ooFF.o_ N _o>oN oocoooo
Poo oo oo N ooo.o ooo.op oNPo.oF _ zoooo ooooooocoo
. . FNo.F opp.._ oNoo.NN N oooooooooo
ooo om PN N ooo.o Noo.o oNFo.oo F oooco co ooooooueoo
. . ooN._ oNF.o oooo.op N Nooom
ooo om om N ooN.o NNN.oP oNo_.oN P cowozoa co oooooooeoo
xuwpwnoooca soommLm mopo>uu eocem cowaow>mo com: oooew mpnowgm>
Proouozp No wmcmmo ucoucopm vcoucoum

 

 

.omooooo. Loo moopo>-o ooooocooowo

.op m¢m<k

59

Since most of the larger markets are not in proximity with the South-
west United States, one cannot deduce conclusively that "Spanish" would
not be successful in such a market size. In fact, in Los Angeles, the
third largest market (whose data are not a part of this study), "Spanish"
garnered a 5.4 percent share in the April/May 1980 Arbitron sweep of that

market.8

A share which is above the national average for this format.

A t-test treatment was performed for the "Religious/Gospel" format
with similar marketing groupings: Group 1 was those markets that displayed
the highest national shares for the "Religious/Gospel" format; Group 2
was the remaining cases in the analysis. Table 11 shows the significant
t-values at the .05 level.

As was the case for the "MOR/Adult Contemporary" format t-test results,
"Religious/Gospel" does not appear capable of segmenting any particular
audience that could be distinguishable by the variable parameters used in
this tudy. Besides the variable representing, "Share of Religious/Gospel
format" all other significant t-values are negative, and thus would be
indicating non-associations with the characteristics of this format's
audience.

For this format and the "MOR/Adult Contemporary" format, it is
entirely possible that the objectives of this study in general may not be
possible with the variables currently in use for analysis if these trends
repeat in ensuing methods. I

A final separate t-test for the "Classical" format included markets
grouped as follows: Group 1 contained those markets that had the highest
share for the "Classical" format; Group 2 was the remaining markets in

the study. Table 12 lists the significant t-values at the .05 alpha level.

60

 

 

 

. . mop. omP.F mmON.F N Hosea;
ooo om om __ oNo. oNo.N oNoo.N _ Foomoonooooo_oN co oeoom
Noo. we oo.N- Pem.mmN Pmo.nmo~ Fppm.momm N
Pom.mmm ouo.moop ooom.mpoo P mcomcma ooo.oo_ Loo mwswcu
. . . Noo.F one.o— —omm.om N
oNo om oN N ooo.o opo.op ooNo.oN F ooooo_oooo oooco oooooooooo
. . - mpm.P mNm.oF Nomw.mp N
oNo oo oo N ooo.. oNo.o Nooo.o N Nooom ooooooo ooooooozoo
xuwpmnonoco souwmcm ma—o>-u Loecm cowuoo>mo com: moooew mPnowco>
meuuozo yo mmemwo neoucoum ueoucoum

 

 

opumwoo\moowmwpmm= Loo mmopo>lu pcoummwcmom

.PP m4m<h

61

 

. . - NNo. _oo.o oFNo.o_ N mooooposo +oop
ooo om oN N ooo._ oNF.N . oNNF.o N no.3 .oooz co oooooooooo
_oo. oo oo.m- ooo. ooo.o oNoN.oo N 6626: oo_ooooo

oNo.N ooo.op NNNN.oo P coozo o. ooopoooeoo
ooo. om oN.o- ooo.. _oo.o. oPNN.oN N moeoooozom ooo:
oNo.o oNo.NP NNNN.om P ooo o. oooooo oooooooooo

. . ooo.. ooo.N N_oo.op N
Noo oo no N oNo.P NNN.m oooo.NN _ oeoooo oposoo +ooo.o_

. . - ooo. oo_.o oNNo.F_ N po>oo
ooo oo op N ooo. ooo., ooo_.N P Noooooo oopoo oooooooomo

. . oN.oo oN.Noo oN.NoNo N
ooo om on o No.oop oN.oFo oN.NoNo N oowooo ooo oeoooo

. . opo.P ooo.o~ NNoo.oN N ooooopoooo
ooo om N_ o NNN.P moo.m mono.No _ oooco oooooooooo

. . oNo.NNo No.ooNo oo.ooNo_ N
Npo oo oo N oNo.ooo oo.mNoP NN.ooooF P oooooz ooo

. . Noo., ooo.o. oNoo.o_ N
oNo oo oo N oNF.N NNo.o oNNo.FN F ooooo_oooo oooocoo oooooooeoo

. . oo.oooo__ oNo.oooNoo oN.Nooooo N
ooo oo oo o No.oooooo oNP.ooo_ooN P_.PPNo_oo F ooooo_oooo Foooh

Eu. —. _. waQOLn— EOUmOLn— m=~m>lv c~OLLw :OwHMw>mD 5mm: mQZOLG mpnmem>
F_ou-o3h mo mmgmmo ugoucoum ugovcopm

 

 

._ooommo_o. Loo moo_o>-o oooowooooom

.NP m¢m<h

62

 

 

 

.. NNNH .NNNH N... N

.. NNNH NNNH. Noo. N ..
_oo. oo oo.o WWW” WWW W ”Nam m m moooooa _ootmmooo
ooo. om oo.o MW“ N “Wm M ommmmum_ m mooseoo oomoo\xoo_o
ooo. om N_.o WWW” mmmu_ “mm“ N m mooogoo Npoo\mzoz
.. NNNN oNo oNo N ......NN.NNN.
ooo. om oN.N ”WM” mmm W mmamnw m mooEooo oomoz Focoooooo
ooo. om _o.o ”mm P “Wm m NWMW M m mooeeoa Nooo
.. NNNH NNNN NNNNNN N .....NNNNNNN.

Apopwnonoea genome; mopo>up cogem cowuow>mo com: manage apnowgo>
pwouuozh we mmemwo ucovcoum ugoucoum

 

 

omszmhzoo .NF m4m<h

63

Similar to the "News/Talk" format t-test results, the results
of the "Classical" format as listed in Table 12 infers larger market
characteristics. This speculation stems from the t-values for;
"percentage urban population," "percent in one unit structures,"
and "percent in owner occupied homes." Again as well, a higher level
of overall income is inferred from the mean comparisons from
"EBI household median," "income per capita," "percent below poverty level,"
and "15,000+ family income."

The general market characteristics of these formats, News/Talk
and Classical, basically remains consistent across each other. It may
be possible that the variables that differentiate these two large market
formats from each other may not be among those variable parameters chosen

fbr this analysis by either the propercflassification or representation.

Test of Hypotheses
As mentioned at the outset of this section, the t-test would also
provide the ability to test the hypotheses stated in Chapter III. .This
sub-section reflects the results of the tests for each of the nine
major hypotheses as interpreted from the t-test on each format which was
discussed in the preceeding section. For all of this discussion, Table
13 will be referenced as it contains the t-values and one-tail probability
estimates for each variable that was the subject of a hypothesis. Recall
that the significance level used throughout this section is the .05 alpha
level.
H]: Markets that show an abnormally high share for the
"Rock/AOR/Contemporary" format will also be signifi-
cantly greater than the remaining markets for the

variable representing the 18-24 age group.

64

 

 

 

 

 

66> ooo. No.o mmmm mm m ooooo ooo +ooo.oN o:
oz ohm. Nm. wwmmnflm m case: we mmopcmugma w:
mo> moo. mo.N mmflm m” m _o>oN Noeoooo zo_oo N:
mo> ooo. oN.F “WWW w m ooooo Noo ooo.o-ooo.o o:
mo> _oo. N_.o mmwm m” m oooeo ooo oo-oo o:
no» oNo. NN.N mm mew w oooooo ooo oeoooo o:
oz NoN. oN. mmmm mm m ooooo ooo +oo m:
or emo. —o.P wmmmnwm m :meoz we mmoucmuemo NI
6.» Noo. om.N mmmm m” m ooooo ooo oN-o_ .1
mo. A.p NMWmfiMwmwca mopowuu meow: oooeu mWMMMMWo> mwmmzpoo»:
_ooo-ooo
mo_oowco> momoooooxmlco momoo .op ooo<h

65

This hypothesis implies a generally larger mean proportion

of the 18-24 age group within those markets that have the highest

audience shares for the specified format, or,

H = G

1 > G

l 2
where G1 represents the percentage of the market within the 18-24

age category for those markets that have the highest shares for

 

"Rock/AOR/Contemporary," and G2 is the percentage within the 18-24
age group for the remaining markets.

Table 13 shows that the t-value provided by the t—test of group
means is significant at the .05 level. Thus, the null hypothesis,
indicative of no difference between group means, is rejected.

The research hypothesis of this question, H], is accepted by the

results of this analysis. It can be said then, that those markets which
have the highest share of audience for the "Rock/AOR/Contemporary"
format also has a significantly greater percentage of persons within

the 18-24 age category.

H2: Markets that show an abnormally high share for the
"Beautiful Music" format will also be significantly
greater than the remaining markets for the variable
representing the percentage of women within the
market.

This hypothesis infers a larger mean proportion of women in the

market group that consists of the highest shares for the "Beautiful

Music" format, or,

where W1 is the mean percentage of women in those markets representing

66

high format shares for "Beautiful Music," and W2 is the mean percentage
of women in the remaining markets.

The notion that this hypothesis might be, in fact, true
originated from the fact that the national mean of Arbitron survey data
shows this format's audience to be skewed heavily towards women. The
actual figures showed male listenership nationally at 41.2 percent, and
female listenership nationally at 58.8 percent.9

Table 13 shows that the t-value calculated from the t-test of
markets in this format is not significant at the .05 level. The null
hypothesis (H0) is then not rejected, while H2 is rejected.

One should fully understand the inferrence made in this second
hypothesis to illustrate the nature of the acceptance of the null
hypothesis and the rejection of the research hypothesis. It is perhaps
confusing that a format skewed so heavily towards a segment of the
population as shown above could cause the acceptance of the null hypothesis.
The inference made for H2 is that a percentage of the entire population
is greater than that percentage of the population for a different group
of markets. So, it is a comparison between markets. The figures stated
above where the heavy skew towards females is shown for this format
would result from diary reports of preference of the "Beautiful Music"
format within markets. To summarize the basic difference, this analysis
attempts to illustrate differences in formats based on between market
differences, while Arbitron will show the differences of format listen-
ership within markets. Thus, it is the hypothesis question which was in
error. The assumption of a proportional difference between markets

based on the specified variable originated from results from a different

67

sampling frame. In other words, different sampling frames will provide
different results
H3: Markets that show an abnormally high share for the
"MOR/Adult Contemporary" format will also be
significantly greater than the remaining markets
for the variable representing the 50+ age group.
The inferrence made from this hypothesis is that the mean for the

50+ age group is significantly larger for those markets with high audience

shares for this format, or,

where A1 is the mean of the 50+ age group in those markets with a very
high "MOR/Adult Contemporary" format share, and A2 is the remaining market's
mean for this variable.

Common belief is that this format typically attracts the "older"
audience group. Arbitron estimates indicate that the 50+ age group
accounts for 52.6 percent of listeners for this format as the national
mean.10 It is theorized, then, that this differential also exists for those
market groups specified in this t-test.

The result shown in Table 13 shows that the t-value computed for
this format in accordance with the 50+ age group variable indicates no
significance at the .05 level. Thus, by these results, H3 is rejected,

while H is accepted.

0
While "MOR/Adult Contemporary" may indeed attract generally older
audience groups as it's core audience, this cannot be inferred to also
mean that those markets that have the highest shares overall for this
format will also have a proportionally larger number of 50+ age group

audience members.

68

H4: Markets that show an abnormally high share for the
"Country" format will also be significantly less than
the remaining markets for the variable representing
income per capita.

This hypothesis infers a generally lower income level for those

markets that have the highest shares for the “Country" format, or,

where, 11 is the mean income per capita level for those markets that
have the highest share for the “Country" format, while 12 is the mean
income per capita level for the remaining markets.

The significance t-value shown in Table 13 confirms the fourth
hypothesis (H4) as being true. Thus, as H4 is accepted, H0 is subsequently
rejected for this question. The confirmation of this hypothesis shows that
markets where "Country" acquires it's highest audience shares also have
generally lower income levels on a per capita basis for the market as
a whole.

H5: Markets that show an abnormally high share for the

"News/Talk" format will also be significantly
greater than the remaining formats for the variable
representing the 35-49 age group.

This variable infers a greater proportion of persons 35-49 within
those markets that hold the highest shares for the "News/Talk" format, or,

H = A

5 1>A

2
where A1 is the mean proportion of persons 35-49 within markets that excell
in audience share of the "News/Talk" format, and A2 is the mean proportion

of the 35-49 age group for the remaining markets.

69

Table 13 is referenced to find the t-value of this variable and
the associated one-tail probability estimate. This referral shows that
the t-value is significant at the .05 level. H0 is then rejected, and
H5 is accepted. This means that for this format, there is a statistically
larger percentage of persons 35-49 where this format shows it's highest
share of total audience.
Notice the means of the two tested groups as shown in Table 13.
It may seem as if the approximately 1.2 percent difference in the population
means of these market groups could not actually be significant. In other
words, the face validity of this conclusion is questionable. However,
one must recall that most of the markets that show the highest share for
this format are relatively large. Appendix 8 lists those markets that
comprised the t-test Group 1, those markets that were highest in market
share for this format. The mean for total population of these markets is
approximately 3.8 million persons. Thus, a seemingly small percentage
difference among these markets would invariably be a consideration of
literally tens of thousands of potential audience members
H6: Markets that show an abnormally high share for the
"Black/Disco" format will also be significantly greater
than the remaining markets for the variable representing
the 8,000-9,999 effective buying income group.
This implies that the markets with the highest share for "Black/Disco"
wil also have a larger proportion of households within the specified EBI

group, or,

where I1 is the mean proportion of households within the highest "Black/
Disco" share markets, and 12 is the mean proportion of this EBI category

for the remaining markets.

70

Table 13 shows the t-value computed for this variable in the t-test
of this format to be significant at the .05 level. This is, however,
not a strong significance, as is so from some of the other variables
in Table 13. Nevertheless, H6 is accepted under‘the parameters of this
analysis, while H0 is not accepted. Thus, it can be said that due to
confirmation of this hypothesis, markets that show high shares for this
formatalso have a generally higher proportion of households within
the 8,000-9,999 EBI category. The basic meaning of this is the presence
of a larger proportion of lower income households within the same markets
where "Black/Disco" is most successful in gaining audience share.
H7: Markets that show an abnormally high share for the
"Spanish" format will also be significantly greater
than the remaining markets for the variable
representing percentage below poverty level.
The inference of this hypothesis is a greater occurrance of poverty
within those markets that excell in gaining audience share with the "Spanish"

format, or,

where P.I is the mean percentage of persons living below the poverty level
within the markets that hold the highest shares for the "Spanish" format,
and P2 is the mean percentage for the same specified variable in the
remaining markets. Table 13 displays the t-value for this tested variable.
The significance level is shown to be statistically significant at the

.05 level. The research hypothesis (H7) is thus accepted, while the

null hypothesis (H0) is rejected. It can then be said by consequence of
this confirmation, that those markets that have the highest share of

the market in the "Spanish" format also possess a larger percentage of

71

persons living below the poverty level. Obviously, this would indicate
a probable lower level of total spendable income throughout such
markets, a consideration of importance to both retailers and local radio
stations.
H8: Markets that show an abnormally high share for the
"Religious/Gospel" format will also be significantly
greater than the remaining markets for the variable
representing percentage of women within the market.
This infers a high percentage of women within those markets that
possess the highest audience share in the "Religious/Gospel" format, or,

H = W 3> W

8 l 2

where W1 is the mean percentage of the women within those markets that
have the highest audience shares in the "Religious/Gospel" format, and
W2 is the mean percentage of women within the remaining markets.

The idea that such a relationship might be true originates from
Arbitron estimates which shows that the listenership of this format to be
skewed heavily towards women. The actual national means for both sexes
as the audience for "Religious/Gospel" is; men, 31.8 percent; women, 68.2

11

percent. It was thought that this large difference of means could be

reflected as well between market groups.

Table 13 shows the resultant values computed in the t-test fOr
this format and the tested variable. The t-value and subsequent probab-
ility estimate is not significant at the .05 level. The conclusion of this
rejection of the research hypothesis (H8) is that there is no difference in
how the percentage of women is distributed among the two market groups.

In other words, the null hypothesis is confirmed.

('(‘1'114LII ‘ .- J‘1|¢-JJJI

72

As;confronted in the second hypothesis (Hz), the sample frame
and nature of the hypothesis itself differed in the basic sense of the
anticipated results. The Arbitron figures which inspired the content
of the two hypothesis (H2, H8), were based on diary reports, and were
thus indicative of listenership of formats within markets. This analysis
conversely attempts to show differences between markets and market groups.
The data to properly test the validity of the Arbitron reports did not
exist in this analysis in the corresponding representations. Again,
different sample frames will yield different results.
H9: Markets that show an abnormally high share for the
"Classical" format will also be significantly greater
than the remaining markets for the variable represent-
ing the 25,000+ effective buying income group.
The inference of this hypothsis is a significantly greater proportion
of households within the 25,000+ EBI group for those markets that show

the highest audience share for the "Classical" format, or,

where I1 is the mean percentage of households within the specified variable
for the markets showing the highest share for the "Classical" format,

arid I2 is the mean percentage for the specified variable for the
remaining markets. .

Table 13 shows the corresponding figures for the t-test that is the
test of this hypothesis. It shows that the t-value for this variable is
highly significant at the .05 level. The statistical result is summarized
by saying that markets which excell in audience share in the "Classical"

format also have a higher proportion of households within the 25,000+

73

EBI category. The means shown for the two respective groups confirm
the stronger significance level shown. Such strong significance may
mean a generally higher level of spendable income for those markets

that exhibit higher audience shares for "classical."

Discriminant Analysis Results

 

This section reports the results of treatment of the refined data
with discriminant function analysis. Each table is comprised of two
sections which contain information about the derivation of the function
itself. Shown with the variable name are: two unstandardized discrim-

inant function coefficients, one for each of the two group§=(i.e., best

 

markets, worst markets) which are used for the computation of disdfiminant
scores; Wilk's Lambda, which is a measure of a group discrimination; and,
the probability estimate of the contribution of the variable's discrim-
ination capabilities as measured by Wilk's Lambda. The second part of
the table gives figures relevant to the effectiveness of the functions
as a discriminatory tool. The eigenvalue and it's associated canonical
correlation show the ability of the functions to separate the two groups.
To interpret thediscriminating power with Wilk's Lambda, it must be
remembered that it is an inverse measure, or in other words, the
{ larger lambda is, the less discriminating power is present.12
The objective of the usage of discriminant analysis is to determine
the linear functions of two groups defined as follows; Group 1 are the

markets whose audience share for format X are the highest in the country,.

and; Group 2 are the markets whose audience share for format X are the

74

lowest in the country. By deriving these coefficients, the classification
of markets previously unclassified is possible. Thus, for example,

should a programmer be interested in the favorability of his market
towards a given format, this type of analysis would take into account the
factors that have contributed to the success and failure of that format,
form a range of scores, and classify markets as either "favorable" or
"unfavorable" for that format. 1

With the purpose of classifying markets whose potential for success
for a format is unknown, it is then most important to compute the means for
the two groups described earlier, then to compute the scores fOr the unknown
groups. These unknown scores can then be placed on a continuum along with
the known scores. The "favorable," or "unfavorable" classification will
then depend on the scores most similar to the unknown market's score.

It is the classification of these unknown markets which is the goal
of this section. The computation of the discriminant functions is the
means by which this classification may be achieved. The information generated
by this method may give some statistical insight into the potential of
markets for certain formats relative to other markets.

This section's analysis is limited to four of the nine formats. The
formats used here are those with the highest national average audience
share. Referral to Appendix B show that these would be: Rock/AOR/
Contemporary, Beautiful Music, MOR/Adult Contemporary, and Country.

These are thus shown to be the most popular formats based nationally. The
remaining five formats are the less popular nationally as shown within
Appendix B, and as such, these esoteric formats have their popularity based
on a more local, or regional criteria. AA grouping such as that of Group

2 would be impossible because there are so many markets with zero audience

75

share for these formats that calculation of variance and implementation
of these results would be very misleading. So, since the presence of
these formats are generally inconsistent across markets, the following
five markets are precluded from this section's analysis:

-News/Talk
-Black/Disco
-Religion/Gospe1
-Spanish
-Classical

Table 14 shows the relevant information regarding the function
derived for markets grouped according to their success with the "Rock/AOR/
Contemporary" format. This grouping as well as the other four market
groupings discussed later, is formed with Group 1 being those markets with
the highest audience shares for this format, and Group 2 being those
markets with the lowest audience shares for this format. The table shows
the discriminating power of this function. Notice that before the function
is derived, Wilk's Lambda is .0219127, meaning that a large amount of

discriminating power exists within the variables being used. Since there

are only two groups by which functions may be formed, only one discriminant N,

function is pgggibleofor each format in this section's analysis. The

 

eigenvalue and the canonical correlation listed in Table 14 mean that the
ability of this function to separate the two groups is exceptionally good.
The classification of the grouped markets was 100 percent, so this function
should have credibility in the separation of the ungrouped markets into
"favorablel and "unfavorable" classifications.

The contribution of the individual variables to the function are
shown by the standardized discriminant coefficients. Interpreted similarly

as correlations or regression coefficients, these values show the relative

 

 

mNmmwmm. cop mmmmm.¢¢ _
ooo. o FmF.Fo «NpmpNo. o ,,

 

mucoowwwcmwm no mcoocmlwsu ounEoN comuucom cowuopmgeou mucoweo> mzpo>cmmwm cowguczu

 

 

 

76

 

6.xpoz Looco Fouoooooo N
ooo.oo- Pooo.oo_- Noooomooov
ooo. ,Nooow.‘ NxopopNom. opoooo.N ooNoN.F_ mom mo_om _ooooo
, x/lll\\\ \ acmscmwpnoumm mezuwcgou
ooo. _NoNo.N- Nmoooo. oooNoo.o oNoooo.N- mo_om _Noooo
mmwccosuemz Focwcmo
ooo. ooopo._- oo_Noo. oopooo.N oooNooo. - mopom .ooooo
ooooemooooomo oooo
ooo. oNoNF._- momNoNowp FoooNo.o oooooo.o- oNo mooseoo N_oo\mzoz
ooo. Ammmwmdao .«mawooflpp NomNoo.o- oooooo.N oeoom oooooooo
.yrllnHMx ;:;wu:; ANV xcocoasmucoU\mo<\¥uom
ooo. .Nooonmw ..NoNMNoo. oNFoFo.N- NoFNoP.o AHMWV mooosomopooomo oooo
oooootccooo
comuuczm onnsoo N ozoeu P oaocm
.oom oooooooeomoo m.NF.3 ooooooccooo oooooooa ooooowccooo oooooooo opoooeoo
umuwveoucwum ucocwEPLUmFo ucocwsweumwo

 

 

compouwwwmmopu umxeoz oagocoosmpco0\mo<\xuomw Loe-mucmcomsou compucom pcocwsmomwo .wp upon»

77

Discriminatory power that the variable adds to the discriminant function.13

Taking the absolute value for comparison, it can be seen that "general
merchandise retail sales" and "Rock/AOR/Contemporary audience share"
are approximately the same in their contribution to the function. These
are followed by "food establishments," "food establishment retail sales,"
"News/Talk formats," and "furniture establishment retail sales" in that
order. ‘

With these discriminant function coefficients, it is then possible

to compute the discriminant scores for markets. This score is computed,

 

 

RXflTElElEleflfl_EEEflg9‘SCElW‘"atlng variable by_1t s correspgnglngwgggjflg;

l4

ient and adding these_productswtogether along with the 9995t99t° In

.-.
_,.-0 Ii.“——-OJ .-

U" ‘--q>-« mn,‘ j -..—cu: -

Table 15, separate discriminant scores have been determined for each
previously ungrouped market. These scores are in standard form, meaning
that over all cases in this analysis, the scores will have a mean of zero
and a standard deviation of one. These single scores, then, represent 1
the number of standard deviations away the market is from the mean of

all cases, which is given at the top of the table.

The categorization of markets into "favorable" and "unfavorable"
groups is dependent upon the relative distance in either direction from
the grand mean along the discriminant function continuum. To avoid possible
misclassification of markets into the wrong category, a "neutral zone" of
classification has been determined, the parameters of which are shown
also at the top of the table. This constitutes a 10 percent area on
either side of the grand mean where erroneous classifications are most
likely to occur. Cases whose scores fall within this "neutral zone" are
thus left unclassified, being essentially too close to call for inferren-

tial statistics.

78

Table 15. Classifications of Ungrouped Markets for the "Rock/AOR/
Contemporary? Format

 

Group Centroids (Means)
Function 1

Group 1 6.66506
Group 2 -6.05914

Mean Score of all Cases: .30296

20 Percent Deviation Around Mean: 1.57537 to -.96946
Discriminant Score Range: 7.9957 to -7.7601

Percent of Cases Classified as Neutral: 20.7%
Percent of Grouped Cases Correctly Classified: 100%

 

Highest Discriminant Favorable/Unfavorable or

 

Market Probability Score 'Neutral (1 10%)
Group Classification
Albany 2 -2.0850 Unfavorable
New York 2 -5.9488 Unfavorable
Atlantic City 1 1.2991 Neutral
Grand Rapids 2 -.2996 Neutral
Bloomington, IL 2 -.7490 Neutral
Pittsburgh 2 -6.6144 Unfavorable
Syracuse 2 -2.1128 Unfavorable
Duluth 2 -4.6965 Unfavorable
Cincinnati 2 -2.7906 Unfavorable
Fort Wayne 1 .3941 Neutral
Eugene l .7080 Neutral
Wichita Falls 2 -2.3222 Unfavorable
Springfield, MO 2 -l.2064 Unfavorable
Knoxville 2 -18902 Unfavorable
Shreveport 2 -l.9220 Unfavorable
Austin 2 -2.4498 Unfavorable
Johnstown 2 - .2516 Neutral
Lubbock 1 3.6361 Favorable
St. Louis 2 -7.4084 Unfavorable
Youngstown 2 -3.5928 Unfavorable
Boston 2 -3.2741 Unfavorable

Table 15 (continued)

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (1 10%)

Group ClassifiCation
Houston 2 -4.609O Unfavorable
Birmingham 1 .4221 Neutral
Savannhan 1 2.8477 Favorable
Tallahassee 1 4.1902 Favorable
Pensacola 2 - .7678 Neutral
McAllan 2 -2.3842 Unfavorable
San Antonio 2 -l.6207 Unfavorable
Bakersfield 2 - .8020 Neutral
Lancaster 1 3.4179 Favorable
Pueblo 2 -2.9767 Unfavorable
Denver 2 -2.3352 Unfavorable
Seattle 1 .3298 Neutral
Reno 2 .1663 Neutral
Mobile 1 .6030 Neutral
Springfield, MA 2 -l.8877 Unfavorable

 

 

80

The grouped markets used to derive the functions on which the
discriminant scores are based do not appear in the tables, but suffice
it to say that in all formats, the markets within these groups were
correctly classified 100 percent of the time. In other words, no
misclassifications occurred in the computed analysis of these grouped
markets.

In Table 15, a majority of the listed markets were classified as
either unfavorable or neutral (89%). The "unfavorable" classification for
markets here means that the characteristics of these "unfavorable"
markets are most like that of the Group 2 markets, the worst "Rock/AOR/
Contemporary" markets in the country. However, it is nevertheless
possible that an "unfavorable" market for this format may still garner a
leading format share in the market. The deception lies in the fact that
this format is the dominant format in the country. This is why infer-
rences made from these results must be made with some prior knowledge of
this format in mind. Appendix B shows the national average share for
this format to be 33.04 percent, or about twice as much as the next popular
format. The worst "Rock/AOR/Contemporary" market in the country, Dallas,
still attains a 21.1 format share of the total audience.15 The market that
is closest to the grand mean in Table 15, Fort Wayne, has an actual audience
share for this format of 37.4 percent, the dominant share in that market.16
So, it is likely that most "unfavorable" classifications in Table 15 would
have PTOJECtEd audience shares of somewhere between that of Dallas and Fort
Wayne. This is good relative to other formats, but still not as dom-

inating as the 49+ format shares held by the Group 1 markets in this

format.

81

The number of "unfavorable" classifications is likely to mean
that these markets are saturated with formats in this category. Thus,
better opportunities may lie with other formats. To repeat, the terms,
"favorable" and "unfavorable," are based on the relative discriminant
score comparison of all markets in this study fOr only the format analyzed.

Table 16 shows the information associated with the discriminant
function developed for classification into favorable and unfavorable
groups for the "Beautiful Music" format. Again, the Wilk's Lambda of
.0139569 shows that the discriminating power of the variables listed is
extremely high. The eigenvalue of 70.65 and canonical correlation of
.993 shows a high degree of separation produced by the variables. The
best (Group 1) and worst (Group 2) markets for "Beautiful Music" were 100
percent correct classifications. Obviously, the ability of these
variables to classify these markets correctly is excellent.

The standardized discfiminant function coefficients show that
"Beautiful Music audience share," and "35-49 age group" are approximately
equal in importance to the discriminating power of the function. These
are followed by. "Country audience share," "furniture establishments." and
"Beautiful Music formats.“

Table 17 shows the classification of markets previously ungrouped.

In the column marked "highest probability group, the "1's" mean that the
highest probability of classification for that market is into Gnoup 1,
while the "2's" mean that the highest probability of classification
would be into Group 2. The table shows that 42 percent of the listed
markets were classified as "unfavorable," meaning that these markets
are most similar to those markets in Group 2, the worst audience shares

for this format. The range of audience share for markets included in

82

 

 

 

 

 

 

 

FNooNoo. oo_ ooooo.oN P
ooo. o ooooopo. o
mucouweocmwm ma mgoacmlwgo munEwN coouuczu cowuopmgeou mocoweo> mzpo>cmmwu comuoczu
6.xpoz coooo Foooooooo N
ooo_.oom- ooo._opp- Noooomooov
ooo. oooNo.P ooooNo. NNoNo_.N- oooNF.F_- Amwv oeoom oooowooo ocooooo
ooo. oopoo.N- NNoooo. NoooN.o_ Pomoo.oN x; oeoem
fxumucmwu=< oomoz Pomouoomm
ooo. NNoNo. - ooopmp. ooNNN.NN ooNom.oo mmw mooseoa oomoz Pootooooo
ooo. ooooo.P- Nooopo. ooooo.No ooNoo.oo me mooosomo_oopmo oeoooocod
ooo. Noooo.N- ooomoo. oomop.om mooNN.NN oooco ooo oo-mo
ooooooccooo
cowuuczm ownsoo N osocw _ ooocw
.oom oooowsooomoo m.NFo3 ooooooccooo oooooood ooooooccooo oooooooa opoooco>
um~_ueov:opm ucmcwewgumpo acocoewgumwo

 

 

cowuoummwmmopu pmxgoz oupmoz Fo$wuzomm= Lo» mucmcoqsou cappuczm acocwEwLUmwo .mp mpnop

Table 17.

83

Classification of Ungrouped Markets for the "Beautiful Music“
Format

 

Group Centroids (Means)

Function 1
Group 1 -10.15039
Group 2 6.09023

Mean Score of all Cases:

-2.03008

20 Percent Deviation Around Mean: -.406018 to -3.654l42

Discriminant Score Range:

8.8605 to -15.1933

Percent of Cases Classified as Neutral:
Percent of Grouped Cases Correctly Classified: 100%

 

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (1 10%)
Group Classification

Anchorage 1 -8.8233 Favorable
Great Falls 2 6.3214 Unfavorable
Lansing 2 2.3163 Unfavorable
Billings 2 1.1657 Unfavorable
Terre Haute 2 4.7183 Unfavorable
Kalamazoo 1 -2.5706 Neutral

New York 1 -15.1933 Favorable
Pittsburgh 1 -7.6041 Favorable
Syracuse 1 -4.9673 Favorable
Cincinnati 1 -3.6261 Neutral
Milwaukee 1 -6.3555 Favorable
Fort Wayne 1 -2.7575 Neutral
Eugene 2 1.8947 Unfavorable
Asheville 2 6.0189 Unfavorable
Wichita Falls 2 7.5384 Unfavorable
Springfield, MO 2 8.3878 Unfavorable
Knoxville 2 .6085 Unfavorable
Shreveport 2 8.1035 Unfavorable
Austin 2 8.8605 Unfavorable

84

Table 17 (continued)
Favorable/Unfavorable or

 

Highest Discriminant +
'Market Progagz;ity Score Ng¥2g21f§2alggg
Johnstown 2 6.4336 Unfavorable
Lubbock 2 5.7042 Unfavorable
St. Louis 1 -4.0888 Favorable
Philadelphia 1 -6.1681 Favorable
San Francisco 1 -9.3982 Favorable
Youngstown 1 -6.1706 Favorable
Boston 1 -7.4075 Favorable
Chicago 1 -3.7728 Favorable
Houston 2 - .4837 Neutral
Dallas 2 1.1606 Unfavorable
Jackson, MS 2 - .6470 Neutral
Birmingham 2 .2963 Unfavorable
Savannah 2 1.6270 Unfavorable
Tallahassee 2 4.4574 Unfavorable
Pensacola 2 3.6426 Unfavorable
Baltimore 1 -7.9168 Favorable
Bakersfield 2 -l.87761 Neutral
Denver 1 -9.3007 Favorable
Seattle 1 -5.7948 Favorable
Reno 2 -l.951l Neutral
Sacremento 1 -8.3980 Favorable
Sioux Falls 2 - .9494 Neutral
Springfield, MA 1 -10.1825 Favorable

 

 

85

this analysis is between 4.2 and 32.6 percent. The market closest

to the grand mean in Table 17 is Reno, whose actual audience share is
16.1 percent, which coincidently approximates the national share average
for this format.

It can be interpreted that the strong negative values in Table 17,
New York, Denver, and San Francisco for example, are most similar to those
Group 1 markets shown in Appendix B, and thus may be capable of high
audience shares for this format.

Table 18 includes the function developed to assist in the classification
of favorable and unfavorable markets for the "MOR/Adult Contemporary"
format. The varibles listed interact to be effective in the discrimination
among the two groups. The standardized discriminant function coefficients
show that "MOR/Adult Contemporary audience share" is a strong contributor
to the function's power to separate the groups. After that, "income per
capita," "eating and drinking establishment retail sales," "35-49 age
group," and "Rock/AOR/Contemporary audience share" contribute to the
function in a similar degree to each other.

The bottom of Table 18 shows that the variables described above have
considerable discriminating power, as indicated by Wilk's Lambda of
.0244518. The variables also produced a high degree of separation as
shown by the eigenvalue of 39.89 and canonical correlation of .988.

This function was used to compute the individual discriminant scores
shown on Table 19. Here, "unfavorable" is in the positive direction.
About 58 percent of the classifications were "unfavorable," as these
groupsare thus statistically shown to be most similar to the Group 2

markets. The nine markets categorized as "favorable" would be most like

 

 

mmmomwm. cop mmomm.mm p

 

 

 

 

86

 

ooo. m wwm.me mpmceNo. o
mucoummmcmwm mo meooamuogu ounEoN cowpucou coowo_mgcou mocovgo> mapo>cmmmm cowuucom
m.xpwz cmpw< Fouwcocou a
oomm.ommu mo¢¢.¢eo Aucoumcouv
ooo. Noonm.p mmoemo. mmmNo.mP nmmoo.om mopom Fwoumm
pewssmwpaopmm
mcpxcwco coo mcwuom
ooo. FNONo.N onomP. momNmm.N Nnumm.NP mcogm mucmmcaq
zeoganmucou »P=v<\moz
ooo. om~e_.F NmooNo. oomwm.m mpoNem.o weogm mucmwu=<
agoLoQEmucoU\mo<\xuom
ooo. Nwmmmpp ommmmo. NNomN.mm uNmnm.me oooeu mm< melmm
ooo. NNNPN.~- oppmmo. Po-mmPNmowpm.- FeumNmeme.- ouwoou goo msouem
pcmpowwemou
cowuucom ounsoo N oooco F ooogu
.mwm acocwsmcumwo m.xpm3 ucmmuwmemou cowuucou pcmmuwmwmou compucou mpnoveo>
umNPucoccoum pcocwewcumwo pcocwsmcummn

 

 

cowuouPNmeopuomeozoxuocoagmucou upou<\moz= to» mpcmcoaaou coouucoo acocwsweumwo .w— «poop

87

Table 19. Classification of Ungrouped Markets for the "MOR/Adult
Contemporary" Format

 

Group Centroids
Function 1

Group 1 5.59389
Group 2 -6.29313

Mean Score of All Casts: -.34962

20 Percent Deviation Around Mean: .839082 to -1.538322
Discriminant Score Range: -8.6426 to 7.2540

Percent of Cases Classified as Neutral: 13.8%

Percent of Grouped Cases Correctly Classified: 100%

 

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (1 10%)
Group Classification
Lafayette 1 5.1524 Favorable
Anchorage 2 -8.6426 Unfavorable
Lansing 2 - .4860 Neutral
Billings 2 -3.8260 Unfavorable
Kalamazoo 1 .8120 Neutral
Augusta 2 -4.2572 Unfavorable
Huntsville 1 - .3216 Neutral
Tampa 2 -6.5418 Unfavorable
Miami 2 -4.5565 Unfavorable
Albany 1 1.1950 Favorable
Atlantic City 1 3.0658 Favorable
Grand Rapids 1 .8377 Neutral
Asheville 2 -4.7598 Unfavorable
Wichita Falls 2 -3.5191 Unfavorable
Springfield, MO 2 -5.2203 Unfavorable
Austin 2 -3.5925 Unfavorable
Johnstown 1 - .0302 Neutral
Lubbock 2 -5.9039 Unfavorable
St. Louis 2 -6.4247 Unfavorable

88

Table 19 (continued)

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (1 10%)

Group Classification
Philadelphia 1 1.7479 Favorable
San Francisco 2 -1.8091 Unfavorable
Youngstown 1 2.0950 Favorable
Boston 1 2.0950 Favorable
Chicago 2 -2.3501 Unfavorable
Dallas 2 -3.2335 Unfavorable
Jackson, MS 2 -2.1970 Unfavorable
Birmingham 2 -5.5787 Unfavorable
Savannah 2 -5.9696 Unfavorable
Pensacola 2 -2.6755 Unfavorable
Baltimore 1 4.6832 Favorable
San Antonio 2 ~3.846O Unfavorable
Bakersfield l .1894 Neutral
Lancaster 2 -2.4631 Unfavorable
Pueblo 2 -2.1127 Unfavorable
Denver 2 -4.6958 Unfavorable
Seattle 1 .1312 Neutral
Reno 2 -2.6648 Unfavorable
Mobile 2 - .7606 Neutral
Sacramento 2 -1.9520 Unfavorable
Sioux Falls 1 .9859 Favorable
Springfield, MA 1 2.1837 Favorable

 

 

89

those Group 1 markets that are listed in Appendix B. The market closest
to the grand mean in this sub-analysis is Huntsville, whose actual
audience share for this format is 8.1 percent. The variables them-
selves are somewhat varied in discriminatory power. Obviously, the
variables "percent over 65," and "Country audience share," are large
contributors within the single discriminant function. “News/Talk
audience share" and "Rock/AOR/Contemporary audience share" are somewhat
less imporant than the first two variables based on the standardized
coefficient.

The function itself reveals outstainding group separation capabilities
as shown by a 181.11 eigenvalue and a .997 canonical correlation. The
variables discussed earlier show exceptional discriminatory power as
measured by a Wilk's Lambda of .0054912. This function serves to achieve
the widest separation of Group 1 and Group 2 markets among the four sub-
analyses. Not surprisingly with such separation, correct classifications
were 100 percent.

The coefficients of the variables were multiplied by the variables
raw values in order to determine the discriminant scores shown in Table 21.
Sixty percent of these listed markets are classified as favorable. This
is the highest occurrance of favorable classifications among the four
format analyses. It may be that "Country" has the best potential within
markets, more so than others. Sacramento, whose actual share for this
format is 7.9 percent, is the market whose discriminant score is closest
to the grand mean.

To summarize the purpose and objective of this section, classifica-
tion of markets for statistical substantiation of probable success for a

format was the goal. Prediction of audience shares themselves may be better

9O

 

 

oomNmmm. cop wnopp.Fw~ P

 

 

 

 

 

ooo. o Nem.m¢ Npmemoo. o
mucouwmocawm no mgoocm-w;u mango; cowpucoo :owuopmgcoo mucowgo> mopo>cmmwu coopucoo
m.xpwz qu$< Pouwcocou a
oooo.o_o- NNNN.NN_- Noooomoooo
ooo. mNNmN.— Fmomoo. oummm.m_ Nonmmmm. meocm
mucmoco< x—oe\m3mz
ooo. Nommm.N- NnmNmp. wcpoeN.N- oowmme.¢ meogm mucmwuo< acucoou
ooo. .oome. - mmeoeo. Nmpmom.P oompmo.N meocm mucmouo<
homeooEmpcou\ao<\xuom
ooo. Npmnm.N mmoNFO. omomN.mN owmuoo.—- mo Lm>o ucmucma
pcmwuwwwmoo
cowaucom ouasoo . N ooocw _ oooew
.mwm poocwswcommo m.x_w3 powwoweemou cowuucoo pcwouwmmwou eowuucou mpnomco>
cmeucoucoum poocoeweumwo poocwsmgomwo

 

 

mcovuouwmwmmopu umxgoz oxcocoou= to» mucmcanou cappucou pcocwswgomwo .ON mpaow

91

Table 21. Classification of Ungrouped Markets for the "Country" Format

 

Group Centroids (Means)
Function 1

Group 1 -8.25288
Group 2 18.56898

Mean Score of all Cases: 5.15805

20% Deviation Around Mean: 7.840236 to 2.475864
Discriminant Score Range: -9.7676 to 18.8739
Percent of Case Classified as Neutral: 15.5%
Percent of Grouped Cases Correctly Classified: 100%

 

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (1 10%)

Group Classification
Lafayette 1 -7.7904 Favorable
Anchorage 1 -6.0508 Favorable
Great Falls 1 -7.1105 Favorable
Lansing 1 -3.4914 Favorable
Billings 1 -5.6222 Favorable
Fayetteville 1 -9.7676 Favorable
Terre Haute l .0825 Favorable
Kalamazoo 1 1.7234 Favorable
Augusta 1 -1.2855 Favorable
Tampa 2 15.8755 Unfavorable
Albany 2 10.2729 Unfavorable
Grand Rapids 1 2.5175 Neutral
Bloomington, IL 1 .2089 Favorable
Pittsburgh 2 12.3026 Unfavorable
Syracuse 1 3.7347 Neutral
Duluth l .4855 Favorable
Cincinnati 1 2.2896 Favorable
Milwaukee 1 3.2717 Neutral
Fort Wayne 1 1.3976 Favorable
Euguene 1 - .9412 Favorable

92

Table 21 (continued)

 

Highest Discriminant Favorable/Unfavorable or
Market Probability Score Neutral (t 10%)

Group Classification
St. Louis 2 16.5079 Unfavorable
San Francisco 2 13.7376 Unfavorable
Youngstown 2 12.6637 Unfavorable
Chicago 2 8.7181 Unfavorable
Houston 1 2.4581 Favorable
Dallas 1 .9350 Favorable
Jackson, MS 1 3.1430 Neutral
Birmingham 1 1.4569 Favorable
Savannah 1 - .4906 Favorable
Tallassee l - 1.1645 Favorable
Pensacola 1 - 5.3901 Favorable
McAllan 1 3.7751 Neutral
Baltimore 1 4.9232 Neutral
San Antonio 1 - .0609 Favorable
Bakersfield l - 2.6446 Favorable
Lancaster 2 6.1829 Netural
Pueblo l - 1.8323 Favorable
Denver 2 5.4681 Neutral
Seattle 2 8.6674 Unfavorable
Reno 1 - 2.1378 Favorable
Mobile 1 - 3.3074 Favorable
Sacramento 2 5.2695 Neutral
Sioux Falls 1 .7830 Favorable

 

 

93

achieved with the multiple regression equations provided in the following
section. Discriminant analysis has the ability to predict group membership,
but specific inferrences beyond that may be misleading. The results shown
are applicable to this study exclusively because of the quality of data.

The point to be made is that this sort of analysis may be useful if
modified and applied frequently.

It is also important to understand the distinction and neanings of the
"favorable/unfavorable" classifications. These are relative only to the
other markets status for that particular format. It may be equated with
z-scores for a number of student's performances on an examination. An
"unfavorable" classification for one format does not give insight into
interactions with other formats within the market. It can only be inter-
preted as a degree of similarity to those market groups which were the
basis of the derivation of the discriminant function.

Classification of new materials can be achieved by computing a discrim-
inant score, then comparing it to the group means and the grand mean of the

market groups shown earlier. The accuracy of this process is, however, unknown.

Multiple Regression Results

 

This section reports the result of the treatment of the data with
multiple regression analysis techniques. Figures relevant to this discussion
are displayed in Tables 22 through 30. Listed in these tables are: the
variable name; the standardized beta; the unstandardized beta, or regression
coefficient B; the standard error of B; the F statistic of the variable;
and, the probablity estimate of the regression coefficient for the variable.
Also included in the tables are the analysis of variance statistics which
show information relevant to the text for R2, the overall test for goodness

of fit of the regression equation.

94

As mentioned in Chapter III, the regression equations are generated
using step-wise procedures where the independent variable explaining the
most variance with the dependent variable will be entered first, the indep-
endent variable explaining the next most variance entered second, and so
on. The goal is the identification of the best combination of independent
variables which enable the highest level of prediction of the respective
dependent variable. The predictive capacity of the equation is determined
by the R2 statistic in the analysis of variance table. For example Table
22 shows a coefficient of determination, R2, of .737. This means that
approximately 74 percent of the variance in the predicted dependent
variable "Rock/AOR/Contemporary audience share" is accounted f9? by the
combined effects of the independent variables also listed in Table 22.

The remaining 26 percent of variance is unaccounted for by the variables
entered into the regression analysis. The "unexplained" variance could
be either measurement error of the variables, of simple unmeasurable
effects upon the dependent variable.

The regression model shown in Table 22 displays the independent
variable predictive of the dependent variable, "Rock/AOR/Contemporary
audience share." Those variables which contribute the most to this equation
can be seen by interpretation of the column showing the standardized betas
for the respective variables. Squaring these values show the amount of
variance contribution relative to the other variables shown in the table.
With this in mind, predictors of special notice can be interpreted to be,
"50+ age group," "death rate per 1,000 persons," and, "Rock/AOR/Contem-
porary formats." These variables have the largest impact on the predictive
ability of this equation while the remaining variables contribute to a

lessor degree. Generally, this set of independent variables show strong

95

 

 

 

 

 

oooN o
No.PP co
ooNoN.oF o
ooooo. o opooopoz
o_NoN. No
ooo. _PNoNo.op oo_o.Np _oNo.ooN<v Noooomooov
mop. ooooNNo.N Noooooo.P mo_oooN.N- Nooo.- mooscoa Foooammopo
Noo. oooopNo.o NNoooNN.F ooNoooP.N NNNP. moanooo_oooao oaoomozoa
ooo. NNNoooo.N oFNooooo. oooooNo._- FNNN.- moosaoa
_ oLoLooaooooo opoo<Nooz
o_o. NooNooP.N PNFoomo.F FNooNoo.o oo_o. moomzoo ooo? zoo oooo ooooo
ooo. NNNNoN.oP oooNoNo.P opoooNN.N- oo_o.- ooooo Noo ooo.o-ooo.o
ooo. NNPoNom.o PNoNNNNo. oNNoPoo.N- oFNN.- apostoo Noomoo\mooootpoo
ooo. oPNmo_.N_ oNoooooo. onoooo.N oNNm. mooseoa
xcocooswucou\mo<\xuom
_oo. opooNo.N_ ooopoo_._ oNNNoNo.o- oNoo.- mopoo
Foooom ooaooo ooooaoaoa
Poo. oooo_o.N_ oNFooooo. oNoo_No.P opoo. oooeo Noo ooo.o_-ooo.oF
ooo. oooNoo.NF oomooooo. omooooo.2- omoo.- ooooo ooo +oo
ooo. NPNFoo.oo oooooooo. oPNoNoo.o- oooo.- apostoa oomooNNoopo
.o. a o Lotto oaooooom o oooo opoooeoo

 

 

mpnmwgo> pcmucmamo mg» mo mgogm mocmmu=< Nuocanmucouwmo<\xoom saw: cowmmmcmmm mpmwopsz .NN mpan

96

 

 

 

 

ooo. a

mo.o to

.oooo.o a

NoooN. o opooopoz

ooooo. No

ooo. oNo_omN._ ooop.o NooN.oP-N<o Nooaomeooo
ooo. _oooooN.o NNooo_No. NNNNNNN.P- ooNo.- moomzao

ooo, ooo oooo ooooo

Noo. ooooNoo.o opNoNooo. oooooooo.- NFFN.- oooco ooo oo-oo
ooo. ooNoooo.o oooNoNom. FooNNooo. moFN. mooseoa

oomoz Poooooooo
No_. oNoooNN.N oopoooom. oNNooooo. opop. moosaod

Fmomoo\mzommmpmm

ooo. oNNmooN.o PoopoooN. ooNNooo._- oFoN.- moostoa _ooomaopo
ooo. ooNNFoo.o FoNoNooF. Nooooooo. - Noo..- mopom Pompom

mmmucmsugmz Foemcmw

Noo. oooNoN.FN mo-ooopopop_. mo-ooFoNoooo. oo-ooooo. ooooaz Noo

ooo. oooNoo.op NoNo_No. NooomNo.F oooo. ooo oooooz
.o« u m Logan ucovcmpm m oumm wpnmweo>

 

 

mpnooco> acmvcmaoo mg» mo mgonm mucowco< ummoz Fomwuoowm now: cowmmmemmm mpqupoz .mN mpno»

97

predictive capability as indicated by the analysis of variance statistics
at the bottom of the table.

Table 23 features the variable "Beautiful Music audience share"
as the dependent variable. Obviously, the independent variable, "median
age" in overall predictive capacity. The analysis of variance table shows
this equation to be a good predictor of audience share for this format
as the combined effects account for approximately 64 percent of the
explained variance.

Table 24 shows the regression coefficients and related figures for
the dependent variable, "MOR/Adult Contemporary audience share." The best
variables for variance explanation in this relationship are, "MOR/Adult
Contemporary formats," and "News/Talk formats.“ In the case of this
dependent variable, comparatively few demographic predictors appear.
Recall that during the results of the t-tests that only a select few var-
iables were significant there as well. These two occurrances support the
inference that this format fails to segregate audiences that can be ident-
ified by a number of measurable social characteristics. The notion that
the "MOR/Adult Contemporary" format is the "mass appeal" format can thus
be generally accepted.

The regression coefficients shown in Table 25 comprise the prediction
equation for the dependent variable, "Country audience share." Here is
a host of moderate to very good predictors whose additive effects are
statistically shown to be an excellent predictor of the specified dependent
variable. Among the best predictors are: "Percentage foreign stock,"
"Percentage Black," and "8,000-9,999 EBI group." The analysis of variance
statistics show that approximately 85 percent of the variance in the pre-
dicted dependent variable is accounted for by this equation, thus making

this model a good approximation of the actual relationship between

 

 

.000. a

 

98

 

_No.o to
Noooo.op a
ooopo. o opooo_oz
oo_oo. No

ooo. oNNooo.op ooo.o oNoo.ooN<o Nooaomooov

ooo. opoooo_.o oNoooNoo. NoNoooN.P o_FN. moasaoo

. pamoz Focoooooo

Noo. ooooNoo.o oo_o_moo. NNooo_No..- Poo_.- mooosoma_ooomo oooa

oNo. oooNoNo.o ooooNooo. ooooNNo.N- oooN.- ooooo

Noo ooo.op-ooo.o_

N_o. N_ooooo.o oooooooo. NoNNooN.P- FooN.- mooeaoo Ngooooo

ooo. ooooNo.o_ Noooooo._ opoooNo.o- oNoo.- mooseoo N_oo\mzoz

ooo. ooooNo.oo Noo_oooo. oooo_oo.N Nopo. moasooo

Agmgonsmucoo quu<\moz

 

o a o Lotto ooooooom o ouoo opoaoaoo

 

 

mpnowgo> ucmccmomo men mo mcogm mucmmvo< agoeanmpcou upzu<\moz sow: :owmmmemmm wpawupaz .eN mpno»

99

 

 

 

 

 

ooo. o
No.F_ to
_oooo.oN a
oNoPo. o opooopoz
ooooo. No
Noo. ooNoo_._P oopooo.N_ NoPNNo.oo-N<o Nooooaooov
ooo. No_ooNN.o oooNoNNo. ooooooom. Nooa. ooozo oo< oN-NF
ooo. ooNoPoo.N _o-oooooooNN. NoNNoNFN. NoNN. ooooo oooozao
ooo. NoNooo.oP opooomoo. ooo_NoN.o Nooo. ooo ooaooz
ooo. ooopoN.oN PooNoooN. oopoNoN.N NoNo. ooooo Noo ooo.o-ooo.o
ooo. NoNooN.oo NNPoNN_P. NoooNooo. - oooo.- Nooom
:mwmcom amoucmogmo
oNo. oooNoop.o NoooooN_. NooNoooN. - oooo.- mooFNEoo osoooo +ooo.op
ooo. ooppoo.oo No-oooooooN. oo-oooooompo. oooo. Poooh ooo
ooo. ooooooo.o ooNoome. NoooFNNo. - NNNN.- moooeoo
aeogoasmucou\mo<\xuom
ooo. oNoooo.oo ooNNPooo. Fopoooo.N Napo. moosaoo ooooooo
ooo. o_NNooN.o oooo_oo. ooooooN.P opop. mooo25mopooomo
empomo ou=<
o a o Lotto ocooooom o ooao oooooooo

 

 

m_nowco> pcmccmamo mg“ mo meogm wucmwc=< muucsou

ooo: ooommoaooo opooopoz .mN opoop

100

 

 

ooo. o

 

 

 

oo.o to
oNoPo.No a
mmmNm. m mpowupaz
ooooo. No

NNN. ooooopoo. N__opoo.N ooNoNoo.NN<o Noooomooov

Noo. oooooNo.N Popooooo. oo_ooooo.- Noop.- mooseoo omwooom

ooo. PoooNoN.o Po-NoomooNoo. Nooomoop. oPoN. Nooom

cmwmeom mmoucmucmm

Foo. oNooNo.NF ooooNoN. oopooooo. - NoNo.- Noseoo

oooaoosooooo oPoo<\ooz

_oo. oNooNo.oF moNoooNF. oNNooooo. - oFNo.- _o>oo Noao>oo

30 En ucmugmn—

ooo. oNooPN.NN oNooooo_. oooooooo. oooo. mooosomt_ooomo oooa

ooo. ooooooo.N _ooNNooN. oooooNo.P- oNPN.- oooeo Noo ooo.o-ooo.o

ooo. oNoooF.o_ ooooNNoN. ooooooN.F- _ooo.- apostoa

utmoz Pocoooooo

ooo. ooooo.NNN oNooNooo. NNoppo_.o oooo. mooseoo NPoNNmzoz

.o. a o Lotto ocooooom o ooao oaoaoeoo

 

 

anowLo> ucmncmamo mg» mo acogm mucmwuz< xpo»\m3mz sow: cowmmuumum wrawupzz oN upon»

101

the additive effects of these independent variables and the dependent
variable.

The variables in Table 26 are the regression elements which best
predict the value for the dependent variable, "News/Talk audience share."
Other large contributors aside from the "News/Talk formats" variable
include: "percentage food establishments," and "percentage below poverty
level." The statistics included in the analysis of variance table shown
below the regression coefficients, most specifically the R2 figure, prove
that this equation closely approximates the actual outcome of the dependent
variable value, the audience share garnered by this format. The variance
explained by this combination of independent variables is shown to be
almost 86 percent.

In Table 27, the regression coefficients computed are given for the
prediction of the dependent variable, "Black/Disco audience share."

As expected, the variable, "percentage Black," would be a key predictor
variable here. Other larger contributors to explained variance include,
"Black/Disco formats," "eating and drinking establishment retail sales,"
and "35-49 age group." Most of the independent variables listed in

Table 27 prove to add a relatively small amount of the explained variance,
however, the additive effects of this model are shown to account for
approximately 94 percent of the total variance as shown by the figures

in the analysis of variance table. What was surprising in this result was
that no income categories appeared in the step-wise procedure. Recall
that a generally lower level of income existed where this format had it's
best total audience share.

Table 28 shows the step-wise regression results with "Spanish audience
share" as the dependent variable. The variables, "percent below poverty

level" and "Spanish formats," are the best predictors of the audience share

 

 

ooo. a

 

 

102

 

Po.NF to
NoNoo.oo a
oooNo. o a_ooo_oz
oopoo. No
ooo. oo_ooooN. oNNPNo.N_ oNNoNo.oF-N<o Noooomoooo
Nop. moooo_N.N oooooNNF. oNoooNoN. ooNo. moomooo
ooo.p goo mpom wugo>wo
ooo. ooNooNP.o NooNooo_. oNoooNoN. ooNo. moPoo _ooooo
mm .3523ng pmsmcmw
oNo. NoooN_F.o oNoooNNo. ooooNoNN. Popp. ooeoz oooooooooo
ooo. oooNoN_.o NoNoNNNF. ooooNoNo. Nooa. moosooa
Ngoeooemucou\mo<\xuom
ooo. Noooooo.N oomNoooo. oooNoooN. oNNo. moooEomopooomo ooooooaoa
ooo. oooooNo.o oooNNooo. Nooooooo. - Noo_.- moooooo N_oo\ozoz
ooo. Nooo_om.o oooooooN. oNoooN_N. - ooo_.- mooooow_.oooao ooo<
ooo. PooNooo.o NNoNoooN. ooooNNoN. Noop. ooooo ooo oo-oN
ooo. o_oopo.oN o_oooooo. Npooooo.,- oooN.- ooooo oo< oo-oo
ooo. NoNooo.oo NoNoNoNN. N_ooooo.p- NNoo.- mo_om Paoooo
‘ .ooomo oooxooto o oo_ooo
ooo. oooooo.oN _oooNFoo. NooooNo.P ooNo. moascoo oomao\xoo_o
ooo. ooooo.mop Po-oNooNo__o. oNo_mNoo. oooo. Noopo oooooooooo
.o.o a o totem ooooooom o oooo apooaeoo

 

 

mPaowLo> pcmucmomo mcu mo mcocm mucmwu=< oumwo\xuo—m sup: cowmmmgmum mpawupoz .NN apnow

103

 

 

 

 

 

ooo. o
Fo.NF co
oopop.oo a
PmNoo. o opooopoz
ooooo. No
ooo. oNNooo.oo oNoooNo.o oooNNo.oo-N<o Noooomooov
Noo. NNNNoNo.o oooooooN. ooooooPm. oooo. mooosomopoooao ocooooooo
oNo. NNPFNNN.N Fo-NoFoNoopo. Fo-ooooPNNoo. oNoo. oooeo Noo +ooo.oN
ooo. ooNPooo.o oooooFN_. ooooooNo. ooN_. mo aooo oooooooeoo
ooo. oopooo.o_ oopoNoop. oooNoooo. PoNN. ooozo ooo ooo.oN-ooo.oP
oNo. Nooooo_.N NNNooooN. onoNNoN. Nooo. moosaoa Poomoonoooo._mo
ooo. PoooNoo.o ooNNooNF. oNoooooo. - oNoo.- moostoa Naooooo
ooo. ooNoo_.oN .o-NoNoFoooo. NpoooooN. oooN. moo_osoa osoooo +ooo.op
ooo. oNooN_.oo No-ooNo_oooN. oo-ooooooNoP. - NooN.- Pooop Noo
_oo. oooNoo.N_ Fo-ooNNooooo. oNPPNoo_. Poop. Nooom oo_ozoo ooooooozoo
ooo. NoFoFN.oP .o-ooNoooNNm. oNNNooop. - ooop.- Noopo oooooooooo
ooo. ooooo.Nm_ Fo-o_ooopooo. ooooNoo._ oooo. _o>oo
zucm>oa zapmm ucmucwa
ooo. oooPN.N.N .oNoNooN. ooo_oPN.o Pooo. moastoa omaooom
a, a o cooao ozoooaom o oooo opoooea>

 

 

mpnowco> pcmucmnmo ms» mo mgogm mucmwu=< smocoaw saw: cowmmmgmmm mpawupoz .NN mpnmh

104

for this format. The others listed in Table 28 in combination with
those mentioned above enable this model to account for about 97 percent
of the variance in the predicted dependent variable.

The above model, along with the model for "Black/Disco audience
share," exhibit outstanding predictive capacities. Since these two
formats are typically "minority" formats whose intent is to capture the
minority as it's core audience, it may be theorized that this sort of
format is more easily identifiable and predictable through statistical
methods than the more generalized formats included in this study. Such a
conjecture would be of great assistance to researchers of broadcast trends
and to programmers considering a change to a minority format if there were,
in fact, true.

The partial regression coefficients computed by the step-wise
procedure with the dependent variable being "Religious/Gospel audience
share," are listed in Table 29. The best contributors to overall variance
explanationare, "Religious/Gospel formats," "automobile dealer retail
sales," and "eating and drinking establishment retail sales." These along
with the remaining independent variables shown in Table 21 combine to
account for about 55 percent of the total variance in the predicted value
for the dependent variable. Once again, as with the "MOR/Adult Contempor-
ary" format, this result coincides with inferences made for this format
during t-test results. The similarities between this format and that of
"MOR/Adult Contemporary" seem to be that no specific demographic is a
characteristic variable, or is a reliable predictor by itself, for these
two formats. This possibility is somewhat confirmed by the t-test results
and the regression models developed for the prediction of the format's

audience share. In those results, it was found that few measurable items

105

 

 

ooo. a

 

 

 

mo.o co
ooNoo.o a
women. m mpawapzz
oNNoo No
Npo. oNNFooo.o ooooNNN.N oNooooo.oN<o Noooomoooo
oNo. oooooNo.N _o-ooooNFNoo. Fo-No_oooNNo.- ooNN.- Nooom
:mmmgom muopcmuemo
ooo. ooFNNoo.o mo-opooooNF. Fo-ooNoFo_N. ooop. oo_ooo too Now
ooo. NoNNoPo.o ooomNNoo. NNoooooN. NoNN. mooom Pooooo
ucmE;mw_nopmm mgouwccom
ooo. oNoNN_N.o ooNoooo_. oooPPNoo.- ooNo.- mo_oo P_oooo
.ooomo ooaxooto ooo ooaooo
_oo. oooNoN._F .o-NoooooooN. ooNNN_oN. Nooo.- mopom
Fwopmm empomo ouz<
ooo. NoNNNoo.N Fo-ooooo_Noo. oNpoooo_.- NNN_.- m._oo _ooooo
ammucozucmz ngmcmo
_o_. ooNooo_.N ooooomoN. oNoooPo.- o_oN.- moooeoo Npoo\mooz
ooo. oNFNoo.oF NNooo_oN. _oooooN._ _oNo. mooseoo PaoaooNWooooopoo
.o. a o Lotto ooaoooom o oooo opoootoo

 

 

apnomeo> ucmccmomo ms» mo mcogm mucmwu=< Pmamow\w=owmwpmm new: commmmemmm mpaoupoz .mN m—no»

106

were significant at the .05 level, or that the combination of variables
resulted in a smaller amount of total variance explained for the regression
equation. A research challenge would be to attempt to ascertain specific,
measurable variables that would be reliable predictors of audience shares
for these two formats that would also increase tht total amount of explained
variance, thus increasing the predictive accuracy of the regression equations.

Table 30 includes the statistics relevant to the regression procedure‘
with "Classical audience share" as the dependent variable. The independent
variables listed combine to account for approximately 84 percent of the
total variance. This model features a comparatively small number of
predictors that additively do a reasonably effective job in predicting the
audience share for this format.

In summary of the step-wise regression procedure for prediction of
the audience share of the nine formats, the method used provided the best
combinations of variables for the most accurate prediction possible. The
range of variance in the predicted values for the dependent variables was
from 55 percent (Religious/Gospel) to 97 percent (Spanish), and was better
than expected. No single variable was found to be a predictor for all
nine formats, but as expected, in each format case, the variable repre-
senting the number of formats was a predictor of the dependent variable.
Generally, the results achieved were not surprising in that a vastly
different set of variables was found to be predictive of each of the nine
formats.

The success of these regression equations in actual prediction of
audience shares for the respective formats within markets will be
illustrated in Chapter V. There, the comparison will be made between the

values generated by these prediction equations and the actual April-May

107

 

 

 

 

 

 

ooo. a

No.o co

Noooo.No a

NNNNo. o opoaoooz

N_ooo. No

Noo. ooPN_o.oF FoNooooo. NNoooNN.N-N<o Nooooaoooo

ooo. ooooooo.N oo-ooo_oooo. No-oNoNoop.- ooN_.- Fooo» ooo

ooo. PooNNoo.m _o-oooooo_oo. _o-oNooNoo. oooo. oooeo ooo oo-oN

Noo. oooooo..o Po-o_ooopooo. oooooooN. oNoN. moosooo Noao\mzoz

ooo. oNooNoo.N mo-NNooooooF. oo-oo.oooN. NNoP. oooooo zoo osoooo
_oo. NoooNo.NP No-oNooooooN. No-ooo_opooo. oNoN. oooo

Log muom mmowccoz

ooo. oNp_o.oo oNooooop. NPNNNNo. Nooo. Pooommo_o
-o a . o totem oeooooom o oooo a_oooao>

mrnomco> ucmccmnoo mg» mo meozm mucmwcz< FoUNmmoFu guwz commmoemmm mpawupzz .em mpno»

108

1980 audience shares for some markets. The residual amount between
the predicted and actual values will show the equation's general

ability to approximate a format's audience share for a market.

109

CHAPTER IV—-NOTES

1N.H. Nie, C.H. Hull, J.G. Jenkins, K. Steinbrenner, D.H. Bent.

Statistical Package for the Social Sciences, 2nd ed., (New York: McGraw-

 

Hill, 1975), p. 340.

2J.C. Nunnally. Psychometric Theory, (New York: McGraw-Hill,

 

1978), p. 128.

3Nie, op. cit., p. 267.

41bid.

5Ibid.

61bid, p. 268.

7J. Duncan. American Radio, Vol. V. lst ed. (Kalamazoo: Gilmore

 

Advertising, Spring 1980), p. A-27.

8Ibid., section B.

91bid., p. A-53.

101616., p. A-51.

H1616., p. A-59.

'ZNie, op. cit., p. 469.

13Nunnally, op. cit., p. 337.

CHAPTER V

SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

This chapter examines the use of the regression models developed
in previous chapters. The standard multiple regression equation is:

fY = a + b]X1 + b2X2 + .......... + prp -

where the predicted value YT is computed by adding the products of the
regression coefficients (bp) and the independent variable (Xp), together
along with the conStant (a). Values for Y. will be generated for some
markets to illustrate how these equations approximate the actual audience
shares of formats within markets. The comparison of predicted and
actual audience shares are shown for all nine formats in Tables 31 through 39.
The interpretation of the residual amount, the difference between
actual and predicted values, could vary due to a number of reasons.
Excluding measurement error, some additional, more identifiable reasons
could be: seasonal fluctuations in radio listenership reflected in the
Arbitron estimates of general program popularity; comparatively poor
management of existing facilities within the market; or, inability of the
existing facilities utilizing a common format to service the potential
audience for that format. All of these possibilities must be critically

considered individually, but the latter consideration may be the most

110

111

important should programming opportunities be sought by radio stations.

Examining Tables 31 through 39, it can be seen that opportunities
for additional "MOR/Adult Contemporary" formats may exist within Youngstown,
Lubbock and Atlantic City.

Higher negative values are likely to mean that opportunities do not
exist for a specific format. The tables show that some markets are
clearly saturated beyond the predicted potential audience whose parameters
are defined by the regression model. A good example of this situation
is found in Table 34, where programming additional "Country" formats
within Lansing, Mobile and Wichita Falls could be futile due to the
exceptionally large, negative residual between predicted and actual values
for this market. Another example is found on Table 36, where the actual
audience share for the "Black/Disco" format far exceeds the predicted value
in Augusta.

To discover the effect of additional stations as they program a
particular format within a market, one need only substitute the desired
number of formats for the actual number of formats. The result would be
a hypothetical situation within a market. For example, if one wished
to know the effect of an additional "Beautiful Music" format in the
Johnstown or Asheville market, changing the value for the variable
"Beautiful Music formats" from one to two would serve to achieve the
simulated effect. The exact effect upon the predicted value is shown
below. Notice that the mere presence of the additional "Beautiful Music"
fOrmat outlet increases the predicted audience share.

Predicted Shares for Beautiful Music
1 station (actual) 2 stations (simulated)

Johnstown 16.4 17.1
Asheville 18.7 23.4

 

112

Table 31. Example Prediction Values for "Rock/AOR/Contemporary"
Audience Share

 

 

Predicted April-May '80
Market Share (Y ) Actual Share (Y) Residual (Y'-Y)
Albany 35.6 31.1 +2.5
New York 25.6 29.1 -3.5
Atlantic City 33.5 41.8 -8.3
Grand Rapids 37.7 37.8 - .l
Duluth 33.2 30.2 +2.0
Cincinnati 32.8 31.4 +1.4
Fort Wayne 36.4 37.4 -l.0
Eugene 34.6 34.1 + .5
Knoxville 34.8 34.4 + .4
Austin 37.5 34.4 +3.3
St. Louis 35.3 33.1 +2.2
Houston 31.1 30.2 + .9
Savannah 33.6 30.9 +2.7
McAllan 40.8 38.0 +2.8
Seattle 43.2 40.2 +3.0

 

 

113

Table 32. Example Prediction Values for "Beautiful Music" Audience

 

 

 

Share
Predicted April-May '80 .
Market Share (Y') Actual Share (Y) Residual (Y - Y)
Anchorage 14.3 21.0 -6.7
Lansing 12.3 12.7 - .4
Terre Haute 13.3 16.5 -3.2
Pittsburgh 18.9 15.1 +3.8
Milwaukee 16.9 16.3 + .6
Asheville 18.7 16.3 +2.4
Shreveport 11.2 10.3 + .9
Johnstown 16.4 10.5 +5.9
Philadelphia 13.1 12.5 + .6
Boston 17.8 13.5 +4 3
Jackson 9.8 14.9 -5.1
Birmingham 16.7 10.4 +6.3
Pensacola 10.1 11.5 -l.4
Denver 15.3 18.5 -3.2
Reno 15.0 16.1 -1.1

 

 

114

Table 33. Example Prediction Values for "MOR/Adult Contemporary"
Audience Share

 

 

Predicted April-May '80
Market Share (Y') Acutal Share (Y) Residual (Y'-Y)
Lafayette 9.4 9.8 - .4
Kalamazoo 16.0 19.2 -3.2
Huntsville 4.1 8.1 -3.8
Atlantic City 28.7 22.7 +6.0
Wichita Falls 11.7 12.3 - .6
Springfield, M0 8.6 7.2 +1.4
Lubbock 12.6 6.3 +6.1
San Francisco 14.9 17.2 -2.3
Youngstown 32.1 22.5 +9.6
Boston 28.8 26.6 +2.2
Chicago 14.9 19.1 -4.2
Baltimore 21.0 28.9 -7.9
Lancaster 14.7 14.0 + .7
Sioux Falls 17.4 19.6 -2.2

 

 

115

Table 34 . Exgple Prediction Values for "Country" Audience Share

 

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)

Lansing 9.8 16.7 -6.7
Fayetteville 21.3 22.0 - .7
Augusta 16.9 12.5 +4.4
Bloomington, IL 17.8 16.5 +1-3
St. Louis 10.1 10.1 --

Houston 15.7 16.2 - .5
Dallas 21.4 25.2 -3.8
Jackson, MS 10.9 7.9 +3.0
Savannah 13.7 16.8 -3.1
McAllan 6.4 6.4 --

San Antonio 18.5 19.4 - .9
Pueblo 27.2 24.3 +2.9
Mobile 17.1 22.8 -5.7
Wichita Falls 34.8 43.1 -8.3
Fort Wayne 17.6 12.6 +5.0

 

 

Table 35. Example

116

Prediction Values for ”News/Talk Audience" Shares

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)
Philadelphia 25.1 21.4 +3.7
Miami 10.9 9.9 +1.0
Grand Rapids 4.5 2.1 +2.4
Dallas 8.6 11.8 -3.2
Springfield, M0 2.7 2.5 + .2
New York 17.8 20.1 -2.3
St. Louis 17.5 25.5 -8.0
Youngstown 8.5 15.8 -7.3
Baltimore 1.6 2.4 - .8
San Antonio 4.2 6.0 -l.8
Bakersfield 2.0 1.4 + .6
Denver 9.7 11.5 -l.8
Milwaukee 5.7 1.9 +3.8
Seattle 12.2 10.9 +1.3
Springfield, MA 8.5 3.4 +5.1

 

 

117

Table 36. Example Prediction Values for "Black/Disco" Audience Share

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)
Jackson, MS 29.0 31.4 -2.4
Birmingham 23.3 23.9 - .6
New York 16.8 18.0 -l.2
Cincinnati 8.2 7.1 +1.1
Tallahassee 19.0 23.4 -4.4
Augusta 9.0 20.3 -1l.1
St. Louis 11.5 10.5 +1.0
Philadelphia 11.8 12.9 -l.l
Miami 6.6 7.4 - .8
Mobile 19.5 16.2 +3.3
Atlantic City 6.8 1.4 +5.4
Houston 16.3 17.5 -l.2
Asheville .6 O * + .6
Shreveport 22.3 16.3 +6.0
Lafayette 2.0 0 * +2.0
Chicago 17.8 14.0 +3.8

 

 

* No "Black/Disco" formats within the market.

118

Table 37. Example Prediction Values for "Spanish" Audience Shares

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)

New York 5.7 4.2 +1.5
San Antonio 11.8 17.1 -5.3
Pueblo 4,2 2,7 +1.5
Austin .4 4.7 -4.3
Houston 7.3 5.8 +1.5
Chicago 2.2 2.2 --

Miami 30.8 31.1 - .3
McAllan 46.2 47.9 -1.7
Lubbock 5.8 6.8 -l.O
Bakersfield 7.1 6.2 + .9

 

 

Table 38. Example Prediction Values for "Religious/Goagel" Audience Share

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)

Syracuse 3.4 2.6 + .8
Milwaukee 1.2 .7 + .5
Springfield, M0 3.4 3.4 --

Duluth .9 1.4 - .5
Birmingham 4.7 11.0 -6.3
Lancaster 5.8 9.1 -3.3
Asheville 3.0 4.9 -l.9
Anchorage 3.8 4.2 - .4
Seattle 2.4 3.7 -1.3
Pittsburgh 2.1 1.3 + .8

 

 

119

Table 39. Example Prediction Values for "Classical" Audience Shares

 

 

Predicted April-May '80
Market Share (Y') Actual Share (Y) Residual (Y'-Y)

Denver 2.6 3.9 -l.3
San Francisco 5.4 3.4 +2.0
Seattle 3.6 3.3 + .3
New York 3.7 2.5 +112
Houston 2.6 2.0 - .6
Milwaukee 2.3 1.1 - 8
Dallas 2.5 1.7 + 8
St. Louis 1.2 0 * +1 2
Chicago 7 2.7 --

Philadelphia 3.2 2.6 + .6
Baltimore .6 0 * + 6
Boston 1.7 1.0 + .7

 

 

* No "Classical" formats within the market.

120

Of course, the ability of any radio facility to acheive such "potential"
within a market is only as good as it's management's success in executing
the policy necessary to maximize the performance of the employees. For
this reason, deductions made from these calculations should be made with
some caution. More times than not, the management can make the difference
in any market situation. A good manager can make the best out of any bad
situation, and produce optimum performance regardless of the market, the
facility, or the format employed. While a bad manager is likely not to
have success even in the best of situations.

The regression models developed here should aid the reader in relating
the potential of formats comparatively with other markets. Differences
between these caluclations and real situations may be due to uncontrollable
or unexplained factors that vary the audience share for a format. It is
held with confidence, however, that these regression models would be the

most accurate predictors of audience share over time.

Conclusions and Recommendations

 

This paper attempted to derive prediction and classification models
from data that was representative of some demographic characteristics of
58 radio markets. The original data was refined somewhat, then subjected to
two major treatments in order that the results would provide programmers
with helpful information about the similarity of markets, the prediction of
format success within markets, and the variables which constitute predictors
of a format's audience share.

Early in Chapter'IV t-tests were performed to discover common charac-
teristics held by markets that do extraordinarily/well with certain formats.

The results showed that the high shares for a format and some demographic

121

characteristics were held in common. Some formats had more of these

unique characteristics than others, but it leaves a void of information

as to how these characteristics are related to the success of audience

shares or even the number of outlets for a particular format within a

market. It was suspected that some of these characteristics would again
appear in either the regression or discriminant analysis models as predictors
for classification of audience share, but this did not occur. Therefore, it
may be valuable to discover whether these characteristics are the key to

the high degree of success for the format. Is this highest tier of audience
share markets somehow dependent on a certain combination of these variables?

Discriminant function analysis was later utilized to classify
markets as being "most similar" to the best markets or the poorest markets
for a format. The accuracy of the classifications themselves is high, and
the use of this treatment enables a researcher to discover the similarity
of markets. A programmer may wish to find out if his market is most similar
with the group of markets that excells in audience share for a particular
format. This way, an inferrence of "favorableness" can be derived. However,
applications of discriminant analysis to actuality may be difficult, and the
utility of it questionable. The analysis of markets using discriminant
functions borders on subjectivity, except for the fact the discriminant
score which represents a market was derived using a reputable statistical
method.

Obviously, the regression models show superiority in both application
and accuracy. Nevertheless, the same data which was used to derive the
regression models was also used to calculate the discriminant scores.

It would be not only interesting, but informative to programmers to discover

whether discriminant scores can be equated with audience share of a format

122

within a market. Since these scores are placed along a continuum, does
this not mean that a range of format success can be derived with each
market somewhere along that continuum? Can a discriminant score imply
the degree of success for a format within a market?

The regression models computed and shown in Chapter V are easily the
best method by which programmers may estimate the success of a particular
format within a market. The models in Tables 31 through 39 are capable
of prediction of the approximate audience share of a format for any of the
nine formats examined in this study. The multiple regression prediction
method is a relatively uncomplicated yet more accurate, process in
comparison to application of other statistical methods such as discriminant
analysis and is genuinely useable by anyone with only basic knowledge of
statistics at hand.

The examples shown in Chapter V indicate that regression models are
generally successful in audience share prediction. Further research utilizing
multiple regression should be directed towards the discovery of the regression
models that would predict the optimum number of format outlets within a
market. For example, what would be the optimum number of radio stations
that would be programming the "Country" format in Cincinnati? When are
there too many "Beautiful Music" stations in San Francisco? Such a
research effort in combination with the information generated in this thesis
could have the potential to be of extreme assistance to broadcasters
nationwide.

The general objective of this study was achieved. That is, to
discover market characteristics beyond age or sex that are predictive of
a format's audience share. Generally, with regards to such predictive

variables, there was no single variable found to be a predictor for all

123

nine formats. Each format required differing combinations of variables
for accurate prediction or classification within markets. This last
fact was expected at the outset, and proven during the course of the

study.

The Passage of Time

 

How the accuracy of the statistics generated here will apply to the
evolutionary changes that are bound to happen to formats and markets
is unknown. Concurrently, populations will shift as success of formats
within markets fluctuate. Almost assuredly, replication of this
prediction idea is needed to compensate for the demographic shifts
within markets. Such demographic factors were proven to be predictors
of format audience share, and thus the monitoring of these variables is
imperative should application of this study be reliable and consistent
with current population demography. A final consideration here would
be the changes in what constitutes the formats themselves. The model
predictive of "Rock" today may not predict accurately years later simply
because of the change in what l'Rock" has become. Consider that the "Rock"
format today is considerably different than the "Rock" format of fifteen
or even ten years ago.

However the circumstances vary, the knowledge of a format's probable
success within a market will remain a constant goal to programmers of

radio stations at any time in the future.

APPENDIX A

NNNNNNNNNN—l—I—J-H—l-d—J-d—J—J
omme'I-bWN—‘OSDWNO‘UW-hWN-HONO

mummth—a

APPENDIX A

124

Markets Included in the Computation

Lafayette, Louisiana
Shreveport, Louisiana
Great Falls, Montana
Billings, Montana

Lansing, Michigan

Grand Rapids, Michigan
Kalamazoo, Michigan
Augusta, Georgia

Savannah, Georgia
Huntsville, Alabama
Mobile, Alabama
Birmingham, Alabama
Dallas-Fort Worth, Texas
San Antonio, Texas

Wichita Falls, Texas
McAllan-Brownsville, Texas
Houston, Texas

Austin, Texas

Lubbock, Texas
Fayetteville, North Carolina
Sheville, North Carolina
Terre Haute, Indiana

Fort Wayne, Indiana
Tampa-St. Petersburg, Florida
Miama, Florida
Tallahassee, Florida
Pensacola, Florida
Milwaukee, Wisconsin
Jackson, Mississippi

30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.

Chicago, Illinois
Bloomington, Illinois
Philadelphia, Pennsylvania
Pittsburgh, Pennsylvania
Johnstown, Pennsylvania
Lancaster, Pennsylvania
Sacremento, California

San Francisco, California
Bakersfield, California
Baltimore, Maryland
Atlantic City, New Jersey
Albany-Schenectady-Troy, N.Y.
New York, New York
Syracuse, New York

Duluth, Minnesota

Pueblo, Colorado

Denver, Colorado
Cincinnati, Ohio
Youngstown, Ohio
Eugene-Springfield, Oregon
Knoxville, Tennessee
Springfield, Massachusetts
Boston, Basschusettts

St. Louis, Missouri
Springfield, Missouri
Anchorage, Alaska

Seattle, Washington

Reno, Nevada

Sioux Falls, South Dakota

APPENDIX B

125

APPENDIX B

Format's Total Share of Audience by Market*

 

 

Rock/AORZContemporary (National Average: 33.04%) Share Percent
Lafayette, Lousiana 58.8%
Anchorage, Alaska 59.8%
Great Falls, Montana 58.3%
Lansing, Michigan 54.6%
Billings, Montana 54.3%
Fayetteville, North Carolina 53.3%
Terre Haute, Indiana 52.6%
Kalamazoo, Michigan 52.3%
Augusta, Georgia 51.3%
Sioux Falls, South Dakota 51.0%
Huntsville, Alabama 49.5%

Beautiful Music (National Average: 15.58%)

 

Tampa-St. Petersburg, Florida 32.6%
Miami, Florida . 26.9%
Albany,-Schenectady-Troy, New York 25.7%
Lancaster, Pennsylvania 25.0%
New York, New York 24.5%
Atlantic City, New Jersey 24.5%
Grand Rapids, Michigan 23.9%

MOR/Adult Contemporary (National Average: 16.11%)

 

Bloomington, Illinois 41.2%
Pittsburgh, Pennsylvania 39.5%
Syracuse, New York 36.9%
Duluth, Minnesota 36.8%
Cincinnati, Ohio 33.8%
Milwaukee, Wisconsin 33.2%
Fort Wayne, Indiana 31.9%

Eugene-Springfield, Oregon 30.4%

126

 

Country (National Average: 10.26%) Share Percent
Asheville, North Carolina 47.3%
Wichita Falls, Texas 43.1%
Springfield, Missouri 40.9%
Knoxville, Tennessee 36.4%
Shreveport, Louisiana 34.7%
Austin, Texas 31.8%
Huntsville, Alabama 31.8%
Johnstown, Pennsylvania 30.9%
Lubbock, Texas 29.3%

News/Talk (National Average: 9.68%)

St. Louis, Missouri 25.5%
Philadelphia, Pennsylvania 21.4%
New York, New York 20.1%
San Francisco, California 16.3%
Youngstown, Ohio 15.8%
Boston, Massachusetts 12.6%
Chicago, Illinois 12.4%
Houston, Texas 12.0%
Dallas-Fort Worth, Texas 11.8%

Black/Disco (National Average: 9.76%)

 

Jackson, Mississippi 31.4%
Birmingham, Alabama 23.9%
Savannah, Georgia 23.9%
Tallahassee, Florida 23.4%
Augusta, Georgia 20.3%
Pensacola, Florida 18.8%

Spanish (National Average: 2.02%)

McAllan-Brownsville, Texas 47.9%
Miami, Florida 31.1%
San Antonio, Texas 17.1%

Lubbock, Texas 6.8%

127

 

Spanish (continued) Share Percent
Bakersfield, California 6.2%
Houston, Texas 5.8%
Austin, Texas 4.7%

Religious/Gospel (National Average: 1.24%)

Birmingham, Alabama 11.0%
Savannah, Georgia 9.8%
Lancaster, Pennsylvania 9.1%
Shreveport, Louisiana 7.7%
Pueblo, Colorado 6.1%
Asheville, North Carolina 4.9%

4.2%

Anchorage, Arkansas

Classical (National Average: 1.27%)
Denver, Colorado
San Francisco, Dalifornia
Seattle, Washington
Reno, Nevada
Milwaukee, Wisconsin
Chicago, Illinois
Philadelphia, Pennsylvania

NNNwwwww
N
R

New York, New York

 

*As reported in the Spring 1980 American Radio

 

APPENDIX C

oowmmbwm—t

“NNNNNNNNNN—‘d—J—J—J—I—I—d—l—J
oooowmmbwmfoooowmmth-doxo

128

APPENDIX C

Variable List

Total population of market

Black percentage of population

Foreign stock percentage of population

Median age of market population

18-24 persons in market

25-34 persons in market

35-49 persons in market

50+ persons in market

Total households in market

Effective buying income per capita

Effective buying income household median

8,000-9,999 effective buying income group
l0,000-l4,999 effective buying income group
15,000-24,999 effective buying income group

25,000+ effective buying income group

Percentage of population 65 years or older

Percentage of population living in urban areas

Public school enrollment

Total income per capita

Percentage of population below the poverty level
Percentage of population with 15,000+ family income
Total bank deposits for the market

Percentage of population living in one unit structures
Total occupied housing units in the market

Percentage living in owner occupied homes

Total manufacturing establishments within the market
Percentage of manufacturing establishments with 100 or more employees
New capitol expenditures

Total retail trade establishments

Food establishments as percentage of total establishments

31.
32.
33.
34.
35.
36.

37.

38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.

129

Auto dealers as percentage of total establishments
General merchandise places as percentage of total
Eating and drinking places as a percentage of total
Gasoline service stations as percentage of total
Furniture dealers as percentage of total establishments

Percentage from total receipts for hotels, motels, trailer
parks and camps

Percentage from total receipts for amusement and recreation
facilities including motion pictures'

Total wholesale establishments

Total wholesale sales

Total monthly social security payments

Marriage rate per 1,000 persons

Divorce rate per 1,000 persons

Birth rate per 1,000 persons

Death rate per 1,000 persons

Serious crime rate per 100,000 persons

Percentage of women in the population

Normalized 18-24 age group

Normalized 24-34 age group

Normalized 35-49 age group

Normalized 50+ age group

Normalized 8,000-9,999 effective buying income group
Normalized 10,000-14,999 effective buying income group
Normalized 15,000-24,999 effective buying income group
Normalized 25,000+ effective buying income group
Number of Rock/AOR/Contemporary formats within the market
Number of Beautiful Music formats within the market
Number of MOR formats within the market

Number of Country formats within the market

Number of News/Talk formats within the market

Number of Black/Disco formats within the market

Number of Spanish formats within the market

Number of Religious/Gospel formats within the market
Number of Classical formats with the market

64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.

Normalized
Normalized
Normalized
Normalized
Normalized
Normalized
Normalized
Normalized
Normalized

number
number
number
number
number
number
number
number
number

of
of
of
of
of
of
of
of
of

130

"Rock/AOR/Contemporary" formats in the market
"Beautiful Music" formats in the market
"MOR/Adult Contemporary" formats in the market
"Country" formats in the market

"News/Talk" formats in the market
"Black/Disco" formats in the market

"Spanish" formats in the market
"Religious/Gospel" formats in the market
“Classical" formats in the market

Market total retail sales

Total food

sales

Total eating and drinking place sales
Total general merchandise sales

Total furniture outlet sales

Total automobile dealer sales

sales

food sales

eating and drinking place sales
general merchandise sales
furniture outlet sales
automobile dealer sales

drug store sales

buying power index

Total drug
Normalized
Normal ized
Normalized
Normalized
Normalized
Normal i zed

store
total
total
total
total
total
total

Market graduated
Market total effective buying income
Market national advertising dollars

APPENDIX 0

FINAL CORRELATION MATRIX

131

N—mo.u
covm.u
nape.
«FPO.
Nmmo.u

vNON.

Femo.u
omON.n

NmNP.n

momo.n
ommo.

aaogo
oo< +oo

oNFN.

«mFF.

nmqp.u

m—cp.u

Paco. _

mFMN.u

mem.a
Fpm¢.

mmom.n

Fame.
wnPN.

N¢o~.-

oooooz
o_ooomooo Noo

ome.
muop.u
ammo.
mvmp.
mono.

noo_.u

nmmo.a
mmPN.

ammo.u

mNpN.
memo.

mono.u
mopm.u

oooooo
Loo Noo

momN.u
mmvm.u
em—o.u
ommN.u
mvmo.n

vam.u

mmom.u
«mmp.

ammo.s

mepN.
mpmw.

mmpm.
emmp.

xuopm
zoomeom N

o¢m¢.l

Pmon.u

expo.

omno.

Nomo.

m—mo.

mFmN.u

ovoo.

oomm.:

momN.
comp.

comm.

m0¢o.n

Nmmo.
momm.

mm<
cowumz

mpop.
mmoo.
NNNF.
ompo.u
umo_.u

mmnp.u

mnop.
mmop.u
wNmm.

mmmp.u
ammo.

nano.n
NMNm.u
momN.

nomm.u

mmmo.u

xuo_m
mmoucwuemo

ooogo ma<
vmumN cmNPposgoz

macaw wm<
«NTNN uwNwFoELoz

mcomgmo ooo.~
Loo oooo eoooo

mcomemo ooo._
two mama mugo>wo

mcomgmo ooo.~
Loo mpom mmowegoz

mmso: vmwoauuo
gonzo cw mmoucmuemo

mmeouuoeum “we:
moo cw amoucmugwo

ooogw wsoucH
+ooo.mF mmoucwucmo

Fm>m4 xuem>oo
zapmm mmoucmugmo

ouwooo Loo msoucH

cowuo_:ooo
cone: mmoucmuemo

macaw mm< +om
cowumz u_o;mm:o: Hmu
oooooo ooo Noo

gooum
cmwmcom mmopcmugmo

mm< cowcmz
xoo_m mmoucmugmo

132

mme.
ammm.
omeo.

mnem.
Fouo.n

pomp.u
PomN.a
«app.
Nmmm.n
PmNo.
meo.
«mm—.1
mmcm.

ommo.n

oooeo
oo< +om

NomN.

Pomp.

NNNN.

MNmo.u
mpmo.

NpmN.n

opmm.u

wepe.u

memo.

wmpN.

NNNm.u

menw.n

mmmo.:

NFmP.

oooooz
opooomooo Noo

mpoo.
Nmmo.u
ONmo.

mpoo.
mNNF.

mmpp.u
mmNp.
oNPF.
NmuN.
meo.u
FmNN.u
mm¢_.n
mmmo.:

vmvm.

aooooo
too Now

eva.
mmmm.
m—Vm.

mump.
ommp.u

mmpm.n
omo¢.i
mmpo.
wqpo.
come.
ammo.
Nmeo.
mmmm.
NNmP.

xuopm
cmwmgou N

omwe.
«Now.
nmNN.

ompm.
me~.n

mmmm.u
emFH.u
mmmo.
P¢m_.n
PmoF.
oomo.u
wmmm.u
Fmem.

MNmN.

mm<
cowuwz

movN.- manage; .osmucou
“Pou<\moz oo L3:52

oPNo.- muosgoo coma:
Powwoooom mo gmasoz

.mpp.- manage; .oEmucou
\No<\xuom No twosoz
mmoo. :meoz mmopcmugmo

_¢mo.- .paopmm empomo
mgouocgou 4o twosoz

mmNP. .pooumm empomo
mow eo gonzoz

moNF.- .Fnoumm empomo
opo< we eonsoz

NNFo. mucmssmm_noumm
too; we eonsoz

pooo.- oooeo ooo
+ooo.mN umNPPoELoz

mmom.- ozoeo Hmu mam.eN
loco.m~ umNoposLoz

oNNo. oooeo Noo ooo.o_
looo.op umNNFmELoz

ooN_. oooeo ooo ooo.o
-ooo.m umNoFoEeoz

NNNF.- noose wm<
+om umNo—oeeoz

mmec.n ooogo mm<
molmm umNoposLoz

xuo_m
mmopcmogmm

133

NONF. NmPN. omoo. PFmp. NNoN. mNPN.u .Fnoumm _ouop
mo N mo .pnoumm mags

N_N_.- vmmr.- meop. mmmm.- moPN.- Nomo.u .Faoumm _ouoh
mo N mo .Fnoumm opz<

ommo. PmmP. mmpN.u Noeo. quo.- NmNo.- .Faopmm pouch No N.
mo .pnopmm meopwceoo

mmoo. memp.- mmoo.- mmmo.u oomo. mNmN. .pamumm

_ooo» oo N no Fooomo
mmwccosusmz Fogmcmo

NFNF. Foqfi. memo. coop. memo. opvp. ._aoumm Pooch No N mo
.Paopmu games new you

mqoo. memo.- ommo.- NPNN. mace. ommp. .Fnoumm Pouch No
N mo .Fnoumm uoom
empp. oNuN. PNNF. anew. FNmN. Nmmp. Pouch Ham

N—mo. PPNN. NFmP. Name. NNON. move. apogee;
Powummmopu mo gonzo:

. .- . .u . . muosgou Nmomow
ammo NomN mmm_ emNN camp NNNN \moowmvpmm No twosoz

eNmo. mono. enno.- make. Pmpo. Nmoo. mousse;
gmwcoom No gmasoz

Nmmp. omoo. FmNN. mwmp. oooN. mFVm. manage;
oomwo\xuopm mo Lonsoz

NNNN. NNoN. mNmo.- mmNo. momN. mNNo.- muoELou
NFoN\mzmz mo sensoz

mmN_.- momN.- ¢_NN. comm. mom~.- ammo. muoeeom

Newcoou mo amass:

ooogw cowuwz opwoou xuoum mm< xuopm
mm< +om upogmmoox Hmm to; Hmm cmwmeoo N cowum: mmoucmugmo

134

who—.1

mmmp.

mump.

mewp.

memo.u

mmmn.

mmeowuaeum
awe: moo cw N

nmNm.

mpvo.a

mnmN.u
veoo.
oomo.

c¢m¢.n

momm.a

cacao msoucH
+ooo.mp N

NQON.-

mono.

empo.l
memo.u
Ncmo.o

vap.

come.

momm.u

Pm>m4
Nugw>oo zopmm N

mev.

coop.u

mmmo.:
ommp.
upmp.

mmom.l

omwm.u
oven.

oomn.u

oooooo
Loo meoucm

«mmp.

epnp.n

Nmuo.u

ano.

Povo.

Nemm.u

ov¢¢.u

wmmm.

mvmp.u

ommm.

co_uopoooo

cone: N

oaogo mm<
omumN umNNPoEcoz

oooeo mm<
NN-N. umNNPoeLoz

moomemo ooo.,
goo muom gamma

mcomemo ooo._
goo muom wugo>wo

mcomgmo ooo.F
two mama mmowcgmz

mmsoz nm_o:uuo
cmczo cw mmoucmugmo

mmcaposeum “we:
moo :N mmopomucma

ooogu msoucH
+ooo.op oooooooeoo

Nm>m4 Nuem>oo
zo_mm mmoucoucmo

opwooo goo mEoucH

cowuo_oooa
cone: mmoucmuemo

Quota mm< +om

oooooz opoeoaooz Noo

muvnmu Lug Hmm

xuoum
cmwmgom mmoucmuemo

mm< cowumz
xuopm mmoucmucmo

135

Ncom.u

omno.-

mmmv.u

NNo¢.1
ommo.

omvn.

omen.

mep.

opmm.l

mwm—.u

Nm—N.

voom.

oomp.n

mva.u

mucouuoepm
awe: mco cw N

momm.
uuvm.
Nva.

Nooo.
“Pep.

mnw¢.u

eva.u

mmem.n

o—Nm.

qup.
oNFm.n

NNOm.u

mNm_.l

Fuov.

ooocu msoucH
+ooo.mp N

«mmm.n
mmwN.u
ONmm.u

eNPo.u
mmN.Fn

NFom.

mNmN.

m¢~m.

momm.n

mN0m.u
Novm.
N—vo.
NONN.1

qum.u

~m>mo
Nugm>oo 3ome N

Nmom.
aN—v.
mmvv.

mmvo.u
NmpN.

mmom.u

NNm¢.n

NNNm.n

¢omm.

ommN.
NNN¢.1

NFNo.u

mmmo.

«vac.

oooooo
Lao msoucfi

NmNP.
empv.
Nope.

ovpo.
mnpo.n

mmem.n
mmpm.n

oNPF.u

mmNN.

Femo.l
«mop.n
me~.u
Pomo.

MNmN.

cowuopoooo
cone: N

muoEcoo .oEmucou
“Fou<\moz No Lmoosz

muoELoo uwmoz
Pomwuoomm No cmnsoz

maoseoo .oEmucou
\No<\xoom mo Lmasoz
cmeoz mmmucmugma

._ooomo eo_ooo
meouwceou mo gmnEoz

.Faoumu empomo
mow No gmnsaz

.Fnoomu empomo
ooo< No cmoEoz

moooEomo_ooomo
noon No emosoz

aaogu Hmm
+ooo.mN umNNFmEEoz

azoxu Hmm mom.vN
nooo.mF umNNPmEgoz

azogu Hmm mam.vp
nooo.op umNN—mexoz

ooocw Hmm mmm.m
-ooo.m anNPoEeoz
oooew mm<
+om nmNNFoeLoz

ooocu mm<
molmm umNN_oELoz

136

mcmp.n

mfimv.

owes.

—o¢N.

onN.u

mocp.u

comm.a
ommm.u

NNmF.

Romp.u

wmmm.u

wmm¢.u

cumm.

mmcopusgpm
uwcs mco cw N

mnmN.
Fpmo.u

mFFo.

mmNN.u
NpmN.
mmmo.n

F—mc.
mmme.

mnoo.n
¢N~o.u
opmm.
momv.

NONF.1

macaw esoucH
+ooo.m_ N

NoeN.u
mmvo.u

mFeo.u

mNmm.
pmmp.u
mFeN.

mNmN.u
mmNN.u

nevo.
ONom.
Ncco.u
mNON.u

comp.

pm>mN
apem>om zepmm N

momN.
nmmp.t

wo—o.

«meN.u
mMNN.
mNm~.u

mee.
mmNm.

omoo.u
Nmoo.
Popm.
ocmm.

wooN.u

oooooo
Loo meoucH

Nome.
omPF.n

ommN.

mmoo.
comp.
vomo.

ovwm.
mNnm.

mwmo.n
mmvv.
mmmq.
mmpo.

anp.

comuopoooo
cone: N

.Nnoumm pouch
No N mo .pnoumm moan

.Naoumm pouch
No N no .—ooamw ouo<

.Nooomm Noooh No N
mo .Pooumm mcouwceoo

.Nooomo
_oooN No N mo _ooomo
mmwccogugmz Fogmcmo

.anumm —mu0h N0 N mm

.pnmumm xcwgo can you

.Nooomm Pouch No
N No .Nnoumm uoou
Pouch Ham
muoecom
NoNuNmmoFu No LmoEoz
mooseou _momow
\moonN—mm No emasoz
mNoELoN
gmwcoom No twosoz
monsoon
oomwo\xuopm No Logan:
muoecoo
NNoN\mzmz No twosoz

muosgom
Newcoou No gmne:z

137

mpmm. Nmo¢.-
omom.u
oaoeo mm< ooogw mm< .memo ooo._

emnmN .Egoz eNan .Egoz \mpwm gamma

va—. Pwvo.u

mwno. mNmo.u

omFN. mpno.
FNPN.

mcomemm ooo.— mcomgmo ooo.~
\muom mugo>wo \muom mNNNLLoz

wham.n azogo mm<
cmumm umNNFmELoz

oooo.- oooeo ooo
anmp chN—mELoz

«FMN. mcomgma ooo.~
goo muom gamma

wmmo. mcomem oooop
Lao mama mogo>mo

Fwno.l mcomLma oooop

two mama mmowggoz
mmeoz cowoouuo

gonzo cw mmoucmuema
mmeouuocum «we:

mco cm mmoucmugma

ooogw mEou:H
+ooo.m_ mmopcmugmo

Nm>mN Noem>oo
zopmm omoNcmuema

ouwoou Loo msoucm

:o_uo_oooo
cone: mmopcmuemo

ooocu mm< +om
cowuw: upozmmooz Hmm
oooooo too Now

xuoum
cmwmeou mmoucmuemo

mm< combo:
NUNFN mmoucmugmo

Nose: umwoouuo
gonzo cw N

138

mNMN.u
Pomo.u
opep.

mmee.n
mmmN.

oeo_.u
mwmo.
PmNN.u
mmme.

FmNo.

mmoN.u
mNNP.u
mpom.u
mONm.

ozone mm<
emumN .Eeoz

opmm.u meNp. mmmN.u mNNF.u
mmom.u ammo. mmmN.u mnmo.n
mmup.u mmm.u Nmmp.u PNPF.1
Nmom.n oeNm. memp.u mmop.u
oPem. nmeN.n memo. meom.u
umem. nmmo. eomfi. omep.
NmNN. ONmo. memm. mmpp.
nom_.u mmmp. mmmN.u mmmN.u
Foeo. mom_.u Pmmo. mmmo.
ommo.u meeo.u meo.u omeo.u
emmp. mmo0.1 OONP. momo.u
meNN. mmmo.u Nmpo. NF—o.n
N—em.u meem. ammo. mmmo.
mmmm.n eomo. omeo. Femo.

ooogm mm< .mgmo ooo._ mcomamo ooo.F mcomgmo ooo.~
eNump .Eeoz \muom momma \muom moeo>wo \mpom mmomggoz

mNoN. mposgou .asmacou
Npoe<\moz No Longs:

FNem.- apogee; uNmoz
Nomwuoomm No twosoz

mmmm.- muoELou .oemucou
\mo<\xuom No Lmoosz

mmmp. cmeoz mmoucwuemm

some. .pnmpmm memmo
mgzpwcgam we Lwnssz

mvpm. .anumm gmpmmo
mmo $0 LmnE:z

omee. .anumm gmpmwo

ou:< mo Nongoz

oooo. moeoENNNNooomo
too; No LmnEoz

omoo.- oooeo Noo
+ooo.oN ooNNNoeeoz

mmmp. azogm Hmm mom.eN
loco.m~ mmNNFNELoz

oNoo. oooeo Noo ooo.o_
-ooo.oN ooNN_oeLoz

ooNo.- oooeo Noo ooo.o
-ooo.o ooNNNNENoz

oeNF. oooem mm<
+om emepoeeoz
mmON.- oooeu mm<

meumm mwNNPmELoz

mesa: umwoouuo
gmczo :N N

139

mmep. mNHN.u Nmmp. mmoo.u
ommp. memp. mnmo.u mmom.
NNmo. mmON. mmmo.- oopp.
Nm~—.u enmo.u NFFF. mmo—.n
mmmo. mnmp. emNo.n mmmp.u
momp.u mmnp.a mmoo.u mmn~.u
memp. mmmN.u mmeo. mmmp.u
meow. mmm~.u Fmpo. mpmo.
eomp. meON.u omoo.u mmNo.
memo.n mmep.u mmoo. mmNo.
mmmp. oeom.t mmmp. mumo.n
emmo. opom.u mm—o. Nem—.u
mmmo. mmmo. mNmo. mmNN.

ooogm mm< oooew one .mewo ooo.F mcomemo ooo.~
emumN .Egoz eNst .Egoz \mpom spoon \muom mueo>wo

mm~—.u
oomo.

mmmo.n

ee_~.u
anp.n
mNmo.a

mnmo.u
mmmo.

emnp.n
mmeo.n
mmpp.u
o~e~.-

mpeo.u

mcomgmo ooo.~
\muom mooweeoz

ooo_.- .Naoumu _ouoe
No N mo .Nnopmm mono

_ooo.- ._ooomo Pooch
to N no .Nooomo oooo

NeMNo.u .anumm Pouch mo N
mo .Nooumm mgouwcmou

.Nnoumm
Nemp. Pouch No N mo powwow
mmwuoomung Noemcmu

FNmN.uNoonm Pouch mo N mo
.Pamumm Newgo new pom

memp.u .anNmN Pouch mo
N mo .Nnoumm cool
mmom.u _ouoh Hmm

nmmm.| mumEgOm
NoNuNNNNFU No LmoEoz

Nomo.- muoELoN Nmomom
\moowmwpmm No gonzoz

Nmem.u muoEeoN
gmwcoom No gonzo:

mmme.- muoEcoo
oomwo\xuopm mo gmnsoz

mmoe.u muoELoN
xNoN\mzmz No monsoz

oomN. mNoELoN

Neucoou No emoEoz

mmso: emwoouuo
mezzo :N N

140

mppe.n
mmpo.u
PFmN.u

mpeN.u
Pump.u

meNN.
mmmN.

emop.
mpmm.u

ommo.u

geese Hmm mmm.e_
loco.o_ emNNFeELez

NNme.n
FeNN.u
NmmN.u

Nmmppu
ommp.a

mmmN.
NpmN.
mFNm.
eemm.n
PemN.u

Pmmm.

oooeo ooo ooo.o
-ooo.o ooNNNoeeoz

meee.
mNmm.
Nmm—.

mmme.
Nmm~.u

NFNN.1
NomN.u
mmpp.
opmN.u
NNN_.
Fmoo.

FFeN.u

eeegm em<
+om eeNNNescez

mNmN.
omNF.
mmmm.

Nmmo.
ome_.-

oNpe.-
NmNp.l
mmmo.u
mem.

mumm.

mNNm.u
NmeN.l

mmmo.

eeeem em<
melmm emNNNeELez

muoeceu .oEepcem
upoe<\moz Ne Lmesoz

mNeELeN e_moz
Noyweoeem we Leesoz

mooseeo .oeeucem
\No<\xeem Ne Lmesoz
cmeez moouceegeo

.Neeumm Le_eeo
eeouwceou Ne tensoz

.Neeumm Lepemo
Now we ceeEoz

.Neeumm empemo
epo< me censoz

moooENNNNooomo
eeeN Ne eeoEoz

oooeo Noo
+ooo.oN ooNNNoeeoz

aaoem Hmm mmm.eN
uooo.mp emNNPeELoz

azogm Hmm mmm.e—
loco.op emNN—eEgoz

ooeem Hm“ mom.m
-ooo.m emNNPoEeez
noeem mm<
+om emNNFoELez

enema em<
melmm emNNNeELez

141

ONPN.:

mNNw.

memo.

mwoc.
wwwN.u
wmep.

emmN.n
mmwm.l

momp.

mNoo.u
FNNP.1
omwN.n

mpop.u

oooeo ooo ooo.o_
-ooo.o_ ooNNNNENoz

mmNN.u

oopo.u

mme_.:

omNo.
PNON.1
mNmP.

opmN.n
NaHN.u

mmmw.u
omep.

mMNN.n
mpm_.u

eNHo.

oooto ooo ooo.o
-ooo.o ooNNNNELoz

ewew.

mmNN.n

NmNo.u

Nmoo.u
mNoF.
mmmo.

._ emN—.
mopp.

NNmo.
emmo.
mmmw.
ewNN.
Nmm_.u

eoeem one
+om eenwwesgez

momN.

omoo.n

mmmN.u

wmmo.
mmm—.u
mNNo.

mmNm.
mem.

FmNp.
meNo.
mwNm.
eomm.
Nmno.l

eoewm em<
meumm eerweELez

.weeumm weuew
we N we .weeumm mega

.weoemm Peeew
we N we .weoomm euoe

.—neumm Peach we N
we .wneumm meeuwccsu

.Neeamm
woeee we N we pneumm
mmwecomeeez Newecmm
.weoumm weuew we N we
.weepmm Neweo eco pom
.weeumm Nope» we
N we .weoumm eeem
wouew Hmm
muescew
wowewmmewu we awesoz
muoEceN Neomem
\moewmwwmm we awesoz
mNoEeem
moweeom we LeeEoz

merLeN

eemwo\xeewm we censoz
mNeELeN

xwee\mzmz we geese:
mpoeweo

Newcoem we Leesoz

eNme.u

mpmm.n

mem.s

NNN~.1
mNMN.

142

.weeumm geweeo
mom we cmesoz

oooo.- oooo.- ooNN.
NoNo.- oooN. owoo.
oooo.- NNoo. moNN.
oooo.- NNNN. oooo.-
oooo. Nooo.- Npow.
oooo. oooo.- Nomo.-
oo__.- oooN.-
NooN.-
.wooomo towooo NooEomwwoonN oooeo Noo

ooo< wo monsoz oooo we eooEoz .ooo.oN oonwoeeoz

mmmm.u
ome—.
mwNm.

eopo.u
pmeo.

mmep.u
mmmo.u
wNem.u

wwmo.u

oooeo ooo ooo.oN
-ooo.o, oonNoeeoz

mNeELeN .oswuceu
Nwoe<\moz we LeeEoz

mueegeo ewmoz
Powwooeem we awesoz

mNoELeN .eseeceu
\mo<\xeem we Leesoz
nose: emeuoeewee

.weoomm Leweeo
eeopwecou we tensoz

.weoumm Lewemo
New we Leesoz

.weopmm eewoeo
eeoq we Leosoz

moeo55ow_ooomo
eeeo we awesoz

ezowm Hmm
+ooo.mN emNNPNELoz

uaogm Hmm mom.eN
loco.mp emuwweewoz

ooecm Hmm mmm.ew
-ooo.ow eerFeELez
ooeem Nmm mom.m
-ooo.m eerweELez
ooecm mm<
+om eeroneez

eoegw emq
oelmm eerwoewez

143

mmNN.u mmmN.u
cone. eNom.
”Nap. NNPF.
mnmo. Fooo.n
mmmN.u opme.u
mnmN.n meN—.u
Pmmm.u memm.u
opmm.u nmue.n
Nmmo. Fem—.
mmpN.u Pmmo.u
Feee.u Feme.u
m—mm.n eeme.u
mNme. mmwe.

.Neeemm eewoeo .weepmm Leweeo
mom we geese: eu=< we geese:

NP—N.u eNFN.
mnmN.l ammo.u
emmp.u Npmo.u
ommm. Pmpp.u
NmNo. mmmp.
enme. Nem_.u
Pemp. mFeH.
enmo. mnnm.
ammo. mmmp.u
mmep. mmmo.
Nmmm. Romp.
empo.u NNFN.
NN—F. mnoo.
muee55mwweeumu eoewm Nmm

oooo wo tensoz +ooo.mN oononLoz

memo.
oemw.

mFNw.

oFNN.u

omeo.u
emmo\n

mm—o.
upmo.u

mmoo.u

eppe.n

mwmo.u

mmmo.

emmw.u

eeewm Hmm mma.eN
nooo.mF eerPmELez

.weeemm weeew
we N we .weopmu memo

.weoemm wouew
we N we .weoumm eeo<

.weeumu weeew we N
we .Neeumm eeouwceoo

.pneumm
Pouch we N we Pompom
moweeogewez woemeee

.wnepmm Peach we N we

.Pnepmm xcwwo ece new

.weoumm Page» we
N we .weeumm eeem
weueh Hmm
muoeLeN
wowewmmewm we Lensez
mNeELeN Peomem
\moewmwwem we Lamasz

mooseem
gmweeem we geese:

mposeem

eemwo\xeewm we cmeEoz
mNoELeN

xwow\mzez we eeoEoz
mNoELeN

Neucoeu we geese:

144

menace; Ngeeeoemucem
Nw3e<\zoz we Lmesez

memm.

muoseeu ewmez
wowwuzemm we Leesez

omme.

mmfim.

mooeeoo .oooo\oo<
\xeem we geese:

FNNN. mmmo.u

NFmN. ammo.n
mmeo. mmmo.u

Nmeo.u

eesez
emeucmeemo

mueewem .eEmuceu
Nw=e<\moz we Le952

mNeEeem uwmoz
wowwuoeem we eeeEoz

mposeeu .oEmuceu
\mo<\xeem we eoesoz
ceeez emeeeeeeeo

.weeumm Leweeo
eeouwegou we Leesoz

.weoumm Lewoeo
moo we LeoEoz

.weepmm empeeo

euo< we awesoz
mueesnmwweoemm

eeeN we emeEoz
ooecm Hmm

+ooo.mN eerwoecez
eoewm Hmm mom.eN

-ooo.mw eeronLez
ooeeo Nmm mma.ew

-ooe.ow eerweELez
eoewm Hmm mmm.m

-ooo.m eerweELez
eoeem mm<

+om eerwesLez
ooeeu em<

melmm eeNNPNELez

.weeumm Leweeo
.ceem we Leesoz

145

NNeo.

mmm_.n

memo.

woow.
emON.
Nomp.

NmmN.
MNPN.

mmoo.n
momp.u
mNmF.
mpmm.

ommw.n

mpeELeo Ngeweeeeeeeo
awoe<\moz we Lmesoz

ONoo. N—mp. ompp.
mmme.u NmmN.n Neew.u
eoow. emNF. mnpp.u
NmNN.n Pmoo.u moop.n
epee. moon. eNNN.
mmmm. eNON. mono.
NNFm. mem. mpmp.
nmme. nmmm. mmmo.
omop. omop. Nmmo.
ommN. NmMN. emmo.
mmme. Nmpe. nmmw.
Poem. Nnmm. omMN.
onm.u omm~.n mono.u
mueseeu ewmez pesto; .Nceo\mo< nose:

wowwuaeem we awesoz \xeem we Leesoz emeueeewmo

wwoo. .Nooomm woeew
we N we .weoemm mote

mwoN. ._ooomo Nooow
wo N no .wooomo oooo

. .Faeumu Pouch we N
mmee mo .weeumm eeouwcgoo

.weepmm
NNMN.-_euew we N we weonN
emweeogewez wooecmo

eemw.._ooumm weeew we N we
.weeumm xcwwo one you

emmm.- .woopmm weeew we
N we .Neoemm eeeN
mmeo.- wouew “mm

mNmo. mooseeo
Nowewmmewe we censoz

mooo. mooseoa Noomoo
\moewmwwem we LeeEoZ

ammo.- mueeceo
coweeom we Leesoz

mNm_.- meoELeo
eemwo\xeowm we LeeEoz

wmmo. meoeceo
xwow\m3ez we tensoz

emow. mooseeo

Neucooo we geese:

.weeumm Lewemo
.cgow we emesoz

146

emoo.n

emNo.u

mFeN.

Nmow.
eoow.
omop.

nmNm.
epem.

Nomo.a

mpesgem
coweeom we Leesoz

wap.

emo_.u

mooo.

Bump.
oooN.
mmNN.

mmom.
em—m.

mNmP.

NmmN.

mueseeu eemwo
\Noowo wo tango:

ooNo. owoo.L
oooN.- NooN.
NNNN. oNNo.
_oow.- oooN.
oooN. mooo.-
oNoN. ooN_.-
oNoo. oooN.-
Nome. oooo.-
mooo. oooN.
mooo. oooo.-
_oNo. ooNo.-
Nooo.-
muesgeu xwew maeseeo

\mzmz we weesaz Ngucoeo we Leesoz

.weopmm weuew
we N we .weeemm mega

._ooomo _ooow
wo N mo .Nooomo oeoo

.Fneumm quep we N
we .Peeumu mgopwegoo

.wooemo
~euew we N we Pneumm
emweoomeeez wegmcmo

.wnewmm weuew we N we

.Paepmm xcwwo use new

.weeemm weuew we
N we .weeumm eeew
wepew How
mueeeeo
_ewewmme_m we geese:
mueELeN Fmomeo
\moewmwwmm we Lease:
mueELeN
coweoom we geese:
mueEeeN
eemwo\xeowm we geese:
mueeLeN
xwew\m3mz we L65:52

mooseeo
Neueoeo we Nongoz

147

memo.u

umN.u

mnmw.u

mmmw.u

mmeo.u

.mweeomo Neeew
we N we ..eeomo eeee

mmNF.

moom.u

mmmo.

eeeo.u

NNmN.

pweF.

Peach Hmm

mPoN.

moNN.u

Nepp.

ommo.u

mmoN.

oNeo.
mNmN.

muesgew
weewmmepo we geese:

momo.

oomw.

oNNo.

oNeo.
oomo.u

NmNP

NoNo.
mmNo.

mueseeo weomeo
\moewmwpem we weesez

.Neeemm weuew
we N we .Feeemu mogo

.Neeeao Neeew
we N we .Neeomo eooo

.Neeooo Neeew we N
we .weoemm eeouwceoo

.wnepmm
Neoew we N we Neeeeo
emwecomeeez Noeeeem

.Neeemo Neeew we N we

.weepmm Neweo ece New
._eoemm woeew we

N we .weoomm eeem

Peeew Hmm
meoeeeo

wowewmmewu we geese:
mNeELeN weomeo

\Noewmwwmm we geese:
mooELeo

cmwcoom we awesoz
meeeceo

eemwo\xeowm we LeeEoz
meoeceo

Nwow\m3ez we LeeEoz

mueeLeo
Newcaeo we geesez

ommp.u

148

.weeomo weoew we
N we .weeeeo eooo

pmo~.u eemp.n
oneN. mooo.:
mmmw.u

nmmo. .weopmm weuew
we N we .weeemm mowo
memN.- .weeemm weeew
we N we .weeemm euo<
mONw. .weeemm woeew we N
we .weoumu ewoewewom
Nwmo.u .weepmm

Pogo» we N we weeumm
moweeenewmz woweceo

.weeumm weeew we N we
.weeemm xcwwo eco pom
.woeumm woeew we

N we .weoumm eeew
weuew Hmm
mNeEweN

wowewmmewo we Leosoz
mueewem weomeo

\moewmwwmm we weeEoz
meeEeeN

omweoom we geese:
meeEweN

eemwo\xeowm we geese:
mNeELeN

xwoo\m3mz we weesoz

mNoEwew
Ngucoeu we awesoz

..eeemm woeew we N .eeemo wooew we N we .weeomo weeew we N me

me .weeumw ewouwcwow emwecegewez Neweceo

.Pneumm xcwwo a new

LIST OF REFERENCES

LIST OF REFERENCES

Arbitron Radio. Arbitron Radio Audience Estimates for the Lansing-
East Lansing Market. New York: Arbitron Radio, October/November,
1980.

Duncan, J. American Radio. Vol. V., lst ed. Kalamazoo: Gilmore
Advertising, Spring, 1980.

 

Emery, W.B. Broadcasting and Government: Responsibilities and Regulations.
East Lansing: Michigan State University Press, 1971.

Haring, J.R. Competition, Regulation, and Performance in the Radio
Broadcast Industry. Yale University: Ph.D. Dissertation, 1975.

 

Hesbacher, P. "Radio Format Strategies." Journal of Communication
(Winter 1976).

 

Kerlinger, F.N. and E.J. Pedhauzer. Multiple Regression in Behavioral
Research. New York: Holt, Rinehart and Winston, 1973.

Kotlar, P. Marketing Management: Analysis, Planning and Control.
New Jersey: Prentice Hall Publishers, 1976.

Lull, J.T., L.M. Johnson and C.E. Sweeney. "Audiences for Contemporary
Radio Formats." Journal of Broadcasting (Fall 1978).

 

Massey, W.F. "Discriminant Analysis of Audience Characteristics."
Journal of Advertising Research (1965).

Neilsen, R.P. and T. Thibodeau. "Applying Market Research Methods for
Formating Decisions for Radio." Feedback (Spring 1978).

Nie, N. H. ,C. H. Hull, J. G. Jenkins, K. Steinbrenner and D.H. Bent.
Statistical Package for the Social Sciences. 2nd ed. New York:
McGraw—H111,l975.

Nunnally, J.C. Psychometric Theory. New York: McGraw-Hill, 1978.

Quaal, W.L. and J.A. Brown. BrOadcast Management. New York: Hastings
House Publishers, 1976.

 

Routt, E., J. McGrath and F. Weiss. 'The Radio Format Conundrum. New York:
Hastings House Publishers, 1978.

 

149

"11111111111111111111