UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A factor analytic search for dimensions of audience exposure to a mass medium Jones, Vernon 1975

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1975_A1 J65_4.pdf [ 14.88MB ]
Metadata
JSON: 831-1.0100118.json
JSON-LD: 831-1.0100118-ld.json
RDF/XML (Pretty): 831-1.0100118-rdf.xml
RDF/JSON: 831-1.0100118-rdf.json
Turtle: 831-1.0100118-turtle.txt
N-Triples: 831-1.0100118-rdf-ntriples.txt
Original Record: 831-1.0100118-source.json
Full Text
831-1.0100118-fulltext.txt
Citation
831-1.0100118.ris

Full Text

A FACTOR ANALYTIC SEARCH FOR DIMENSIONS OF AUDIENCE EXPOSURE TO A MASS E D I UN by VERNON JAMES JONES B.A., Un i v e r s i t y of B r i t i s h Columbia, 1968 M.B.A., University of B r i t i s h Columbia, 1970 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY ; i n the Faculty of Commerce and Business Administration We accept t h i s thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA June, 1975 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Commerce and. Business A d m i n i s t r a t i o n The University of British Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 ABSTRACT The object of t h i s study has been to investigate dimensions of aggregate audience exposure to a mass medium (a d a i l y newspaper). A mass medium such as a newspaper has a v a r i e t y of content and an organized structure. I n d i v i d u a l audience members are expected to be s e l e c t i v e i n t h e i r exposure to the medium using both i t s content and structure as a basis f o r t h e i r choices. Moreover, audience members are expected to act s i m i l a r l y i n t h e i r s e l e c t i v e use of the medium, r e s u l t i n g i n dimensions of audience exposure which are determined by i t s content and s t r u c t u r a l organization. Such dimensionality has implications f o r the managerial use of the medium as a v e h i c l e or, more p r e c i s e l y , a set of veh i c l e s f o r the del i v e r y of s p e c i f i c news and adv e r t i s i n g information. The p r i n c i p a l research method employed i n t h i s study was factor analysis. A c r i t i c a l review of r e l a t e d f a c t o r a n a l y t i c a p p l i c a t i o n s i n advertising research was undertaken. However, unlike these previous studies, p r i o r expectations concerning the fa c t o r r e s u l t s were introduced into the an a l y s i s . This was accomplished through the construction of a model which stated that s e l e c t i v e exposure i s a function of the content and structure of a medium (or media). The model was used to predict aggregate audience exposure on a content and/or s t r u c t u r a l basis and these predictions were investigated using factor a n a l y s i s . The procedure was r e p l i c a t e d across samples and the r e s u l t s were va l i d a t e d through r e l a t i o n to external v a r i a b l e s thought to be associated with mass media exposure. i i i The r e s u l t s c l e a r l y indicated that the dimensions of audience exposure to the news content of a d a i l y newspaper were determined by the l a t t e r ' s content and s t r u c t u r a l organization. Accordingly, i t was concluded that the i n t e r n a l "managerial" sections of the newspaper represented v e h i c l e s for the de l i v e r y of s p e c i f i c news and promotional information. These r e s u l t s confirmed the argument that conventional audience assessment procedures, presently calculated on the basis of exposure to the en t i r e newspaper, should recognize i n t e r n a l content and s t r u c t u r a l s e l e c t i v i t y of the newspaper's i n t e r n a l "managerial" sections. iv TABLE OF CONTENTS I. INTRODUCTION . . 1 The Factor Analytic Search for Dimensions of Audience Exposure . . 3 Dimensions of Audience Exposure to a Mass Medium -A Daily Newspaper . . 5 The Data and Analytic Procedure . . 8 Importance of the Study . . 9 the managerial importance of the study . . 9 the methodological importance of the study . . 11 Scope of the Study . . 12 Organization of the Study . . 13 II. THE FACTOR ANALYTIC SEARCH FOR DIMENSIONS OF AUDIENCE EXPOSURE . . 14 Factor Analysis . . 14 Rotation . . 18 Varimax Rotation . . 1-9 Applications for Factor Analysis in Media Research . . 20 Summary . . 29 A Note on Reliability and Validity in Factor Analysis . . 31 III. DIMENSIONS OF AUDIENCE EXPOSURE TO A MASS MEDIUM . . 35 The Controversy Surrounding Selective Exposure . . 35 Summary . . 40 Selective Exposure to a Mass Medium: An Economic Approach . . 42 I:. ... ..Dimensions of Audience Exposure to a Da i l y Newspaper . . 49 News Content . . 51 Advertising Content . . 56 Summary . . 61 IV. THE DATA . . 62 Measuring Instrument and Sample . . 62 Content C l a s s i f i c a t i o n . . 64 Audience Descriptor Variables . . 65 V. ANALYTIC PROCEDURE . . 71 Data Extraction and the A P r i o r i C l a s s i f i c a t i o n System . . 72 The Empirical C l a s s i f i c a t i o n System . . 75 i ) Factor Analysis of a Co r r e l a t i o n Matrix Calculated on Dichotomous Data . . 75 i i ) P r i n c i p a l Components Analysis . . 78 i i i ) Number of Factors . . 80 A Test f o r the Independence of the Two C l a s s i f i c a t i o n Systems . . 80 Exploratory Analysis . . 86 Cross V a l i d a t i o n . . 88 External V a l i d a t i o n . . 88 VI: RESULTS OF ANALYSIS ON NEWS DATA . . 92 Panel A . . 93 Panel A - working hypotheses . . 93 data extraction and the a p r i o r i c l a s s i f i c a t i o n system . . 93 the e m p i r i c a l l y determined c l a s s i f i c a t i o n system . . 95 v i test for the independence of the two c l a s s i f i c a t i o n systems . . 96 Panel A - working hypotheses ri, . . 102 data extraction and the a p r i o r i c l a s s i f i c a t i o n system . . 102 the em p i r i c a l l y determined c l a s s i f i c a t i o n system . . 102 test for the independence of the two c l a s s i f i c a t i o n systems . . 103 Panel A - working hypothesis . . 107 data extraction and the a p r i o r i c l a s s i f i c a t i o n system . . 107 the empirical c l a s s i f i c a t i o n system . . 107 discussion . . 108 Panel B . . 114 Panel B - working hypothesis . . 114 Panel B - working hypothesis . . 119 Panel B - working hypothesis . . 122 Panel C . . 127 Panel C - working hypothesis • . . 127 Panel C - working hypothesis . . 136 Panel C - working hypothesis . . 139 External V a l i d a t i o n of Factor Results on News Data . . 143 the regression model . . 143 op e r a t i o n a l i z a t i o n of the model . . 144 the regression r e s u l t s . . 145 VII. RESULTS OF ANALYSIS ON ADVERTISING DATA . . 152 Panel A . . 153 Panel A - working hypothesis H, . . 153 v i i data extraction and the a p r i o r i c l a s s i f i c a t i o n system . . 153 the empirical c l a s s i f i c a t i o n system . . 154 test f o r the independence of the two c l a s s i f i c a t i o n systems . . 154 discussion . . 155 exploratory analysis . . 159 summary and discussion • . 180 Panel B . . 186 Panel B - working hypothesis H^ . . 186 exploratory analysis . . 189 discussion . . 194 Panel C . . 202 Panel C - working hypothesis . . 202 exploratory analysis . . 203 discussion . . 206 External V a l i d a t i o n of Factor Results on Advertising Data . . 219 VIII. SUMMARY AND CONCLUSIONS . . 223 Summary of Results of Analysis on News Data . . 223 Summary of Results of Analysis on Advertising Data . . 230 Conclusions . . 233 on the use of factor analysis . . 233 on the economic model of s e l e c t i v e exposure . . 235 on the re s u l t s of data analysis . . 236 Implications of the Study . . 238 methodological implications . . 239 managerial implications . . 239 Areas f o r Future Study . . 243 BIBLIOGRAPHY 245 v i i i LIST OF TABLES 6-1 Analysis of News Data - Panel A ( a l l issues) . . 99 6-2 (i ) Analysis of News Data - Panel A (issue #3) . . 105 6-2 ( i i ) Analysis of News Data - Panel A (issue #6) . . 106 6-3 (i ) Analysis of Public A f f a i r s Data - Panel A . . 110 6-3 ( i i ) Analysis of Sports Data - Panel A . . I l l 6-•3 ( i i i ) Analysis of Business Data - Panel A . . 112 6-•3 (iv) Analysis of Women's Data - Panel A . . 113 6--4 Analysis of News Data - Panel B ( a l l issues) . . 116 6-•5 (i) Analysis of News Data - Panel B (issue #3) . . 120 6--5 ( i i ) Analysis of News Data - Panel B (issue #5) . . 121 6--6 (i ) Analysis of Public A f f a i r s Data - Panel B . . 123 6--6 ( i i ) Analysis of Sports Data - Panel B . . 124 6--6 ( i i i ) Analysis of Business Data - Panel B . . 125 6--6 (iv) Analysis of Women's Data - Panel B . . 126 6--7 (i) Analysis of News Data - Panel C ( a l l issues) . . 131 6--7 ( i i ) Analysis of News Data - Panel C ( a l l issues) . . 133 6--7 ( i i i ) Analysis of News Data - Panel C (issues 3,4,5,6) . . 135 6--8 (i) Analysis of News Data - Panel C (issue #2) . . 137 6--8 ( i i ) Analysis of News Data - Panel C (issue #6) . . 138 6 -9 (i ) Analysis of Public A f f a i r s Data - Panel C . . 140 6 -9 ( i i ) Analysis of Sports Data - Panel C . . 141 6 -9 ( i i i ) Analysis of Women's Data - Panel C . . 142 6 -10 Regression of Women's Factor Scores . . 148 6 -11 Regression of Sports Factor Scores . . 149 i x 6-12 Regression os Public A f f a i r s Factor Scores . .150 6- 13 Regression of Business Factor Scores . .151 7- 1 Analysis of Advertising Data - Panel A ( a l l issues) . .157 7-1 ( i ) Analysis of Advertising Data (Panel A) - 2 Factor Solution. . 163 7-1 ( i i ) Analysis of Advertising Data (Panel A) - 3 Factor Solution. . 170 7-1 ( i i i ) Analysis of Advertising Data (Panel A) - 5 Factor Solution. . 172 7-1 (iv) Analysis of Advertising Data (Panel A) - 8 Factor Solution. . 178 7-2 ( i ) Analysis of Advertising Data i n Women's Section - Panel A . .184 7-2 ( i i ) Analysis of Advertising Data i n Sports Section - Panel A . . 185 7-3 Analysis of Advertising Data - Panel B ( a l l issues) . .187 7-3 ( i ) Analysis of Advertising Data (Panel B) - 3 Factor Solution. . 196 7-3 ( i i ) Analysis of Advertising Data (Panel B) - 5 Factor Solution. .198 7-3 ( i i i ) Analysis of Advertising Data (Panel B) - 6; Factor Solution. .200 7-4 Analysis of Advertising Data - Panel C ( a l l issues) . .207 7-4 ( i ) Analysis of Advertising Data (Panel C) - 3 Factor Solution. . 213 7-4 ( i i ) Analysis of Advertising Data (Panel C) - 7 Factor Solution. .216 7-5 Regression of "X" Factor Scores . .221 7- 6 Regression of "Y" Factor Scores . .222 8- 1 Summary of Hypothesis Testing . .225 X LIST OF FIGURES 1. Raw Data for Factor Analysis . . 15 2. Swanson's Factors I & II for Nightime T e l e v i s i o n . . 22 3. Relation Between Cost and Value of Search . . 43 4. Selective Exposure to a Mass Medium as a Function of Value and Cost of Information . . 44 5. Sample Questionnaire Page . . 63 6. Rigorous Content C l a s s i f i c a t i o n Results . . 67 7. Relaxed Content C l a s s i f i c a t i o n Results . . 6 8 8. A n a l y t i c Procedure . . 73 9. Maximum Product - Moment Cor r e l a t i o n as a Function of the Marginal S p l i t s of Two Dichotomous Variables . . 76 10. ' A Hypothetical Factor Loadings Matrix . . 82 11. Contingency Table Analysis . . 83 12. Contingency Table Analysis - Panel A . . 97 13. Contingency Tables - Panel A (issue #3) . .103 14. Contingency Table Analysis - Panel B . . 115 15. Contingency Table Analysis - Panel C . . 128 16. Contingency Table Analysis - Panel A . .155 17. Desription of Advertising Quarter Pages - Panel A . .161 18. Contingency Table Analysis - Panel B . . 186 19. Description of Advertising Quarter Pages - Panel B . .191 20. Contingency Table Analysis - Panel C . .202 21. Description of Advertising Quarter Pages - Panel C . . 210 ACKNOWLEDGMENT I p a r t i c u l a r l y wish to acknowledge the assistance and guidance of Dr. F.H. S i l l e r i n the preparation of t h i s d i s s e r t a t i o n . Dr. S i l l e r , the Chairman of my thesis committee, has not only proven invaluable to t h i s project but also o r i g i n a l l y encouraged me to pursue doctoral studies and guided me throughout the program. I am also p a r t i c u l a r l y indebted to Professors D.L. Weiss, J.D. Claxton, S.M. Oberg and W.G. Davenport who contributed i n various ways to the completion of t h i s project. My friends and colleagues i n the Ph.D. program have also contributed to my studies and have served to make l i f e i n t e r e s t i n g i n a number of capacities over the l a s t few years. P a r t i c u l a r l y , I would l i k e to acknowledge the assistance of Dr. W.E. McMullan and h i s " c r e a t i v i t y " sessions. My parents and s i s t e r , Roni, have provided tremendous support throughout my many years at U.B.C. and I would l i k e p a r t i c u l a r l y to thank my wife, Barbara, for her patience and assistance. Also, a number of friends have been most tolerent of my "preoccupation." F i n a l l y , I would l i k e to thank Miss Joan Adair and Mrs. Pam Pryce for t h e i r s k i l l and patience i n typing the document. 1 Chapter I INTRODUCTION The purpose of t h i s study i s to investigate the dimensions of aggregate audience exposure to a s p e c i f i c mass medium: a d a i l y newspaper. A medium such as a newspaper has a v a r i e t y of content and an organized structure. Individual audience members are thought to be s e l e c t i v e i n t h e i r exposure to the medium using both i t s content and structure as a basis for t h e i r choices. Moreover, audience members are also expected to act s i m i l a r l y i n t h e i r s e l e c t i v e use of the medium. This r e s u l t s i n dimensions of aggregate audience exposure which are determined by the content and s t r u c t u r a l organization. Such dimensionality has implications for the use of the medium as a v e h i c l e , or more p r e c i s e l y a set of veh i c l e s , for the d e l i v e r y of s p e c i f i c news and advertising content. In previous research concerning dimensions of audience exposure, the focus has been on find i n g an underlying dimensionality to a set of advertising vehicles such as t e l e v i s i o n programs or magazines.''" It was supposed, usually i m p l i c i t l y , that i n d i v i d u a l s are d e l i b e r a t e l y s e l e c t i v e 1. A reyiew of this l i t e r a t u r e i s the subject of Chapter I I . For a short synopsis r e f e r to J. F. Engel, D. T. Ko l l a t and R. D. Blackwell, Consumer Behaviour (2nd ed.,New York: Holt, Rinehart and Winston, Inc., 1973), p. 384. 2 with respect to the content of such vehicles.''' Hence viewers chose to watch p a r t i c u l a r t e l e v i s i o n programs because of t h e i r content. Accordingly, i t was expected that audience exposure was more highly correlated among vehicles of s i m i l a r content than among veh i c l e s of d i f f e r e n t content. Such a highly i n t e r c o r r e l a t e d group or c l u s t e r was thought of as a program type or a dimension of audience exposure. To use a s i m p l i f i e d example, i f t e l e v i s i o n s i t u a t i o n comedies were highly i n t e r c o r r e l a t e d they would be thought of as a s i t u a t i o n comedy dimension. This research t r a d i t i o n w i l l be extended i n two important ways. F i r s t , the search for aggregate dimensions of audience exposure w i l l be applied to a s p e c i f i c mass medium. Just as i n d i v i d u a l s choose to read p a r t i c u l a r magazines or watch p a r t i c u l a r t e l e v i s i o n programs, i t i s expected with c e r t a i n media such as a newspaper they select content i n t e r n a l l y . That i s , i n d i v i d u a l s do not read the i n t e r n a l content of a newspaper randomly or exhaustively but rather on a deliberate s e l e c t i v e basis. Such s e l e c t i v i t y i s expected to r e s u l t i n dimensions of audience exposure. Second, the influence of the structure of a mass medium w i l l by systematic-a l l y recognized. As stated above, previous research has i m p l i c i t l y assumed that aggregate dimensions of audience exposure r e f l e c t only content s e l e c t i v -i t y . The structure of a mass medium simply means the way i n which the con-tent of the medium i s organized." For example, exposure to t e l e v i s i o n may be as much determined by the time of day or national network as by programming content. This equally applies to a p r i n t medium such as a newspaper where time and p o s i t i o n are expected to be the major s t r u c t u r a l influences. As 1. For example, C. E. Swanson, "The Frequency Structure of T e l e v i s i o n and Magazines," Journal of Advertising Research, VII (June 1967). 3 an example of the former, i t may be that i n d i v i d u a l s vary t h e i r exposure over time, that i s , they may read the newspaper thoroughly one day and not another i n d i c a t i n g that dimensions-., of audience exposure w i l l r e s u l t from the p a r t i c u l a r issues of the newspaper. S i m i l a r l y , i n d i v i d u a l s may read according to the way content i s positioned i n the newspaper; f o r example, they may use indexed sections as a means of s e l e c t i v e exposure. The balance of t h i s chapter serves as an introduction to the organization of the d i s s e r t a t i o n . There i s also a discussion of the importance and scope of the study. The Factor A n a l y t i c Search for Dimensions of Audience Exposure The term "dimensions; of aggregate audience exposure" has a more precise meaning wit h i n the context of t h i s study than a highly i n t e r -correlated group or c l u s t e r of t e l e v i s i o n programs or magazines. This necessarily involves some introduction to the methodology employed. In the search for dimensions of aggregate audience exposure, the research method used i s factor analysis or, more accurately, p r i n c i p a l components analysis. The term " f a c t o r a n a l y s i s " describes a set of techniques where the observable responses of i n d i v i d u a l s on a number of variables are represented i n terms of a reduced set of variables c a l l e d f a c t o r s . ^ 1. For an excellent short summary of factor analysis see R. J. Rummel, "Understanding Factor Analysis," Journal of C o n f l i c t Resolution, XI ( D e c , 1 967 ) . A synopsis of p r i n c i p a l components analysis can be found i n P. E. Green and D. S. Tul.l, Research for Marketing Decisions (Englewood C l i f f s , N.J.: Pr e n t i c e - H a l l , 1 9 7 0 ) , p. 4 02 - 4 3 1 . As such i t i s used as a means to discovering or v e r i f y i n g the underlying dimensionality to a set of r e l a t e d v a r i a b l e s . The r e s u l t i n g factors represent the precise meaning attached to dimensions of aggregate audience exposure within the context of this study. In the s i m p l i f i e d example used above, the viewing scores of a sample of i n d i v i d u a l s on a number of t e l e v i s i o n programs (variables) would be factor analyzed i n an attempt to f i n d a fewer number of factors or dimensions which adequately account for the variance i n the o r i g i n a l scores. A discussion of factor analysis and i t s a p p l i c a t i o n to t h i s aspect of media research i s undertaken i n Chapter I I . A number of studies are considered but i t i s generally concluded that the technique has been inappropriately applied. Factor analysis, i t w i l l be argued, i s almost always capable of discovering some inte r p r e t a b l e underlying dimensionality i n a given data set. Thus i t i s of i n t e r e s t to determine not whether underlying factors exist but whether they can be interpreted i n terms of some a p r i o r i theory or hypothesis. The research reviewed i n Chapter II i s weakened by the f a i l u r e to set up a p r i o r i hypotheses about the expected r e s u l t s of f a c t o r analysis. As a r e s u l t there has been no'."attempt to d e l i m i t e i t h e r the problem or the data to be investigated. Accordingly, i t has been d i f f i c u l t to adequately i n t e r p r e t the r e s u l t s . It i s an important feature of t h i s study that c e r t a i n methodo-l o g i c a l safeguards w i l l be used i n the a p p l i c a t i o n of factor a nalysis. F i r s t , a p r i o r i expectations concerning dimensions of aggregate audience exposure w i l l be s p e c i f i e d as c l e a r l y as possible. This i s accomplished through construction of research hypotheses i n Chapter I I I . Second, i n the data analysis an attempt w i l l be made to r e p l i c a t e the r e s u l t s across 5 d i f f e r e n t samples. F i n a l l y , external v a l i d a t i o n w i l l be attempted by r e l a t i n g the factor r e s u l t s to external variables thought to be related to media exposure. Dimensions of Aggregate Audience Exposure to a Mass Medium  - A Daily Newspaper The existence of dimensions which describe audience exposure to a p a r t i c u l a r medium suggests that i n d i v i d u a l audience members are s e l e c t i v e i n t h e i r exposure to that medium. It further implies that such s e l e c t i v i t y i s re l a t e d among audience members in that they respond to the content and structure of the medium i n s i m i l a r ways. In order to develop a set of p r i o r expectations concerning audience dimensions, Chapter III w i l l specify a micro-model of i n d i v i d u a l behaviour which underlies a set of predictions about aggregate behaviour. Chapter I I I i s divided i n t o three sections. F i r s t , i t i s necessary to review the l i t e r a t u r e on s e l e c t i v e exposure. Generally, research into s e l e c t i v e exposure has developed as a subset of cognitive consistency theory. Cognitive consistency theory i s that branch of psychology which deals with the i n d i v i d u a l ' s drive to maintain consistency among h i s cognitions and behaviour.''' Such an approach, i t w i l l be argued, i s not functional for the purposes of t h i s study. Chapter I II develops, as an a l t e r n a t i v e , an economic model of s e l e c t i v e 1. A review of cognitive consistency theory i s a v a i l a b l e i n M. E. Shaw and P. R. Costanzo, Theories of So c i a l Psychology (New York: McGraw-Hill, 1970), p. 188-218. The r o l e of cognitive consistency theory i n the s e l e c t i v e exposure l i t e r a t u r e i s reviewed by R. P. Abelson et a l . , Theories of Cognitive Consistency (Chicago: Rand McNally and Company, 1968), p. 770 - 800. 6 exposure with respect to mass media. The task of the f i r s t section of Chapter III i s to reconcile the consistency theory approach to s e l e c t i v e exposure with a broader economic approach. The second section of Chapter I I I develops the economic model. Any economic model of behaviour involves c e r t a i n basic assumptions.^ I t assumes a set of preferences for the i n d i v i d u a l concerned, a set of choices i n a scarce environment with prices attached to each choice, and a budget constraint which l i m i t s the extent to which an i n d i v i d u a l can indulge i n these choices. Further, i t assumes the i n d i v i d u a l i s a maximizer of h i s own u t i l i t y . The economic model can be adapted to media exposure i n the following manner. The i n d i v i d u a l has preferences for c e r t a i n content. He i s constrained by the l i m i t e d time i n which he can expose himself to a p a r t i c u l a r medium (budget c o n s t r a i n t ) . He i s then forced to economize or choose among d i f f e r e n t types of content i n order to maximize h i s u t i l i t y . The costs or prices of p a r t i c u l a r choices are represented by the structure of the medium. The structure may support or i n h i b i t the i n d i v i d u a l ' s exposure to selected content. The i n d i v i d u a l ' s content choices can be thought of as providing p o s i t i v e u t i l i t y . The associated structure, the p r i c e s associated with p a r t i c u l a r content, can be thought of as providing negative u t i l i t y . Thus the i n d i v i d u a l maximizes h i s p o s i t i v e u t i l i t y by exposing himself to selected content and he minimizes negative u t i l i t y by using the structure of the medium. 1. A review of "economic man" e s p e c i a l l y with respect to consumer decision processes i s a v a i l a b l e i n F. M. Nicos i a , Consumer  Decision Processes (Englewood C l i f f s , N.J.: Prentice-Hall,1966),p.55-70. 7 With regard to any mass medium, the i n d i v i d u a l maximizes h i s o v e r a l l u t i l i t y by s e l e c t i v e exposure to those combinations of content and supporting structure made available to him. These combinations i n a sense form an array or configuration from which he makes s e l e c t i v e choices. Further, because only a few such combinations of content and structure are a v a i l a b l e , i n d i v i d u a l s respond to the choices i n re l a t e d ways, eit h e r s e l e c t i n g or not s e l e c t i n g , leading to aggregate dimensions of audience exposure. These dimensions are determined by what the medium makes a v a i l a b l e . The t h i r d section of Chapter III adapts the economic model to a p a r t i c u l a r mass medium - a d a i l y newspaper. It i s expected i n t h i s s i t u a t i o n that there are three sources of influence or variance on i n d i v i d u a l exposure: the content of the newspaper, the managerial structure or organization of the newspaper and the time structure due to various issues of the newspaper over time. However, i t i s further expected,as concerns the news or nonadvertising content of the news-paper, that i t i s the array of content and managerial structure which determines i n d i v i d u a l s e l e c t i v e exposure. In simpler terms, the news content i s organized into d i s t i n c t managerial sections and i n d i v i d u a l s e l e c t i v e exposure i s on t h i s basis. Thus aggregate dimensions of audience exposure over time are determined by the managerial content and s t r u c t u r a l organization of the newspaper. The above expectation concerns the news or nonadvertising content of the newspaper. However t h i s study i s conducted at least p a r t i a l l y within the context of ad v e r t i s i n g research. It i s of 8 fundamental importance to t h i s study to inv e s t i g a t e whether the dimensions of audience exposure to advertising are determined by the managerial content and s t r u c t u r a l organization of the news. This w i l l e s t a b l i s h the usefulness of the managerial sections as advertising vehicles which are i n t e r n a l to the newspaper. The culmination of Chapter I I I then i s a set of p r i o r expectations, referred to as working hypotheses, f o r the i n v e s t i g a t i o n of dimensions of aggregate audience exposure to the news and advertising of a d a i l y newspaper. The Data and Ana l y t i c Procedure The data used i n t h i s study were obtained from a newspaper readership survey. The instrument used for measuring audience exposure was a self-administered aided r e c a l l device. The units of measurement were i n d i v i d u a l pages of the newspaper divided into four quarters. The research instrument and data are described i n Chapter IV. The a n a l y t i c procedure, described i n Chapter V, seeks to operationalize the working hypotheses developed i n Chapter I I I with respect to the data a v a i l a b l e . Each working hypothesis determines the s e l e c t i o n of a set of quarter pages from the data f i l e and c l a s s i f i e s them according to content and/or s t r u c t u r a l c r i t e r i a . This i s referred to i n the text as the a p r i o r i c l a s s i f i c a t i o n system. The same subset of quarter pages with respondent scores i s then subjected to p r i n c i p a l component analysis using the number of a p r i o r i categories to determine the number of factors. The 9 object here i s to use the loadings of quarter pages on p a r t i c u l a r factors as a means to c l a s s i f y i n g those quarter pages, hence i t i s ref e r r e d to as the empirical c l a s s i f i c a t i o n system. The working hypothesis of i n t e r e s t i s then operationalized through a s t a t i s t i c a l test of independence between the r e s u l t s of a p r i o r i and empirical c l a s s i f i c a t i o n systems. Chapter V further s p e c i f i e s procedures for both r e p l i c a t i o n and external v a l i d a t i o n . Where the working hypothesis under i n v e s t i g a t i o n i s rejected factor analysis i s then used i n i t s more f a m i l i a r r o l e as a data exploration mechanism. Importance of the Study This study w i l l make two general contributions to media and adv e r t i s i n g research. F i r s t , i t w i l l demonstrate there i s a meaningful underlying dimensionality to the audience of a d a i l y newspaper from a managerial viewpoint. Second, i t w i l l e x p l i cate a c e r t a i n methodological approach to the search f o r audience dimensions i n media research. the managerial importance, of .the .study. This study i s directed towards the v e r i f i c a t i o n of the i n t e r n a l content and s t r u c t u r a l organization of a mass medium as the p r i n c i p a l determinant of aggregate dimensions of audience exposure. With respect to a s p e c i f i c newspaper,it i s directed f i r s t at e s t a b l i s h i n g the dimen-s i o n a l i t y of exposure to "news" as determined by the managerial content/ s t r u c t u r a l organization (managerial sections). This w i l l v e r i f y the existence of veh i c l e s f o r the d e l i v e r y of news content which are i n t e r n a l to the newspaper. The establishment of such i n t e r n a l v e h i c l e s w i l l further confirm the need f o r a re-assessment of audience measurement procedures which currently ignore the p o s s i b i l i t y of s e l e c t i v i t y within a p a r t i c u l a r medium. In research antecedent to t h i s study, S i l l e r was able to demonstrate that most i n d i v i d u a l s are s e l e c t i v e when reading the newspaper and that s e l e c t i o n of c e r t a i n content was to some extent re l a t e d to audience predispositions characterized by demographic, personality and a t t i t u d e data.^ He argued that such s e l e c t i v i t y should be taken into account when assessing the conventional reach and frequency measures applied to newspaper readership. It was subsequently demonstrated, using a s p e c i f i c example, that f a i l u r e to recognize such i n t e r n a l s e l e c t i v i t y could s e r i o u s l y d i s t o r t conven-2 t i o n a l reach and frequency measures. It i s v i r t u a l l y a premis of t h i s current study that i n d i v i d u a l audience members are content s e l e c t i v e . The focus i s on dimensions of aggregate audience exposure recognizing the configuration of content and structure with which the i n d i v i d u a l i s pre-sented. This w i l l further confirm the inadequacy of conventional audience measures and v e r i f y the existence of exposure vehicles within a p a r t i c u l a r newspaper. 1. F. H. S i l l e r , "Newspaper Reading: A Study i n Selective E f f e c t s " (unpublished Ph.D. d i s s e r t a t i o n , School of Business Administration, U n i v e r s i t y of Western Ontario, 1972). 2. F. H. S i l l e r and V. J . Jones, "Newspaper Campaign Audience Segments", Journal of Advertising Research, XIII (June, 1973). The study w i l l also attempt to e s t a b l i s h the extent to which the managerial sections (the configuration of news content and structure) determine the dimensions of aggregate exposure to advertising. The re s u l t s w i l l i n d i c a t e to what extent the managerial sections presently act as adv e r t i s i n g v e h i c l e s . This has implications for the future use of such managerial sections f o r the appropriate placement of advertising and perhaps for the design of adv e r t i s i n g copy. the methodological importance of the study There are two considerations here. F i r s t , t h i s study seeks to improve upon the a p p l i c a t i o n of factor analysis i n the area of media research by recognizing the l i m i t a t i o n s of the technique and introducing c e r t a i n methodological safeguards i n i t s a p p l i c a t i o n . The technique i s pri m a r i l y used i n t h i s study as a c l a s s i f i c a t i o n device, that i s , the object i s to c l a s s i f y a set of variables e m p i r i c a l l y and then investigate to what extent the r e s u l t s match on a p r i o r i c l a s s i f i c a t i o n of the same var i a b l e s . This approach may i n fact have wider im p l i c a t i o n than the improvement of factor a n a l y t i c applications i n media and ad v e r t i s i n g research. Rummel, for example, i n h i s widely used text on factor analysis suggests, "A second deductive approach (using factor analysis) involves hypothesizing that c e r t a i n patterns e x i s t . The data are then factored to see i f the patterns emerge. Factor analysis has not often been employed to test hypotheses, but r e s t r a i n t i s due to research^ t r a d i t i o n and not to methodological d i f f i c u l t i e s . " 1. R. J. Rummel, Applied Factor Analysis (Evanston: Northwestern Un i v e r s i t y Press, 1970), p. 22. 12 Second, the more deliberate a p p l i c a t i o n of factor analysis requires the development of a model of i n d i v i d u a l s e l e c t i v e exposure to a mass medium. In t h i s model, the d i s t i n c t i o n between content and s t r u c t u r a l influences on i n d i v i d u a l exposure i s s p e c i f i c a l l y recognized. Although structure i s known to influence media exposure, none of the studies to be c i t e d i n Chapter II were able to e x p l i c i t l y account for i t except as a nuisance f a c t o r . The development of an economic model of s e l e c t i v e ex-posure i s a d i f f e r e n t point of departure i n media research which i s nevertheless compatible with more t r a d i t i o n a l approaches. Scope of the Study The data used i n t h i s study were generated as part of a research project c a r r i e d out on a s i n g l e newspaper i n a medium sized c i t y having only the one major d a i l y . Generalizations to other newspapers w i l l depend on the degree to which the others tend to match the c h a r a c t e r i s t i c s of that studied. However, i t i s expected that the models of newspaper ex-posure developed herein are s u f f i c i e n t l y general to allow such t r a n s f e r -a b i l i t y . Generalizations to other mass media of course require the develop-ment of s p e c i f i c models with respect to the c h a r a c t e r i s t i c s of those media. However, i t i s expected that the a p p l i c a t i o n of the factor a n a l y t i c method and the d i s t i n c t i o n between content and structure of media developed i n Chapter I I I are generalizable to most mass media. The i d e n t i f i c a t i o n of par-t i c u l a r media vehicles i s c l e a r l y dependent on the mass medium concerned. 1 3 Organization of the Study As indicated above, the study has the following organization: Chapter II discusses the meaning of factor analysis and outlines the research t r a d i t i o n where factor analysis has been applied to the search for dimensions of audience exposure. Chapter III reviews the l i t e r a t u r e on s e l e c t i v e exposure, develops an economic model of i n d i v i d u a l s e l e c t i v e exposure and generates a set of p r i o r expectations or working hypotheses concerning the dimensions of aggregate audience exposure to a d a i l y newspaper. Chapters IV and V discuss r e s p e c t i v e l y the data bank to be used and the a n a l y t i c procedure to be applied to that data. This involves the o p e r a t i o n a l i z a t i o n of the working hypotheses developed i n Chapter I I I . Chapters VI and VII present the r e s u l t s of analysis on the news and advertising data r e s p e c t i v e l y . These chapters include the d e t a i l s of the a p p l i c a t i o n of the procedure not covered i n Chapter V, the presentation of r e s u l t s i n a seri e s of tables, and where necessary in t e r p r e t a t i o n s of those r e s u l t s . Chapter VIII draws conclusions about the r e s u l t s of Chapters VI and VII and discusses the implications of these fi n d i n g s . 1 4 Chapter II THE FACTOR ANALYTIC SEARCH FOR DIMENSIONS OF AUDIENCE EXPOSURE In order to provide background to t h i s study t h i s chapter w i l l f i r s t undertake a b r i e f review of factor analysis. This w i l l be concerned with methods of extracting and r o t a t i n g factors. The balance of the chapter re-views a number of studies i n the advertising l i t e r a t u r e which apply p r i n c i p a l component analysis to the search for dimensions of audience exposure. F i n a l l y , there i s a review of problems of r e l i a b i l i t y and v a l i d i t y i n factor analysis. Factor Analysis As mentioned i n Chapter I, factor analysis describes a set of tech-niques where the observable responses of objects or i n d i v i d u a l s on a number of variables are represented i n terms of a "reduced" set of variables or factors. The ent i r e set of re l a t i o n s h i p s within a data matrix i s of i n t e r e s t and there i s no p a r t i t i o n i n g of the data into dependent and independent v a r i a b l e s . The object of factor analysis i s to construct a set of va r i a b l e s , c a l l e d f a c t o r s , which are fewer than the o r i g i n a l v a r i a b l e s , are orthogonal, and contain most of the "information" i n the o r i g i n a l data matrix. The usual input into a factor a n a l y t i c program i s a matrix of c o r r e l a t i o n c o e f f i c i e n t s f o r o r i g i n a l variables calculated across i n d i v i d u a l s . Factors 15 are then "extracted" from the c o r r e l a t i o n matrix. However, the procedure i s more e a s i l y understood i n terms of the basic data underlying the c o r r e l a t i o n matrix. This i s the matrix of scores for i n d i v i d u a l s on a number of t e s t s or variables as i n Figure Figure 1: Raw Data for Factor Analysis Variables 1 2 - j m 1 2 X l m Individuals i x. . N XN1 A factor i s simply a l i n e a r combination of the variables i n the o r i g i n a l data where the scores on the variables have been standardized. The matrix of standardized scores on o r i g i n a l variables can i n fact be l i n e a r l y transformed to y i e l d a matrix of (unstandardized) " f a c t o r scores" on a new set of v a r i a b l e s c a l l e d f a c t o r s . If the factor scores are then standardized, the o r i g i n a l scores can be reconstructed by adding proportions of the stand-ardized factor scores. Thus, the o r i g i n a l standard score on a v a r i a b l e i s a " l i n e a r combination of l i n e a r combinations".''' The proportions of standard 1. P.E.Green and D.S.Tull, Research for Marketing Decisions (Engle-wood C l i f f s , N.J.: P r e n t i c e - H a l l , 1970), p.408. Much of t h i s discussion of factor analysis i s adapted from Green and T u l l ' s Chapter 12. Excellent re-ferences r e q u i r i n g knowledge of matrix algebra are W. W.» Cooley and P. R. Lohnes, M u l t i v a r i a t e Data Analysis, (New York: John Wiley & Sons, 1971), and M. M. Tatsuoka, M u l t i v a r i a t e Analysis, (New York:John Wiley & Sons,1971). factor scores, c a l l e d f actor loadings, are selected such that every i n d i -viduals o r i g i n a l standard score can be exactly reconstructed from h i s standardized factor scores as follows: where, z^j = i n d i v i d u a l i ' s standardized score on o r i g i n a l v a r i a b l e j . z,££ = i n d i v i d u a l i ' s standardized score on factor f. a£j = the proportion contributed by factor f to the o r i g i n a l stand-ardized scores on variable j (factor loading). The factor loadings are i n fact c o r r e l a t i o n s between the factor scores and the o r i g i n a l scores (standardized or unstandardized). What i s to be gained by such a complex representation of the o r i g i n a l scores? To answer t h i s question two important features must be added to the transformation of o r i g i n a l scores, i . e . , the "extraction" of facto r s . F i r s t , the factors are chosen so that t h e i r scores, unlike the o r i g i n a l scores, are uncorrelated. The factors are orthogonal. Second, the factors are chosen sequentially, so that the f i r s t one accounts f o r the maximum amount of variance i n the o r i g i n a l data set, the second accounts for the maximum amount of r e s i d u a l variance, and so on. Now, i t w i l l be r e c a l l e d that the major objective of factor analysis i s parsimony, that i s , the construction of a new set of factors which are fewer than the o r i g i n a l variables yet contain most of the i n -formation i n the o r i g i n a l data matrix. The above discussion implies that F z i j ' = ^ a f j z l f f = l there are as many factors as o r i g i n a l variables as a r e s u l t of the analysis but that the extracted f a c t o r s successively account f o r l e s s variance i n the o r i g i n a l matrix. Thus, one can achieve parsimony simply by eliminat-ing those factors deemed not to have accounted for s u f f i c i e n t variance. A fewer number of factors are then r e l i e d upon to reconstruct the o r i g i n a l scores. In p r a c t i c e there are two general procedures governing what i s c a l l e d the extraction of fact o r s : true factor analysis ( c a l l e d common factor analysis by Rummel and p r i n c i p a l components analysis. Common factor analysis extracts only a l i m i t e d number of fa c t o r s , supposedly those explaining the common variance or communality among the o r i g i n a l v a r i a b l e s . This procedure requires communality estimates. P r i n c i p a l components an a l y s i s , however, extracts the f u l l set of fact o r s , i . e . , as many factors as there are var i a b l e s i n the o r i g i n a l data set. In the l a t t e r case only the most important factors are considered. There are some a r b i t r a r y c r i t e r i a for determining exactly how many factors should be considered. Generally, however, these c r i t e r i a are substitutes for the analyst's judgement as to adequate variance accounted for by the t o t a l factor space or the l a s t factor considered. The factor a n a l y t i c pro-cedure used i n t h i s study i s p r i n c i p a l components analysis. Discussion as to the reason f o r t h i s choice i s undertaken i n Chapter V. The main output of i n t e r e s t from factor analysis i s the fac t o r loadings matrix. The loadings for each factor on the o r i g i n a l variables 1. Rummel, Applied Factor Analysis, p. 104. 18 are used to i n t e r p r e t those fa c t o r s . Where a factor i s heavily loaded on a p a r t i c u l a r set of v a r i a b l e s the analyst attempts to define what i s common among those variables and i n t e r p r e t s the factor accordingly. Rotation While p r i n c i p a l components analysis provides a useful t o o l from the standpoint of data reduction, i t does not usually present the best r e s u l t s from an i n t e r p r e t i v e point of view, that i s , the loadings r e s u l t i n g from the extraction of factors are not s u f f i c i e n t l y c l e a r to allow for f a c t o r i n t e r p r e t a t i o n . Thus an often used procedure i s to rotate the axes indicated by the component a n a l y s i s . ^ Because the points i n the o r i g i n a l variables space are now represented as l i n e a r combinations of f a c t o r s , r o t a t i o n has the e f f e c t of r o t a t i n g or s h i f t i n g factor loadings. Rotation has l i t t l e mathematical or s t a t i s t i c a l s i g n i f i c a n c e -there are an i n f i n i t e number of possible rotated solutions. It i s per-formed i n order to stimulate the i n t e r p r e t a t i o n of factors or components. However, while r o t a t i o n does not imply a mathematically unique solu t i o n , c e r t a i n r o t a t i o n c r i t e r i a do. The i d e a l c r i t e r i o n for factor r o t a t i o n i s an " i n t e r p r e t a b l e " s o l u t i o n . This best s e l e c t i o n of axes i s often c a l l e d a "simple structure", 1. P r i n c i p a l components analysis has a simple geometric i n t e r -pretation. E s s e n t i a l l y i t introduces into the o r i g i n a l v a r i a b l e s space an axis (component) for which the variance of scores explained from pr o j e c t i n g the data points onto t h i s axis i s maximized. Successive axes, orthogonal to the f i r s t , are then drawn maximizing the amount of r e s i d u a l variance explained. Each i n turn explains a l e s s e r amount of variance i n the data. This procedure can be extended u n t i l the number of axes equals the number of o r i g i n a l variables and the t o t a l variance i n the o r i g i n a l space i s accounted f o r . a term coined by Thurstone r e f l e c t i n g the idea that each v a r i a b l e should be represented by as few factors as possible (Thurstone's simple structure requirements can be found i n Rummel, Ch. 16^). In the e a r l i e r days of factor analysis r o t a t i o n was performed g r a p h i c a l l y (also discussed ex-2 tens i v e l y by Rummel) . This procedure was purely subjective - the f i n a l l o c a t i o n of the factors being an accumulation of v i s u a l judgements. Dis-s a t i s f a c t i o n with the labour and s u b j e c t i v i t y of graphical r o t a t i o n led to attempts to develop an a n a l y t i c r o t a t i o n technique. This involved e f f o r t s to reduce Thurstone's simple structure c r i t e r i a to the maximizing or minimizing of a mathematical function. This s i m p l i f i c a t i o n involved two major approaches: quartimax and varimax. E a s i l y interpreted component loadings are those i n which c e r t a i n loadings are close to-unity and others are near zero. The quartimax c r i t e r i o n tends to s i m p l i f y v a r i a b l e com-p l e x i t y , e. g., each v a r i a b l e loads high on one or a few factors and near zero on others. The varimax c r i t e r i o n focuses on factor complexity. It i s now generally considered that the varimax c r i t e r i o n more accurately 3 approximates simple structure. Varimax Rotation The varimax c r i t e r i o n i s a function of the variance of the column of factor loadings. As there are more high and low loadings on a f a c t o r , the variance of the squared factor loadings i s l a r g e r . The highest variance i s obtained when the loadings are near zero or unity. Therefore 1. Rummel, Applied Factor A n a l y s i s , p. 380. 2. Ibid., p. 368. 3. Ibid., p. 392. 20 a r o t a t i o n can be computed by maximizing the variance (hence, varimax) of the squared factor loadings.^ It should be noted that the new axes explain i n t o t a l as much of the common variance as explained by the unrotated loading matrix. The varimax r o t a t i o n merely breaks up t h i s variance i n a d i f f e r e n t way. However, successive components no longer account for maximum re s i d u a l variance i n the o r i g i n a l space. That i s , the variance maximizing property of p r i n c i p a l components i s l o s t although the components as a group account for the same proportion of t o t a l explained variance. Applications for Factor Analysis i n Media Research A number of researchers have been interested i n i d e n t i f y i n g dimensions of exposure to media v e h i c l e s . This has led to some i n t e r e s t i n g r e s u l t s and to some controversy over the appropriate use of factor analysis which was the technique employed (more p r e c i s e l y , p r i n c i p a l components a n a l y s i s ) . The o r i g i n a l research was published by Kirsch and Banks, 2 "Program Types Defined by Factor Analysis". In t h i s study, a panel of consumers kept a diary of t h e i r t e l e v i s i o n viewing and tabulations were made on the basis of viewing or not viewing p a r t i c u l a r programs (62 programs were included). On the basis of the 0/1 tabulations c o r r e l a t i o n c o e f f i c i e n t s were calculated among pai r s of programs. The c o r r e l a t i o n 1. Varimax r o t a t i o n i s discussed i n Green and T u l l , Research  for Marketing Decisions, p. 418 - 421. The computational procedure i s given i n Cooley and Lohnes, M u l t i v a r i a t e Data Analysis, p. 145 - 148. 2. A. D. Kirsch and S. Banks, "Program Types Defined by Factor An a l y s i s , " Journal of Advertising Research, II (September, 1962). matrix was fac t o r analyzed using p r i n c i p a l components analysis and a varimax r o t a t i o n performed. Only s i x factors were considered. The c r i t e r i o n was that each factor account for at least two percent of the variance. The factors were interpreted on the basis of t h e i r high loadings on p a r t i c u l a r programs and the researcher's p r i o r knowledge of the common a t t r i b u t e s of such programs. For example, the f i r s t f a c t o r was described as "westerns on the ABC network". It i s i n t e r e s t i n g to note that Kirsch and Banks, as we l l as recog-n i z i n g common a t t r i b u t e s interpretable as western content, also recognize the structure of the t e l e v i s i o n viewing s i t u a t i o n , i . e . , they discover that the t e l e v i s i o n network may play a ro l e i n determining viewing patterns. In fact 'ABC westerns' and 'NBC westerns' emerge as two d i f f e r e n t f a c t o r s . One l o g i c a l l y needs to investigate further the structure of the viewing s i t u a t i o n . For instance, might the day of the week have an influence so that ABC westerns i n fact form a Monday night factor or would the time of day have an influence? A s i m i l a r study by C. E. Swanson reports an analysis of the viewing frequency for nightime television.''' Instead of the viewed/not viewed (0,1) dichotomy, respondents reported whether or not they had seen up to 4 weekly broadcasts of 108 network programs (thus providing a frequency measure over the duration of the study). C o r r e l a t i o n c o e f f i c i e n t s were calculated between a l l p a i r s of programs to demonstrate 1. C. E. Swanson, "The Frequency Structure of T e l e v i s i o n and Magazines," Journal of Advertising Research, VII (June, 1967). whether or not r e l a t i v e l y frequent viewers of one program were . r e l a t i v e l y frequent viewers of other programs. The data were analyzed using p r i n c i p a l components analysis and the r e s u l t i n g factors rotated to an 'objective solution'.'' Twelve of the possible 108 factors were selected on the c r i t e r i a that each accounted for at least 2% of the t o t a l variance and included two or more programs with a minimum loading of i .3. As an example, Figure 2 reproduces Swanson's f i r s t two factors which are i n t e r -preted according to programs where the factors are most heavily loaded. 2 Figure 2: Swanson's Factors I & II for Nightime T e l e v i s i o n Factor I (Donna Reid - Patty Duke Ozzie & Harriet) Donna Reid . 71 Patty Duke .71 Ozzie & Harriet .70 My Three Sons .60 Farmer's Daughter .53 Flin t s t o n e s .50 Addams Family .40 Shindig .47 Bewitched .46 Johnny Quest .46 Factor II (To T e l l the Truth -I've Got a Secret) To T e l l the Truth .80 I've Got a Secret .78 Password .60 Perry Mason .40 what's My Line .35 Andy G r i f f i t h Show .34 It i s implied by naming Factor I the "Donna Reid - Patty Duke -Ozzie & H a r r i e t " factor that there are some common a t t r i b u t e s among these programs. However, Swanson makes only vague reference to such common at t r i b u t e s and i t i s l e f t as more or l e s s self-evident that Patty Duke and Donna Reid are a l i k e . Perhaps, but then why perform the factor analysis? Moreover, how does one know that these programs are not loaded 1. Swartson uses the phrase 'objective s o l u t i o n ' . This probably r e f e r s to the use of a n a l y t i c a l r o t a t i o n e.g. varimax. 2. Swanson, p. 10. heavily together because they are broadcast on the same network at 6:00 and 6:30 re s p e c t i v e l y and opposite the network news on competitive channels? These questions suggest that Swanson needed to do some p r i o r thinking about which variables should be analyzed, the possible r e l a t i o n -ships among them and the factors he expected to f i n d . Given that c e r t a i n s t r u c t u r a l phenomena are thought or known to e x i s t i n t e l e v i s i o n viewing, Swanson's f a i l u r e to systematically employ p r i o r knowledge could be l a b e l l e d 'naive empiricism'. The factor a n a l y t i c search for program types has been c r i t i c i z e d by A. S. C. Ehrenberg. He i s e s s e n t i a l l y concerned about the a r b i t r a r i -ness of the technique i n i n t e r p r e t i n g f a c t o r s . Reviewing Swanson's work he notes that a l l programs i n the f i r s t factor were broadcast by the CBS network and a l l i n the second by the ABC network. Therefore, he argues, Swanson has no basis f o r i n t e r p r e t i n g h i s factor as program types; he could have as well interpreted them as s t r u c t u r a l factors r e f l e c t i n g a tendency of viewers to be network l o y a l (or, le s s c h a r i t a b l y , so over-come by i n e r t i a that they cannot be bothered to change channels). Ehrenberg i s c r i t i c a l of the use of factor a n a l y s i s , arguing that the existence of program types (dimensions) can only be investigated a f t e r accounting for the s t r u c t u r a l aspects i n viewing, i . e . , the channel, day of the week, etc. He summarizes, 1. A. S. C. Ehrenberg, "On Methods: The Factor A n a l y t i c Search for Program Types," Journal of Advertising Research, VIII (March, 1968). "To discover whether there are any s p e c i a l tendencies for people who view one t e l e v i s i o n program also to view another program, one has of course f i r s t to allow for the s t r u c t u r a l factors i n the viewing s i t u a t i o n . They are choice of channel, day of week, r a t i n g l e v e l , etc., as discussed above, and also the lead-in or inheritance or addiction e f f e c t s which obviously operate for viewing patterns on the same day and on days exactly a week apart. It i s only i n the remaining or unexplained ^ variance that one can look for evidence of program types." He suggests, as an a l t e r n a t i v e method, holding s t r u c t u r a l aspects constant and then i n v e s t i g a t i n g the c o r r e l a t i o n matrix for program types. For example, using h i s own data, he investigates c o r r e l a t i o n s among programs shown on the same day and channel. An average c o r r e l a t i o n i s calculated; programs which c o r r e l a t e above the average are found to be closer together i n time. This methodology, Ehrenberg argues, allows the systematic introduction of p r i o r knowledge, the factor a n a l y t i c method does not. He seems to be r e f l e c t i n g on the obvious complexity of conducting a factor analysis on t e l e v i s i o n programs a f t e r they have been con t r o l l e d f or a l l s t r u c t u r a l factors - probably an i m p o s s i b i l i t y . However, Ehrenberg's complete skepticism i s unwarranted. Why, for example, does he presume that one must f i r s t account for s t r u c t u r a l influences before content? Program types may indeed explain the s t r u c t u r a l i n t e r p r e t a t i o n . The fact that the ABC network runs a number of s i t u a t i o n comedies perhaps explains the network l o y a l t y rather than the other way around, as Ehrenberg assumes. If one considers Figure 2, Ehrenberg's g l i b observation that these are CBS and ABC factors may e a s i l y be questioned. As noted above, p r i o r thinking about hypotheses 1. I b i d . , p. 62. perhaps would have allowed Swanson to investigate the importance of both program types and structure as dimensions of audience viewing. Factor analysis i s a t r i g g e r for the development and d e s c r i p t i o n of underlying constructs or dimensions. However, to be u s e f u l , factor analysis should be conducted i n l i g h t of what i s already known about the s i t u a t i o n to be investigated. One means of p a r t i a l l y circumventing the Ehrenberg c r i t i c i s m s i s to analyze preferences for t e l e v i s i o n programming rather than actual viewing. This i s the approach taken by W. D. Wells i n "The Rise and F a l l of T. V. Program Types".^ The objective of t h i s research i s to investigate through factor analysis how program types vary over the period from 1962 to 1968. Reacting to the c r i t i c i s m that while i n d i -v i d u a l programs might change the types of t e l e v i s i o n programming never change, Wells successively factor analyzes viewing preferences from year to year demonstrating that while factors such as "westerns", "news" and " s i t u a t i o n comedies" persisted, others such as "teenage comedies" and "panel shows" disappeared and s t i l l others such as "movies" and "super-n a t u r a l " appeared. The factor r e s u l t s are quite strong. Notice that by analyzing preferences rather than actual viewing, Wells avoids much of the s t r u c t u r a l confounding. For example, with respect to "news", while some people may prefer news programs generally, causing a pre-ference factor to form, the fact that at 6:00 p.m. each night they were forced to choose between Walter Cronkite and Huntley-Brinkley might cause these programs to load on separate network viewing f a c t o r s . 1. W. D. Wells, "The Rise and F a l l of T e l e v i s i o n Program Types," Journal of Advertising Research, IX (September, 1969) . While an i n t e r e s t i n g aspect to the search f o r dimensions, Wells research su f f e r s from some drawbacks. F i r s t , as an aside, he does not properly i n v e s t i g a t e what he intends. Preference factors emerge from what i s offered by the t e l e v i s i o n networks. In other words, he has demon-strated s e l e c t i v e preferences among programs offered. However, i t seems quite p l a u s i b l e that i f programming on t e l e v i s i o n had not changed from 1962 to 1968, the same preference factors would have pe r s i s t e d . If Wells had wished to inve s t i g a t e s e l e c t i v e preferences he might have done so. For purposes of the research question he posed, tabulations from back issues of "T.V. Guide" would have s u f f i c e d . The second drawback of course i s that Wells does not analyze exposure which, i n the l i g h t of the structure of t e l e v i s i o n viewing, may be s u b s t a n t i a l l y d i f f e r e n t from preferences. In t e r e s t i n g l y , he discusses a comparison of h i s "preference f a c t o r s " with actual "viewing f a c t o r s " based on the 1965/66 season and concludes that, "the patterns formed by the viewing reports were very s i m i l a r to patterns formed by preference ratings".''' Unfortunately, t h i s i s v i r t u a l l y a footnote to the research and data are not presented. However, i f Wells' judgement of the data can be accepted, "The influence of time and channel was apparent, but i t was much smaller than the influence of program content".2 In a recent a r t i c l e , A. V. Bruno investigated e x p l i c i t l y some of 3 the s t r u c t u r a l aspects of t e l e v i s i o n viewing. The viewing of 124 network programs by sample audiences i n New York and Grand Rapids was 1. I b i d . , p. 27. 2. I b i d . 3. A. V. Bruno, "The Network Factor i n T.V.Viewing," Journal of  Advertising Research, XIII (October, 19 73). 27 analyzed. The viewing v a r i a b l e for each program was created from respondents' viewing d i a r i e s and was scaled on a 0-5 continuum from "respondent did not view the program i n diary period" to "respondent viewed program f i v e times during diary period". Factor analysis was undertaken to investigate the extent to which the viewing l e v e l s of various programs were r e l a t e d and to determine the difference, i f any, between viewing by respondents i n d i f f e r e n t geographical areas. Both analyses y i e l d e d 5 factors explaining about 31% of the variance, and there were some s i m i l a r i t i e s between the two samples. The factors were inter p r e t a b l e according to both network and content, some seemingly dominated by programs of a s i m i l a r type and others dominated by a p a r t i -cular t e l e v i s i o n network. Bruno, having read Ehrenberg, concludes that i t i s necessary to examine the o r i g i n a l viewing variables to t r y to gain a better under-standing of the f a c t o r v a r i a b l e s . " I t should be possible to int e r p r e t the c o r r e l a t i o n of viewing l e v e l s i n terms of p o t e n t i a l day-and-hour influences, the programs that precede and follow, and the strengths of i n d i v i d u a l networks. In a general way, i t i s known that d i f f e r e n t channels, d i f f e r e n t times of the evening, and d i f f e r e n t program types a f f e c t viewing habits." Bruno was l i m i t e d i n the presentation of h i s analysis and, as a r e s u l t , i s somewhat d i f f i c u l t , to follow. To determine the network e f f e c t upon viewing he 1. Ibid., p. 35. checked to see whether or not the c o r r e l a t i o n of viewing l e v e l s between a program and the one preceding i t on the same channel was higher than i t s c o r r e l a t i o n with programs preceding i t on other channels. If t h i s was the case he recorded a plus, i f not, a minus. He then conducted a signs test under the n u l l hypothesis of an equal d i s t r i b u t i o n of p o s i t i v e and negative signs. Rejecting the hypothesis i n favour of a high number of pluses tends to v e r i f y a network e f f e c t . A l t e r n a t i v e l y , he investigated correlations of viewing l e v e l s for programs i n the same category i n order to in v e s t i g a t e the e f f e c t of show type on viewing behaviour. The r e s u l t s generally were that a strong net-work e f f e c t was i d e n t i f i e d as well as a weaker but s i g n i f i c a n t r e l a t i o n s h i p between viewing of programs of the same type. Although Bruno's r e s u l t s suggest a strong network influence i t does not seem j u s t i f i e d to say that t h i s i s stronger than or overides the influence of programs of s i m i l a r content. However, Bruno has demonstrated that s t r u c t u r a l f a c tors are important and that the f a c t o r a n a l y t i c tech-nique may serve as a basis to some i n t e r e s t i n g i n v e s t i g a t i o n of the c o r r e l a t i o n matrix. Another study by Bass, Pessemier and T i g e r t (concerned with maga-zine readership rather than t e l e v i s i o n viewing), i s important i n the con-text of the current research.'' Respondents were asked to i n d i c a t e t h e i r 1. F. M. Bass, E.A.Pessemier and D. J . T i g e r t , "A Taxonomy of Magazine Readership Applied to Problems in-Marketing Strategy and Media Sel e c t i o n , " Journal of Business, XLII (July, 1969). readership of 44 d i f f e r e n t p r i n t vehicles on a 6 point scale from "never read" to "read almost every issue". The data were subjected to p r i n c i p a l component analysis followed by varimax r o t a t i o n y i e l d i n g 5 major factors accounting for 50% of the variance. As an example, factor 1 loaded highly on the New York Times, Saturday Review, the New Yorker, A t l a n t i c Monthly, Consumer Reports and Esquire. This was interpreted as a c u l t u r a l , i n -t e l l e c t u a l factor. S i m i l a r l y , there were l i g h t reading, fashion, sensa-t i o n a l i s t i c and homemaker fact o r s . This i s somewhat d i f f e r e n t from the studies c i t e d above as one would not expect s t r u c t u r a l factors to be as s i g n i f i c a n t i n an i n d i v i d u a l ' s s e l e c t i o n of magazines. However, what i s of i n t e r e s t i n t h i s study i s that the authors, once they had the f a c t o r s , calculated the factor scores for the sample of respondents and then per-formed a stepwise regression of the scores against various demographic and nondemographic data also c o l l e c t e d from the respondents. The R^ from s i g n i f i c a n t predictor variables ranged from .14 on factor 5 to .31 on f a c t o r 1. This i s an i n t e r e s t i n g use of factor scores although i n other cases they have been used as predictor variables rather than as the c r i -t e r i o n . ^  The stepwise regression of factor scores i s a f i r s t step towards r e l a t i n g factor r e s u l t s to audience predispositions and provides some external v a l i d i t y for the factors. Summary The research c i t e d above i s strongly empirical. The emphasis has 1. For example see W. F. Massy, " T e l e v i s i o n Ownership i n 1950: Re-suits of a Factor Analytic Study," and D. W. Twedt, "A Multiple Factor Analysis of Advertising Readership," Quantitative Techniques i n Marketing  Analysis., ed. R. E. Frank, A. A. Kuehn, and W.F.Massy, (Homewood, 111.: Irwin, 1962). been to investigate whether or not dimensions of audience exposure exist and, i n one case, there has been an e f f o r t to i d e n t i f y audience segments on the basis of these dimensions.^ There have been no hypotheses as to the c r i t e r i a i n d i v i d u a l s use i n s e l e c t i n g information or media type. Further-more, there has been no formulation of hypotheses regarding the dimen-o sions of audience exposure which might be expected. However, as with a l l research, there i s at least some i m p l i c i t i l l - f o r m e d theory or hypothesis which leads one to undertake the analysis. This seems to have been that i n d i v i d u a l s are cons i s t e n t l y s e l e c t i v e with respect to the content of media v e h i c l e s , for example, i n viewing t e l e v i s i o n , i n d i v i d u a l s are ex-pected to sel e c t programs of a p a r t i c u l a r type. As a r e s u l t of t h i s content s e l e c t i v i t y , i t i s expected that the viewing frequency of one program i s associated with«the viewing frequency of other programs of a s i m i l a r type. Hence there are groups or cl u s t e r s of t e l e v i s i o n programs or, more appropriately, dimensions of t e l e v i s i o n viewing. Using h i s p r i o r knowledge of t e l e v i s i o n , the researcher has i d e n t i f i e d the common at t r i b u t e s of the programs within each dimension and concluded that the dimensions of viewing more or l e s s v e r i f y the existence of program types. Why some researchers have expected that audience members w i l l s e l e c t t e l e v i s i o n programs (or other media vehicles) of a p a r t i c u l a r type i s not c l e a r . As has been demonstrated such a premis ignores the f a c t that t e l e v i s i o n viewing may be s u b s t a n t i a l l y influenced by s t r u c t u r a l dimensions. Further, t h i s expect-ation suggests t e l e v i s i o n viewing i s not s u b s t a n t i a l l y influenced by a desire f o r v a r i e t y which to some may be co u n t e r i n t u i t i v e . 1. Bass, Pessemier and Ti g e r t , Journal of Business, XLII. In summary, the f a c t o r a n a l y t i c r e s u l t s of the search for content dimensions are not strong. It appears that there has been i n s u f f i c i e n t concern i n the studies over either the appropriateness of the methodology or the j u s t i f i c a t i o n of various aspects of method. However, the i n a b i l i t y to f i n d strong content dimensions and the exploratory, empirical nature of the technique has led to the recognition of s t r u c t u r a l influences as im-portant to audience exposure. A Note on R e l i a b i l i t y and V a l i d i t y i n Factor Analysis H. J. Einhorn i n h i s a r t i c l e "Alchemy i n the Behavioural Sciences" summarizes methodological weaknesses i n the use of various multivariate methods including factor a n a l y s i s . ' In h i s discussion of the l a t t e r , E i n h o r n r e l i e s on two a r t i c l e s : one by J. S. Armstrong and P. Soelberg who question 2 the u t i l i t y of factor analysis for dealing with r e a l data; the other by Armstrong alone where he attacks the idea that factor analysis can be used 3 to derive theory. In the f i r s t of these a r t i c l e s Armstrong and Soelberg conducted a simulation of a study where f i f t y employees rated t h e i r supervisors on 20 t r a i t s . Pearson product moment co r r e l a t i o n s were calculated among the v a r i a b l e s and factors extracted using p r i n c i p a l components analysis. Those I. H. J. Einhorn, "Alchemy i n the Behavioural Sciences", Public  Opinion Quarterly, XXXVI ( F a l l , 1972). 2. J. S. Armstrong and P. Soelberg, "On the Interpretation of Factor A n a l y s i s , " Psychological B u l l e t i n , LXX (1968). 3. J . S. Armstrong,"Derivation of Theory by Means of Factor Analysis or Tom Swift and His E l e c t r i c Factor Analysis Machine," American  S t a t i s t i c i a n , XXI (December, 1967). factors with eigenvalues greater than 1 were rotated. It was possible to summarize 71% of the variance i n the o r i g i n a l 20 variables with only nine factors. The authors suggest that they could also i n t e r p r e t these factors and that the l i t e r a t u r e i n the area was supportive. The problem was that the responses to the o r i g i n a l 20 variables were not r e a l data but random  normal deviates. They state, "Since i t appears to be rather simple for a researcher to make sense out of the patterns provided by factor analysis, some benchmark or measure of factor r e l i a b i l i t y should be made a requirement for pu b l i c a t i o n . Unfortunately, s t a t i s t i c a l t e s t s for measuring factor r e l i a b i l i t y do not appear to be well developed."1 R e l i a b i l i t y means simply that the same factors would be obtained a new set of responses. Three a l t e r n a t i v e methods are forwarded by Armstrong and Soelberg as means to assessing r e l i a b i l i t y : (1) Cross-validation or s p l i t samples: the analysis i s conducted on d i f f e r e n t samples or, a l t e r n a t i v e l y , the sample i s s p l i t and the analysis conducted on each h a l f to see i f the r e s u l t s are the same. (2) A p r i o r i analysis: p r i o r to the analysis the researcher works out i n as much d e t a i l as possible the solu t i o n he expects to f i n d . "He might, for example, postulate the number of factors he expects to appear, which v a r i a b l e s should load together, or what variables he expects w i l l dominate which factors. His predictions could be based on behavioural models, previous findings re-ported i n the l i t e r a t u r e , or merely on well-educated hunches."2 (3) Monte Carlo Simulation: where sample sizes are so small that i t i s not p r a c t i c a l to s p l i t the sample or one has l i t t l e p r i o r knowledge about the underlying behavioural 1. Armstrong and Soelberg, Psychological B u l l e t i n , LXX, p. 362. 2. Ibid ., p. 363. 33 processes, s u i t a b l e samples of random data could be analyzed using the same procedures and sampling d i s t r i b u t i o n s obtained for comparison with obtained results.1 In addition to these possible measures of r e l i a b i l i t y , Armstrong and Soelberg argue for the d e s i r a b i l i t y of some measure of v a l i d i t y . This simply means that the factors would c o r r e l a t e with some outside c r i t e r i o n . For example, they suggest i t would be u s e f u l to specify one dependent v a r i a b l e which the factor analysis was designed to predict or explain. In the second a r t i c l e c i t e d above, a protagonist named 'Tom Swift' obtained a set of measurements on the a t t r i b u t e s of a group of p h y s i c a l objects where the underlying dimensions were already known. The idea was to see i f factor analysis led to the rediscovery of these dimen-sions. Standard conventions were used i n the a n a l y s i s : only factors with eigenvalues greater than 1 were used y i e l d i n g 3 factors summarizing 90% of the variance i n the o r i g i n a l scores, orthogonal r o t a t i o n was per-formed. However, the r e s u l t s were only vaguely interpretable and cer-t a i n l y did not conform to the known underlying dimensionality of the o r i g i n a l v a r i a b l e s . The author goes on to show that i f Tom had extracted more factors h i s r o t a t i o n would have yielded a better approximation to the known under-l y i n g f a c t o r s . The point i s to r e i n f o r c e the need for p r i o r knowledge i n defining the problem to be analyzed and i n t e r p r e t i n g the r e s u l t s . As Armstrong states, 1. The above i s adapted from Einhorn, Public Opinion Quarterly, XXXVI, and Armstrong and Soelberg, Psychological B u l l e t i n , LXX. "The point i s , however, that without a p r e s p e c i f i e d theory Swift has no way to evaluate h i s r e s u l t s The factor analysis might have been useful i n evaluating theory Tom Swift's work would have been much more valuable i f he had a s p e c i f i e d conceptual model. He would have been able to present a more convincing argument f o r h i s r e s u l t i n g theory had i t agreed with hi s p r i o r model."1 With i s o l a t e d exceptions, there has been l i t t l e systematic a p p l i -cation of r e l i a b i l i t y and v a l i d i t y checks i n the studies c i t e d e a r l i e r i n 2 t h i s chapter. It i s a major objective of t h i s study to implement such safeguards. Both cross v a l i d a t i o n ( r e p l i c a t i o n ) and external v a l i d a t i o n w i l l be conducted where the r e s u l t s of analysis and data a v a i l a b l e allow. Moreover, the next chapter i s devoted e n t i r e l y to developing a behavioural model and using i t to determine the expected r e s u l t s of factor a n a l y s i s . 1. Armstrong, American S t a t i s t i c i a n , XXI, p. 20. 2. Bruno attempts c r o s s - v a l i d a t i o n of h i s r e s u l t s on d i f f e r e n t samples (Journal df Advertising Research, X I I I ) . Bass, Pessemier and Tigert attempt external v a l i d a t i o n through regression of factor scores on external variables (Journal of Business, XLII). Ehrenberg recognizes the need to account for p r i o r knowledge but does not do so within the context of factor analysis (Ehrenberg, Journal of Advertising Research, VIII). 35 Chapter I I I DIMENSIONS OF AUDIENCE EXPOSURE TO A MASS MEDIUM The object of t h i s research i s to investigate dimensions of audience exposure to a mass medium, a d a i l y newspaper. Although the research i s ess-e n t i a l l y a study of aggregate behaviour i t rests on a model of i n d i v i d u a l be-haviour, i . e., i t i s expected that the dimensions of audience exposure can be predicted from an underlying micro-model of i n d i v i d u a l behaviour. The f a i l u r e to develop such a micro-model, as discussed i n Chapter I I , has been a fundamental weakness to the determination of dimensions of exposure i n pre-vious studies. The researcher would make some i m p l i c i t assumption about be-haviour ( i . e . , s e l e c t i v i t y according to content), but would not systemat-i c a l l y introduce t h i s into h i s analysis and as a r e s u l t he was unable to ade-quately i n t e r p r e t h i s r e s u l t s . This chapter w i l l develop a micro-model of i n d i v i d u a l s e l e c t i v e ex-posure with respect to a mass medium and use i t to construct a set of hypo-theses concerning aggregate dimensions of audience exposure. However, be-fore proceeding,a review of the l i t e r a t u r e on i n d i v i d u a l s e l e c t i v e exposure i s i n order. The Controversy Surrounding Sel e c t i v e Exposure One might l o g i c a l l y ask why i t i s reasonable to expect that i n d i v i -duals are s e l e c t i v e i n t h e i r exposure to mass communications. The simple answer i s that an i n d i v i d u a l has l i m i t e d receptive c a p a c i t i e s and as a re-s u l t w i l l expose himself to communication which provides him with u t i l i t y , i . e., that which serves a purpose or has capacity for s a t i s f y i n g some psychological or s o c i a l need. Because i n d i v i d u a l s have varying u t i l i t i e s or preferences for a l t e r n a t i v e communications, they p r a c t i s e s e l e c t i v e ex-posure i n order to maximize t h e i r u t i l i t y . This approach w i l l be pursued at a l a t e r point. The term 'selective, exposure' has t r a d i t i o n a l l y held quite a d i f f e r e n t connotation i n the f i e l d s of psychology and communications and, accordingly, a b r i e f review of the d i r e c t i o n of t h i s l i t e r a t u r e i s undertaken. Sele c t i v e exposure can be most simply defined as "any biased non-random attention"'''. What t h i s means i s that there i s d i f f e r e n t i a l attention to communications and that t h i s attention i s governed by d r i v e s , values, p r i o r reinforcements, etc. However, the term usually r e f e r s more r e s t r i c t -i v e l y to "biased attention s p e c i f i c a l l y programmed to seek out information that confirms one's preconceptions and avoid information that i s discrepant 2 with one's b e l i e f system". The acceptance of t h i s proposition i s consid-ered c e n t r a l to c o g n i t i v e consistency theories. Further, i t i s fundamental to the argument that mass communications do not e a s i l y persuade people to change attitudes and p r a c t i c e s . Communication campaigns, i t i s argued, reach the converted, while others tune out; thus mass communications re -inf o r c e but r a r e l y convert. 1. R. P. Abelson et a l . (ed.), Theories of Cognitive Consistency: Source Book (Chicago: Rand McNally & Co., 1968), p. 769. 2. Ibid. 37 A s t r a i g h t consistency theory approach to s e l e c t i v e exposure, i t can be argued, i s no longer functional for a study i n mass communications. Recent thinking among researchers within the f i e l d of s e l e c t i v e exposure supports such a viewpoint. The issue i s made unclear by the fac t that there i s l i t t l e question among researchers that s e l e c t i v i t y e x i s t s , i . e., that i n d i v i d u a l s are disproportionately exposed to communications which are con-genial to both t h e i r cognitions and att i t u d e s . This i s referred to i n the l i t e r a t u r e as 'de facto s e l e c t i v i t y ' . As a r e s u l t , research has t r a d i t i o n -a l l y focused on s e l e c t i v i t y motivated by the need for consistency and avoid-ance of dissonance among cognitions, attitudes and decisions. However, i n spite of considerable research e f f o r t , the l i n k between s e l e c t i v e exposure and the need to seek confirming information or avoid discrepant information has not been supported. Freedman and Sears i n t h e i r extensive review of the l i t e r a t u r e conclude, "Laboratory evidence does not support the hypothesis that people prefer to be exposed to supportive as opposed to nonsupportive information."! This statement has been challenged. M i l l s reviews the experimental 2 l i t e r a t u r e and to some extent disagrees with the above statement. He suggests the methodology i n many of the experiments i s inappropriate. For example, he suggests s e l e c t i v e exposure i s confounded with c e r t a i n t y , i . e . , the lower the ce r t a i n t y about correctness of choice the greater the i n t e r e s t i n supportive information. Certainty about choice i s a c e n t r a l feature of 1. J . M i l l s , "Interest i n Supporting and Discrepant Information," Theories of Cognitive Consistency, ed. R. P. Abelson et a l . (Chicago: Rand McNally and Company, 1968), p. 771. 2. Ibid ., pp. 771-776 Festinger's theory of cognitive dissonance, i . e . , lower c e r t a i n t y about choice leads to greater post d e c i s i o n dissonance and hence dissonance reduc-1 2 t i o n , possibly through seeking supportive information. ' Thus, M i l l s may be correct that experimental evidence was gathered i n ignorance of fundamental constraints on consistency theory. However, i f M i l l s i s correct his notions s t i l l undermine the generality of the consistency hypothesis by r e s t r i c i n g the circumstances under which i t i s v a l i d . A more d i f i c u l t challenge to Freedman and Sears' conclusion, how-ever, comes from Sears himself, suggesting that i f the consistency hypothesis i s rejected a paradox seems to be implied, "In nature, the patterns appear to be i n general, one of people being exposed to positions with which they already agree. The usual explanation i s that people a c t i v e l y seek supportive information and avoid non-supportive information. If t h i s hypothesis i s rejected as inconsistent with experi-mental data, what explanations of 'de facto' s e l e c t i v i t y remain?"^ Sears discusses possible explanations for de facto s e l e c t i v i t y and even challenges whether de facto s e l e c t i v i t y i s so u n i v e r s a l . He concludes that i t s existence i s probably due to the unusual a v a i l a b i l i t y of supportive information and the l i k e l i h o o d that supportiveness in nature i s correlated with other a t t r a c t i v e features of information such as t r u t h f u l l n e s s and use-f u l l n e s s . 1. L. Festinger, A Theory of Cognitive Dissonance (Evanston: Row, Peterson and Co., 1957). 2. J . F. Engel, D. T. K o l l a t and R. D. Blackwell, Consumer Be-haviour (New York: McGraw-Hill, 1967), p. 504. 3. D. 0. Sears, "The Paradox of De Facto Selective Exposure with-out Preferences for Supportive Information," Theories of Cognitive Con- sistency, ed. R. P. Abelson et a l . , p. 783. 39 Sears and Freedman i n t h e i r review of the experimental l i t e r a t u r e conclude that the consistency approach may be over-worked.'' They suggest that the motive of u t i l i t y i s important, arguing that people want information when i t answers a f e l t need or serves a p r a c t i c a l purpose. When nonsupportive i n -formation i s us e f u l people w i l l prefer i t to supportive information which i s not useful . Katz reaches much the same conclusion i n h i s review of the sur-2 vey l i t e r a t u r e . He describes the Sears and Freedman d e f i n i t i o n of u t i l i t y as something provided by any piece of information that helps i n the perform-ance of a r o l e or successful completion of a task, i . e . , a concept analogous 3 to h i s own functional approach to mass media exposure. However, one must be c a r e f u l with the use of the word ' u t i l i t y ' . As Katz suggests i n r e l a t i o n to the functional approach to s e l e c t i v e exposure, "Some kinds of exposure, however, are 'useful' p r e c i s e l y because they are 'useless' i n that they permit people to avoid, or escape, the performance of a task or r o l e . " ^ Thus, i f one poses the u t i l i t y of information as an a l t e r n a t i v e to the consistency hypothesis i n explaining s e l e c t i v e exposure he i s i n danger of arguing that an individual's drive for consistency does not provide him with u t i l i t y . In fa c t , as Katz points out, the two concepts are not at odds; one merely needs a broader concept of u t i l i t y compared to that of Sears and Freed-man. He argues the need, "to explain s e l e c t i v i t y i n terms of the functional contribution of exposure to some s o c i a l or psychological need, of which selec-1. D. 0. Sears and J.L.Freedman,"Selective Exposure to Information: A C r i t i c a l Review",Public Opinion Quarterly, XXXI (1967),pp.196-213. 2. E.Katz, "On Reopening the Question of S e l e c t i v i t y i n Exposure to Mass Communications", Theories of Cognitive Consistency,ed. R.P.Abelson et al.,pp. 788-796. 3. I b i d . , p. 792. 4. Ib i d . , p. 793 40 t i v i t y f o r the purpose of a t t i t u d e reinforcement may be only a p a r t i c u l a r case".l In other words, Katz does not deny the consistency hypothesis but subsumes i t under what might be c a l l e d the u t i l i t y hypothesis. The d i s t i n c -t i o n between Sears and Freedman's concept of u t i l i t y and the broader concept of Katz i s important.- The former may be l a b e l l e d the usefullness hypothesis and l i k e consistency hypotheses subsumed under the concept of u t i l i t y . Further support f o r the reconsideration of the r o l e of consistency theory i s provided by McGuire. Although he does not espouse Katz' u t i l i t y concept he does focus on s e l e c t i v e exposure as simply biased behaviour r e -s u l t i n g from the i n d i v i d u a l ' s l i m i t e d receptive capacity. "My own f e e l i n g i s that we are l e f t with our i n i t i a l assumption that the information encoding capacity of the i n d i v i d u a l i s quite l i m i t e d , so that considerable s e l e c t i v i t y n e c e s s a r i l y occurs. A study of such s e l -e c t i v i t y would be a f r u i t f u l f i e l d of research but i t seems that defensive avoidance i s not a very powerful f a c t o r i n i n d i v i d u a l s e l e c t i v i t y . The time has come, I think, to turn to the broader question of which are, i n f a c t , the t a c t i c s of perceptual s e l e c t i v i t y , and discontinue the current excessive pre-occupation with t h i s one possible t a c t i c of defensive avoidance".2 Summary The s e l e c t i v e exposure l i t e r a t u r e i n psychology has been strongly grounded i n theory. In f a c t , s e l e c t i v e exposure has received less attention f o r i t s own sake than f o r i t s r o l e i n consistency theory. Recently, there has 1. I b i d . , p. 788. 2. W.J. McGuire, "Selective Exposure: A summing Up", Theories of  Cognitive Consistency, ed. R.P. Abelson et a l . , p. 800. been a movement towards recognizing that the i n d i v i d u a l has l i m i t e d receptive c a p a c i t i e s and that he seeks u t i l i t y from s e l e c t i v e exposure. This focus has ce r t a i n advantages f o r t h i s paper. I t requires l i t t l e t h eorizing concerning the i n d i v i d u a l ' s "black box" but instead focuses on behaviour recognizing that each i n d i v i d u a l has a unique set of preferences or a u t i l i t y function. It allows f u l l y for s e l e c t i v e exposure on the basis of usefullness or con-sistency. Further because of forced s e l e c t i v i t y , i . e . , the need for the i n d i v i d u a l to engage i n economizing behaviour i n exposure to communications, i t i s possible to focus on the s t r u c t u r a l aspects of communication. The im-portance of these s t r u c t u r a l influences i n mass media has been established i n the marketing l i t e r a t u r e c i t e d i n Chapter I I . This t h e o r e t i c a l s h i f t i n the s e l e c t i v e exposure l i t e r a t u r e because i t deals with choices, u t i l i t y and con-s t r a i n t s could be referred to as an economic approach to s e l e c t i v e exposure. The economic approach, allowing for both the u t i l i t y of content as well as structure i n the s e l e c t i v e exposure to media, conveniently integrates the t h e o r e t i c a l d i r e c t i o n of the s e l e c t i v e exposure l i t e r a t u r e with the empirical d i r e c t i o n of the marketing l i t e r a t u r e . As a f i n a l comment, i t should be noted that i n the marketing studies c i t e d there was no..recognition of the notion that i n d i v i d u a l s may seek supportive information or avoid discrepant information. Indeed t h i s i s pro-bably appropriate f o r t e l e v i s i o n because of the general noncontroversial type of programming. Such naive empiricism may have prevented media researchers from getting sidetracked into investigations of the consistency hypothesis. 42 Selec t i v e Exposure to a Mass Medium: An Economic Approach The above discussion concluded that an e f f e c t i v e means of analyzing i n d i v i d u a l s e l e c t i v e exposure was through an economic approach. The c e n t r a l features of such an economic model are that i n d i v i d u a l s have l i m i t e d recep-t i v e c a p a c i t i e s , have unique preferences or varying u t i l i t i e s f o r selected information and seek to maximize t h e i r o v e r a l l u t i l i t y . Information which has u t i l i t y can be defined to include that which i s useful or i n t e r e s t i n g to the i n d i v i d u a l , that which i s consistent with h i s at t i t u d e s , cognitions or decision; i n fact any information which i s perceived to f u l f i l l a psychologi-c a l or s o c i a l need. This u t i l i t y i s provided by the varying content of a mass medium. Media can also be thought of as providing time and place u t i l - . i t y . These are features offered by the structure of a p a r t i c u l a r medium. The two kinds of u t i l i t y , that associated with message or program content and that associated with media structure, can be thought of as being p o s i t i v e and negative r e s p e c t i v e l y . Hence an i n d i v i d u a l may perceive p o s i -t i v e u t i l i t y or value i n c e r t a i n information but also perceive negative u t i l -i t y or cost i n exposing himself to i t . As an example, he may wish to watch the 11:00 p.m. National News on t e l e v i s i o n for the perceived value of the news but be deterred by the time at which i t i s broadcast, the perceived cost. Although structure could be thought of as providing p o s i t i v e u t i l i t y i t ess-e n t i a l l y represents the cost i n time the i n d i v i d u a l "pays" i n order to expose himself to c e r t a i n content. Structure which enables an i n d i v i d u a l to f i n d valuable content quickly and e a s i l y merely minimizes the negative u t i l i t y associated with the exposure. The problem now i s to arrange these con-cepts of p o s i t i v e and negative u t i l i t y into a model of i n d i v i d u a l behaviour. 43 Engel, K o l l a t and Blackwell i n t h e i r general model of i n d i v i d u a l consumer decision processes have developed a paradigm which i s somewhat anal-ogous to the above situation.''" E s s e n t i a l l y , they are concerned with the amount of search a consumer w i l l undertake while gathering information i n or-der to make a purchase. They suggest that the two important elements i n such a d e c i s i o n are the perceived value of the search as measured by the value of the expected purchase and the perceived cost of the search as measured by the time and energy expended. These r e l a t i o n s h i p s are depicted i n Figure 3. 2 Figure 3: Relation Between Cost and Value of Search Perceived Value High Low Perceived High I II Cost Low I I I IV The authors explain the a p p l i c a t i o n of t h i s model as follows: "Search would be most u n l i k e l y i n those s i t u a t i o n s corre-sponding to c e l l I I , where value of search i s low and cost i s high. C e l l IV would be the next most u n l i k e l y s i t u a t i o n to produce search because both the cost and value of search are low. Which of the remaining two s i t u a t i o n s i s most l i k e l y to produce the greatest amount of search i s unclear. It would seem that search would be most intense i n c e l l I I I , where the value of search i s high and the cost i s low. How-ever, there i s considerable evidence that r e l a t i v e l y intense search occurs i n conditions characterizing c e l l I. Thus whether or not cost i s i n fact a deterrent to search when perceived value of search i s high i s yet to be established." 1. Engel, K o l l a t and Blackwell, p. 380-383. 2. I b i d . , p. 381 3. Ibid. 44 This study, of course, i s not concerned s p e c i f i c a l l y with search a c t i v i t y per se but rather with the perceived value and perceived cost of the information r e s u l t i n g from that search a c t i v i t y . In p a r t i c u l a r the study focuses on the degree to which s e l e c t i v e exposure w i l l r e s u l t from the i n t e r r e l a t i o n s h i p s between perceived value and perceived cost of the i n -formation. However, i f Engel et a l . ' s model i s recast i n terms of the selected information (rather than purchase), the r e s u l t i s a useful concept-u a l i z a t i o n of the trade-off between p o s i t i v e and negative u t i l i t y i n selec-t i v e exposure. The Engel model i s redefined i n Figure 4. Figure 4: S e l e c t i v e Exposure to a Mass Medium as a Function of Value  and Cost of Information Perceived Value of Selected  Information High Low Perceived Cost of High . I II Selected Information Low I I I IV The set of expectations f o r the s e l e c t i v e exposure model (Figure 4) i s somewhat d i f f e r e n t from that of the Engel model (Figure 3). F i r s t , con-sider the l e f t column where the perceived value of selected information i s high, i . e . , the .content of a mass medium provides high p o s i t i v e u t i l i t y . In c e l l I where the perceived cost of selected information i s also high, i . e . , the structure of the medium causes high negative u t i l i t y , s e l e c t i v e ex-posure i s a function of both the content and structure of that medium. In c e l l I I I where the perceived cost of selected information i s low, i . e . , the structure of the medium causes only minimal negative u t i l i t y , s e l e c t i v e ex-posure i s a function of only the content of the medium. The expected u t i l i t y of s e l e c t i v e exposure i s maximized. Second, consider the righ t column of Figure 4 where perceived value of selected information i s low. In c e l l I I , where the perceived cost of selected information i s high, s e l e c t i v e exposure i s expected to be zero or due to random sources. The u t i l i t y of s e l e c t i v e exposure i n " c e l l II i s minimized. In c e l l IV, where the perceived cost of selected information i s low, i . e . , the structure of the medium causes minimal negative u t i l i t y , s e l e c t i v e exposure i s a function of only the medium's structure. Notice that each c e l l of the model predicts how an i n d i v i d u a l w i l l behave when faced with a p a r t i c u l a r combination of content and structure i n a medium. It p r e d i c t s , for example, i n c e l l I I I that he w i l l s e l e c t i v e l y expose himself to content of high perceived value. The model implies that the i n d i v i d u a l i s prototypic, i . e . , a l l i n d i v i d u a l s are economizers who maximize u t i l i t y through s e l e c t i v e exposure. However, i t does not suggest that every i n d i v i d u a l s e l e c t s the same content, only that he i s s e l e c t i v e with respect to content. He se l e c t s content which i s of high perceived value to him. The model i s independent of any p a r t i c u l a r medium, assuming only that i t has both content and structure. Now the important question i s how does t h i s model r e l a t e to aggre-gate dimensions of audience exposure. An i n d i v i d u a l ' s s e l e c t i v i t y i s de-pendent on what the p a r t i c u l a r medium of i n t e r e s t makes a v a i l a b l e to him. To predict aggregate dimensions of audience one must then introduce a medium with the expectation that i n d i v i d u a l audience members react to varying con-tent and structure i n s i m i l a r ways. An example would best i l l u s t r a t e t h i s point. One may hypothesize that for a p a r t i c u l a r medium such as t e l e v i s i o n , 46 there are, from the perspective of the i n d i v i d u a l , a few basic program types represented by a large number of programs. It can be further assumed, for purposes of example, that structure provides a minimum of negative u t i l i t y and thus each i n d i v i d u a l ' s behaviour w i l l be characterized by c e l l I I I ( i n -d i v i d u a l audience members select programs of high perceived value). However, the hypothesis indicates there are only a few basic program types from the i n d i v i d u a l ' s perspective and thus i n d i v i d u a l s who s e l e c t a program with p a r t i c u l a r content also tend to select other programs of s i m i l a r content. They may, of course, select d i f f e r e n t programs but exposure to programs of s i m i l a r content w i l l be more highly correlated across i n d i v i d u a l s . ' Hence the dimensions of aggregate exposure w i l l r e f l e c t underlying s i m i l a r i t i e s i n the perceived value of content. The model of Figure 4 suggests nothing about which aggregate dimensions of audience exposure to expect with respect to content or structure. This depends on the medium of i n t e r e s t and hypotheses concerning audience exposure to that medium. what use f u l l n e s s , then, has the model? F i r s t , i t sorts out the influence of content and structure. If the medium has known or hypothesized content or s t r u c t u r a l q u a l i t i e s the model char-a c t e r i z e s the i n d i v i d u a l ' s behaviour and constrains the misinterpretation 1. Note that i f there were as many program types as programs,i.e., i n d i v i d u a l s did not perceive program types but saw a l l programs as d i f f e r e n t according to content, then aggregate dimensions of audience exposure would r e f l e c t i n d i v i d u a l programs. In fact these would not be mutually exclusive r e s u l t s . Both sets of aggregate dimensions could exist depending on the degree of c l u s t e r i n g forced on the variables (programs analyzed). What i s important i s to v e r i f y the dimensions of audience exposure which are of i n -terest - hence, the importance of an hypothesis. Without an hypothesis to guide i n v e s t i g a t i o n an i n f i n i t e v a r i e t y of dimensions of audience exposure could be extracted from the data (using factor analysis) and there would be few c r i t e r i a with which to evaluate the r e s u l t s (see J. S. Armstrong). The l a t t e r would be appropriate only when one knows l i t t l e about h i s data and wishes to dredge i t for possible interpretable structure. of any aggregate dimensions of exposure as one or the other. Second, the model contends that only content provides p o s i t i v e u t i l i t y . Structure provides only degrees of negative u t i l i t y and thus merely ei t h e r supports or confounds s e l e c t i v e exposure to content. Thus, Ehrenberg's i n t e r p r e t a t i o n of Swanson's factors being network factors can be dismissed since networks do not provide p o s i t i v e u t i l i t y to t e l e v i s i o n viewers but only minimize negative u t i l i t y a r i s i n g from structure. Third, the model suggests a methodology for determining the degree to which the content and structure of a medium deter-mine the aggregate dimensions of exposure. For example, i f i n fact such dimensions are content and structure determined, i . e . , the i n d i v i d u a l ' s be-haviour i s characterized by c e l l I, the input to the model can be alt e r e d to eliminate e i t h e r s t r u c t u r a l or content influences. That i s , i f i t i s expect-ed that structure i s confounding the ana l y s i s , t h i s can be eliminated by the appropriate s e l e c t i o n of variables which are not d i f f e r e n t i a l l y influenced by structure. Thus s e l e c t i v e exposure as a function of high perceived cost of selected information i s eliminated, i . e . , the i n d i v i d u a l ' s behaviour i s characterized by c e l l I I I , and any dimensions of audience exposure should strongly r e f l e c t content influences. S i m i l a r l y , one can eliminate that source of variance due to content with the r e s u l t that the i n d i v i d u a l ' s be-haviour i s characterized by c e l l IV and any dimensions of audience exposure w i l l strongly r e f l e c t s t r u c t u r a l influence. Notice, however, that the above i n t e r p r e t a t i o n may represent some-what of an o v e r s i m p l i f i c a t i o n when applied to t e l e v i s i o n viewing. The perceived cost of information may be affected by a number of factors as indicated i n Chapter I I , i . e . , broadcast time, network, day of the week, etc. These have been aggregated into a general perceived cost dimension. If one i s i n v e s t i g a t i n g s e l e c t i v e exposure as a function of content, these struc-t u r a l influences w i l l confound that search to varying degrees. To handle t h i s problem, one could select only those t e l e v i s i o n programs shown on the same network, thus eliminating one major source of s t r u c t u r a l confounding. Further, one could select from t h i s set only programs shown i n prime time thus eliminating another source of s t r u c t u r a l confounding. Presumably, any dimensions of audience exposure r e f l e c t i n g program types would become strong-er as these confounding s t r u c t u r a l influences were eliminated. Taken to the l i m i t , a l l sources of s t r u c t u r a l influence could be eliminated leaving a set of t e l e v i s i o n programs completely co n t r o l l e d f o r s t r u c t u r a l e f f e c t s . Thus, one could f i n a l l y test whether or not dimensions of audience exposure r e f l e c t e d content s e l e c t i v i t y , or were uninterpretable, or s t i l l r e f l e c t e d sources of variance unaccounted f o r . However, i n t h i s case the number of programs being analyzed i s l i k e l y to be s u f f i c i e n t l y small to render the need f or a multivariate technique such as factor analysis superfluous for data analysis. As suggested e a r l i e r , the d i f f i c u l t y i n con-t r o l l i n g f o r s t r u c t u r a l e f f e c t s seems to have been Ehrenberg's l o g i c i n abandoning factor analysis i n favour of analysis of c o r r e l a t i o n s . "*" In summary, the model of s e l e c t i v e exposure suggests the following procedure i n the search for dimensions of audience exposure. (1) e s t a b l i s h as c l e a r l y as possible what major sources of variance are expected to influence i n d i v i d u a l s e l e c t i v e exposure, e.g., program content, network, day of week, etc. 1. p. 19. 49 (2) select those sources of variance or combinations thereof which represent a research question of i n t e r e s t , e.g., are there t e l e v i s i o n program types? (3) select data and conduct analysis to v e r i f y these sources of variance as determinants of audience exposure e l i m i n -a t i n g other major sources of variance which may confound the r e s u l t . The dis c u s s i o n now turns to the a p p l i c a t i o n of the model to the readership of a d a i l y newspaper. Dimensions of Audience Exposure to a D a i l y Newspaper Individual s e l e c t i v e exposure to the i n t e r n a l content of a d a i l y newspaper can be simply conceived as a function of the perceived value of i n -formation and the perceived cost of f i n d i n g information. The object of t h i s study i s not to test t h i s proposition but rather to use i t i n b u i l d i n g a set of expectations about the dimensions of aggregate audience exposure to the newspaper. To t h i s end, the following discussion w i l l develop a set of working hypotheses which both guide the a n a l y t i c procedure of Chapter V and provide a means of evaluating the r e s u l t s of that procedure. A working hypothesis within the confines of t h i s study i s a pro-v i s i o n a l conjecture or formulation to guide the i n v e s t i g a t i o n . It i s not a rigorous hypothesis i n the s t a t i s t i c a l sense and should not be interpreted as such. However, the working hypotheses are grounded i n the l i t e r a t u r e c i t e d i n Chapter II and the above economic model of i n d i v i d u a l s e l e c t i v e exposure. Further, they represent n o n s t a t i s t i c a l predictions concerning the expected r e s u l t s of data analysis. The precise r o l e of the working hypotheses w i l l become more cl e a r i n the dis c u s s i o n of a n a l y t i c procedure i n Chapter V. 50 It should be emphasized at t h i s point that the object of t h i s study i s not p r i m a r i l y to investigate the mechanisms of i n d i v i d u a l s e l e c t i v e ex-posure. I t i s e s s e n t i a l l y to explore aggregate behaviour s i m i l a r to the factor a n a l y t i c studies of Chapter I I . However, i t was emphasized i n that chapter and i n the previous section that there i s not a unique s o l u t i o n to the f a c t o r i n g problem. Accordingly, one needs to investigate the existence of audience dimensions which are of some p a r t i c u l a r i n t e r e s t . In t h i s study, as has been discussed, these are simply i n t e r n a l newspaper vehicles which are open to manipulation by newspaper management and/or advertisers. It i s necessary at t h i s point to separate the analysis of news data from that of adv e r t i s i n g data. The d i s t i n c t i o n i s made on the basis that advertising i s defined as any sponsored communication and news i s anything other than adve r t i s i n g . There i s an important reason for t h i s separation. News and advertising content generally have a d i f f e r e n t set of managerial objectives and dimensions of audience exposure may vary accordingly. It i s fundamental to t h i s study that dimensions of exposure be established as a function of the news content and s t r u c t u r a l organization of the newspaper, that i s , the newspaper's configuration of content and structure. I f t h i s can be established, then the adv e r t i s i n g data can be investigated using such a configuration. I f the advertising data were not to follow t h i s pattern i t would l i k e l y d i s t o r t the i n v e s t i g a t i o n of the news data, i f both news and ad v e r t i s i n g were analyzed together. S i m i l a r l y , the news data might d i s t o r t any i n t e r e s t i n g underlying dimensionality to the adv e r t i s i n g data. Hence the most prudent course i s to consider these major subsets of data separately and use the separate r e s u l t s for comparison purposes. 51 News Content Consider the i n d i v i d u a l newspaper reader. I t i s expected that t h i s i n d i v i d u a l wishes to maximize h i s u t i l i t y i n exposure to news content within the paper. Because he has a unique set of preferences ( u t i l i t y function) there i s high perceived value i n exposure to selected content, that i s , he i s predisposed to c e r t a i n information because i t i s consistent with h i s a t t i -tudes, i s us e f u l , i n t e r e s t i n g or i n some way provides him with u t i l i t y . Further, he w i l l seek to minimize perceived cost which i n t h i s case i s the amount of time spent and f r u s t r a t i o n experienced while searching the news-paper f o r information. The news content of d a i l y newspapers, at le a s t the one used i n t h i s study, i s p a r t i c u l a r l y suited to a high degree of s e l e c t i v e exposure. Such content i s thought by management to f a l l i nto c e r t a i n categories which pro-vide d i f f e r e n t i a l u t i l i t y to audience members. For example, sports or business would normally be thought of as such categories.' The structure of most newspapers usually coincides with such content categories thus the news content i s usually w ell understood to f a l l i nto contiguous content and s t r u c t u r a l categories for the convenience of s e l e c t i v e exposure of audience members. This i s e s s e n t i a l l y the configuration of news content and structure with which the i n d i v i d u a l i s presented. C l e a r l y , such a configuration i s represented by c e l l III of the economic model: high perceived value and low perceived cost of selected information. 1. Such a content c l a s s i f i c a t i o n of news content i s s i m i l a r to an a p r i o r i s p e c i f i c a t i o n of program types. The c l a s s i f i c a t i o n procedure i s c l e a r l y dependent on the data. The reader must accept t h i s pending d i s -cussion i n Chapters IV and V. 52 The content/structure categories of news content can be thought as a p r i o r i 'managerial sections'. It i s expected, as a r e s u l t of the above argument, that dimensions of audience exposure are determined by these sections. For example, where c e r t a i n news content can be c l a s s i f i e d as sports and t h i s i s contained i n a recognizable s t r u c t u r a l section, i t i s hypothesized that there i s s e l e c t i v e exposure to sports content and th i s w i l l emerge as a dimension of the exposure to news content. This i s not a t r i v i a l expectation. Dimensions of audience exposure i n t h i s study are empirically determined which means that i f there i s not sub s t a n t i a l s e l e c t i v e exposure to sports there w i l l not be a sports dimension. In essence, one determines a p r i o r i the managerial sections of the news content and tests whether or not these are r e f l e c t e d as dimensions of audience exposure. However, besides the managerial structure there i s a s t r u c t u r a l i n -fluence due to time, that i s , i f a number of issues of a newspaper are con-sidered simultaneously instead of a single issue, i t i s expected that ex-posure to content w i l l vary from issue to issue. Thus when considering time structure, which confounds exposure to selected content, c e l l I of the model i s appropriate. There are then three primary sources of variance to news exposure: that due to content, that due to managerial structure and that due to time ( i . e . , exposure w i l l vary from issue to i s s u e ) . There are other possible sources of variance such as right or l e f t page or p o s i t i o n on the page but these are considered minor r e l a t i v e to the above. With respect to the news data,however, managerial structure and content coincide, thus there are only two p r i n c i p a l sources of variance. I t i s of i n t e r e s t then to inve s t i g a t e whether or not dimensions of aggregate audience exposure r e f l e c t that source of variance which provides d i f f e r e n t i a l p o s i t i v e u t i l i t y to audience members-the managerial content/structure sections. These sections are known to per-s i s t over time (across issues) which s t i l l allows for a high degree of con-tent s e l e c t i v i t y . The model can be described as follows, Perceived Value of Selected Information  Across Managerial Sections High Low Perceived Cost of High I II Selected Information Across Issues Low III IV C l e a r l y , the i n d i v i d u a l i s i n c e l l I: both time structure and con-tent as represented by the managerial sections a f f e c t s e l e c t i v e exposure.'' However, i t i s only the l a t t e r which provides p o s i t i v e u t i l i t y . Hence the following working hypothesis can be defined: H^: The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper over time are determined by the managerial content/structure sections. Note that t h i s hypothesis predicts the p r i n c i p a l , most important dimensions of audience exposure. While i t i s expected that they would 1. Notice that unlike the general model, t h i s model does not s t r i c t -l y contrast content with structure. In f a c t , a true representation of the s i t u a t i o n here i s three-dimensional: 1 content dimension and 2 s t r u c t u r a l dimensions. However, one s t r u c t u r a l dimension supports content s e l e c t i v i t y . Thus the 3 dimensions can be collapsed i n t o 2 where the content/structure dimension represents content. The legitimacy of t h i s procedure i s v e r i f i e d under working hypothesis H£ discussed below. account for a substantial porportion of the variance i n the o r i g i n a l v a r i -ables, i t i s not expected that they w i l l account for a l l such communality. If i t i s v e r i f i e d , working hypothesis suggests two subsidiary hypotheses. For the f i r s t subhypothesis, the confounding influence of time structure can be eliminated, that i s , only single issues of the newspaper are considered. This model i s as follows: Perceived Value of Selected  Information High Low Perceived Cost of Select- High I II ed Information Across Managerial Sections Low III IV Under H^ i t was assumed that managerial structure and content were coincident. With the elimination of time, s e l e c t i v e exposure as a function of managerial content can be s p e c i f i c a l l y investigated. This i s described by c e l l I I I , i . e . , the structure supports content and hence there i s low perceived cost to selected information. H2: The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper i n a single issue are determined by the managerial content categories. If t h i s hypothesis were confirmed the r e s u l t s w i l l r e p l i c a t e those of H^ but be much stronger, that i s , the r e s u l t s w i l l account for sub-s t a n t i a l l y more variance and exhibit clearer factor loadings. 55 The second subhypothesis would eliminate content as a source of var-iance, that i s , only content that f e l l into a p a r t i c u l a r category over time would be analyzed. Thus the perceived value of selected information would be low. However, the dominant s t r u c t u r a l influence, time, suggests that the perceived cost of selected information over the issues would also be low, i . e . , c e l l IV. This model i s as follows: Perceived Value of Selected  Information High Low Perceived Cost of High I II Selected Information Across Issues Low III IV H^: The p r i n c i p a l dimensions of audience exposure to a p a r t i c u l a r category of news content over time are determined by the various issues of the news-paper . If t h i s hypothesis i s confirmed, i . e . , the various issues determine dimensions of audience exposure, i t suggests l i t t l e or no content s e l e c t i v i t y within a p a r t i c u l a r content category. Note that with neither ^ and H^ w i l l the dimensions account for a l l the variance i n the o r i g i n a l v a r i a b l e s . There are c e r t a i n to be i n -fluences with respect to p o s i t i o n i n g on the newspaper page as mentioned e a r l i e r and a c e r t a i n random influence. In the case of H^ there i s the strong p o s s i b i l i t y of content dimensions within p a r t i c u l a r content categories. There may be s e l e c t i v e exposure to c e r t a i n content over time within the managerial content/structure sections. It i s merely expected that these are 56 r e l a t i v e l y minor sources of variance. Advertising Content As discussed i n Chapter I, within the context of advertising re-search the purpose of e s t a b l i s h i n g the aggregate dimensions of audience exposure to a mass medium i s the i d e n t i f i c a t i o n of advertising v e h i c l e s . The purpose of analyzing the advertising content of a d a i l y newspaper then can be stated as follows: to determine the extent to which the managerial content/ structure sections of the news content (herein referred to as managerial sections) act as vehicles f o r s e l e c t i v e exposure to advertising. The analysis of advertising content i s a more complicated task than analysis of news content. There could w e l l be underlying dimensionality to adv e r t i s i n g content aside from any s t r u c t u r a l considerations, i . e . , high perceived value to selected advertising content. Further, a d v e r t i s i n g content i s s t r u c t u r a l l y organized i n a complicated way. Much of i t i s not organized to the advantage of s e l e c t i v e exposure on a content basis. That i s , much of i t i s scattered about the newspaper and i f an i n d i v i d u a l were content s e l e c t i v e , he might be frustrated by the i n a b i l i t y to f i n d i t e a s i l y . On the other hand, a considerable amount of adv e r t i s i n g i s known to be found i n c e r t a i n sections or, p a r t i c u l a r l y i n t h i s study, on the back pages of "p h y s i c a l " sections. Thus, i f the advertising content of such p o s i t i o n i n g i s reasonably consistent, s e l e c t i v e exposure on a content/structure basis could r e s u l t . These complications require that the problem and data to be i n v e s t i -gated be c a r e f u l l y delimited. It should be r e i t e r a t e d that the primary object of t h i s study with respect to advertising i s the i n v e s t i g a t i o n of the managerial sections as ad v e r t i s i n g v e h i c l e s . This objective immediately allows the elimination of a l l advertising which does not f a l l within these sections, such data being i r r e l e v a n t to the purpose at hand. More s i g n i f i -cantly, an assumption i s required that the perceived value of selected a d v e r t i s i n g content i s low, i . e . , i t i s not suggested that a d v e r t i s i n g content provides low p o s i t i v e u t i l i t y but only that selected advertising content does so. What does t h i s mean? E s s e n t i a l l y i t implies that the i n d i v i d u a l i s s e l e c t i v e l y exposed to advertising as a r e s u l t of his selec-t i v e exposure to news content, i . e . , he reads advertising because i t i s proximate to selected news content or i t f a l l s within a p a r t i c u l a r manager-i a l section. Hence the i n d i v i d u a l ' s exposure to selected advertising con-tent i s determined by the managerial structure of the newspaper. This i s c l e a r l y a questionable assumption. However, i t i s c r i t i c a l to the invest-i g a t i o n of managerial structure as the primary determinant of s e l e c t i v e exposure to advertising. Rejection of the hypotheses developed using t h i s assumption should be interpreted accordingly. There are then three sources of variance with respect to the advertising data: that due to content, that due to managerial structure and that due to time. Again, there are possible sources of variance assumed to be r e l a t i v e l y minor. Similar to the news data, content and managerial structure coincide, thus there are only two primary sources of variance. However, with the news data exposure was a function of selected content which was supported or at l e a s t not confounded by managerial structure. With the advertising data, on the other hand, exposure i s thought to be a function of managerial structure which i s not confounded by selected content because of 58 the latter's low perceived value. As stated previously, managerial structure (as determined by the news data) i s known by the individual to persist over time, i.e., across issues. Recall that only content i s capable of providing positive u t i l i t y . If the managerial sections minimize negative u t i l i t y in exposure to content then time would be expected to confound the process; thus there would be low perceived cost of selected information across managerial sections and high perceived cost to selected information across issues. Aggregate ..dimensions of audience exposure to selected advertising would reflect the organization of that content into managerial sections , i . e. , c e l l II in the model below.''' Perceived Cost of Selected Information  Across Managerial Sections High Low Perceived Cost of High I II Selected Information Across Issues Low III IV To reiterate, two structural influences affect exposure to selected content which i s i t s e l f assumed to have low perceived value. However, i t i s expected that there is low perceived cost to advertising content because of i t s organization into managerial sections and hence proximity to news con-1. Note that each c e l l of the model represents a possible hypothesis for investigation. Cell II i s merely appropriate for the hypotheses of inter-est in this study. tent. The following working hypothesis can be constructed: 59 H^: The p r i n c i p a l dimensions of audience exposure to the advertising content of a d a i l y news-paper over time are determined by the managerial sections (as determined by the news content). The confirmation of t h i s hypothesis tends to v e r i f y s t r u c t u r a l s e l e c t i v i t y of a d v e r t i s i n g content over time. Again, two subhypotheses are suggested. F i r s t , the confounding s t r u c t u r a l influence of time can be removed by examining data from a s i n g l e issue: Perceived Value of Selected  Information High Low Perceived Cost of Selected High I II Information Across Managerial Sections Low III IV Notice that content i s being contrasted with structure. The i n d i v i d u a l i s i n c e l l IV: low perceived value but also low perceived cost to selected 1. Note that t h i s model does not contrast content with structure. A true representation of the model i s 3 dimensional: 1 content dimension and 2 s t r u c t u r a l dimensions. However selected content i s of low perceived value and hence supports s t r u c t u r a l s e l e c t i v i t y according to managerial sections. Thus the model collapses to 2 dimensions where i t i s expected that managerial sections y i e l d greater p o s i t i v e u t i l i t y . Note t h i s does not assume advert-i s i n g provides low u t i l i t y but only that s e l e c t i v i t y with respect to a d v e r t i s i n g content provides low u t i l i t y . 60 content. Selective exposure i s on a s t r u c t u r a l basis. : The p r i n c i p a l dimensions of audience exposure to the a d vertising content of a d a i l y newspaper in a single issue are determined by the managerial sections. The second subhypothesis eliminates the managerial sections as a source of variance, that i s , only advertising content f a l l i n g within a p a r t i c u l a r managerial section over time i s considered. The perceived value of selected content i s s t i l l low but the dominant s t r u c t u r a l influence, time, suggests s t r u c t u r a l s e l e c t i v i t y according to the issues of the newspaper. Again, the s e l e c t i v e exposure of the i n d i v i d u a l i s characterized by c e l l IV. Perceived Value of Selected  Information High Low Perceived Cost of High I II Selected Information Across Issues Low I I I IV Hg: The p r i n c i p a l dimensions of audience exposure to advertising content contained within a p a r t i c u l a r managerial section over time are determined by the various issues of the newspaper. This subhypothesis i s very important. If i n fact dimensions of audience exposure r e f l e c t the issues of the newspaper, then time structure w i l l dominate any content s e l e c t i v i t y within a p a r t i c u l a r managerial section. Such a f i n d i n g would strongly imply l i t t l e s e l e c t i v i t y towards ad v e r t i s i n g on a content basis. 61 Summary This chapter has developed a set of working hypotheses concerning the dimensions of audience exposure to the news and advertising content of a daily newspaper. This has been accomplished using an economic behavioural model of selective exposure as a function of the content and structure of a mass medium. It has been argued that a daily newspaper has a particular configuration of content and structure which then determines the dimensions of aggregate audience exposure. The next chapter describes in greater detail the data bank on which the working hypotheses are to be operationalized. 62 Chapter IV THE DATA The data used i n t h i s study were c o l l e c t e d as part of a project with which t h i s w r i t e r was not involved. It has been analyzed extensively i n re-search r e l a t e d and antecedent to t h i s study. Accordingly, a complete docu-mentation of the o r i g i n a l research design, both objectives and procedures, i s already available.''' The following b r i e f l y summarizes major features of the data which are of importance ot the current study. Further, there i s a d e s c r i p t i o n of a content c l a s s i f i c a t i o n system already applied to the data which i s again useful f o r t h i s study. Measuring Instrument and Sample The data were obtained from a newspaper readership survey. Every Friday for 6 weeks respondents completed a self-administered aided r e c a l l questionnaire which included an abridged, miniaturized reproduction of Thursday's newspaper. The miniature newspaper pages were divided into quarters and respondents reported whether or not they had looked at part-i c u l a r quarter pages. (see Figure 5). The newspaper maintained normal layout and organization during the course of the survey. The respondents were r e c r u i t e d through a two-stage, 1. F. H. S i l l e r , "Newspaper Reading: A Study i n S e l e c t i v e E f f e c t s " . F igure 5.: THE LONDON FREE- PRESS, ihundoy, October 24, 1961 Salmon sauce tops omelette 4 fl™! k g . ^dumi, ^'h'f^ The letter from my sister-In-Uw .v-.-i "We will imvt in ttmi- fur Sunday brunch.'' Thai was Tine but I tract a Ht-C group from 10 to 11 on Sunday mcrningj to had lo have the brunch ready before leaving for the church. 1 bad been keen to try a salmon omelette after a fa-mous chef In a famous hotel bad served it at a targe break-fast party. Before leaving the bouse t made the Salmon Sauce which (tori mer (and between If desiredl the ome-lette and set the table. Salruoat Omeleite Saaee: One 7 1 4-ot. can B.C. salmon, nut drained. 3 La-We spoons butter. 1 table-spoons flour. 7 I cup whole tab.eipoon chopped chives (or fri'rre-drird). 1 tablespoon COOK A TURKEY LEG Delicious and Economical! lickxn meal your tomil/ > FRESH TURKEY PIECES TURK IT LEGS . . . TURKEY 1REAST . TURKEY WINGS .' chopped parsley, IH teaspoon nutmeg. OmHe«e: I egjjs. l'I Cup colli water, 2 4 truspoon salt. 12 teaspoon Tabasco (option-al i. I lea-.iMM.iK Worcester-shire sauce, about i Uble-spiMwis butter. To make Sauce melt butter, blend in flour, ih.-n stir In tnilk or cream un'il thick. Add salmon i n c l u d i n g liquid, chjves. parsley and nutmeg and heal Through starring until broken up and Mary Allen's T"-Kitchen (Jbunterf You wanted to know . 63 beat fggl u'f-il mixed and slightly .'oiiny with water, salt. Tabasco (if used) and Worcestershire sauce. Each omeietie khocld be cooked one at a tune in a heavy stope-sid-ed omeietie pan. .Mine is cast aluminum and measures S l'I" in diameter at lop and slopes to 11 1" in diameter at bottom. Melt 1 tablespoon butter in it until bubhlv anil h-glniung to brown. Im mediately add • LOOKED AT THIS PAGE YES 1 ND 2 (ooduclej by Mary Allea QUESTION Ghana and would like to send nun some baking for Christ-mas—but find it difficult lo lo-cate recipes that wilt survive Che travel period. Last Chnat-mas I sent him same Christ-mas cake and nuts a-xt bnli> sealed in two-pound collee cans. It arrived well—out my • sent cookies and cands large mixing bowl. Wurk with sour csaaa hands until the but-ter i> alt blended inio the sug-ar and fliur. This lakes at Iraii 1 nun of Heady work-Turn mil on lightly floured board. Kru-ad and pr>-ss a-.d shape iiili I you can Ic r.e dcugh to hold together frutn the warmth of your palm* ^ , With lishtiv floured rolling ! Rub Q pin on tightly routed board pepper oil out io a perfect!: level 1 Place ir Roast Beef lo Serve Fifty Twentv io 3 pounds beef nb I to 3 rollsi. 1 ounce (I table-poofwi sail. 1 teaspoon pep-t hi c Cm Bone and roll nh or remove chin; bone and short ribs i your buictier should do roa*i.ng pan. fat th. 1 1 *'• wider rvlrr and apparently they didn't my guile on lop of dough. t M I sris a little. and when bono tip ran. lift sol portion to al-low liquid to How underneath until all it icarly softly set-Tip pan at marp angle and with spatula or rubber scrap-er, roll cjoki-1 sid* near tian-TURKEY NECKS 2 5 « * • TURKEY SACKS Cold Springs Form Stores Market Building - FRIDAY TIL • — FHONt 431-4342 M a r y Alien suggests usins a can^-^.ick-O-Larilcm pumpkin is a bolder for doughnjts lo b r t W M t J out at Halloween. Baking powdej doughnuts good Hallowfln' handouts me. I warned to make yeast doughnuts for Halloween cele-brators but Shirley said. "More oris will make baking powder doughnuts than >east doughnuts. If you want your readers to make your recipes give them baking pow.ter On : i be r i B P wiJ brown m 1 hoard that ftCO! fa stove. roll out half iiB.Mgh (leave re-maining tlaS3 refrigerator) to Vf uBOaVim Cut out three at C S a e with well-[loured do .Ca cutter. Carry prehcal.JSfter and lifting remaining half is rolled cul. Coilert scraps and them and cut them too. If you warn to. try alt balls cut from centres in; of including them* in sci Your very small Hallo-callers will like ibi travel at al! well. The parcels take IVs if 1 months lo react) him by sea mail. I would very much like to send a variety of goodies rini wrth his Christmas cake Lhi> year. Any help you can give me will bo appreciated. JC. ANSWER I am sending you by direct mail our recipes for Banana Bread-Will Travel. Christmas Tree Fruitcakes and Dark Blond Florida Fruit Cake •Jtarp kn,tc I cut enact 1 1 r «juarr» T V * I lifted up. abuui S at a lime, in a row on a spatula to transfer to un buttered cookie -Jveiv i 4' apart. I decorated each wuh ] or 4 silvers ' They could be spnnkled carefully with colored sugar or decorat-ed with little "leaves" mad* of quartered guce red or n chcr Bake I LOOKED AT THIS SECTION YES NO Sprinkle sifted long sugar ilh ecu Li^-aWiuuy"iltde o v e r * h e n * " f""1-. U rot three r aawMatvawWi ••tamJ — SAVE 6c — ASSORTED 'HONt <:.!-s: it MINCE TARTS i*4M«B-2 KOHN l l l l i l l iWKi l i l ! . ^^ I Pkg. of i SAVE Ac — SUBMARINE 6™ 43 R O L L S . 3-25 SAVE'VOV—CHOCOLATE BROWNIES... KOMEMADF. DOUGHNITS i baking powder—30) Two medium • large eggs, fi teaspoon nutmeg. *i cup each ove sugar. 14 cup oil or fat. 3 4 sold and teaspoon gratm orange rind Have i .about 1 medium orange>. J *i cups sJted all-purpose flour. 4 teaspoons baking powder, 1 teaspoon sail. LI cup milk less 1 tablespoon. In bowl beat togeth-er the eggs, nutmeg, sugar, oil and rind. 2 or 3 minute-.. Remove beaters. Sift together sifted flour, baking ponder and salt and sJt over etc miiture alternately wtth milk uLiil butter is blended. .Scrape down sides of bowl, ccver and chill thoroughly one or two Baal oil or fat m f ' pan at 1" drpth to 373 deg. rahr. (A [ be very t CSXtton becomes ttlB side. Z.^i ngM hand a large tray SaSSIatter covered with paper and transfer cooked * , Tl i " r f f ' 'o it with slotted * i » C T r " r l n"t until by blending ter. 13 cup corn syrup. 11 teaspoon vanilla, 14 te.upoon sail, scant 1 tablespoons milk and 1 pound lifted icing sugar. Note: If you want Chocolate Glaze variation, to one third of gtaie, add 2 Liblrapoons co-Coconut squares pasify encrusted Today's riZj3>rought coniai.unt SSiaran-WTapped square ancSTEJiote: "I hope o fond of make ihr _ thtiro". RfJnS M. boi'ght tr a^^ca i bakery. It was a i x C M irtSl soft pas-trj' top aij^fjum and cake (pea.iut butter cookies) One rup butter or half and cold dtp balf and good quabty in GUue mj<ae shortening, t teaspoon vanilla. 1 cup soft hut- 1 rup brown sugar, well packed, 1 cup white sugar. 1 eggs. I cup peanut bu'ter, 3 cups sifted all-purpose flour. 1 teaspoons bakuu* powder, h teaspoon salt. C r e a m butter or butter and shortening. Add sugar* gradually, c r e a m i n g well. Beat in eggs and vanilla, mix-ing thoroughly Add peanut butter and mix well Sift to-gether the flour and baking powder and salL Sur into creamed mixture. (Add 1 cup of chopped daLej or 1 cup of chopped peanuts if desired.1 Roll into small balls, about 1" in diameter, and place on greased cookie sheet. Flatten with tines of a fork that has been dipped in water. Bake at 3TS deg. Fahr. for 10 to 12 minutes . Watch closely. Makes about S dozen. 350 deg Fahr. IB minuics I turned pans and twitdwd Uiem on racks I or 3 limes duni'g baking penod Do not brnwn or burn the^ e. They should be only slightly tinged with gold. QUESTION: Our Uonelte group has been askeJ to cater to a dinner for 1W hungry men. V« would Uke to serve roast beef a.id would appreciate vour hel;>. We would also like a M M lor a cabbage salad for the same meal. Kathleen B. ANSWER: Here are recipes for Roast Beef and Cabbage Salad fur fifty. You will have to double i. In* so that bulb is in centr- of roast and away from bune i- Roast at Wfl deg. Farh un ol therrr.omeer registers that meat has been cooked as desired: Rare—140 deg. Fahr. II io 20 minutes per pound. V , d e g . Fahr.. n to » nvnutes per pound: Weil Done—170 leg. Fahr . 17 io JO minulea per Note: The centre of a large roast will continue cooking af-ter it hns been rtmcvrd from the oven CaMagaj Salad Tor Ftfly Twniy-four cups chilled shredded cabbage. ! cups mild v inegar , : ! 1 tablespoons salt. 3 teaspoons pepper. 4 tea-i : . dry mustard. 1 I cup sugar, \i CUD butter. I well-bea'en eggs, 1 cup ludit bnr.g ihe vinegar, season-ings, .ugnr and butter jus: to the boil. Sluwly i M tins hot vinegar mixture inio the beat-en eg^ s Cock, sumng until m i x t u r e thickens. Remove from heat. 9eal m cream. While it is stilt hot pour over I'-e shredded cabbage. Chill spoon almond extract. 1 egg ifor brushing top), fruit sugar (lot sprinkling lop). Tu rciake Pastry cut butter and icing su a^r with pastry blender unul sue of peas. Dniile cold water o1 in* with fork. Press heiween pjlms. Dmdi Scotch Shortbread 1967 •Our Phone Number Has Changed BUT NOT OUR LOW PRICES!! 432-6315 Is Our New Number THESE ARE OUS LOW, LOW PRICES! ROUND - SIRLOIN - T-60NE WING - RUMP OR TIP STEAKS or ROASTS » MOVIE citu've->v c 1 A N V Of1 BOIC . --Atari ..I'J ,• '-^ l " L i x i ' n ' 1 l [ . | l , ' | ( 1 ' | - < A16 B14C12 64 area-type, p r o b a b i l i t y sampling design from a parent population of a l l per-sons who l i v e d i n the home deli v e r y zone for the newspaper and who were 15 years or older. There were, i n t o t a l , 1220 respondents at the end of the survey divided into 3 panels: A, B, and C with 402, 404 and 414 respondents respectively. These panels were balanced demographically and were organized according to the de l i v e r y routes of the newspaper. The object was to make possible the d e l i v e r y of a s l i g h t l y d i f f e r e n t newspaper to members of each panel, so that c e r t a i n experimental manipulations could be conducted. These are not of i n t e r e s t to t h i s study and as such the 3 panels are somewhat of a nuisance because they d i d not receive the same newspaper. However, t h i s does represent an .unusually good opportunity for cross v a l i d a t i o n . Content C l a s s i f i c a t i o n As may beralready c l e a r , i t i s important to determine the informa-t i o n a l content of each quarter page of the questionnaire. This has been done i n previous research with the data bank under the assumption that s e l e c t i v e newspaper reading i s a function of the kinds of information that a reader chooses.'' Categories of advertising versus news were f i r s t distinguished. An advertisement was characterized as having an i d e n t i f i a b l e sponsor who paid for the newspaper space containing the message about a product,service or i n -s t i t u t i o n . A l l material not c l a s s i f i e d as a d v e r t i s i n g was c l a s s i f i e d as news. There followed a complete l i s t i n g of a l l types of information that appeared i n the research newspapers. Some categories i n the l i s t i n g were derived from the newspaper index; others were defined i n consultation with newspaper personnel. 1. I b i d . , p. 82. 65 An exhaustive cataloguing of a l l content within the newspaper yielded 13 news categories, e.g., business, government news, sports, etc. and 19 ad v e r t i s i n g categories, e.g., furni t u r e and appliances, groceries, etc. The content analysis a l l o c a t e d items from each quarter page to the above categories. There arose two d i f f i c u l t i e s . F i r s t , some content was not c l a s s i f i a b l e . Hence the quarter page where i t appeared was not c l a s s i f i a b l e and must be eliminated for purposes of analyzing content. Secondly, various quarter pages contained a l t e r n a t i v e l y c l a s s i f i a b l e items, i . e . , a quarter page might have a furni t u r e advertisement and a p o l i t i c a l a r t i c l e . Two c l a s s i f i c a t i o n schemes were developed to handle t h i s problem. Under the r i g -orous c l a s s i f i c a t i o n scheme, i n order for a quarter page to belong to a cer-t a i n category, a l l content i n the quarter must be c l a s s i f i a b l e into that cat-egory. Under the relaxed scheme i t was only necessary for such content to s u b s t a n t i a l l y predominate the quarter page. In p r a c t i c e , the relaxed scheme was only s l i g h t l y l e s s demanding than the rigorous scheme. It depended on the s i z e of the a r t i c l e or advertisement within the quarter page, the nature of i n f r i n g i n g material and the placement of the question s t i c k e r (Figure 5). A l l these had to support the c l a s s i f i c a t i o n of the quarter page. The c r i t i c a l requirement of t h i s procedure was to be consistent. Substantial i n t e r n a l comparison from page to page and issue to issue was used i n attempting to maintain consistency i n the a p p l i c a t i o n of the relaxed scheme. The r e s u l t of both content analysis schemes was that there was not enough quarter pages to support the o r i g i n a l exhaustive set of content cate-gories. As a r e s u l t of the rigorous scheme only 3 news categories were l e f t : 66 ^business, government news, sports; and only 7 advertising categories: f u r n i -ture and appliances, groceries, l i q u o r , tobacco products, women's clothing and jewellry, automobiles and accessories, c l a s s i f i e d s . Further, even these categories were not supported across a l l three panels. The relaxed scheme allowed 4 news categories and 10 a d v e r t i s i n g . Figures 6 and 7 summarize the r e s u l t s of the content a n a l y s i s . Contained i n each c e l l i s the number of quarter pages c l a s s i f i e d i n t o each content category. Because the newspaper presented to each of the three panels was somewhat d i f f e r e n t , the number of quarters i n each category varies across panels. These figures present the r e -s u l t s from the rigorous and relaxed schemes r e s p e c t i v e l y . The categories of news content are the a p r i o r i content sections or categories which are expected to determine dimensions of audience exposure as discussed i n Chapter I I I . However, t h i s research has the premis that news-paper structure also influences dimensions of audience exposure. In accor-dance with economic model of Chapter III i t i s necessary then to examine the quarter pages c l a s s i f i e d i n the above scheme and determine whether structure supports or i n h i b i t s t h e i r access by the reader. This i s part of the a n a l y t i c procedure of Chapter V. E s s e n t i a l l y each quarter page must be examined to see i f i t i s juxtaposed to other content, e i t h e r s i m i l a r or d i s s i m i l a r which would be recognized by the reader as a s t r u c t u r a l s e c t i o n . The exact s t r u c t u r a l r e -quirements are determined by the working hypothesis under i n v e s t i g a t i o n . Audience. Descriptor Variables Chapter II discussed the d e s i r a b i l i t y of e s t a b l i s h i n g the external Figure 6: Rigorous Content C l a s s i f i c a t i o n Results 67 Content C l a s s i f i c a t i o n  Category Panel A Panel B Panel C Tota l I . News 1. business 20 2. government 8 3. sports — I I . Advertising 1. f u r n i t u r e and appliances 20 2. groceries 6 3. l i q u o r — 4. tobacco products 12 5. women's clo t h i n g / j e w e l l r y 13 6. automobiles & accessories 11 7. c l a s s i f i e d s — 13 7 11 18 14 8 19 6 7 14 28 14 33 15 11 57 12 7 12 41 47 14 Totals 90 71 88 249 1. I b i d . , p. 93. 68 Figure 7; Relaxed Content C l a s s i f i c a t i o n Resultsl Content C l a s s i f i c a t i o n Category Panel A Panel B Panel C Total I . News 1. business 2. government 3. sports 4. women's world I I . Advertising 1. furniture and appliances 2. groceries 3. movie theatres 4. l i q u o r 5. cigars 6. cigarettes 7. women's clothing/jewellry 8. men's clothing/jewellry 9. automobile & accessories 10. c l a s s i f i e d s Totals 28 26 10 14 20 25 19 23 17 10 15 48 68 39 52 24 7 18 6 6 12 10 16 25 6 10 14 17 18 28 9 7 19 13 9 36 15 77 16 13 47 6 .6 39 30 70 15 177 171 178 526 1. I b i d . , p. 95. 69 v a l i d i t y of f a c t o r r e s u l t s , i . e . , that these r e s u l t s correlate with some out-side c r i t e r i a . Fortunately, the data bank has a v a i l a b l e a number of measures c o l l e c t e d on the respondents which were thought at the time to be possibly r e -la t e d to newspaper readership. Again a complete and d e t a i l e d d e s c r i p t i o n of these descriptor v a r i a b l e s i s a v a i l a b l e elsewhere.-'- The following i s a summary of those variables which are possibly u s e f u l to the study f o r the purpose of external v a l i d a t i o n : P ersonality T r a i t Measures cognitive structure s o c i a l recognition unders tanding manifest anxiety general self-confidence Opinion Related Measures newspaper r a t i n g a dvertising r a t i n g l i b e r a l i s m newspaper.coverage newspaper source 1. I b i d . , p. 271-278. • a person's desire f o r completeness and structure i n information. • a person's concern about h i s reputation and what other people think of him. • a person's wish to understand many areas -.of knowledge. a person's general l e v e l of uneasiness, concern, tension, apprehensiveness or worry. a person's b e l i e f i n h i s a b i l i t y to be generally successful. - a person's attitu d e towards the news-paper as a mass communications medium. - a person's a t t i t u d e towards various a t t r i b u t e s of newspaper advertising (favourable to unfavourable). - a person's a t t i t u d e s on a general l i b e r -alism/conservatism dimension. - a person's a t t i t u d e towards the news-- paper's news coverage. - a person's a t t i t u d e towards the newspaper as a source of information on various t o p i c s . 70 newspaper personality Leisure Interest Measures hobbies sports Demographics age sex income education a person's attitude towards the person-a l i t y image of the newspaper. a person's interest i n various hobbies, a person's interest i n various sports. 71 Chapter V ANALYTIC PROCEDURE The methocUto be employed i n t h i s research i s one of i n v e s t i g a t i n g the strength of a s s o c i a t i o n among a set of variables through the a p p l i c a t i o n of p r i n c i p a l components analysis. Unlike the marketing studies described i n Chapter II there w i l l be systematic introduction of three safety features as recommended by Armstrong and Soelberg:^ a) there w i l l be introduction of p r i o r knowledge and expectations into the anaylsis by means of the working hypotheses developed i n Chapter I I I . b) there w i l l be c r o s s - v a l i d a t i o n of r e s u l t s across three panels of respondents(replication). c) where possible there w i l l be external v a l i d a t i o n through the stepwise regression of factor scores on variables thought to be r e l a t e d to newspaper readership. Where the expected r e s u l t s cannot be v e r i f i e d an exploratory analysis using factor analysis w i l l be undertaken. The nature of t h i s i n v e s t i g a t i o n i s an unfolding one,i.e.,although factor analysis w i l l be conducted where possible i n the l i g h t of p r i o r expect-ation, the r e s u l t s may indicate a l t e r n a t i v e d i r e c t i o n s . These new d i r e c t i o n s are d i f f i c u l t to discuss without reference to p a r t i c u l a r data. How-1. See Chapter I I , Armstrong and Soelberg,Psychological B u l l e t i n , LXX, p. 363. ever, the procedure i s charted as clearly as possible i n Figure 8. The dis-cussion primarily i s concerned with the means of introducing prior expecta-tion into the factor analytic procedure. The object is to minimize a p r i o r i the amount of judgement undertaken by the analyst at each step. Data Extraction and the A P r i o r i Classification System The raw data for this study are organized into 3 panels of respond-ent scores on a series of quarter pages. Each score is either 1 or 0 re-presenting whether the respondent has or has not been exposed to the content of the quarter page in question. The f i r s t step is to extract subsets of the data that are manageable. There are two d i f f i c u l t i e s here. F i r s t , i f one attempts to analyze a l l quarter pages at once there are far too many for a typical factor analytic program. Second, i f one attempts to analyze a l l re-spondents together, the problem arises that not a l l quarter pages were pre-sented to each panel. Furthermore, the opportunity would be lost to replicate findings across matched samples. It i s reasonable therefore to conduct the analysis separately for each panel. The second step is the selection of quarter pages which have some research significance. This is of course c r i t i c a l l y important to the re-sults and explicates the use of prior expectation. The quarter pages are se-lected according to the working hypotheses of Chapter III. Suppose for ex-ample, i t is the f i r s t working hypothesis concerning news content which is of interest. Figure 8.: A n a l y t i c Procedure Working hypothesis Newspaper reach questionnaire xlr-Miniaturized newspaper page 73 3L C l a s s i f i c a t i o n system 1 Respondent by h page;scores C l a s s i f i a b l e h pages Respondent by c l a s s i f i a b l e h page scores C o r r e l a t i o n of h pages across subjects ± P r i n c i p a l component analysis with r o t a t i o n AL A p r i o r i c l a s s i f i c a t i o n system Em p i r i c a l l y determined c l a s s i f i c a t i o n system f/a + -+ a/p A B C D Bi Yes i s i t N possible to i n t e r p r e t ^factor analysis r e s u l t s No Hypothesis : independence of c l a s s i f i c a t i o n systems Ho : x2 = 0 Hi c2> 0 Accept Ho Reject Ho External T (regressic on factoi / a l i d a t i o n m analysis r scores). \ 1 Cross v a l i d a t i o n Investigate loadings mal a l t e r n a t i v e factor t r i x f o r dimensions / ^Exploratory 'ASfatLysls- • \ 1 Cross v a l i d a t i o n 74 H^: The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper over time are determined by the managerial content/structure sections. It may be necessary to r e - i t e r a t e p r e c i s e l y what i s meant by a man-a g e r i a l content/structure section. As discussed i n Chapter I I I , i t i s one with high perceived value and low perceived cost.'' More simply, i t i s a d i s -t i n c t organizational section of the newspaper. With respect to the news data, i t i s a section where most of the news content has some underlying s i m i l a r i t y , e.g., a sports section i s e a s i l y i d e n t i f i a b l e s t r u c t u r a l l y and contains news predominantly dealing with the sports. C l e a r l y , the working hypothesis has the.assumption that quarter pages content analyzed as government, sports, bus-iness and women's f a l l i nto d i s t i n c t s t r u c t u r a l sections. This assumption i s investigated i n the next chapter. The procedure i s to c o l l e c t a l l the quarter pages i n the i n d i v i d u a l issues of the newspaper which have been c l a s s i f i e d into a content category. The relaxed c l a s s i f i c a t i o n system described i n Chapter IV i s used i n order to provide s u f f i c i e n t quarter pages for analysis. The quarter pages are then checked to see i f they f a l l into s t r u c t u r a l sec-tions as i d e n t i f i e d by the newspaper index. It i s expected (and v e r i f i e d i n Chapter V) that c l a s s i f i a b l e quarter pages of news data w i l l f a l l i nto physi-ca l sections which are coincident with content categories or be c l e a r l y iden-t i f i a b l e from the newspaper index as f a l l i n g into sections. It .should be noted that i t i s not an assumption of the hypotheses concerning advertising that a l l content c l a s s i f i e d quarter pages f a l l i nto managerial sections. 1. By "low perceived cost" i s meant a minimum of expected negative u t i l i t y caused by f r u s t r a t i o n and time spent i n search for content (see Chapter I I I ) . 75 H^, then, indicates that a l l quarter pages f a l l i n g into the manager-i a l sections over the various issues of the study are to be extracted for an-a l y s i s . Notice the c l a s s i f i c a t i o n of the quarter pages into s t r u c t u r a l sec-tions has nothing to do with what respondents have reported reading. The quarter pages are selected according to a procedure developed by the research-er i n accordance with a c l a s s i f i c a t i o n scheme and the working hypothesis. No reference i s made to respondent scores. Hence i t can be l a b e l l e d the "a p r i o r i c l a s s i f i c a t i o n system". The Empirical C l a s s i f i c a t i o n System The next major step i n the a n a l y t i c procedure i s to extract from the o r i g i n a l data matrix the set of respondent scores on the quarter pages of i n -t e r e s t . Each p a i r of quarter pages can be i n t e r c o r r e l a t e d across respondents y i e l d i n g a c o r r e l a t i o n matrix of quarter pages. Usually where c o r r e l a t i o n s are calculated between dichotomously scored v a r i a b l e s a phi c o e f f i c i e n t i s used. However, t h i s i s a shortcut method which i s exactly equivalent to applying the formula for a Pearson product moment c o r r e l a t i o n c o e f f i c i e n t . Each c o r r e l a t i o n i s i n essence a measure of whether or not readers of one quarter page are readers of a second quarter page. However, c o r r e l a t i o n s based on dichotomous data represent a methodological d i f f i c u l t y for fac t o r analysis - a point r e q u i r i n g some discussion. i ) Factor Analysis of a Co r r e l a t i o n Matrix Calculated on Dichotomous Data Where the data being factor analyzed are dichotomous,the proportion 76 of cases with one value of the dichotomy for any p a r t i c u l a r v a r i a b l e i s re-ferred to as the marginal s p l i t . Such s p l i t s or proportions, i f they vary from one var i a b l e to another, can produce a r t i f a c t u a l f a c t o r s . The e f f e c t of di f f e r e n t marginal s p l i t s i s to r e s t r i c t the maximum to which the Pearson r -(phi c o e f f i c i e n t ) can range. To demonstrate t h i s point, Figure 9 presents the maximum ' r ' for a set of contrasting s p l i t s . Figure 9: Maximum Product - Moment Co r r e l a t i o n as a Function of the Marginal S p l i t s of Two Dichotomous Variables ^ Marginal  S p l i t s .1/.9 .2/.8 .5/.5 .7/.3 .9/.1 ..1/.9 1.00 .2/.8 .66 1.00 .5/.5 .33 .50 1.00 .7/.3 .27 .33 .66 1.00 .9/.1 .11 .17 .33 .51 1.00 Examination of Figure 9 indicates that the more c l o s e l y aligned are the marginal s p l i t s the closer the maximum possible range of ' r 1 i s to 1. The point i s that variables with s i m i l a r s p l i t s w i l l c o r r e l a t e more with each other than variables with less s i m i l a r s p l i t s even where they a l l measure the same thing. As a r e s u l t f a c t o r i n g data where va r i a b l e s have d i f f e r e n t s p l i t s can lead to spurious f a c t o r s . The variables which c o r r e l a t e more highly be-cause of t h e i r marginal s p l i t s may produce a d i s t i n c t factor - referred to i n 2 the l i t e r a t u r e as a d i f f i c u l t y f a c t o r / 1. R. L. Gorsuch, Factor Analysis (Philadelphia: W.B.Saunders and Co., 1974) , p. 260. .2. G. A. Ferguson, "The F a c t o r i a l Interpretation of Test D i f f i c u l t y , " Psychometrika, VI (1941), p. 323-29. 77 D i f f i c u l t y factors are recognized by examining the mean scores on var i a b l e s . When a fac t o r brings together only variables with s i m i l a r mean scores, then i t can be suspected to be a d i f f i c u l t y f a ctor. In terms of newspaper exposure, quarters which are highly read may be brought together i n a factor opposed to those which are lowly read, regardless of s i m i l a r under-l y i n g content or structure. These could be a r t i f a c t u a l or d i f f i c u l t y factors. For example, assume there are 10 quarter pages dealing with automobile ad-v e r t i s i n g but 3 of these are i n colour. Further, assume that the quarter pages i n colour have higher mean scores. The d i f f e r e n t s p l i t between colour and noncolour quarters w i l l cause t h e i r i n t e r c o r r e l a t i o n s to be lower as i n Figure 9. This may lead to separate factors i n spite of the fact that the content of these quarters as defined by the c l a s s i f i c a t i o n system i s the same. How i s the problem of d i f f e r e n t s p l i t s to be handled? It can be seen from the above example that i n the context of t h i s research the so- c a l l e d a r t i f a c t u a l factors are i n themselves i n t e r e s t i n g . The fac t that c e r t a i n quarters are more highly or lowly read despite underlying s i m i l a r i t y i s not information which should be necessarily avoided. Aside from t h i s , the i d e a l s i t u a t i o n would be to analyze only quarter pages with the same s p l i t . This i s of course impossible and some judgement may have to be made on the i n c l u s i o n or exclusion of varia b l e s . According to Gorsuch, "The best recommendation appears to be to avoid variables with extreme skews. In addition r e s u l t s should be examined with the p o s s i b i l i t y of d i f f i c u l t y factors i n mind."l 1. . Gorsuch, Factor Analysis, p. 260. 78 This problem i s pursued as part of the data analysis. Recognizing that large differences among marginal s p l i t s may a f f e c t the r e s u l t s , i t may be necessary, as indicated i n Figure 8, to remove variables from the analysis which are skewed r a d i c a l l y , i . e . , analyze quarter pages which as nearly as possible have the same mean readership (the data allows considerable f l e x -i b i l i t y i n t h i s regard - although s u f f i c i e n t quarters cannot be found which have exactly the same s p l i t , the range for most i s not very l a r g e ) . i i ) P r i n c i p a l Components Analysis P r i n c i p a l components analysis was discussed i n Chapter II. An a l t -ernative technique of factor analysis (or more appropriately, set of tech-niques) i s common factor analysis. The e s s e n t i a l d i f f e r e n c e between p r i n c i -p a l components analysis and common factor analysis i s that the l a t t e r reduces the rank of the c o r r e l a t i o n matrix by i n s e r t i n g numbers i n the diagonal which are l e s s than unity, causing fewer factors to be extracted than the o r i g i n a l number of v a r i a b l e s . The inserted numbers are c a l l e d communalities and re-present the amount of variance which a v a r i a b l e has i n common with a l l other v a r i a b l e s . The c r i t i c a l feature of common factor analysis i s the estimation of communalities. Although common factor analysis may be a legitimate means to invest-igate the data, i t can be seen from the l a s t chapter that the dimensions being investigated are not n e c e s s a r i l y expected to account for a l l the common variance. I t i s usually the case that only the most important dimensions are being investigated. Thus, for example, when i n v e s t i g a t i n g working hypothesis H,, the managerial sections over time are expected to determine the most im-79 portant dimensions of audience exposure. These dimensions w i l l account for a su b s t a n t i a l proportion of the variance among the o r i g i n a l variables, and most of the common variance or communality. However, variance due to p a r t i -cular issues may be expected to contribute to communality. This means that appropriate communality estimates are very complicated i f at a l l possible. An exception to th i s i s H2. Here, the common variance i n the o r i g i n a l var-i a b l e space i s expected to be accounted for by the predicted dimensions. One of the routine procedures could be used to estimate communalities and a common factor analysis performed.''' Overall at t h i s point a superior method of f i t t i n g a model to the data i s suppression of the number of factors extracted by p r i n c i p a l compo-nents analysis - to be discussed below. This i s consistent with discovering the factors which account f o r the largest amount of variance. P r i n c i p a l components analysis also has the feature of providing a mathematically unique so l u t i o n i n a confusion of techniques, the issue of communality estimation 2 far from having been s e t t l e d . Further i t seems to have been without ex-ception the technique used i n the marketing studies discussed i n Chapter I I . 1. This i s not pursued i n the data analysis as the differences between p r i n c i p a l components analysis and common factor analysis are not of major concern here. However, i t i s i n t e r e s t i n g to note that experi-mentation with common factor analysis on H2 led to almost the i d e n t i c a l r e s u l t s as the p r i n c i p a l components approach. 2. Some p r a c t i c a l observations on problems of communality estima-t i o n are discussed by .P. E.' Green and D. S. T u l l , Research for Marketing  Decisions (Englewood C l i f f s , N.J.: P r e n t i c e - H a l l , 1970), p. 423. 80 i i i ) Number of Factors Because p r i n c i p a l components analysis extracts a number of factors equal to the number of o r i g i n a l v a r i a b l e s a means must be found to l i m i t t h e i r number. Here the a p r i o r i c l a s s i f i c a t i o n system i s h e l p f u l . It i s ex-pected that the most important dimensions of newspaper exposure are those determined by the managerial content/structure sections. It follows that the number of factors can be suppressed to exactly equal the number of a p r i o r i sections. A varimax ro t a t i o n i s performed minimizing the number of variables i n which any one factor occurs. The r e s u l t of the procedure i s a quarter page by factor matrix of factor loadings. It i s to be expected that these factors account for a s u b s t a n t i a l amount of variance among the o r i g i n a l v a r i a b l e s . They can be interpreted according to the strength of t h e i r load-ings on the a p r i o r i c l a s s i f i e d quarters. The test i s a) to see i f the factors can be interpreted, and b) to see i f the factors match the a p r i o r i defined managerial sections. It i s expected that c e r t a i n loadings on each factor w i l l be high and others near zero. The high loadings on each factor should be associated with quarter pages of s i m i l a r a p r i o r i content and each factor should be c l e a r l y i n t e r p r e t a b l e i n terms of one of the a p r i o r i sections. A Test for the Independence of the Two C l a s s i f i c a t i o n Systems The d i f f e r e n c e between the above use of factor analysis and the normal procedure i s that here i t has been s p e c i f i e d i n advance exactly what  i t i s the v a r i a bles (quarters) have i n common. They have been selected on the basis that they do i n fact represent dimensions of exposure. The usual 81 factor a n a l y t i c method uses the d i s t r i b u t i o n of factor loadings as a t r i g g e r to i n t e r p r e t a t i o n i n a search for underlying dimensionality. Here an a p r i o r i c l a s s i f i c a t i o n system i s being tested against an empirical one. Consider, as an example, hypothetical r e s u l t s for working hypothesis ^ where the factors are expected to c l e a r l y be determined by the a p r i o r i managerial sections. Suppose there were 15 quarter pages (quarters) to be analyzed, which were a p r i o r i c l a s s i f i e d as 5 of women's content, 4 of busi -ness content and 6 of sports content. Figure 10 represents a possible factor loadings matrix where 'x' i s a high loading and 'o' i s a low loading. A quarter i s em p i r i c a l l y c l a s s i f i e d according to i t s highest loading. It re-ceives an 'x' wherever i t loads highest and an 'o' elsewhere. In t h i s man-ner the factor loadings matrix can be used as an em p i r i c a l l y determined c l a s s i f i c a t i o n system. The loading of any p a r t i c u l a r quarter page i s not of paramount i n t e r e s t i n t h i s study as any one quarter could evoke an un-usual l e v e l of exposure. What i s of i n t e r e s t i s the o v e r a l l dimensions of exposure i n the data matrix as determined by the factor a n a l ysis. The pro-cedure i s objective i n t h i s sense: a) each quarter page i s included i n the empirical factor where i t loads highest ( i n the case of a t i e a quarter page i s considered not to have been em p i r i c a l l y c l a s s i f i e d as indicated by the a p r i o r i system - a conservative judgement. b) each factor i s interpreted according to the highest number of quarters i t has of a p a r t i c u l a r a p r i o r i c l a s s i f i c a t i o n . 82 Thus i n Figure 10: Factor I i s c a l l e d a women's factor as i t includes a l l the women's quarters loading highly; Factor II i s c a l l e d a business factor be-cause i t includes predominantly business quarters; and accordingly Factor III i s l a b e l l e d sports. (Notice the factor analysis has loaded two sports quarters i n manner not expected a p r i o r i ) . Figure 10 1: A Hypothetical Factor Loadings Matrix A P r i o r i Empirical C l a s s i f i c a t i o n System C l a s s i f i c a t i o n System I II I I I KW) x(W) o o 2(W) x(W) o o 3(W) x(W) o o 4(W) x(W) o o 5(W) x(W) o o 6(B) o x(B) o 7(B) o x(B) o 8(B) o K(B) o 9(B) 0 X(B) o 10(S) o X(S) o IKS) o x(s) o 12(S) 6 o x(S) 13(S) 0 o x(S) 14(S) 0 p x(S) 15(S) 0 o x(S) Where a l l the quarters load as the a p r i o r i system p r e d i c t s , as with the women's fac t o r , there i s ' l i t t l e question that t h i s i s a women's dimension. But what of a s i t u a t i o n where a l l quarters do not load as pre-dicted? Can i t be c e r t a i n that factors II and I I I represent business and sports dimensions respectively? Again, because no p a r t i c u l a r quarter page i s of i n t e r e s t the following test can be made: are the two c l a s s i f i c a t i o n systems independent? A nonparametric t e s t can be designed which w i l l test whether or not such an overlap of quarters could have occurred by chance, that i s , given that a l l quarters are c l a s s i f i e d under each system, i s i t 83 l i k e l y that two c l a s s i f i c a t i o n systems which a l l o c a t e quarters completely independently i n fact a l l o c a t e them i n the observed manner by chance? Such a test involves the use of a simple four f o l d contingency table analysis. The dimensions of the table are the a p r i o r i and empirical c l a s s i f i c a t i o n systems. In the upper l e f t c e l l are a l l quarters which are included under both systems, i . e . , those which are both a p r i o r i c l a s s i f i a b l e and emerge with high loadings. In the upper r i g h t c e l l are a l l quarters included by the a p r i o r i system but not the empirical system. In the lower l e f t c e l l are quarters included by the empirical system but not the a p r i o r i system. In the lower r i g h t c e l l are quarters not included by either system. Figure 11 presents contingency tables constructed from the above example. Figure 11: Contingency Table Analysis  Women's: Empirical System A P r i o r i  System Business: A P r i o r i  System + — + 5 0 5 ' P = 5.' 10! 5! 10! 15! 5! 0! 10! 0! - 0 10 10 = .001 5 10 15 Empirical System + -+ 4 0 4 - 2 9 11 P = .011 6 9 15 84 Sports: A P r i o r i  System Empirical + System + 4 2 6 - 0 9 9 4 11 15 p = .011 Each working hypothesis then can be operationalized through a 2 x 2 contingency table where the n u l l hypothesis i s one of independence between the a p r i o r i and empirical c l a s s i f i c a t i o n systems. However, the s t a t i s t i c a l t e s t i n g of these tables must recognize c e r t a i n l i m i t a t i o n s . Where N, the 2 t o t a l number of observations i n the table i s large, a X t e s t i s normally 1 2 applied, corrected for continuity. I f X with 1 degree of freedom i s found to be s i g i n i f i c a n t at a low l e v e l , the independence hypothesis can be reject-ed, that i s , the c l a s s i f i c a t i o n of quarter pages under the empirical system cannot be considered independent of the a p r i o r i c l a s s i f i c a t i o n . Where N i s small the exact p r o b a b i l i t y of the p a r t i c u l a r occurrence can be calculated. However, t h i s can e a s i l y involve an extended process. The exact p r o b a b i l i t y of a p a r t i c u l a r set of frequencies i n a 2 x 2 table i s given by the hypergeometric d i s t r i b u t i o n . In i t s reduced form t h i s i s equivalent to the following formula, X 2 = NG ' . A D - B C |-N/ 2) 2 (A+B)(C+D)(A+C)(B+D) where, A A+B c - -D • C+D A+C B+D N "' P = (A+B)'. (C+D) ! (A+C)'. (B+D)! N: A: B: C C . 85 However, to tes t the p r o b a b i l i t y of two c l a s s i f i c a t i o n systems y i e l d i n g such an u n l i k e l y table of frequencies by chance one must c a l c u l a t e not only the p r o b a b i l i t y of that p a r t i c u l a r occurrence but also the t o t a l of t h i s p r o b a b i l i t y and the p r o b a b i l i t i e s of more extreme occurrences, ( c a l l e d the Fisher Test) In pr a c t i c e t h i s can become a substantial task. For-tunately, there are tables which can be used to ca l c u l a t e whether or not a p a r t i c u l a r set of frequencies i n a contengency table would be s i g n i f i c a n t at s p e c i f i e d levels.'' When exact p r o b a b i l i t i e s are not required, as i n t h i s study, the use of these tables i s appropriate. Their usefullness i s l i m i t e d however, by constraints of N ^ 30 and no marginal t o t a l greater than 15. To govern the a p p l i c a t i o n of these t e s t s , S i e g e l has developed the following set 2 of r u l e s . These w i l l be followed accordingly. 2 1. When N<40, use X corrected for continuity. 2 2. When N i s between 20 and 40, the X tes t may be used i f a l l expected frequencies are 5 or more. I f the smallest frequency i s l e s s than 5 use the Fisher test i n a l l cases. 3. When N< 20 use the Fisher test i n a l l cases. The exact p r o b a b i l i t y of each occurrence has been calculated i n Figure 11. Whereas the n u l l hypothesis would be c l e a r l y rejected i n the women's case, the p r o b a b i l i t y i n the sports and business cases i s consider-1. S. Sie g e l , Nonparametric S t a t i s t i c s (New York: McGraw-Hill, 1956) , p. 99, p. 256. 2. Ibid., p. 110. 86 ably higher. The procedure i n Figure 8 i s applicable to a l l of the content and s t r u c t u r a l hypotheses as outlined i n Chapter I I I . One need only select the quarters indicated by the working hypotheses and suppress the number of factors to equal the number of a p r i o r i dimensions. However, what i f the a p r i o r i dimensions cannot be supported i n t h i s manner? This w i l l mean that the a p r i o r i content c l a s s i f i c a t i o n system does not determine the dimensions of exposure, a s i g n i f i c a n t f i n d i n g i n i t s own r i g h t . It must be noted that t h i s does not mean the dimensions do not e x i s t , only that they are not the most important dimensions as hypothesized. Exploratory Analysis Chapter I II strongly emphasized the interdependent nature of the working hypotheses. The v e r i f i c a t i o n of H^,for example,suggested the sub-hypotheses H2 and H.j. However, i t i s H^ and H^ which are the c r i t i c a l hy-potheses. It i s these which determine whether the hypothesized dimensions of audience exposure p e r s i s t over time. If rejected the pursuit of IL, and H^ as well as H c and H, becomes superfluous. Even i f the l a t t e r were v e r i f i e d D o they would i n no way confirm deliberate s e l e c t i v i t y according to the manager-i a l sections. They serve p r i m a r i l y to confirm the respective models of s e l e c t i v e exposure. Accordingly, i f or H^ i s rejected an exploratory analysis of the rel a t i o n s h i p s among the selected variables must be undertaken. The normal procedure at th i s point i s to allow a factor analytic program to determine the dimensions or factors. Where p r i n c i p a l components analysis i s concerned t h i s usually involves the use of an arbitrary device such as interpretation of only those factors with eigenvalues greater than 1.^ Less often i t i n -volves the use of the scree test which investigates the amount of variance accounted for by an additional factor compared to previous factors. Where the rate of increase i n variance explained becomes quite gradual, the extrac-tion of factors i s terminated. The eigenvalue c r i t e r i o n and the scree test are completely arbitrary devices. They concentrate on the variance accounted for by in d i v i d u a l factors and the t o t a l factor space. In exploratory anal-ysis conducted i n th i s study these c r i t e r i a are de-emphasized. What i s considered important i s the s t a b i l i t y of certain factors as the factor, space i s expanded. The procedure continues to re l y on suppression of the number of factors regardless of variance accounted for as i n the above hypothesis t e s t -ing. However, as the size of the factor space i s altered, the investigation w i l l concentrate on those factors which remain r e l a t i v e l y stable. These stable factors are interpreted according to the researcher's prior know-ledge of the content and structure of the variables (quarter pages) involved. As with a l l exploratory factor analysis the procedure i s highly dependent on the data and further discussion w i l l be undertaken when necessary i n l a t e r chapters. 1. An eigenvalue i s equal to the variance accounted for by a factor. An eigenvalue greater than 1 merely indicates that the associated factor accounts for more variance than an o r i g i n a l variable - each o r i g i n a l variable has variance equal to 1 because of standardization. 88 Cross V a l i d a t i o n (Replication) As discussed previously, t h i s study involved three separate panels of respondents each of which were presented with a somewhat d i f f e r e n t set of variables or quarter pages. Accordingly, the r e s u l t s are not s t r i c t l y comparable as each panel was not manipulated i n p r e c i s e l y the same manner. However, the managerial sections determined by the content and structure of the news content are for the most part adequately represented by quarter pages. As dimensions of audience exposure are expected to r e f l e c t these managerial sections and not be influenced by the varying content and structure of i n d i v i d u a l quarter pages within the sections, i t i s completely l e g i t i -mate to attempt cross v a l i d a t i o n of r e s u l t s across the three panels of respondents. In fact t h i s represents, a stronger t e s t of the r e s u l t s than i f each panel were presented with exactly the same quarter pages. If the l a t t e r were the case the dimensions might be suspected to r e s u l t from an a r t i f a c t of the p a r t i c u l a r v a r i a b l e s , though t h i s suspicion should be mitigated by the use of p r i o r expectation. External V a l i d a t i o n Given any matrix of factor loadings i t i s possible to c a l c u l a t e a respondents-by-factors matrix of factor scores. A respondent's factor score under p r i n c i p a l components analysis i s a l i n e a r combination of h i s o r i g i n a l s cores,i.e., the sum of the respondent's score on the o r i g i n a l quarter pages where each quarter page i s weighted by i t s con t r i b u t i o n to the f a c t o r ( s t a n -dardized factor loading). Working hypothesis H^ (and rL,) suggests dimensions (factors) of audience exposure are r e l a t e d to high perceived value of selected content. Hence scores on these dimensions ought to be r e l a t e d to predi s p o s i -tions within the audience ('audience bias' i n terms of the s e l e c t i v e exposure 89 l i t e r a t u r e ) . That i s , i t ought to be possible to r e l a t e factor scores to various c h a r a c t e r i s t i c s of the audience. For example,where a sports dimension has been discovered, the respondent factor scores are calculated on t h i s d i -mension and association between these scores and audience c h a r a c t e r i s t i c s i n -vestigated. A high r e l a t i o n s h i p might be found between sports scores.and sex for example (high sports scores associated with male readers and low with f e -male readers). If such r e l a t i o n s h i p s were found t h i s would add external v a l -i d i t y to the existence of dimensions of audience exposure and suggest the ex-istence of audience segments .related to s e l e c t i v e exposure to i n t e r n a l news-paper content. An appropriate means for i n v e s t i g a t i n g these r e l a t i o n s h i p s i s step-wise regression a n a l y s i s , i . e . , the regression of factor scores on external variables which may be r e l a t e d to s e l e c t i v i t y i n newspaper exposure.^ Chapter IV described the a v a i l a b l e external data, a set of respondent scores on a num-ber of v a r i a b l e s that might predispose audience members to c e r t a i n content. Where there are i l l - d e f i n e d or weak a p r i o r i convictions about such r e l a t i o n -2 ships stepwise regression i s considered appropriate. This technique adds successive v a r i a b l e s to the regression equation according to the v a r i a b l e ' s a b i l i t y to account for the largest amount of incremental variance i n the equa-t i o n . The r e s u l t i s a subset of independent variables which are maximally 1. Although factor scores are often used i n regression analysis, i t i s usually as predictor v a r i a b l e s rather than c r i t e r i o n . The advantage i s to reduce the number of predictor variables and to avoid the problem of multi-c o l l i n e a r i t y through the orthogonal nature of the f a c t o r s . However,as describ-ed i n Chapter I I , Bass, Pessemier and Tigert have conducted a regression ana-l y s i s using magazine readership factor scores as the c r i t e r i o n and a set of p e r s o n a l i t y and demographic variables as predictors. 2. R. E. Frank, W. F. Massy and Y. Wind, Market Segmentation (Englewood C l i f f s , N.J.: P r e n t i c e - H a l l , 1972), p. 145, 149. 90 correlated with the dependent v a r i a b l e . The model assumed i n the development of regression analysis needs to be considered i n the context of the above a p p l i c a t i o n . I t can b r i e f l y be described as follows: F i r s t , f o r each value of any p r e d i c t o r v a r i a b l e x]_, a normal d i s t r i b u t i o n of the c r i t e r i o n v a r i a b l e y i s assumed. Second, the model i s l i n e a r , i . e . , the conditional means of a l l these normal d i s t r i b u t i o n s of the dependent v a r i a b l e f o r any p r e d i c t o r v a r i a b l e x-^ , l i e i n a s t r a i g h t l i n e with slope . And, f i n a l l y , the normal d i s t r i b u t i o n s of the dependent v a r i -able a l l have equal variances. The assumptions about the normality of the conditional d i s t r i b u t i o n s of the c r i t e r i o n v a r i a b l e are not necessary to the performance of regression a n a l y s i s . However, i f such normal d i s t r i b u t i o n s can be assumed, then the c o e f f i c i e n t s of the regression equation are also normally d i s t r i b u t e d , enabling the t e s t of various hypotheses about the universe values of the c o e f f i c i e n t s and the construction of confidence i n t e r v a l s around the c o e f f i c i e n t s . An issue i n the a p p l i c a t i o n of regression analysis using f a c t o r scores as the c r i t e r i o n i s whether or not the conditional d i s t r i b u t i o n s of such f a c t o r scores are normally d i s t r i b u t e d . The construction of f a c t o r scores, e s p e c i a l l y from dichotomous data, gives no reason to expect this to be the case. It may be that s t a t i s t i c a l inference to a universe population i s not j u s t i f i e d . Fortunately, i n t h i s study the three panels.of respondents provide the opportun-i t y f o r c r o s s - v a l i d a t i o n of the regression r e s u l t s . I f r e l a t i o n s h i p s found as a r e s u l t of stepwise regression are stable, inferences concerning audience pre-d i s p o s i t i o n s and exposure to selected content w i l l be s u b s t a n t i a l l y supported. It should be noted that audience bias i s expected to be re l a t e d to content and not structure, i . e . , that which provides p o s i t i v e u t i l i t y . Accordingly, external v a l i d a t i o n i s most important with respect to the news data where content has high perceived value. Dimensions of audience exposure, on the other hand, to a d v e r t i s i n g are expected to r e s u l t from structure. However, the managerial sections determined by the news are hypothesized to determine dimensions of audience exposure to advertising. If the working hypotheses concerning advertising are confirmed any s i g n i f i c a n t predictor v a r i a b l e s for r e s u l t i n g factor scores should at least be consistent with those of the news data. This chapter has summarized the a n a l y t i c procedure as described i n Figure 8. The object has been to minimize the amount of subjective judge-ment introduced by the analyst during data a n a l y s i s . However, many decisions are inherently dependent upon the data. For example, the s p e c i f i c regression analyses to be performed for external v a l i d a t i o n depend upon the r e s u l t s of factor analysis. The execution of subhypotheses and H^ depends on the successful confirmation of H^ as does any exploratory analysis. S i m i l a r l y H^ and H^ depend on H^. Chapters VI and VII, accordingly, not only present the r e s u l t s of data analysis but also f i l l i n a number of d e t a i l s i n the a n a l y t i c procedure. 92 Chapter VI RESULTS OF ANALYSIS ON NEWS DATA This chapter presents the r e s u l t s of analysis on the news data. As indicated i n Chapter V t h i s involves the o p e r a t i o n a l i z a t i o n of successive work-ing hypotheses in v o l v i n g d i f f e r i n g subsets of the data bank. The r e s u l t s are presented separately f o r the three panels of respondents: A, B and C. 93 Panel A Panel A - working hypothesis H-^ : The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper over time ;are determined by the managerial content/structure sections. data extraction and the a p r i o r i c l a s s i f i c a t i o n system The data analyzed under t h i s working hypothesis are :"the scores of the 402 Panel A respondents to the news content over the s i x issues, of the study. I t i s f i r s t necessary to define p r e c i s e l y the meaning of the content/structure sections with respect to the data. This was not done i n Chapter V although the c r i t e r i a were in d i c a t e d . The quarter pages have been c l a s s i f i e d according to content i n previous research. The r e s u l t s of t h i s procedure were presented i n Figures 6 and 7. However, the working hypothesis has the assumption that the newspaper i s organized i n such a way as to support s e l e c t i v e exposure to news content. I t i s necessary, then, to extract from the quarter pages of panel A a subset which not only represents the news content but also the supporting structure of the issue. Simply, t h i s means v e r i f y i n g that the content quarter pages f a l l i nto i d e n t i f i a b l e s t r u c t u r a l sections. It i s assumed that the news-paper has been organized i n t h i s way, hence the term managerial content/struc-ture s e c t i o n . Before proceeding t h i s assumption i s investigated as follows. Using the relaxed c l a s s i f i c a t i o n , system described i n Chapter IV, 78 quarter pages were extracted from the s i x issues presented to panel A repre-94 senting four content categories: business/finance (28), government (26), sports (14), women's (10). The content c l a s s i f i c a t i o n system would not sup-port any further news content categories for panel A or any other panel as indicated i n Figure 7. The device used to investigate the structure of the newspaper i s the index found on the front page. The index l i s t s the various sections of content within the paper. The procedure i s to see i f the indexed content sec-tions match the above content categories, i . e . , are content quarter pages of a p a r t i c u l a r type to be found within indexed content sections. The s i x a v a i l -able issues of the newspaper indicate a consistent s t r u c t u r a l organization. There i s a large f i r s t section dealing with l o c a l , national and i n t e r n a t i o n a l news which could be loosely l a b e l l e d a "public a f f a i r s " section. Although oc-ca s i o n a l l y p u b l i c a f f a i r s content i s found outside these pages, t h i s i s c l e a r -l y an e a s i l y i d e n t i f i a b l e s t r u c t u r a l section. Of the 26 quarter pages content, analyzed as "government" 23 f a l l within t h i s p u b l i c a f f a i r s section. More c l e a r l y , these are always s p e c i f i e d as i n the index sports, women's and f i n -ance/farm sections with associated pages. Occasionally, immediately adjacent to the finance/farm section, there are further indexed a r t i c l e s concerning business. These plus the finance/farm pages are a cl e a r s t r u c t u r a l section. A l l the respective quarter pages presented to Panel A f a l l into t h e i r respec-t i v e "sports", "women's" and "business" sections. There i s further usually a s t r u c t u r a l section which could be described as "entertainment". However, there are no news quarter pages c l a s s i f i e d thus under the content c l a s s i f i c a t i o n system. It can be concluded that of the 78 quarter pages indicated by the 95 content c l a s s i f i c a t i o n system 75 are representative of the managerial content/ structure sections. However, on a check of the content c l a s s i f i c a t i o n , one further public a f f a i r s quarter page was found to be mistakenly c l a s s i f i e d as news when, i n f a c t , i t was advertising, hence reducing the number to 74. A t o t a l of 74 out of 77 quarter pages then c l e a r l y supports the assumption that news content i s s t r u c t u r a l l y organized i n t o managerial content/structure sec-tions . These content/structure sections over the s i x issues of the study represent the a p r i o r i c l a s s i f i c a t i o n system i n the i n v e s t i g a t i o n of working hypothesis H-^ . The r e s u l t s of the a p r i o r i system are indicated on the l e f t hand side of Table 6 - 1 . the e m p i r i c a l l y determined c l a s s i f i c a t i o n system The next step i s to analyze the matrix of respondent scores to the 74 quarter pages and to c l a s s i f y the quarter pages according to re l a t i o n s h i p s among these scores. The exposure to these quarter pages i s i n t e r c o r r e l a t e d across a l l respondents of panel A. The a p r i o r i system indicates the expec-t a t i o n that the 74 quarter pages w i l l f a l l i n t o 4 d i s t i n c t categories. Accord-i n g l y , i t i s expected that 4 basic dimensions w i l l account f o r a s u b s t a n t i a l proportion of the variance among the o r i g i n a l v a r i a b l e s and, more importantly, w i l l be c l e a r l y i n t e r p r e t a b l e in. terms of the a p r i o r i system. A p r i n c i p a l components analysis was conducted suppressing the number 96 of extracted factors to 4 and a varimax r o t a t i o n performed. The r e s u l t i n g f a c t o r loadings matrix i s presented under the empirical c l a s s i f i c a t i o n system i n Table 6-1. The f a c t o r space accounted f o r 50.4%r.of the.variance i n the o r i g i n a l v a r i a b l e s . Moving down the matrix of Table 6-1, an 'x-' has been placed beside the highest loading i n each row and each 'x' loading defined according to the a p r i o r i system. Thus the factors can v e r y - c l e a r l y be interpreted as business, public a f f a i r s , sports and women's r e s p e c t i v e l y . In f a c t , inspection of the matrix indicates that only 3 of the 74 variables do not load as predicted: #19, 20, 48. Further, with the l e t t e r exceptions a l l 'x' loadings and only 'x' loadings exceed .4 which i s an a r b i t r a r y but often used cutoff mark for the i n c l u s i o n of variables to be used i n factor i n t e r p r e t a t i o n . t e s t f o r the independence of the two c l a s s i f i c a t i o n systems Althought the strength of i n d i v i d u a l loadings i s of some i n t e r e s t , the basic concern here i s with the fa c t o r analysis as a c l a s s i f i c a t i o n device. This i s the means by which working hypothesis H-^  i s operationalized as a s t a t i s t i c a l hypothesis: does the empirical c l a s s i f i c a t i o n of variables based on the readership matrix match the a p r i o r i c l a s s i f i c a t i o n based on managerial content/structure sections? As described i n Chapter V, a contingency table analysis can be con-ducted under the n u l l hypothesis of independence of the two c l a s s i f i c a t i o n systems. The contingency tables f o r each dimension are given i n Figure 12: 97 FIGURE 12: Contingency Table Analysis - Panel :A Business a/l + empirical + -26 2 28 x: 2 = 0 46 46 26 48 74  K(.l AD-BC.'.-N/2>2 : (A+B)(C+D) (A+C)(B+D) = 61.84 Pub l i c A f f a i r s . e mpirical + a/p_ Sports + LLP. + — 25 0 25 2 47 49 27 47 74 empirical + 14 0 14 1 59 60 15 59 74 X<2 = 61.65 XK2 = 61.97 98 Figure 1 2 Con't Women1 s a/p empirical + — + 10 0 1 0 - 0 6 4 6 4 1 0 6 4 74 :v2 = 6 5 . 6 9 In each of the above cases the n u l l hypothesis of independence be-tween the two c l a s s i f i c a t i o n s systems can be stated as follows: H 0 : ; X 2 = 0 Hi .X2 > 0 The c r i t i c a l value of X with 1 degree of freedom i s 1 0 . 8 3 at prob-a b i l i t y l e v e l . 0 0 1 . Thus the HQ i n each case i s rejected and i t can be con-cluded that the a p r i o r i and empirical c l a s s i f i c a t i o n systems are not inde-pendent. Hence the working hypothesis i s supported. 99 Table 6-1: Analysis of News Data - Panel A ( a l l issues) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I IV 1 (Sports/issue #1) -.24 -.16 (S./l)x .65 -.12 2 (S/l) -.23 -.17 (S/l)x .62 -.09 3 (Business/issue #1) (B/l)x-.64 -.02 .20 .04 4 (B/l) (B/l)x-.74 -.11 .19 .02 5 (B/l) (B/l)x-.59 -.16 216 .04 6 (B/l) (B/l)x-.60 -.13 .15 .05 7 (Women's/issue #1) -.09 -.06 -.02 (W/l)x .72 8 (Public A f f a i r s / i s s u e #2) -.11 (P/2) x/. 48 .12 .22 9 (P/2) -.25 (P/2)x-.55 .11 .15 10 (P/2) -.33 (P/2)x-.41 .08 .16 11 (S/2) -.15 -.25 (S/2)x .50 -.06 12 (S/2) -.23 -.29 (S/2)x .42 .12 13 (B/2) (B/2)x-.74 -.19 .03 .01 14 (B/2) (B/2)x-.75 -.13 .06 -.03 15 (B/2) (B/2)x-.73 -.11 .06 .00 16 (B/2) (B/2)x-.62 -.25 .07 -.02 17 (B/2) (B/2)x-.59 -.22 .04 .05 18 (B/2) (B/2)x-.63 -.04 .01 .08 19 (B/2) -.27 (B/2)x-.47 .07 .16 20 (B/2) -.27 (B/2)x-.48 .15 .02 21 (W/2) -.04 -.01 .03 (W/2)x .78 22 (W/2) -.06 -.16 .07 (W/2)x .76 23 (P/iss #3) -.30 (P/3)x-.68 .04 -.00 24 (P/3) -.32 (P/3)x-.68 .01 -.06 25 (P/3) -.23 (P/3)x-.69 .01 .05 26 (P/3) -.28 (P/3)x-.67 .01 .06 27 (S/3) -.15 -.32 (S/3)x .48 .02 28 (S/3) -.26 -.20 (S/3)x .58 -.05 29 (S/3) -.25 -.31 (S/3)x .47 .08 30 (B/3) (B/3)x-.62 -.28 .19 .04 31 (B/3) (B/3)x-.71 -.21 .15 .05 32 (B/3) (B/3)x-.72 -.19 .13 .05 33 (B/3) (B/3)x-.61 -.32 .14 .04 34 (B/3) (B/3)x-.60 -.31 .15 .03 35 (W/3) -.08 -.17 .06 (W/3)x .79 100 Table 6-1 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I IV 36 (W/3) -.06 .18 .04 (W/3)x .78' 37 (W/3) -.03 -.27 -.01 (W/3)x .69 38 (P/4) -.05 (P/4)x-.58 .28 .15 39 (P/4) -.07 (P/4)x-.58 .22 ^23 40 (P/4) .02 (P/4)x-.61 .29 .14. 41 (P/4) -.01 (P/4)x-.59 .28 .11 42 (P/4) -.18 (P/4)x-.47 .13 • .31 43 (S/4) -.09 -.27 (S/4)x .59 -.00^  44 (S/4) -.23 -.20 (S/4)x .42 (W/4)x .07" 45 (W/4) -.01 -.22 .05 .69" 46 (P/5) -.16 (P/5)x-.49 .34 .19 47 (P/5) -.13 (P/5)x-.50 .35 .21 48 (P/5) -.24 .26 (P/5)x .31 .26 49 (S/5) -.21 .09 (S/5)x .72 -.03 50 (B/5) (B/5)x-.70 .09 .26 \05 51 (B/5) (B/5)x-.75 .05 .22 .06 52 (B/5) (B/5)x-.76 .03 .22 .08-53 (B/5) (B/5)x-.61 .17 .24 .07 54 (B/5) (B/5)x-.69 .10 .28 .13 55 (W/5) -.04 .16 .05 (W/5)x .80 56 (W/5) -.05 .15 .07 (W/5)x .79 57 (P/6) -.07 (P/6)x-.52 .29 .22.) 58 (P/6) -.15 (P/6)x-.57 .30 .20 59 (P/6) -.11 (P/6)x-.52 .31 .26; 60 (P/6) -.09 (P/6)x-.59 .30 _ _ .•32 61 (P/6) -.09 (P/6)x-.51 .28 .36 62 (P/6) -.13 (P/6)x-.49 .29 .33 63 (P/6) -.18 (P/6)x-.53 .38 .21 64 (S/5) -.19 -.13 (S/6)x .73 .01 65 (S/5) -.24 -.12 (S/6)x .70 .09 66 (S/5) -.27 -.07 (S/6)x .69 Tor 67 (S/5) -.29 -.07 (S/6)x .66 .02' 68 (B/6) (B/6)x-.77 -.11 .18 .07 69 (B/6) (B/6)x-.80 -.08 .20 .07 70 (B/6) (B/6)x-.77 -.10 .19 '.08 71 (B/6) (B/6)x-.66 -.19 .23 .08 72 (B/6) (B/6)x-.66 -.16 .24 '.10' 73 (B/6) (B/6)x-171 -.11 .26 .07 74 (W/6) -.01 -.18 .04 (W/6)x .75 101 Table 6-1 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system i i i i n IV interpretation business public sports women's a f f a i r s p r o b a b i l i t y <.001 <.001 <.00i <.001 cumulative proportion of t o t a l variance accounted for by factor space = 50.4%. 102 Panel A - working hypothesis l i , : The principal dimensions of audience exposure to the news content of a daily newspaper in a single issue are determined by the managerial content categories. data extraction and the a p r i o r i classification system The object of this working hypothesis is to eliminate the source of variance due to time, i.e., to investigate particular issues independent-ly with the expectation that the variance w i l l be largely accounted for by the managerial content/structure sections. Two issues were selected for analysis: issue #3 and issue #6. These were chosen because they had the largest number of variables with representation from a l l sections. Issue #2 was not selected because of the misclassification of two variables as indi-cated i n Table 6-1. There were then 15 quarter pages in issue #3: public affairs (4), sports (3), business (5), and women's (3); and 18 in issue #6: public affairs (7), sports (4), business (6) and women's (1). These have, of course, already been screened against content and structural c r i t e r i a . The results of the a pr i o r i classification are indicated on the l e f t hand side of Tables 6-2(i) and 6-2(ii). the empirically determined classification system The procedure was followed as indicated in Chapter V under working hypothesis rL^. In both issues 3 and 6 the number of factors was suppressed to 4 and the resulting factor spaces presented in Table 6-2(i) and ( i i ) account for 75.8% and 73.3% of the respective total variance. Using the 'x' procedure 103 and l a b e l l i n g accordingly, the factors i n issue #3 can.be interpreted as business, public a f f a i r s , women's and sports respectively. In issue #6 bus-iness, public a f f a i r s , sports and women's factors can be Interpreted. In neither case was there any discrepancy between the a p r i o r i and empirical systems. Further, differences between 'x' and non 'x' loadings are exception-a l l y clear. test for independence between the two c l a s s i f i c a t i o n systems A 2 x 2 contingency table can be constructed for each of the fac-tors i n Table 6-2(1) and ( i i ) . As an example, Figure 13 presents such tables for each factor i n Table 6-2(i). FIGURE 13: Contingency Tables - Panel A (issue #3) business (factor 1)  empirical a/p + — + 5 0 5 - 0 10 10 •5 10 15 women's (factor 3) empirical a/p + — + 3 0 3 0 12 12 3 12 15 public a f f a i r s (factor 2)  empirical a/p + -+ 4 0 4 - 0 11 11 4 11 15 Sports (factor 4)  empirical lis: + — + 3 0 3 - 0 12 12 3 12 .15 104 C r i t e r i a c i t e d e a r l i e r i n d i c a t e that i t i s not appropriate to test f o r the i n -2 dependence of the c l a s s i f i c a t i o n s systems using X where N <20 and where ex-pected frequencies are small. Hence i t becomes necessary to c a l c u l a t e the exact p r o b a b i l i t y of each r e s u l t . However, as t h i s process i s cumbersome i t was decided to inspect at what l e v e l each r e s u l t was s i g n i f i c a n t rather than calculate exact p r o b a b i l i t i e s . Where N i s l e s s than 30 and no marginal t o t a l i s greater than 15 t h i s can be done through Siegal (p. 256). In Table 6-2(i) and a l l subsequent tables where t h i s procedure i s followed the l e v e l of s i g -n i f i c a n c e f o r the r e s u l t i s indicated '<p'. Hence i n Table 6-2(i) the prob-a b i l i t y for such concurrence between the a p r i o r i and empirical c l a s s i f i c a -tions systems under the n u l l hypothesis of independence i s l e s s than .005 for each of the 4 fa c t o r s . In c e r t a i n cases throughout t h i s chapter, i t i s not possible to calculate l e v e l s of s i g n i f i c a n c e through the use of tables. This i s usually due to marginal t o t a l s being greater than 15. In these cases exact probab-i l i t i e s are cal c u l a t e d and are indicated according to an a s t e r i s k i n the tables. P r o b a b i l i t i e s are also calculated i n cases where factors are i n t e r -pretable but the p r o b a b i l i t y of occurrence exceeds the maximum l e v e l of s i g -n i f i c a n c e as indicated i n the tables (.05). This, for instance, occurs i n Table 6 - 2 ( i i ) factor 4. Here the factor i s c l e a r l y i n t e r p r e t a b l e as a women's factor but there i s only a single v a r i a b l e with which to test the i n t e r p r e t a t i o n . The p r o b a b i l i t y of such a c r o s s - c l a s s i f i c a t i o n occuring by chance i n a t o t a l of 18 variables i s equal to .055 which exceeds the required l e v e l of s i g n i f i c a n c e . In a l l such cases, where factors can be interpreted, p r o b a b i l i t i e s are reported whether or not they exceed .05. 105 Table 6 - 2 ( i ) : Analysis of News Data - Panel A (issue #3) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I IV 1 P u b l i c A f f a i r s .24 x-.85 .09 .13 2 P .25 x-.84 .03 .16 3 P .16 x-.84 .15 .17 4 P .20 x-.82 .16 .16 5 Sports .11 -.27 .10 X .73 6 S .30 -.09 -.01 X .79 7 S .26 -.15 .11 X .76 8 Business x.78 -.17 .07 .25 9 B x.85 -.14 .08 .16 10 B x.82 -.13 .08 .15 11 B x.79 -.25 -.00 .18 12 B x.77 -.25 .02 .14 13 Women's .07 -.07 x .94 .05 14 W .08 -.08 x .94 .02 15 W .03 -.19 x .82 .11 i n t e r p r e t a t i o n business p u b l i c women's sports a f f a i r s p r o b a b i l i t y <.005 <.005 <.005 <.005 cumulative proportion of t o t a l variance accounted f o r by factor space = 75.8%. 106 Table 6 — 2 ( i i ) : Analysis of News Data - Panel A (issue #6) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i b r i system system ••>!•• II 1 T 1 . . i v 1 . Public A f f a i r s .09 X .68 -.15 .14 2 P .20 X .71 -.15 .11 3 P .18 X .78 -.14 .08 4 P .12 X .84 -.15 .09 5 P .14 X .81 -.12 .16 6 P .18 X .80 -.08 .02 7 P .21 X .78 -.23 -.08 8 Sports .15 .23 x-,77 .00 9 S .17 .19 x-.81 .04 10 S .23 .15 x-.78 -.03 11 S .27 .14 x-.77 .02 12 Business x .86 .18 -.17 -.04 13 B x .89 .16 -.15 -.04 14 B x .85 .19 -.13 -.08 15 B x .81 .22 -.20 .15 16 B x .82 .21 -.20 .16 17 B x .83 .13 -.27 .05 18 Women1s .05 .33 -.00 x .90 i n t e r p r e t a t i o n business public sports women' s a f f a i r s p r o b a b i l i t y <.005 <.005 <.005 .055* cumulative proportion of variance accounted f o r = 73.3%. * - exact p r o b a b i l i t y 107 Panel A - working hypothesis H~: The p r i n c i p a l dimensions of audience exposure to a p a r t i c u l a r category of news content over time are determined by the various issues of the newspaper. data extraction and the a p r i o r i c l a s s i f i c a t i o n system , The object of t h i s working hypothesis i s to eliminate the source of variance due to content, i . e . , to investigate p a r t i c u l a r content over time with the expectation that the variance w i l l l a r g e l y be accounted for by the d i f f e r e n t issues or time structure. This involves analyzing the business, p u b l i c a f f a i r s , sports and women's var i a b l e s separately. It was decided to investigate a l l 4 categories under the expectation that each might produce d i f f e r e n t r e s u l t s . The v a r i a b l e s are those indicated in Table 6-1 with the 3 m i s c l a s s i f i e d variables removed. The r e s u l t s of the a p r i o r i c l a s s i f i c a t i o n system are presented on the l e f t side of' Table 6 - 3 ( i ) , ( i i ) , ( i i i ) , ( i v ) . the empirical c l a s s i f i c a t i o n system In each case the number of factors was suppressed to equal the number of issues i n which the variables occur. Table 6-3 ( i ) : The f i v e factors can be interpreted as issues #6, #3, #4, #5, and #2 respectively. However, variables 3 and 12 do not load as predicted by the a p r i o r i system and i t may be inappropriate to i n t e r p r e t f a c t o r 4 as issue #5. 108 Table 6 - 3 ( i i ) : The f i r s t three factors can be interpreted as issues #6, #1 and #3. Factors 5 and 6 can be interpreted as issues #2 and #4. However, 2 of the 14 variables load i n c o r r e c t l y ; issue #5 which i s re-presented by a single v a r i a b l e i s not represented by a factor, and factor 4 i s uninterpretable according to the a p r i o r i system. Table 6 - 3 ( i i i ) : The factors can be interpreted as issues #5, #6, #3, #1 and #2 res p e c t i v e l y . Again two v a r i a b l e s , 6 and 7, do not load as pre-dicted. Further, i f one momentarily ignores the 'x' procedure, a number of loadings can be seen to exceed .4 though they are not 'x' loadings. This re-inforces the i n t e r p r e t a t i o n of factor 5 as issue #2 but suggests that factor 1 might be a more general factor, and factor 4 at least p a r t i a l l y represents issue ill. Table 6-3(iv): The s i x factors can be c l e a r l y interpreted as issues #3, #5, #2, #1, #4 and #6 respectively. A l l variables loaded as pre-dicted by the a p r i o r i system. discussion The r e s u l t s concerning working hypothesis H^ are c l e a r l y mixed. A l -though support f o r the hypothesis does seem to be there, the loadings are not unequivocal. The amount of variance explained by these factors i s also ques-tionable. For example, i n 6-3(iv) where a l l v a r i a b l e s load c o r r e c t l y and 92.2% of the variance i s accounted for by the 6 f a c t o r s , the r e s u l t i s les s encouraging with the understanding that there are only 10 variables and thus 109 on the average any one variable could account for 10% of the variance. These results await further investigation on Panel B but they suggest possible grounds for content s e l e c t i v i t y within p a r t i c u l a r content categories. n o Table 6-3 ( i ) : Analysis of Public A f f a i r s Data - Panel A a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system . system I II III IV V 1 iss#2 .26 .09 -.19 .05 x.81 2 2 .20 .22 -.17 .31 x.71 3 2. .04 .30 -.05 x.59 .40 4 iss#3 .18 x.86 -.14 .09 .09 5 3 .13 x.86 -.14 .07 .13 6 3 .18 x.82 -.13 .23 .06 7 3 .21 x.82 -.10 .13 .10 8 iss#4 ' .24 .11 x-.71 .11 .29 9 4 .22 .13 x-.70 .14 .29 10 4 .21 .17 x-,86 .14 .02 11 4 .24 .11 x-,86 .14 .02 12 4 .25 .23 -.41 x.50 -.07 13 iss#5 .38 .15 -.21 x.51 .14 14 5 .36 .14 -.30 x.59 .21 15 iss#6 x.64 .18 -.30 -.12 .27 16 6 x.65 .23 -.26 .10 .21 17 6 x.78 .15 -.13 .13 .14 18 6 x.79 .14 -.27 .14 .15 19 6 x.80 • .12 -.21 .09 .08 20 6 x.75 .12 -.12 .37 .01 21 6 x.71 .17 -.13 .14 .07 interpretation iss#6 iss#3 iss#4 iss#5 iss#2 probability <.005 .0001* .0008* .03* .014* cumulative proportion of t o t a l variance accounted for = 70.8% I l l Table 6 - 3 ( i i ) : Analysis of Sports Data - Panel A a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system sys tem I -•- I I I I I IV ". VI 1* is s # l -.30 x-.87 -.19 -.11 -.12 .10 2 1 -.26 x-.88 -.18 -.10 -.16 -.09 3 iss#2 -.14 -.45 -.02 -.31 x-.65 .04 4 2 -.22 -.07 -.25 -.02 x-.82 .21 5 iss#3 -.24 -.10 -.33 x-.77 -.14 .03 6 3 -.30 -.33 x-.68 -.20 -.13 .11 7 3 -.21 -.12 x-.82 -.21 -.16 .16 8 iss#4 -.22 -.19 , -.17 -.56 -.07 x .58 9 4 -.23 -.07 -.14 -.01 -.17 x .87 10 iss#5 x-.58 -.34 -.07 -.43 -.13 .05 11 iss#6 x-.68 -.21 -.09 -.27 -.22 .19 12 6 x-.78 -.15 -.19 -.17 -.08 .14 13 6 x-.77 -.19 -.20 -.07 -.06 .18 14 6 x-.79 -.15 -.15 -.04 -.15 .08 interpretation iss#6 iss#7 iss#3 probability <.05 <.025 '.14* cumulative proportion of variance accounted iss#2 n/s <.025 for= 78% iss#4 <.025 112 Table 6 - 3 ( i i i ) : Analysis of Business Data - Panel A a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I I I I I I IV V 1 iss# l -.15 .28 .21 x.74 .13 2 1 -.32 .30 .23 x.71 .18 3 1 .01 .43 .33 x.53 .22 4 1 .03 .42 .36 x.53 .20 5 iss//2 -.43 .11 .12 .46 x.57 6 2 x-.54 .06 .10 .41 .46 7 2 -.53 .07 .02 x.57 .44 8 2 -.14 .27 .32 .10 x.75 9 2 -.13 .26 .22 .09 x.78 10 2 -.23 .18 .06 .28 x.69 11 iss#3 -.34 .19 x.65 .29 .10 12 3 -.41 .24 x.57 .39 .11 13 3 -.47 .23 x.52 .38 .08 14 3 -.27 .18 x.79 .10 .23 15 3 -.18 .19 x.80 .16 .23 16 iss#5 x-.72 .30 .27 .12 .17 17 5 x-.80 .27 .20 .21 .15 18 5 x-.71 .36 .24 .20 .18 19 5 x-.62 .30 .31 -.00 .21 20 5 x-.62 .34 .34 .10 .20 21 iss#6 -.42 x. 66 .22 .29 .13 22 6 -.45 x.67 .21 .30 .14 23 6 -.42 x.64 .19 .31 .15 24 6 -.21 x.81 .17 .17 .22 25 6 -.22 x.82 .18 .14 .19 26 6 -.32 x.75 .15 .17 .24 interpretation iss//5 iss#6 iss#3 iss # l iss#2 probability .0001* .0000* .0000* .0003* .002* cumulative proportion of variance accounted for = 74.3% 113 Table 6-3(iv) : Analysis of Women's Data - Panel A a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system I , II II I IV V VI 1 i s s # l .27 .22 .30 x.85 -.17 .13 2 iss#2 .24 .23 x.76 .30 -.13 .24 3 2 .31 .24 x.81 .14 -.20 .14 4 iss#3 x.84 .29 .30 .14 -.12 .07 5 3 x.84 .32 .25 .12 -.12 .13 6 3 x.70 .08 .14 .29 -.27 .38 7 iss#4 .24 .30 .24 .17 x-.85 .13 8 iss//5 .31 x.84 . .22 .13 -.17 .21 9 5 .24 x.85 .23 .17 -.23 .18 10 iss#6 .25 .37 .30 .14 -.14 x.78 interpre tation iss#3 iss#5 iss//2 i s s # l iss#4 iss#6 p r o b a b i l i t y <.0i <.025 <.025 .1* .1* .1* cumulative proportion of t o t a l variance accounted for = 92.2% 114 Panel B Tables 6 - 4 , 6 - 5 and 6 - 6 represent the ap p l i c a t i o n of working hypo-theses H-^ , H.2 and H3 i n the same manner as was discussed i n d e t a i l under Pane A. However, some discussion i s made necessary due to the d i f f e r e n t news-paper content presented to panel B and, of course, the somewhat d i f f e r e n t r e s u l t s . Panel B - working hypothesis H]_: Figure 7 indicates there were 8 7 news quarter pages presented to panel B. Screening f o r s t r u c t u r a l c r i t e r i a as described under Panel A lead t elimination of 3 . Of the remaining 8 4 variables spread over 6 issues of the newspaper 1 9 were c l a s s i f i e d as women's, 2 3 as sports, 2 2 as pu b l i c a f f a i r s , and 2 0 as business. There were 4 0 4 respondents i n Panel B. The r e s u l t s of working hy-pothesis H-L are presented i n Table 6 - 4 . The fa c t o r space accounts f o r 4 4 . 9 % of the variance i n the o r i g i n a l v a r i a b l e s . S i x of the 8 4 variables are not c l a s s i f i e d as predicted by the a p r i o r i system: # 1 6 , 1 7 , 1 8 , 3 4 , 4 2 , 4 3 . The contingency table analysis i s presented i n Figure 1 4 . FIGURE 14: Contingency Table Analysis - Panel B  Women's empirical 115 a/ + — + 19 0 19 - 0 65 65 19 65 84 = 78.38 Sports empirical + -+ 22 1 23 - 2 59 61 24 60 84 jy2 = 65.3c Public A f f a i r s empirical + a/p Business j/p_ + — 20 2 22 4 58 62 24 60 84 empirical + 17 3 20 0 64 64 17 67 84 •X' -2 = 52.69 X2 = 63.04 116 Table 6-4: Analysis of News Data - Panel B ( a l l issues) a p r i o r i c l a s s i f i c a t i o n emp t i r i c a l c l a s s i f i c a t i o n system s ystem I .11 ' I I I " IV -1 (S/l) -.20 (S/l ) x .58 -.10 .19 2 (S/l) -.24 (S/l ) x .57 -.13 .19 3 (B/l) -.08 .13 -.26 (B/l)x .50 4 (B/l) -.10 .03 -.10 (B/l)x .64 5 (B/l) -.11 -.02 -.19 (B/l)x .48 6 (B/l) -.06 -.05 -.14 (B/l)x .57 7 (W/l) (W/l)x .67 .02 -.03 .02 8 (W/l) (W/l)x .59 .01 -.07 .07 9 (P/2), .14 .09 (P/2)x-.59 .09 10 (P/2) .04 .06 (P/2)x-.56 .12 11 (P/2) .09 .06 (P/2)x-.62 .05 12 (P/2) .04 .09 (P/2)x-.60 .11 13 (P/2) .06 .07-, (E/2)x-.45 .17 14 (S/2) -.06 (S/2)x .52 -.30 .04 15 (S/2) -.09 (S/2)x .53 -.31 .06 16 (S/2) .09 .37 (S/2)x-.39 -.01 17 (B/2) .13 .21 (B/2)x-.42 .12 18 (B/2) -.02 .18 (B/2)x-.48 .13 19 (W/2) (W/2)x .68 -.11 -.14 .01 20 (W/2) (W/2)x .56 -.08 -.19 .00 21 (W/2) (W/2)x .72 -.04 -.15 -.02 22 (P/3) .03 .21 ;(P/3)x-.55 .30 23 (P/3) .06 .19 (P/3)x-.54 .37 24 (P/3) .13 .17 (P/3)x-.60 .23 25 (P/3) .13 .18 (P/3)x-.60 .26 26 (S/3) .02 (S/3)x .57 -.35 .04 27 (S/3) .06 (S/3)x .54 -.38 .07 28 (S/3) -.11 (S/3)x .58 -.23 .16 29 (S/3) .07 (S/3)x .53 -.31 .11 30 (S/3) .08 (S/3)x .46 -.08 .18 31 (B/3) .05 .11 -.29 (B/3)x .58 32 (B/3) .10 .07 -.18 (B/3)x .66 33 (B/3) .06 .11 -.18 (B/3)x .67 34 (B/3) .03 .16 (B/3)x-.40 .39 35 (W/3) (W/3)x .78 -.02 -.12 .03 117 Table 6-4 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II III IV 36 (W/3) (W/3)x .77 .03. -.12 .02 37 (W/3) (W/3)x .76 .05 -.10 .04 38 (W/3) (W/3)x .75 .03. -.18 .06 39 (W/3) (W/3)x .59 .01 -.25 -.07 40 (P/4) .23 .24 (P/4)x -.47 .15 41 (P/4) .26 .31 (P/4)x -.39 .09 42 (P/4) .27 (P/4)x .38 -.38 .06 43 (P/4) .24 (P/4)x .41 -.38 .07 44 '(S/4) .07 (S/4)x .60 -.24 .09 45 (S/4) .06 (S/4)x .63 -.07 .12 46 (S/4) .01 (S/4)x .57 -.13 .08 47 , (S/4) .11 (S/4)x .59 -.26 .06 48 (B/4) .07 .26 -.02 (B/4)x .70 49 (B/4) .05 .27 -.07 (B/4)x .66 50 (B/4) .09 .28 -.05 (B/4)x .65 51 (B/4) .12 .31 -.17 (B/4)x .53 52 (B/4) .16 .28 -.04 (B/4)x .60 53 (W/4) (W/4)x .78 .05 -.08 .03 54 (W/4) (W/4)x .75 .09 -.09 .02' 55 (P/5) .19 .21 (P/5)x -.45 .20 56 (P/5) .21 .16 (P/5)x -.48 .18 57 (S/5) -.10 (S/5)x .63 -.14 .17 58 (S/5) .03 (S/5)x .62 -.07 .17 59 (S/5) .04 (S/5)x .61 +.08 .24 60 (B/5) .01 .11 - -.07 (B/5)x .73 61 (B/5) .09 .09 -.04 (B/5)x .75 62 (B/5) .04 .12 -.21 (B/5)x .58 63 (B/5) -.00 .16 -.11 (B/5)x .64 64 (B/5) -.02 .12 -.11 (B/5)x .66 65 (W/5) (W/5)x .85 -.01 -.06 -.01 66 (W/5) (W/5)x .86 -.00 -.05 -.02 67 (W/5) (W/5)x .79 -.05 -.11 .00 68 (W/5) (W/5)x .76 -.05 -.13 .03 69 (W/5)x .70 .01 -.18 .06 70 (P/6) .17 .20 (P/6)x -.42 .15 71 (P/6) .18 .29 (P/6)x -.56 .16 72 . (P/6) .38 .17 (P/6)x -.51 .03 73 (P/6) .25 .18 (P/6)x -.16 .02 74 (P/6) .31 .19 (P/6)x -.58 .03 75 (P/6) .25 .26 (P/6)x -.58 .03 118 Table 6-4 Gon't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II III • IV 76 (P/6) .22 .22 (P/6)x-.64 .05 77 (S/6) -.13 (S/6)x .67 -.23 .17 78 (S/6) -.08 (S/6)x .70 -.17 .10 79 (S/6) .06 (S/6)x .65 -.16 .08 80 (S/6) -.02 (S/6)x .72 -.13 .04 81 (S/6) -.05 (S/6)x .66 -.09 .12 82 (S/6) -.08 (S/6)x .(67 -.09 .12 83 (W/6) (W/6)x .78 .02 -.13 -.00 84 (W/6) (W/6)x .70 .05 -.18 .06 i n t e r p r e t a t i o n women 's sports p u b l i c business a f f a i r s p r o b a b i l i t y <.001 <.001 <.001 <.00i cumulative proportion of variance accounted f o r = 44.9% 119 Panel B - working hypothesis H2: Again two p a r t i c u l a r issues were selected f o r an a l y s i s : issue #3 and issue #5. The re s u l t s are presented i n Tables 6-5(i) and 6 - 5 ( i i ) . The 4 factors accounted for 71.2% and 73.7% i n issues #3 and #5 re s p e c t i v e l y . Using the 'x' procedure and l a b e l l i n g accordingly, the factors i n issue #3 can be interpreted as women's, public a f f a i r s , sports and business; and i n issue #5 as business, women's, sports and public a f f a i r s . In neither issue was there any discrepancy between the a p r i o r i and empirical c l a s s i f i c a t i o n systems. Further,'x' loadings are exceptionally strong. Table 6 - 5 ( i ) : Analysis of News Data - Panel B (issue #3) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I I I I I I IV 1 (public a f f a i r s ) .06 x .81 .17 -.18 2 (P) .08 x .80 .20 -.22 3 (P) .17 x .78 .20 -.16 4 (P) .16 x .78 .22 -.14 5 (Sports) .04 .28 X .79 -.03 6 (S) .08 .37 X .79 -.02 7 (S) -.06 .15 X .79 -.21 8 (S) .14 .18 X .77 -.18 9 (S) .07 .00 X .67 -.24 10 (Business) .06 .22 .17 x-.79 11 (B) .09 .11 .12 x-.89 12 (B) .04 .15 .17 x-.85 13 (B) .04 .38 .22 x-.56 14 (Women's) X .89 .05 .05 -.11 15 (W) X .90 .06 .04 -.05 16 (w) X .89 .06 .03 -.05 17 . (w) X .87 .11 .04 -.09 18 (W) X .73 .18 .08 .04 i n t e r p r e t a t i o n : women's public sports business a f f a i r s p r o b a b i l i t y : «.005 <.005 <.005 <.005 cumulative proportion of t o t a l variance accounted f o r : 71.2% 121 Table 6 - 5 ( i i ) : Analysis of News Data - Panel B (issue #5) system I II III IV 1 , (public a f f a i r s ) .19 .16 -.22 x .80 2 (P) .16 .21 -.17 x .81 3 (sports) .18 •.09 X-.79 .17 4 (S) .12 .06 x-.80 .15 5 (S) .26 .01 x-,77 .05 6 (Business) x .81 .01 -.16 .02 7 (B) x .82 .07 -.13 .02 8 (B) x .71 .03 -.09 .35 9 (B) x .81 .02 -.19 .12 10 (B) x .83 .01 -.13 .09 11 (Women's) -.00 X .93 -.01 .02 12 (W) -.01 X .94 -.01 .02 13 (W) .04 X .89 -.03 .11 14 (W) .04 X .87 -.00 .12 15 (w) .08 X .79 -.02 .22 i n t e r p r e t a t i o n : business women's sports p u b l i c a f f a i r s p r o b a b i l i t y : <.005 <.005 <.005 <.01 cumulative proportion of t o t a l variance accounted f o r = 73.7% 122 Panel B - working hypothesis H3: The 4 content/structure sections are investigated independently. The variables are those indicated i n Table 6-4 with the 6 m i s c l a s s i f i e d v a r i -ables removed. The r e s u l t s of the a p r i o r i c l a s s i f i c a t i o n system are on the l e f t side of Tables 6-6 ( i ) , ( i i ) , ( i i i ) , ( i v ) . The same procedure was followed as with panel A. Table 6-6(i); The f i v e factors can c l e a r l y be interpreted as issues #6, #3, #2, #4, and #5 re s p e c t i v e l y . A l l variables load as predicted by the a p r i o r i system; however, variables 14 and 15 while loading c o r r e c t l y under factor 1 have quite high loadings on factor 4 as w e l l . Table 6-6 ( i i ) : The f i r s t f i v e factors can be interpreted as issues #6, #3, #4, #2 and #1, r e s p e c t i v e l y . However, 4 of 22 variables including a l l of issue #5 (variables 14, 15, 16) do not f i t the predicted pattern. There \ i s no f a c t o r representing issue #5 and factor #6 cannot be interpreted i n terms of the a p r i o r i system. Table 6 - 6 ( i i i ) : The four factors can be c l e a r l y interpreted as issues #4, #5, #1 and #2 r e s p e c t i v e l y . A l l variables loaded as predicted by the a p r i o r i system. Table 6-6(iv): The s i x factors can be c l e a r l y interpreted as issues #5, #3, #2, #1, #4 and #6 r e s p e c t i v e l y . A l l variables loaded as pre-dicted by the a p r i o r i system. 123 Table 6-6(i): Analysis of Public A f f a i r s Data - Panel B a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I I I I I I IV V 1 issue #2 .09 -.12 x-.67 -.39 .11 2 2 .00 -.13 x-.69 -.30 .14 3 2 .24 -.16 x-.77 .04 .05 4 2 .18 -.25 x-.75 .02 .05 5 2 .16 -.11 x-.55 -.07 .10 6 issue #3 .14 x-.82 -.17 -.14 .09 7 3 .14 X T . 84 -.16 -.08 .12 8 3 .19 x-.78 -.24 -.11 .09 9 3 .24 x-,76 -.16 -.14 .18 10 issue #4 .23 -.21 -.19 x-.71 .20 11 4 .22 -.14 -.16 x-.82 .01 12 issue #5 .21 -.20 -.13 -.20 x.81 13 5 .32 -.19 -.23 -.03 x.74 14 issue #6 x.56 -.05 -.04 -.43 .08 15 6 x.55 -.26 -.10 -.35 .28 16 6 x.75 -.11 -.15 -.14 .15 17 6 x.85 -.12 -.11 -.11 .08 18 6 x.84 -.17 -.13 -.15 .03 19 6 x.79 -.17 -.18 -.07 .17 20 6 x.74 -.22 -.25 -.03 .21 interpretation: iss#6 iss#3 iss#2 iss#4 iss#5 pro b a b i l i t y : < 005 .0002* .005 .005* .005* 'cuffliilaiSiv.e- proportion: of variance accounted ...for ,.= 124 Table 6 - 6 ( i i ) : Analysis of Sports Data - Panel B a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system sys tern I I I I I I IV V VI 1 issue #1 .22 -.16 .14 -.18 x-.88 -.12 2 1 .25 -.17 .12 -.17 x-.88 -.10 3 issue #2 .19 -.20 .14 x-.88 -.14 -.05 4 2 .20 -.18 .14 x-,86 -.19 -.12 5 issue #3 .27 x-.67 .36 -.15 -.15 .16 6 3 .30 x-.64 .36 -.15 -.11 .16 7 3 .27 x-.58 .17 . -.31 -.25 -.04 8 3 .12 x-.77 .21 -.19 -.09 -.21 9 3 .11 x-.75 -.00 -.04 -.04 -.45 10 issue #4 .25 -.24 x .65 -.22 -.12 .01 11 4 .30 -.13 .48 -.10 -.06 x-.52 12 4 .24 -.18 x .63 -.00 -.06 -.29 13 4 .15 -.23 x .81 -.11 -.10 -.10 14 issue #5 x .46 -.12 .12 -.33 -.31 -.34 15 5 x .47 -.01 .38 -.11 -.23 -.27 16 5 .33 -.09 .19 -.14 -.19 x-.67 17 issue #6 x .76 -.22 .16 -.20 -.18 .08 18 6 x .78 -.22 .18 -.16 -.11 .04 19 6 x .68 -.13 .23 -.06 -.04 -.25 20 6 x .74 -.18 .10 -.12 -.12 -.19 21 6 x .72 -.07 .14 -.12 -.11 -.22 22 6 x .71 -.17 .15 - -.04 -.18 -.14 interpretation: iss#6 iss#3 iss#4 iss#2 iss#l iss#5 probability: < .005 .0007* .003* .004* .004* .25* cumulative proportion of variance accounted for = 70.3% 125 Table 6 - 6 ( i i i ) : .Analysis of Business Data - Panel B a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I IV 1 issue #1 .24 .16 x .72 -.04 2 1 .21 .22 x .70 -.21 3 1 .04 .11 x .81 -.14 4 1 .06 .16 x .82 -.17 5 issue .#3 .23 .15 .18 x-.80 6 3 .21 .22 .17 x-.86 7 3 .23 .23 .20 x-.80 8 issue #4 x .81 .27 .13 -.17 9 4 x .84 .19 .17 -.13 10 4 x .83 .25 .10 -.12 11 4 x .73 .11 .13 -.23 12 4 x .78 .15 .12 -.18 13 issue #5 .29 X .68 .14 -.29 14 5 .36 X .69 .12 -.24 15 5 .20 X .74 .10 -.14 16 5 .13 X .84 .23 -.05 17 5 .12 X .82 .22 -.14 interpretation: iss//4 iss#5 is s # l iss#3 pro b a b i l i t y : <.005 <.005 <.005 <.005 cumulative proportion of variance accounted for = 71.6% 126 Table 6-6(iv): Analysis of Women's Data - Panel B a p r i o r i c l a s s i f i - empirical c l a s s i f i c a t i o n system cation system I II III IV V VI 1 issue #1 - -.-24 .21 -.27 x .74 -.28 .03 2 1 -.18 .20 -.13 x .85 -.07 .20 3 issue #2 -.28 .24 x-.78 .15 -.14 .11 4 2 -.13 .20 x-.84 .10 -.08 .09 5 2 -.24 .25 x-.76 .20 -.21 .15 6 issue #3 -.28 x .74 -.22 .12 -.33 .09 7 3 -.26 x .76 -.18 .17 -.35 .06 8 3 -.26 x .78 -.23 .13 -.22 .07 9 3 -.32 x .76 -.20 .16 -.10 .13 10 3 . -.20 x .69 -.22 . 17 -.17 .31 11 issue #4 -.35 .27 -.29 .21 x-.63 .18 12 4 -.25 .28 -.19 .24 x-.64 .31 13 issue #5 x-.79 .28 -.21 . 15 .31 .08 14 5 x-.79 .30 -.23 .15 .28 .07 15 5 x-.80 .26 -.16 .21 .12 .16 16 5 x-.79 .22 -.23 .18 .03 .23 17 .5 x-.68 .28 -.12 .05 .11 .31 18 issue #6 -.38 .24 -.25 .19 .40 x .57 19 6 -.38 .19 -.19 .18 .21 x .73 i n t e r p r e t a t i o n : iss#5 iss#3 iss#2 i s s # l iss#4 iss#6 p r o b a b i l i t y : <T.005 <.005 .001* .0058* .0058* .0058* cumulative proportion of variance accounted for = 81.5% 127 Panel C Panel C - working Hypothesis H-^ : Figure 7 indicates there were only 42 news quarter pages presented to Panel C. Screening f o r s t r u c t u r a l c r i t e r i a l e d to elimination of 1 quarter. Of the remaining 41 variables spread over 6 issues of the newspaper 16 were c l a s s i f i e d as p u b l i c a f f a i r s , 15 were c l a s s i f i e d as sports, and 10 as women's. There were 414 respondents i n Panel C. The r e s u l t s regarding work-ing hypothesis H^ are presented i n Tables 6-7( i ) , ( i i ) , ( i i i ) . Table 6-7 (i) : The r e s u l t s from the s t r i c t a p p l i c a t i o n of the a n a l y t i c procedure are presented i n t h i s table. As only 3 content/structure sections are presented the number of factors are suppressed to 3. The fa c t o r space accounts f o r 49% of the variance i n the o r i g i n a l v a r i a b l e s . The re s u l t s found with Panels A and B are c l e a r l y not as strong with Panel C. The factors can be i n t e r p r e t -ed as follows. Factor 1 contains a l l women's variables as w e l l as some weaker loadings on p u b l i c a f f a i r s v a r i a b l e s . I t i s most strongly a women's fa c t o r . S i m i l a r l y , f a c t o r 2 i s most strongly a sports f a c t o r . Factor 3 represents mostly p u b l i c a f f a i r s v a r i a b l e s from issues #1 and #2 but because of high sports loadings could be interpreted as an issue #1,2 f a c t o r . However, by de-f a u l t i t i s interpreted as a public a f f a i r s f a c t o r at l e a s t f o r purposes of testing the c l a s s i f i c a t i o n systems. Under such l a b e l l i n g no fewer than 10 128 of the 41 variables are not c l a s s i f i e d as predicted by the a p r i o r i system (#12, 13, 15, 21, 24, 27, 31, 32, 33, 34). The contingency table analysis i s presented i n Figure 15. FIGURE 15: Contingency Table Analysis - Panel C  Women's empirical a/p + — . + 10 0 10 - 6 25 31 16 25 41 Xc2 = 17.42 Sports a/p empirical a/p_ + -+ 12 3 15 - 1 25 26 13 28 41 A f f a i r s empirical + + 9 7 16 - 3 22 25 12 29 41 & 2 = 22.08 XK2 = 7.21 Although the n u l l hypothesis i s rejected i n a l l cases (in f a c t o r 3 129 only at the .01 l e v e l of s i g n i f i c a n c e ) some further consideration of H-^  i n th i s context i s required. Experience with data manipulation suggests some pos-s i b l e explanation f o r these r e l a t i v e l y poor r e s u l t s . Consider the empirical c l a s s i f i c a t i o n of fa c t o r 3. This factor, e s p e c i a l l y i f high loadings on variables 3, 4 and 14 are considered, could be interpreted as an issue #1,2 factor f o r sports and p u b l i c a f f a i r s . The empirical c l a s s i f i c a t i o n of P variables (Public a f f a i r s v a r i ables) not loading on f a c t o r 3 i s p r i m a r i l y with f a c t o r 1 and i n one case f a c t o r 2 though the loadings are not strong.^ This suggests expansion of the factor space to 4 factors with the expectation that the f i r s t 3 factors w i l l remain r e l a t i v e l y stable and the 4th f a c t o r w i l l r e -present P variables i n the l a t e r issues. The r e s u l t s from this experiment are presented i n Table 6 - 7 ( i i ) . Inspection of this table v e r i f i e s the expected r e s u l t . ' iMdJcgoryer^cfa ZJg^§e&&[kB8Gb$. .#i 5 p »a^ d>p v.b l i c r a f f airs7is^sues;_# 4's 5 P,6 Now, i f one had sta r t e d with t h i s l a t t e r f a c t o r space and hypothesized how the r e s u l t might look with 3 fa c t o r s , he would probably expect the P v a r i -ables to combine. The fact that t h i s does not happen, but rather the P and S variables from issues #1 and #2 combine, suggests there i s a stronger r e l a t i o n -ship among these l a t t e r v a r i a b l e s . Why has t h i s r e s u l t emerged i n Panel C and not Panels A and B? F i r s t , i f one inspects panel B (Table 6-4), i t can be seen that 3 of the m i s c l a s s i f i c a t i o n s concern issue #2 var i a b l e s ; and i n Panel A (Table 6-1), although the variables are c o r r e c t l y c l a s s i f i e d there i s some 1. In subsequent discussion p u b l i c a f f a i r s variables or quarter pages are often referred to as P variables or quarters. S i m i l a r l y , these are S, W and B variables representing sports, women's and business. 130 evidence of higher loadings of S va r i a b l e s on the wrong f a c t o r s . Second, S and P variables/issues #1,2 represent 10 of the 41 variables analyzed under Panel C suggesting that i f they are highly correlated they may tend to emerge as a s i n g l e f a c t o r . This i s less l i k e l y under Panels A and B where there are far more variables analyzed. Third, inspection of newspaper content reveals that the sports content of issues #1 and #2 l a r g e l y deal with World Series Baseball coverage which may be unduly re l a t e d to normal pu b l i c a f f a i r s content. Such a seasonal abberation may suggest an unusually high degree of related ex-posure among the sports, public a f f a i r s and perhaps business content. While overshadowed by a large number of variables i n panels A and B, t h i s r e l a t i o n -ship perhaps was discovered i n Panel C and hinted at i n panels A and B. As an i n v e s t i g a t i o n of such speculation i t was decided to remove issues #1 and #2 variables and repeat f o r variables drawn only from issues #3 to #6. The r e s u l t s are presented i n 6 - 7 ( i i i ) . Issues #1 and #2 included 18 v a r i a b l e s , leaving only 23 f o r furth e r a n a l y s i s . The fa c t o r space accounts f o r 57.4% of the variance and, as inspection i n d i c a t e s , the f a c t o r r e s u l t s are much c l e a r e r . Each v a r i a b l e loads strongly and c o r r e c t l y on women's, sports and pu b l i c a f f a i r s f a c t ors r e s p e c t i v e l y . 131 Table 6-7 ( i ) : Analysis of News Data - Panel C ( a l l Issues) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I 1 (Public A f f a i r s / i s s u e #1) .08 .04 (P/l)x-.67 2 ( P 7 D .06 .05 (P/l)x-.62 3 (Sports/issue #1) .17 (S/l)x .48 -.46 4 (S/l) .17 (S/l ) x .47 -.42 5 (S/l) -.10 (S/l)x .34 -.25 6 (Public A f f a i r s / i s s u e #2) .25 .15 (P/2)x-.65 7 (P/2) .18 .09 (P/2)x-.70 8 (P/2) .18 .19 (P/2)x-.64 9 (P/2) .30 .10 (P/2)x-.63 10 (P/2) .32 .11 (P/2)x-.58 11 (P/2) .32 .12 (P/2)x-.66 12 (S/2) .06 .43 (S/2)x-.52 13 (S/2) .00 .49 (S/2)x-.52 14 (S/2) .04 (S/2)x .51 -.49 15 (S/2) .13 .32 (S/2)x-.47 16 (Women's/issue #2) (W/2)x .74 -.09 -.22 17 (W/2) (W/2)x .69 .06 -.24 18 (W/2) (W/2)x .73 -.09 -.24 19 (S/3) -.04 (S/3)x .65 -.29 20 (S/3) .19 (S/3)x .50 -.21 21 (P/4) (P/4)x .40 .36 -.38 22 (P/4) .31 .27 (P/4)x-.36 23 (P/4) (P/4)x .47 .20 -.28 24 (W/4) (W/4)x .81 .05 -.07 25 (W/4) (W/4)x .74 .08 -.13 26 (W/4) (W/4)x .78 .08 -.06 27 (P/5) (P/5)x .37 .34 -.35 28 (S/5) .11 (S/5)x .60 -.13 29 (W/5) (W/5)x .82 .02 -.06 30 (W/5) (W/5)x .86 .02 -.06 31 (P/6) .33 (P/6)x .39 -.22 32 (P/6) (P/6)x .42 .38 -.24 33 (P/6) (P/6)x .38 .30 -.35 34 (P/6) (P/6)x .38 .36 -.27 35 (S/6) .04 (S/6)x .79 -.10 36 (S/6) .03 (S/6)x .81 -.11 37 (S/6) .11 (S/6)x .75 -.05 38 (S/6) .11 (S/6)x .76 -.06 Table 6-7(i) Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I 39 (S/6) .05 (S/6)x .71 .05 40 (W/6) (W/6)x .78 .08 -.08 41 (W/6) (W/6)x .73 .14 -.04 interpretation: women's sports publ-icxaf f a i r s p r o b a b i l i t y : ^001 <.001 <.01 cumulative proportion of variance accounted for = 49% 133 Table 6 - 7 ( i i ) : Analysis of News Data - Panel C ( a l l issues) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II III IV 1 (P/D .06 -.08 (P/l)x .66 -.11 2 (P/D .04 -.07 (P/l)x .61 -.12 3 (S/l) -.06 (S/l)x-.61 .45 .17 4 (S/l) -.07 (S/l)x-.60 .41 .17 5 (S/l) -.07 (S/l)x-.39 .23 .02 6 (P/2) .18 -.13 (P/2)x .62 -.31 7 (P/2) .11 -.07 (P/2)x .68 -.29 8 (P/2) .14 -.18 (P/2)x .61 -.24 9 (P/2) .23 -.08 (P/2)x .60 -.31 10 (P/2) .25 -.08 (P/2)x .54 -.32 11 (P/2) .27 -.12 (P/2)x .63 -.25 12 (S/2) .10 (S/2)x-.50 . .49 -.03 13 (S/2) .07 (S/2)x-.58 .50 .02 14 (S/2) .10 (S/2)x-.59 .46 .00 15 (S/2) .06 -.37 (S/2)x .45 -.06 16 (W/2) (W/2)x .78 .04 .24 -.03 17 (W/2) (W/2)x .72 -.04 .24 -.08 18 (W/2) (W/2)x .79 .03 .26 -.01 19 (S/3) -.03 (S/3)x-.68 .24 -.09 20 (S/3) .16 (S/3)x-.48 .16 -.24 21 (P/4) .27 -.24 .31 (P/4)x-.53 22 (P/4) .17 -.15 .30 (P/4)x-.50 23 (P/4) .36 -.11 .23 (P/4)x-.42 24 (W/4) (W/4)x .80 -.03 .06 -.20 25 (W/4) (W/4)x .75 -.09 .12 -.13 26 (W/4) (W/4)x .78 -.07 .05 -.17 27 (P/5) .23 -.22 .28 (P/5)x-.53 28 (S/5) .08 (S/5)x-.57 .08 -.23 29 (W/5) (W/5)x .81 .03 .05 -.18 30 (W/5) (W/5)x .84 .05 .05 -.22 31 (P/6) .17 -.24 .14 (P/6)x-.57 32 (P/6) .27 -.24 .17 (P/6)x-.56 33 (P/6) .19 -.14 .27 (P/6)x-.63 34 (P/6) .17 -.17 .18 (P/6)x-.68 35 (S/6) .01 (S/6)x-.75 .03 -.25 36 (S/6) .01 (S/6)x-.77 .03 -.28 37 (S/6) .05 (S/6)x-.68 -.03 -.34 38 (S/6) .07 ( S/6)x-.71 -.01 -.28 134 Table 6-7(ii) Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I I I I I I IV 39 (S/6) .02 (S/6)x-.66 -.12 -.24 40 (W/6) (W/6)x.77 -.07 .07 -.1§ 41 (W/6) (W/6)x.70 -.10 .01 -.24 interpretation: women's sports public public a f f a i r s a f f a i r s (4,5,6) (1,2) cumulative proportion of variance accounted for = 53.4% 135 Table 6 - 7 ( i i i ) : Analysis of News Data - Panel C (is s u e s 3,4,5,6) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I 1 (S/3) .06 (S/3)x-.60 -.28 2 (S/3) .16 (S/3)x-.48 -.29 3 (P/4) .26 -.22 (P/4)x-.67 4 (P/4) .14 -.10 (P/4)x-.68 5 (P/4) .39 -.09 (P/4)x-.51 6 (W/4) (W/4)x.. ,. 83' -.01 -.23 7 (W/4) (W/4)x •'.77 -.07 -.19 8 (W/4) (W/4)x .80 -.06 .19 9 (P/5) .22 -.22 (P/5)x .63 10 (S/5) .08 (S/4)x-.61 .18 11 (W/5) (W/5)x .83 .02 .17 12 (W/5) (W/5)x .85 .03 .21 13 (P/6) .15 -.22 (P/6)x .62 14 (P/6) .23 -.22 (P/6)x .64 15 (P/6) .17 -.17 (P/6)x .68 16 (P/6) -.17 -.21 (P/6)x .68 17 (S/6) -.00 (S/6)x-.80 .18 18 (S/6) .02 (S/6)x-.83 .18 19 (S/6) .08 (S/6)x-.78 .15 20 (S'/6) .07 (S/6)x-.79 .15 21 (S/6) .04 (S/6)x-.73 .06 22 (W/6) (W/6)x .79 -.08 .18 23 (W/6) (W/6)x .72 -.11 .21 i n t e r p r e t a t i o n women's sports cumulative proportion of variance accounted f o r = pu b l i c a f f a i r s 57.5% 136 Panel C - working hypothesis H2: Two issues were selected f o r ana l y s i s : issue #2 and issue #6. It was expected that these might make an i n t e r e s t i n g contrast to H£ under Panels A and B because of the mixed re s u l t s indicated i n Table 6-7 ( i ) . How-ever, Table 6-7 ( i i ) does i n d i c a t e that 3 factors (although not the same f a c -tors) tend to account f o r the variance i n each case. The only v a r i a b l e which may not load as indicated i s var i a b l e 15 which loaded i n c o r r e c t l y i n Table 6 - 7 ( i i ) . The r e s u l t s are presented i n Tables 6-8(i) and 6 - 8 ( i i ) . The 3 factors accounted f o r 67% and 68.2% i n issues #2 and #6 re s p e c t i v e l y . Using the 'x' procedure and l a b e l l i n g accordingly, the factors i n issue #2 can be interpreted as public a f f a i r s , sports and women's; and i n issue #6 as sports, p u b l i c a f f a i r s and women's. In neither issue was there any discrepancy be-tween the a p r i o r i and empirical c l a s s i f i c a t i o n systems. The 'x' loadings are strong except f o r v a r i a b l e 10 i n Table 6-8(i) which corresponds to var i a b l e 15 i n Table 6 - 7 ( i i ) . 137 Table 6-8 ( i ) : Analysis of News Data - Panel C (issue #2) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system System I II I I I 1 (P) x-.62 .34 .18 2 (P) x-.68 .27 .10 3 (P) x-.73 .27 .08 4 (P) x-.77 .13 .18 5 (P) x-.76 .11 .22 6 (P) x-.74 .21 .26 7 (S) -.18 X .86 .11 8 (S) -.20 X .85 .06 9 (S) -.26 X .79 .04 10 (S) -.37 X .49 .14 11 (W) -.18 .08 X .90 12 (W) -.23 .11 X .82 13 (W) -.19 .08 X .90 i n t e r p r e t a t i o n p u b l i c sports women!s a f f a i r s p r o b a b i l i t y <;.005 < 005 <.005 ' cumulative proportion of variance accounted f o r = 67% 138 Table 6 - 8 ( i i ) : Analysis of News Data - Panel C (issue #6) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II II 1 (P) .24 • (P)x .61 .22 2 (P) .23 (P)x .61 .30 3 (P) .12 (P)x .86 .10 4 (P) .16 (P)x .85 .11 5 (S) (S)x .82 .22 .01 6 (S) (S)x .85 .20 -.01 7 (S) (S)x .79 .19 .10 8 (S) (S)x .81 .17 .08 9 (S) (S)x .74 .07 .10 10 (W) .04 .20 i (W)x .88 11 (W) .07 .24 (W)x .85 in t e r p r e t a t i o n sports public women1 s a f f a i r s p r o b a b i l i t y <005 <.005 <.025 cumulative proportions of variance accounted f o r = 68.2% 139 Panel C - working hypothesis H^: Only 3 content/structure sections are a v a i l a b l e for i n v e s t i g a t i o n with Panel C. These are discussed separately as follows: Table 6-9(i) : As indicated under H-^ , a number of public a f f a i r s v a r i a b l e s did not load on the same f a c t o r . However, because they a l l did load onto two f a c t o r s as indicated i n 6-7(ii) and under H3 the e f f e c t of other content i s removed, i t was decided to investigate a l l 16 p u b l i c a f f a i r s v a r i -ables. Factors 1 and 2 can be interpreted as issues #2 and #4; both factors 3 and 4 can be interpreted as factor 6; f a c t o r 1 can be interpreted as issue #1. The only v a r i a b l e representing issue #5 does not load on a separate factor but as f a c t o r 3. Table 6 - 9 ( i i ) : The s i n g l e m i s c l a s s i f i e d v a r i a b l e as i n d i c a t e d i n 6 - 7 ( i i ) i s removed leaving 14 sports variables i n 5 f a c t o r s . Factors 1, 2, and 4 can be interpreted as issues #6, #2 and #3 r e s p e c t i v e l y . Variables from issue #1 load both on factors 3 and 5. Table 6-9(111): The four factors can be interpreted as issues #1, 2, 3, and 4 r e s p e c t i v e l y . A l l variables load as predicted by the a p r i o r i system. Table 6-9 ( i ) : Analysis of Public A f f a i r s Data - Panel C a p r i o r i c l a s s i f i c a t i o n emp i r i c a l c l a s s i f i c a t i o n system system I I I I I I ... i v V 1 issue #1 .28 -.11 -.07 -.11 x .84 2 1 .21 -.10 -.08 -.10 x .87 3 issue #2 X .61 -.37 -.26 .01 .12 4 2 X .64 -.11 -.34 -.02 .26 5 2 X .71 -.15 -.06 -.10 .21 6 2 X .71 -.17 -.07 -.14 .27 7 2 X .78 -.10 -.09 -.24 .00 8 2 X .78 -.16 -.09 -.17 .12 9 issue #4 .28 x-.72 -.31 -.12 .05 10 4 .16 x-.76 -.25 -.07 .08 11 4 .16 x-.72 -.06 -.23 .13 12 issue #5 .27 -.31 x-.49 -.30 .08 13 issue #6 .11 -.16 . X-.81 -.17 .09 14 6 .16 -.23 x-.79 -.19 .04 15 6 .23 -.18 -.21 x-.83 .14 16 6 .17 -.18 -.26 x-.85 .10 interpretation iss#2 iss#4 iss#6 iss#6 iss#l probability < 005 < 005 .13* .05* <.0i cumulative proportion of variance accounted for = 69.6% 141 Table 6 - 9 ( i l ) ; Analysis of Sports Data - Panel C a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system sys tem I II I I I IV V 1 issue #1 .17 .28 x.88 .11 .12 2 1 .18 .20 x.90 .11 .15 3 1 .14 .09 .22 .11 x .90 4 issue #2 .14 x.88 .13 .15 -.02 5 2 .22 x.84 .24 .09 .05 6 2 .26 x.71 .21 .16 .17 7 issue #3 .30 .39 .33 x.55 .10 8 3 .21 .12 .10 x.89 .05 9 issue #5 x.47 .23 -.02 ;\41 .24 10 issue #6 x.77 .26 .23 .12 -.14 11 6 x.80 .27 .18 .16 -.07 12 6 x.79 .12 .04 .16 .17 13 6 x.79 .17 .12 .13 .08 14 6 x.72 .04 .11 .10 .22 i n t e r p r e t a t i o n iss#6 iss#2 i s s # l iss#3 i s s # l p r o b a b i l i t y <005 <.005 <.05 <.025 .21* cumulative proportion of variance accounted f o r = 75.9% 142 Table 6 - 9 ( i i l ) : Analysis of Women's Data - Panel C a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II I I I IV 1 issue #1 x-.81 -.20 -.27 .28 2 1 x-.78 -.35 -.19 .10 3 1 x-.81 -.22 -.26 .27 4 issue #2 -.24 x-.75 -.28 .36 5 2 -.29 x-.82 -.14 .24 6 2 -.27 x-.72 -.39 .16 7 issue #3 -.31 -.29 x-.83 .26 8 3 -.34 -.32 x-.79 .30 9 issue #4 -.27 -.43 -.30 x .65 10 4 -.25 -.25 -.25 x .82 i n t e r p r e t a t i o n i s s # l iss#2 iss#3 p r o b a b i l i t y <-:01 <.01 <.01 cumulative proportion of variance accounted f o r = 84.1% iss#4 <.01 External Validation of Factor Results,on News Data As discussed i n Chapter V, external v a l i d a t i o n of the factor results i s attempted using stepwise regression analysis, i . e . i t i s of interest to v a l i d -ate the interpretation of factor results by r e l a t i n g factor scores to audience predispositions. the regression model A linear regression model can be used to express the relationship between factor scores and audience predispositions: F = a + 3- x + B x + + B x + e 1 1 2 2 n n where, F = factor scores calculated as a linear combination of o r i g i n a l scores. X to = the independent variable measures. a, 3 to 3 = the population parameters to be estimated from the sample 1 information. e = the error term Each co e f f i c i e n t i s tested for i t s significance under the n u l l hypothesis that i t i s equal to zero. H : 3 = 0 0 H : 3 > 0 1 ' The significance test of each co e f f i c i e n t uses the F-di s t r i b u t i o n , that i s , an F-ratio i s calculated for each co e f f i c i e n t and the associated probab-i l i t y i s determined for such a value under the n u l l hypothesis that the coeff-i c i e n t i s equal to 0. In stepwise regression only those variables are included i n the f i n a l equation which add s i g n i f i c a n t l y to the t o t a l variance explained i n the dep-endent variable by a subset of independent variables. The inclusion or exclusion of an independent variable i n the equation i s determined by the 144 s p e c i f i c a t i o n of a s i g n i f i c a n c e l e v e l for values of the F - d i s t r i b u t i o n . In the r e s u l t s reported i n t h i s study independent v a r i a b l e s are included i f the F value has a p r o b a b i l i t y l e s s than or equal to .1. The F test i s also used to test the strength.of the r e l a t i o n s h i p i n the o v e r a l l regression equation. o p e r a t i o n a l i z a t i o n of the model I t was argued i n Chapter I I I that p o s i t i v e u t i l i t y i n s e l e c t i v e exposure was provided only by media content. Hence only content dimensions could be expected to be r e l a t e d to audience a t t r i b u t e s or p r e d i s p o s i t i o n s . However under the working hypotheses concerning the news data, dimensions of audience exposure should be r e l a t e d to such predispositions due to the f a c t that the structure of the medium supports content s e l e c t i v i t y . There are two possible sets of factor scores a v a i l a b l e f or external v a l i d a t i o n : those associated with H , i . e . dimensions of audience exposure over time, and those associated with H , i . e . dimensions of audience exposure i n a s i n g l e issue. Although 2 the factor loadings i n the l a t t e r case are extremely c l e a r , i t i s possible that the dimensionality, within a p a r t i c u l a r issue may r e f l e c t some unique c h a r a c t e r i s t i c s of that issue. On the other hand, factor scores from dimen-sions of audience exposure over time are c l e a r l y more v a l i d . A high score on a sports factor across a number of issues c e r t a i n l y r e f l e c t s a high degree of deliberate exposure to sports content over time. Hence, using the f a c t o r loadings of Table 6-1 a matrix of factor scores was calculated f or panel A, i . e . , the score of each panel A respondent on each of four f a c t o r s : Business, Pu b l i c A f f a i r s , Women's and Sports. S i m i l a r l y , matrices of factor scores were calculated f or panels B and C. In the case of panel C where working hypothesis H was not as strongly confirmed, the factor loadings matrix using only issues 3 to 6 was selected for further analysis [as i n Table 6-7 ( i i i ) ] . Each respondent has a f a c t o r score on each f a c t o r . However not a l l respondents completed the questionnaires designed to measure the various 145 a t t r i b u t e or p r e d i s p o s i t i o n a l data discussed i n chapter IV. In other words, the analysis i s disrupted by missing data i n the independent v a r i a b l e set. The simplest procedure for handling t h i s problem, e s p e c i a l l y i n l i g h t of the large sample s i z e , was to delete from the analysis those respondents with incomplete data sets. Accordingly, the sample s i z e f or panel A was reduced from 402 to 282 respondents, for panel B from 404 to 282 respondents, and for panel C from 414 to 258 respondents. the regression r e s u l t s Tables 6-10 to 6-13 present the r e s u l t s of the stepwise regression proc-edure. For ease of comparison the r e s u l t s are presented for each type of factor across the d i f f e r e n t panels. For example, Table 6-10 presents the regression of Women's factor scores for panels A, B and C. On the l e f t side of each table the independent v a r i a b l e s are l i s t e d and i n each c e l l can be found the associated regression c o e f f i c i e n t where s i g n i f i c a n t . In brackets i s the p r o b a b i l i t y of the associated value under the F d i s t r i b u t i o n . Only v a r i a b l e s with F p r o b a b i l i t i e s l e s s than or equal to .1 are included. At the bottom of each table i s the R 2 for each regression equation and i t s associated F probability.''' The object of the regression analysis i s to provide external v a l i d i t y to the r e s u l t s of factor a n a l y s i s . There are not a p r i o r i any strong convictions as to which audience a t t r i b u t e s or predispositions are l i k e l y r e l a t e d to low or high scores on p a r t i c u l a r f a c t o r s . This i s p r i m a r i l y a search procedure. Fortunately, there i s the opportunity to repeat the analysis on three d i f f e r -ent samples. I t i s of primary i n t e r e s t to investigate whether or not any As w i l l have been noted, a number of factors have strong negative rather than p o s i t i v e loadings implying negative factor scores. Hence p o s i t i v e r e l a t i o n s h i p s between fa c t o r scores and independent v a r i a b l e s could appear negative. To s i m p l i f y presentation, a l l factor scores where the high load-ings are negative have been reversed i n sign. Hence, p o s i t i v e c o e f f i c i e n t s i n a l l cases i n d i c a t e d i r e c t r e l a t i o n s h i p s and negative c o e f f i c i e n t s indicate inverse r e l a t i o n s h i p s . 146 s i g n i f i c a n t r e l a t i o n s h i p s can be found and whether these tend to confirm or contradict the factor r e s u l t s . Most p a r t i c u l a r l y , i t i s of i n t e r e s t whether s i g n i f i c a n t r e l a t i o n s h i p s can be val i d a t e d across at le a s t two of the panels. Table 6-10: The regression of Women's factor scores indicates a p o s i t i v e r e l a t i o n s h i p to the female sex, an i n t e r e s t i n hobbies and a favourable a t t i t -ude towards various a t t r i b u t e s of newspaper adver t i s i n g . C e r t a i n l y , the f i r s t of these r e l a t i o n s h i p s could have been expected and the l a t t e r i s encour-aging to the use of the Women's section as an adv e r t i s i n g v e h i c l e . A p o s i t i v e r e l a t i o n s h i p with age and a.negative r e l a t i o n s h i p with an i n t e r e s t i n sports i s demonstrated i n two of the three panels. Neither of these are inc o n s i s -tent with what might have been expected. The remaining r e l a t i o n s h i p s may have some foundation but only occur i n one of three panels. Table 6-11: S i m i l a r l y , the regression of sports factor scores lends v a l i d i t y to the factor r e s u l t s . However, there i s only a si n g l e r e l a t i o n -ship across a l l three panels, i . e . , a negative r e l a t i o n s h i p with sex i n d i c -ating a high degree of male readership. An i n t e r e s t i n sports i s found curiously i n only two of three cases. Again, the advertising r a t i n g score i s r e l a t e d to higher f a c t o r scores. Table 6-12: The most disappointing r e s u l t s are those of the Public A f f a i r s Section. Only the opinion r e l a t e d measures concerning newspaper coverage, source and personality demonstrate very systematic r e l a t i o n s h i p s . If interpreted accordingly, i t would seem that the readers of the Public A f f a i r s Section have a low opinion of the q u a l i t y of the newspaper. Although seemingly paradoxical, t h i s could quite possibly be the case. There i s further some i n d i c a t i o n of female readership. Table 6-13: The regression of the Business factor scores ind i c a t e the most stable r e l a t i o n s h i p s i n s p i t e of the fa c t that only panels A and B provided adequate business quarter pages f or ana l y s i s . The r e s u l t s i n d i c a t e a p o s i t i v e r e l a t i o n s h i p with a need f o r understanding and age while a negative . 147 r e l a t i o n s h i p with l i b e r a l i s m and sex, i . e . , p r i m a r i l y male readership. However, no r e l a t i o n s h i p s are indicated for either' income or education as might have been expected. In general terms the above r e s u l t s are encouraging from the point of view of external v a l i d a t i o n . In a l l cases the R 2 i s low although i t i s somewhat higher for the Women's fa c t o r s . However, t h i s merely indicates that the independent v a r i a b l e s cannot account for a large proportion of the variance i n the dependent v a r i a b l e . Although higher R 2 would c e r t a i n l y be preferable i t i s not necessary to the use of regression analysis i n t h i s l i m i t e d context. Table 6-10: Regression of Women's Factor Scores 148 Independent va r i a b l e s Panel A Panel B Panel C Personality t r a i t v a r i a b l e s Cognitive structure S o c i a l recognition Understanding Manifest anxiety General self-confidence -1.06 (.00) .68 (.02) Opinion r e l a t e d measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality .99 (.01) -.37 (.05) .67 (.08) .10 (.06) .97 (.00) .95 (.00) Leisure i n t e r e s t measures Hobbies Sports .37 (.01) -.29 (.01) .45 (.00) -.26 (.01) .17 (.03) Demographics Age Sex (female = '+') Income Education .69 (.00) .66 (.00) .85 (.00) .13 (.00) .86 (.00) -.82 (.07) R 2 .32 (.00) .42 (.00) .32 (.00) Table 6-11: Regression of Sports Factor Scores 149 Independent v a r i a b l e s Panel A Panel B Panel C Personality t r a i t v a r i a b l e s Cognitive structure S o c i a l recognition Understanding Manifest anxiety General self-confidence -1.00 (.01) .54 (.08) Opinion r e l a t e d measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality .17 (.02) -.35 (.09) .81 (.04) -.42 (.02) .51 (.03) .91 (.01) Leisure i n t e r e s t measures Hobbies Sports .46 (.00) .19 (.03) Demographics Age Sex (female = '+') Income Education -.48 (.00) -.53 (.00) -.78 (.04) -.43 (.03) R 2 .17 (.00) .18 (.00) .11 (.00) Table 6-12; Regression of Public A f f a i r s Factor Scores Independent va r i a b l e s Panel A Panel B Panel C Personality t r a i t v a r i a b l e s Cognitive structure S o c i a l recognition Understanding Manifest anxiety General self-confidence .66 (.07) Opinion re l a t e d measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality -.14 (.00) -.49 (.01) -.15 (.00) -.36 (.04) -.46 (.03) -.17 (.00) -.51 (.02) -.89 (.03) Leisure i n t e r e s t measures Hobbies Sports .43 (.00) -.18 (.06) Demographics Age Sex (female = '+') Income Education -.05 (.01) -.12 (.00) .31 (.01) .29 (.02) R 2 .19 (.00) .20 (.00) .08 (.00) Table 6-13: Regression of Business Factor Scores Independent va r i a b l e s Panel A Panel B Personality t r a i t measures Cognitive structure S o c i a l recognition Understanding Manifest anxiety General self-confidence 1.03 (.00) 1.10 (.00) -.47 (.07) Opinion re l a t e d measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality -.27 (.04) -.39 (.01) Leisure i n t e r e s t measures Hobbies Sports Demographics Age Sex (female = '+') Income Education .11 (.00) -.57 (.00) .15 (.00) -.27 (.02) R 2 .15 (.00) .17 (.00) 152 Chapter VII RESULTS OF ANALYSIS ON ADVERTISING DATA This chapter presents the r e s u l t s of analysis on the advertising data. This involves the op e r a t i o n a l i z a t i o n of working hypothesis(,H^ on the three ) panels of respondents. However, as there i s at best only marginal support f o r H^, much of the data reported results from exploratory analysis. 153 Panel A Panel A - working hypothesis H^: The p r i n c i p a l dimensions of audience exposure-to the advertising content of a d a i l y newspaper over time %i^iji^t^iM^%Q4iy^^%^ge^VSl^8 truc'turM ^ sections. data e x t r a c t i o n and the a p r i o r i c l a s s i f i c a t i o n system As discussed previously, the s t r u c t u r a l sections used to analyze the advertising data are p r e c i s e l y those.used to analyze the news data. I t i s necessary then to define exactly which advertising quarter pages f a l l within the p a r t i c u l a r sections. Figure 7 indicates that a t o t a l of 9 9 advertising quarter pages c l a s s i f i a b l e into.7 content categories were presented to Panel A over the s i x issues of the study. However, only 67 of these f e l l within the s t r u c t u r a l sections: 15 i n pub l i c a f f a i r s , 31 i n sports, 4 i n business, 17 i n women's. The remainder were e i t h e r on the back pages of various sec-tions or otherwise separate from the s t r u c t u r a l sections. (Note that unlike the news data, i t was not assumed that a l l advertising variables f a l l within the s t r u c t u r a l sections - the object i s to investigate the s t r u c t u r a l sections as advertising v e h i c l e s ) . The s t r u c t u r a l sections into which the advertising quarter pages f a l l represent the a p r i o r i c l a s s i f i c a t i o n system i n the i n v e s t i g a t i o n of working hypothesis H^. The re s u l t s of t h i s a p r i o r i c l a s s i f i c a t i o n are i n d i -cated on the l e f t side of Table 7-1. 154 the empirical c l a s s i f i c a t i o n system The a p r i o r i system in d i c a t e s 4 s t r u c t u r a l dimensions of a d v e r t i s -i n g . Hence the number of factors was suppressed to 4. The factor space accounted f o r 47% of the variance i n the o r i g i n a l v a r i a b l e s . The same 'x' pro-cedure was followed as with the news data. Inspection of Table 7-1 indicates mixed r e s u l t s . Factor 1 i s dom-inated by women's quarter pages but also includes a number of pub l i c a f f a i r s quarters. Factor 2 i s a mixed factor representing issues 1, 2 and 3 but dom-inated by sports quarter pages. Factors 3 and 4 s i m i l a r l y represent issues 4,5 and issue 6 r e s p e c t i v e l y . C l e a r l y , the o v e r a l l empirical system does not match the a p r i o r i system. While factors 1 and 2 best represent the women's and sports sections, p u b l i c a f f a i r s and business are not represented by inde-pendent f a c t o r s . t e s t f o r the independence of the two c l a s s i f i c a t i o n systems A contingency table analysis for factors 1 and 2 as women's and sports i s presented i n Figure 16. Here, as with working hypothesis H-p a x> t e s t can be applied to the independence of the two c l a s s i f i c a t i o n systems. 155 FIGURE 16: Contingency Table Analysis - Panel A= women's a/2 empirical + -+ 17 8 25 - 0 42 42 17 50 67 -X-2 = 34.16 (p <.001) sports + -+ 13 18 31 - 6 30 36 19 48 67 <W = 4.06 (p <.05) discussion Although the working hypothesis has some equivocal support, i t must be concluded that the o v e r a l l f a c t o r r e s u l t s do not confirm the a p r i o r i c l a s s -i f i c a t i o n of quarter pages. There are two reasons f o r t h i s conclusion. F i r s t , two of the a p r i o r i c l a s s i f i c a t i o n s are not at a l l supported. While t h i s i s s i g n i f i c a n t i n i t s e l f , i t also throws i n t o jeopardy the legitimacy of the 4 fa c t o r s o l u t i o n . The f a c t o r space was suppressed to 4 on the basis of the num-ber of c l a s s i f i c a t i o n s . If these c l a s s i f i c a t i o n s are not supported there i s no longer any reason a p r i o r i to expect 4 factors to adequately dimensionalize 156 the data. Perhaps a 3 or a 5 factor s o l u t i o n would be better. This i s e s p e c i a l l y true because of the few number of business quarter pages i n the a n a l y s i s . A 3 f a c t o r s o l u t i o n representing women's, sports and p u b l i c a f f a i r s could be expected based on the p o s s i b i l i t y that the business variables do not represent an adequate amount of variance to emerge as an independent f a c t o r . Second, i n s p i t e of the s t a t i s t i c a l tests of factors 1 and 2, i t i s possible that there are other i n t e r p r e t a t i o n s of these f a c t o r s . This i s p a r t i c u l a r l y important because the content of the advertising quarter pages does not n e c e s s a r i l y coincide with the structure of the newspaper as with the news data. Hence factor 1 may be i n t e r p r e t a b l e as a f a c t o r where quarter pages i n the women's section and some i n the p u b l i c a f f a i r s section have something i n common aside from the s t r u c t u r a l organization of the newspaper, i . e . , they have s i m i l a r underlying content. In summary, one can make very few conclusions about the r e s u l t s i n Table 7-1. I t i s necessary then to experiment with the data with two expec-t a t i o n s . F i r s t , a d i f f e r e n t factor s o l u t i o n may more adequately represent the structure i n the data. Second, d i f f e r e n t i n t e r p r e t a t i o n s of the f a c t o r r e -s u l t s may be possible e s p e c i a l l y on a content b a s i s . 1 5 7 Table 7-1: Analysis of Advertising Data - Panel A ( a l l " i s s u e s ) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II III IV 1 (Public af f a i r s / i s s # l ) .34 (P/l)x-.44 -.02 .03 2 (P/D .28 (P/l)x-.42 -.02 .05 3 (Sports/iss#l) .13 (S/l)x-.54 -.19 .16 4 (Women's/iss#l) (W/l)x .58 -.23 -.04 .01 5 (W/l) (W/l)x .63 -.21 -.13 -.02 6 (W/l) (W/l)x .53 -.21 -.02 -.04 7 (W/l) (W/l)x .53 -.25 -.21 .05 8 (W/l) ^ (W/l)x .54 -.25 -.22 .04 9 (P/iss#2) .39 (P/2)x-.46 -.15 .14 10 (S/2) .02 (S/2)x-.55 -.34 .14 11 .'-(S72) -.06 (S/2)x-.30 -.07 .13 12 (S/2) .25 (S/2)x-.49 -.27 .13 13 . (S/2) .21 (S/2)x-.52 -.29 .13 14 (S/2) .24 (S/2)x-.50 -.23 .11 15 (Business/iss#2) .27 (B/2)x-.55 -.16 .08 16 (W/2) (W/2)x .60 -.21 -.14 .14 17 ,iW/2) (W/2)x .72 -.11 -.10 .09 18 • (W/2) (W/2)x .45 -.29 -.17 .17 19 (P/iss#3) .46 (P/3)x-.51 .00 .21 20 (P/3) (P/3)x .47 -.44 -.08 .27 21 (P/3) (P/3)x .52 -.34 -.06 .20 22 (S/3) .22 (S/3)x-.49 -.32 .23 23 (S/3) .10 (S/3)x-.60 -.30 .21 24 (S/3) .08 (S/3)x-.60 -.34 .25 25 (S/3) .23 (S/3)x-.62 -.18 .29 26 (S/3) .16 (S/3)x-.66 -.15 .26 27 (B/3) .14 (B/3)x-.48 -.23 .19 28 (W/3) (W/3)x .64 -.14 -.03 .17 29 (W/3) (W/3)x .60 -.15 -.05 .09 30 (P/4) (P/4)x .37 -.27 -.34 .24 31 (P/4) (P/4)x .47 -.17 -.25 .31 32 (P/4) (P/4)x .14 -.37 -.17 .22 33 (P/4) (P/4)x .43 -.40 -.18 .26 34 (S/4) .18 . (S/4)x-.49 -.36 .25 35 (S/4) .08 -.36 (S/4)x-.58 .19 36 (S/4) .12 -.32 ( S/4)x-.64 .21 37 (S/4) .19 -.39 (S/4)x-.58 .11 38 (S/4) .15 -.38 \ (S/4)x-.60 .20 158 Table 7-1 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I I I I I I IV 39 ; (S/4) .21 (S/4)x-.43 -.39 .18 40 (S/4) .31 -.20 (S/4)X-.50 .12 41 (W/4) (W/4)x .64 -.04 -.24 .25 42 (P/5) (P/5)x .56 -.22 -.35 .18 43 (P/5) (P/5)x .44 -.16 -.42 .22 44 (S/5) .20 -.30 (S/5)x-.56 .15 45 (S/5) .17 -.28 (S/5)x-.58 .24 46 (S/5) .08 -.13 (S/5)x-.78 .15 47 (S/5) .11 -.13 (S/5)x-.73 .18 48 (B/5) .13 -.21 (B/5)x-.53 .17 49 -.(B/5) .22 -.23 (B/5)x-.60 .31 50 (W/5) (W/5)x .63 .03 -.14 .20 51 (W/5) (W/5)x .67 -.017' -.27 .28 52 (P/5) (P/5)x .56 .07 -.39 .28 53 (P/5) (P/5)x .56 .06 -.40 .28 54 (P/6) .43 -.22 -.15 (P/6)x .45 55 (P/6) .37 -.22 -.22 (P/6)x .60 56 (P/6) .42 -.15 -.21 (P/6)x. .58 57 (S/6) -.00 -.41 -.23 (S/6)x .57 58 (S/6) .11 -.36 -.24 (S/6)x .01 59 (S/6) .37 -.22 -.08 (S/6)x .58 60 (S/6) .39 -.25 -.12 (S/6)x .60 61 (S/6) .20 -.32 -.17 (S/6)x .53 62 (S/6) .16 -.21 -.33 (S/6)x .61 63 (S/6) .17 -.23 -.33 (S/6)x .65 64 (S/6) .15 -.31 -.30 (S/6)x .63 65 .(S/6) .13 -.22 -.38 ("sV6)x .64 66 (W/6) (W/6)x .57 -.09 -.10 .38 67 (W/6) (W/6)x .62 -.08 -.07 .47 interpretation women's sports probability <.001 <.05 cumulative proportion of variance accounted for = 47% 159 exploratory analysis The object i n this section i s to vary the size of the factor space i n search of interpretable r e s u l t s . ?A.>^o^|^ings;tmigKfi beT,-expected. :fiom ex- 71 \ amination of Table 7-1. F i r s t , that i f the number of factors i s supressed, factors 2, 3 and 4 w i l l tend to combine. This i s expected as a result of these factors being dominated by S variables and factor 1 by W variables. Second, the factor space as i t i s expanded w i l l tend to break down i n a systematic man-ner, i . e . , the factor 3 solution w i l l make sense i n terms of the factor 2 and factor 4 solutions, etc. This w i l l become clear through discussion of the data. The results i n Table 7-1 were obtained under the hypothesis that the managerial structure of the newspaper was most fundamental to dimensions of audience exposure. In an exploratory analysis where the object i s to find some underlying dimensionality which may or may not be associated with manager-i a l structure, a more complete description of the quarter pages must be under-taken. Figure 17 tabulates a number of characteristics which could perhaps contribute to communality among the quarter pages. As indicated these include the content of the advertising material, the s t r u c t u r a l section i n which i t i s found, the issue i n which i t i s found, whether i t i s found on the right or l e f t page, and i t s positioning on the newspaper page, i . e . , f i r s t , second, t h i r d or fourth quadrant. (l=top l e f t , 2=top r i g h t , 3=lower l e f t , 4=lower r i g h t ) . The l a t t e r two characteristics concerning page and positioning are not ex-pected to contribute the interpretation of r e s u l t s . I t can be expected that these are r e l a t i v e l y minor sources of variance\compared to content, managerial 160 structure and time structure. However, th i s information i s easily tabulated and may contribute to the analysis. Tables 7-1(i) and 7-1(ii) represent a systematic expansion of the factor space. The following discussion attempts to interpret these results successively given the information i n Figure 17. There i s a d i f f i c u l t y here i n finding an adequate means to present the results. One must examine the loadings and interpret according to the coding of the characteristics. Because t h i s i s not an hypothesis testing s i t u a t i o n , the 'x' procedure i s not followed. A l l reasonably high loadings are of interest whether mixed or not. A l l l a b e l l i n g i s done i n terms of managerial structure, i . e . , P, B, S, W, as this i s the most manageable device with respect to the data as w i l l become evident. Interpretation according to content and other characteristics i s integrated, where appropriate, into the discussion of each table. Table 7-1(i): The 2-factor solution accounts for 41% of the variance i n the o r i g i n a l 67 variables. The two factors are dominated respectively by quarter pages i n the sports section and by quarter pages i n the women's section (referred to subsequently as S and W variables). There are at least three possible interpretations: a) that the sports and women's sections have a strong st r u c t u r a l influence b) that the advertising quarter pages found i n sports and a l t e r n a t i v e l y women's have underlying,similarity i n content - quarter pages i n public a f f a i r s and 1 6 1 FIGURE 17: Description of Advertising Quarter Pages - Panel A quarter pages 1 2 3 4 5 6 7 8 content coding CG CGR W/C G W/C F F section P P S W W w w w issue 1 1 1 1 1 1 1 1 P a8 e L L L R L R R quadrant 3 4 3 4 1 2 4 9 10 11 12 13 14 15 16 17 18 CG L CGR A L L L WN G F P S S s s s B w w w R R R L L R L L R L 4 3 4 3 4 4 4 3 4 3 19 20 21 22 23 24 25 26 27 28 29 CG M/C F CGR A/A A/A L L L G F P P P S S S S S B W w 3 3 3 3 3 3 3 3 3 3 3 L L L R R R L R R R R 1 3 3 4 2 4 3 4 4 4 3 30 31 32 33 34 35 36 37 38 39 40 41 CG F A A CGR A/A A/A A L L A/A G P P P P S S S S S S S W 4 4 4 4 4 4 4 4 4 4 4 4 L R R R R R R L L R L R 3 4 2 4 4 2 4 3 4 4 2 4 162 FIGURE 17 Con't quarter pages 42 43 44 45 46 47 48 49 50 51 52 53 content coding CG A L L A/A A/ A L CGR G G F F section P P S S s s B B W w w w xssue 5 5 5 5 5 5 5 5 5 5 5 5 page L R R L R R . R L L R R R quadrant 3 4 4 3 2 4 4 4 3 4 1 2 54 55 56 57 58 59 60 61 62 63 64 65 66 67 CG M/C F L CGR M/C M/C L • A/A A/A L A •WN G P P P S S S S S S S S S W w 6 6 6 6 6 6 6 6 6 6 6 6 6 6 L R R L R L L R L L R R L R 3 3 4 3 4 1 3 4 2 4 3 4 3 4 CG - cigarettes CGR - cigars W/C - women's cl o t h i n g G - groceries F - fu r n i t u r e A - automobiles A/A - automobile accessories M/C -men's clo t h i n g WN - wine Table 7 - l ( i ) : Analysis of Advertising Data (Panel A)-2 Factor Solution quarter ONT SN page des c r i p t i o n I PG Q factor loadings I matr II 1 CG P 1 L 3 .27 P .37 2 CGR P 1 L 4 .27 3 S 1 S .52 .19 4 W/C W 1 L 3 .11 W .58 5 G W 1 R 4 .13 W .62 6 W/C W 1 L 1 .07 w .52 7 > F w 1 R 2 .25 w .54 8 S F w 1 R 4 .25 w .55 9 CG p 2 R 4 .40 p .43 10 \ L s 2 R 3 S .63 .07 11 CGR s 2 R 4 S .30 .01 12 A s 2 L 3 S .51 13 L s 2 L 4 S .54 .26 14 L s 2 R 4 S .48 .28 15 L B 2 L 4 B .45 .31 16 WN w 2 L 3 .21 w .62 17 G W 2 R 4 .08 w .72 18 F W 2 L 3 .30 w .49 19 CG p 3 L 1 .35 p .52 20 M/C p 3 L 3 .39 p .54 21 F p 3 L 3 .27 p .57 22 CGR s 3 R 4 S .58 .29 23 A/A s 3 R 2 S .65 .17 24 A/A s 3 R 4 S .69 .16 25 L s 3 L 3 S .60 .32 26 L s 3 R 4 S .61 .25 27 L B 3 R 4 B .51 .21 28 G w 3 R 4 .10 w .67 29 F w 3 R 3 .09 w .61 30 CG p 4 L 3 P .44 .43 31 F p 4 R 4 .33 p .53 32 A p 4 R 2 .38 p .47 33 A p 4 R 4 .42 p .50 34 CGR s 4 R 4 S .62 .26 35 A/A s 4 R 2 S .67 .14 36 A/A s 4 R 4 S .68 .18 37 A s 4 L 3 S .64 .23 38 L s 4 L 4 S .69 .21 39 L s 4 R 4 S. .57 .27 40 A/A s 4 L 2 S .46 .33 41 G w 4 R 4 .20 w .68 Table.; 7 - l ( i ) Con't CNT SN I • PG I II 42 CG P 5 L 3 .37 P .60 43 A P 5 R 4 .40 P .48 44 L S 5 R 4 S .59 .24 45 L s 5 L 3 S .63 .23 46 A/A s 5 R 2 S .64 .12 47 A/A s 5 R 4 S .61 .14 48 L B 5 R 4 B .50 .34 49 CGR B 5 L 4 B .63 .29 50 G w 5 L 3 .07 W .65 51 G w . 5 R 4 .21 W .71 52 F w 5 R 1 .24 w .61 53 F w 5 R 2 .26 w .61 54 CG p 6 L 3 .37 p .54 55 M/C p 6 R 3 .48 p .51 56 F p 6 R 4 .41 p .54 57 L s 6 L . 3 S .64 .15 58 CGR s 6 R 4 S .62 .27 59 M/C s 6 L 1 .38 s .51 60 M/C s 6 L 3 .43 s .52 61 L s 6 R 4 S .50 .34 62 A/A s 6 L 2 S .57 .30 63 A/A s 6 L 4 S .60 .32 64 L •s 6 R 3 S .62 .30 65 A s 6 R 4 S .63 .28 66 WN w 6 L 3 .21 w .64 67 G w 6 R 4 .21 w .71 i n t e r p r e t a t i o n X cumulative proportion of variance accounted f o r = 41% 165 business are picked up on t h i s b a s i s . c) that both content and s t r u c t u r a l influences are at work. Any of these i n t e r p r e t a t i o n s are po s s i b l e . Note that the r e s u l t i s forced to two f a c t o r s . Thus sports and women's may be.dominant s t r u c t u r a l influences f o r c i n g the P and B variables into one or the other because they are less important dimensions. This i s supported by the less cle a r loadings of P and B v a r i a b l e s , i . e . , generally they load somewhat on both factors 1 and 2. However despite these mixed re s u l t s a l l the B variables are picked up by factor 1 and a l l but one of P variables by f a c t o r 2, suggesting "some underly-ing s i m i l a r i t y i n content overiding s t r u c t u r a l i n f l u e n c e . This p o s s i b i l i t y can be investigated by inspection of the content of the quarter pages with regard, to what i s common between the P and W variables and what i s common between the S and B v a r i a b l e s . For t h i s Figure 17 must be examined. C l a s s i f y i n g these variables according to theitfr highest loading the 2 factors have the following content: Factor 1 q/p content q/p content q/p content coding coding _ coding 3 27 L (P) 47 A/A 10 L 30 CG (B) 48 L (B) 11 CGR 34 CGR 49 CGR (B) 12 A 35 A/A 57 L 13 L 36 A/A 58 CGR 14 L ! J 37 A 61 L 15 L (B) 38 L 62 A/A 22 CGR 39 L 63 A/A 23 A/A 40 A/A 64 L 24 A/A 44 L 65 A 25 L 45 L 26 L 46 A/A 166 Factor 2 q/p content q/p content q/p content — coding coding coding i ' CG 19 CG (P) 50 G 2 CGR 20 M/C (P) 51 G 4 W/C (P) 21 F (P) 52 F 5 G (P) 28 G 53 F 6 W/C 29 F 54 CG (P) 7 F 31 F (P) 55 M/C (P) 8 F 32 A (P) 56 F (P) 9 CG (P) 33 A (P) 59 M/C (S) 16 WN 41 G 60 M/C (S) 17 G 42 CG (P) 66 WN 18 F 43 A (P) 67 G Factor 1 i s c l e a r l y dominated by l i q u o r , c i g a r , automobile and auto-accessory advertisements and f a c t o r 2 by women's and men's clothing, f u r n i t u r e , grocery and wine advertisements (where the variables are from p u b l i c a f f a i r s or business a P or B i s indicated, otherwise a l l variables of factor 1 are S and those of factor 2 are W) . The most obvious content i n t e r p r e t a t i o n of factors 1 and 2 i s that they are male and female f a c t o r s , i . e . , the quarter pages group i n t o those of primary i n t e r e s t to men and those of primary i n t e r e s t to women ( i t w i l l be r e -c a l l e d that sex was the cle a r e s t discriminator between readership of the sports and women's news). Some of the above variables do not i n t u i t i v e l y seem to f i t such a pattern, e.g., #2-El Producto c i g a r s , or #32 and #34-Shell gasoline, which load with f a c t o r 2 on the basis of proximity to other P v a r i a b l e s . In-t e r e s t i n g l y , v a r i a b l e 20 i n the pu b l i c a f f a i r s section and variables 59 and 60 i n the sports section load on f a c t o r 2 with the W variables - t h e i r content i s men's clot h i n g which might be expected to load with S var i a b l e s . This could i n d i c a t e that the content i n t e r p r e t a t i o n i s incor r e c t or, a l t e r n a t i v e l y , 167 that men's clothing advertisements may hold a greater i n t e r e s t for women than men. I f , i n f a c t , the male/female content i n t e r p r e t a t i o n i s correct then sex should be strongly r e l a t e d to f a c t o r scores. This w i l l be i n v e s t i -gated at a l a t e r point. I t should be r e - i t e r a t e d that the above content i n t e r p r e t a t i o n i s a purely tentative one. Underlying s t r u c t u r a l influence (S and W) may w e l l be at work with content acting to force the P and B variables to one factor or another. What can be concluded? This r e s u l t alone does not allow a concrete conclusion, however: a) The factors could be explained by both content and managerial structure - to some extent these coincide ; and. thisjhismlthe^fios.tlllkelyr jhiSerptetation. b) What i s i n t e r e s t i n g i s that an i n t e r p r e t a b l e pattern e x i s t s where 2 factors account f o r 41% of the variance i n 67 v a r i a b l e s . Further, i t i s reasonable to expect that whether due to s i m i l a r i t y i n underlying content or proximity to d i f f e r e n t kinds of news content, sex may be strongly r e l a t e d to each dimension. For future reference i t i s convenient to l a b e l factors 1 and 2 as ''X' and 'Y'. The procedure, as indicated, i s to expand the number of factors i n v e s t i g a t i n g f o r possible emergence of a P (or even B) f a c t o r , for the clearer emergence of S and W f a c t o r s , f o r possible s p l i t t i n g of the X and Y factors according to time structure (that i s , factors i n t e r p r e t a b l e as issues) or the appearance 168 of possible content f a c t o r s . Table 7 - l ( i i ) : The 3 factor s o l u t i o n accounts for approximately 44% of the variance i n the o r i g i n a l variables (only a 3% improvement over the two f a c t o r s o l u t i o n ) . Except for the disproportionate number of S variables, i f the p u b l i c a f f a i r s section were i n fact an underlying s t r u c t u r a l dimension i t i s reasonable to expect i t s emergence as a t h i r d f a c t o r a f t e r X and Y which could then be more c l e a r l y i n t e r p r e t a b l e as sports and women's s t r u c t u r a l dimensions. This to some degree i s a better t e s t of working hypothesis than the 4 f a c t o r s o l u t i o n because of the few number of B v a r i a b l e s . There i s no i n d i c a t i o n of such a r e s u l t . Instead the large X factor s p l i t s approx-imately i n t o X]_ 2 3 and x 4 , 5 , 6 seemingly r e f l e c t i n g proximity i n time, i . e . , the s t r u c t u r a l e f f e c t s of the issues emerge. Note that the P variables are s p l i t among Y, X]_ 2,3 and X4,5,6 with highly mixed loadings. There i s l i t t l e systematic about t h i s except that the P variables are s t i l l strongly associated with Y and that there i s an i n d i c a t i o n of d r i f t , ( d r i f t , i n the context of t h i s paper, refers to the variables having mixed loadings across the factor space and the tendancy for these to change as the factor space i s expanded). Note that unlike the 2 f a c t o r s o l u t i o n not a l l P v a r i a b l e s load p r i m a r i l y with the W va ri ab les on Y. B remains strongly associated with X, s p l i t t i n g evenly between X-^2,3 and ^ts,6' This r e s u l t more completely undermines the s t r u c -t u r a l hypothesis concerning p u b l i c a f f a i r s : the P variables seem more strong-l y associated with S and W variables than with each other. Table 7-l(reconsidered): The four f a c t o r s o l u t i o n was of course presented e a r l i e r and represents the test of the working hypothesis on mana-g e r i a l s t r u c t u r e . Within the context of the exploratory analysis t h i s r e s u l t 169 makes considerably more sense. C l e a r l y again the p u b l i c a f f a i r s f a c t o r has not emerged and X now s p l i t s i n t o 3 f a c t o r s . The P variables continue to d r i f t across the f a c t o r s . The loadings are such to instance, with variables 66 and 6 7 factors 1 and 4, f a c t o r 4 could be the Y fa c t o r continues to stand up the P variables d r i f t out. give added emphasis to time structure. For s t a r t i n g to d r i f t , i . e . , loading on both interpreted as issue #6. In t e r e s t i n g l y , but i s more dominated by W variables as It can be conjectured at t h i s point the strength of the Y fa c t o r indicates the p o s s i b i l i t y of strong cross issue content c o r r e l a t i o n . A l t e r n a -t i v e l y , the continuing s p l i t t i n g of the X fa c t o r could i n d i c a t e strong within issue c o r r e l a t i o n s . Table 7 - l ( j i i ) : The f i v e f a c t o r s o l u t i o n i s i n t e r e s t i n g p r i m a r i l y because the fa c t o r space remains st a b l e . Factor 1 i s s t i l l best interpreted as Y. Factors 2 and 3 represent s p l i t s i n X: x2,3 4 a n c * x4 5' Factor 4 remains stable as issue #6. The added f i f t h f a c t o r strongly indicates d r i f t i n the W va r i a b l e s and i s best interpreted as issue #1. Thus s t r u c t u r a l prox-imity due to the issues i s emerging strongly as the major explanatory device for added f a c t o r s . Table 7 - l ( i v ) : This table presents the 8 factor s o l u t i o n . The 6 and 7 fa c t o r solutions are omitted. In these l a t t e r factors the basic s p l i t t i n g 170 Table 7 - l ( i i ) : Analysis of Advertising Data (Panel A) - 3 Factor Solution a/p ' factor loadings matrix • I II III 1 -.34 .02 P-.45 2 -.28 .03 P-.43 3 -.14 .24 S-.54 4 W-.57 -.00 -.25 5 W-.60 .04 -.25 6 W-.51 -.05 -.24 7 W-.52 .15 -.28 8 W-.52 .16 -.28 9 -.39 .18 P-.47 10 -.01 .36 S-.55 11 -.05 .14 S-.29 12 -.25 .27 S-.50 13 -.21 .29 S-.53 14 -.23 .23 S-.51 15 -.26 .16 B-.56. 16 W-.60 .16 -.23 17 W-.71 .08 -.13 18 W-.46 .20 -.30 19 P-;48 .10 P-.50 20 P-.50 .20 P-.44 21 P-.54 .13 -.34 22 -.23 .38 S-.49 23 1 -.11 .36 S-.60. 24 -.09 .41 S-.60 25 -.26 .30 S-.61 26 -.19 .27 S-.65 27 -.16 .29 S-.48 28 W-.66 .08 -.15 29 . W-.60 .05 -.17 30 P- r39 P .39 -.28 31 P-^O P .35 -.18 32 F-.43 .24 -.37 33 P-.46 .27 -.40 34 -.20 .42 S-.50 35 -.08 S .57 -.39 36 -.12 S .62 -.35 37 -.17 S .50 -.43 38 -.14 S .58 -.41 39 -.21 .40 S-.45 40 -.29 S .44 -.23 41 W-.66 .29 -.06 171 Table 7-1 ( i i ) Con't a/p f a c t o r loadings matrix I II III 42 P-.56 .34 -.24 43 P-.45 P .43 -.18 44 -.19 S .51 -.33 45 -.17 S .59 -.30 46 -.06 S .70 -.17 47 -.09 S .67 -.17 48 -.30 B .49 -.24 49 -.23 B .64 -.25 50 W-.64 .18 .02 51 W-.69 .33 -.03 52 W-.58 .43 .05 53 W-.58 .44 -.03 54 P-.50 .37 -.20 55 P-.46 P .52 -.20 56 P-.51 P .50 -.12 57 -.09 S .54 -.37 58 -.21 S .56 -.32 59 S-.47 S .40 -.18 60 S-.48 S .44 -.22 61 -.29 S .45 -.29 62 -.25 S .63 -.18 63 -.27 S .65 -.20 64 -.24 S .61 -.27 65 -.22 S .69 -.20 66 W-.62 .28 -.09 67 W-.70 .31 -.07 i n t e r p r e t a t i o n Y X4,5,6 X l , 2 , 3 cumulative proportion of variance accounted for = 44%. 172 Table 7 - 1 ( i i i ) : Analysis of Advertising Data (Panel A) - 5 Factor Solution I I III IV 1 .12 .27 .03 .13 P .57 2 .09 .27 .03 .14 P .50 3 -.02 S' .43 .19 .24 .39 4 W .41 .12 .04 .05 W .51 5 W .46 .11 .14 .00 W .51 6 W .35 .09 .03 .01 W .52 7 .28 .05 .24 .16 W .68 8 .29 .05 .25 .15 W .68 9 .30 P .43 .12 .14 .31 10 .03 S .59 .30 .12 .02 11 -.09 S .27 .0.6 .16 .08 12 .25 S .54 .22 .09 .11 13 .21 S .57 .24 .09 .10 14 .19 S .49 .20 .10 .21 15 .20 B .54 .13 .07 .23 16 W .57 .23 .11 .10 .23 17 W .66 .11 .07 .04 .30 18 W .37 .25 .15 .17 .31 19 P .44 P .55 -.06 .16 .17 20 P .41 P .43 .05 .25 .27 21 P .53 P .40 .01 .12 .13 22 .25 S .55 .28 .18 .05 23 -.01 S .53 .29 .26 .31 24 -.02 S .53 .33 .31 .31 25 .27 S .70 .11 .23 .04 26 .15 S .68 .11 .24 .13 27 .16 B .52 .19 .15 .05 28 W .67 .22 -.02 .07 .11 29 W .59 .20 .01 .02 .17 30 P .37 .30 .31 .20 .13 31 P .49 .21 .21 .26 .13 32 P .38 P .38 .14 .19 .20 33 P .39 P .39 .15 .24 .24 34 .20 S .55 .31 .21 .06 35 .03 .33 S .58 .23 .21 36 .07 .28 S .63 .25 .22 37 .20 S .44 S .54 .08 .09 38 .17 S .43 S .57 .17 .05 39 .23 S .49 .34 .14 .06 40 .23 .15 S .50 .15 .29 41 W .69 .13 .20 .15 .08 Table 7-1 ( i i i ) Con't I II 42 P .55 .25 43 P .44 .19 44 .24 .37 45 .20 .34 46 .07 .13 47 .08 .12 48 .34 .28 49 .31 .34 50 W .67 .04 51 W .70 .08 52 W .63 .02 53 W .64 .03 III IV V .32 .13 .21 .39 .19 .15 S .53 .11 .04 S .55 .21 .03 S .78 .18 .11 S .73 .21 .13 B .49 .12 .05 B .56 .24 -.08 .10 .11 .07 .23 .20 .14 .35 .19 .03 .36 .19 .02 54 P .45 .24 .12 P .41 .13 55 .36 .20 .20 P .60 .19 56 P .42 .14 .20 P .57 .17 57 .03 S .42 .20 S .57 .02 58 .15 .37 • .21 S .60 .05 59 .34 .'17 .07 S .59 .24 60 .36 .20 .11 S .60 .24 61 .24 .34 .14 S .51 .05 62 .12 .13 .34 S .67 .22 63 .14 .16 .33 S .69 .21 64 .24 .37 .26 S .58 -.06 65 .23 .29 .35 S .60 -.07 66 W .60 .14 .07 .32 .12 67 W .62 .09 .05 W .43 .21 i n t e r p r e t a t i o n : Y x2,3 4 x4,5 Iss#6 Iss#l cumulative proportion of variance accounted for = 50%. 174 of the X factor continues while the Y factor remains quite strong. According-l y , i t might have been expected that ,in the 7 f a c t o r solution,the l a s t 6 f a c -tors would r e f l e c t the s i x issues of the study. In f a c t t h i s i s almost the case. Three factors can be loosely interpreted as issues 2,3 and 4, (factors r e f l e c t i n g issues 1 and 6 remain strong). However, issue 5 does not emerge; instead there i s a vague factor more or less representing a s e r i e s of snowtire advertisements. In expanding the space to 8 and 9 factors however i t was found that the "snowtire" factor disappeared while a f a c t o r r e f l e c t i n g issue 5 did emerge and remain st a b l e . As the 8 f a c t o r s o l u t i o n represents the most int e r p r e t a b l e and stable f a c t o r structure at t h i s point, i t i s presented i n Table 7 - l ( i v ) . Further, a d e s c r i p t i o n of each f a c t o r i n terms of content, managerial section, issue, page and quadrant presented as follows: Factor 1 q/p content managerial xssue page quadrant coding section 5 G W 1 R 4 16 WN W 2 L 3 17 G w 2 R 4 18 F w 2 L 3 21 F p . 3 L 3 28 G w 3 R 4 29 F w 3 R 3 31 F p 4 R 4 41 G w 4 R 4 42 CG p 5 L 3 43 A p 5 R 4 50 G w 5 L 3 51 G w 5 R 4 52 F . w 5 R 1 53 F w 5 R 2 54 CG p 6 L 3 55 M/C p 6 R 3 56 F p 6 R 4 66 WN w 6 L 3 67 G w 6 R 4 Factor 2 q/p content managerial coding section 54 CG P 55 M/C P 56 F P 57 L S 58 CGR S 59 M/C S 60 M/C S 61 L S 62 A/A S 63 A/A S 64 L S 65 A S 66 WN W 67 G W issue page quadrant 6 L 3 6 R 3 6 R 4 6 L 3 6 R 4 6 L 1 6 L 3 6 R 4 6 L 2 6 L 4 6 R 3 6 R 4 6 L 3 6 R 4 Factor 3 9 CG P 10 L S 11 CGR S 12 A S 13 L S 14 L S 15 L B 23 A/A S 24 A/A S 26 L S 27 L S 2 R 4 2 R 3 2 R 4 2 L 3 2 L 4 2 R 4 2 L 4 3 R 2 3 R 4 3 R 4 3 R 4 Factor 4 30 CG P 31 F P 34 CGR S 35 A/A S 36 A/A S 37 A S 38 L S 39 L S 40 A/A S 4 L 3 4 R 4 4 R 4 4 R 2 4 R 4 4 L 3 4 L 4 4 R 4 4 L 2 176 Factor 5 q/p. content managerial coding "section 19 CG P 20 M/C P 21 F P 22 CGR S 24 A/A S 25 L S 26 L S 27 L B 32 A P 33 A P issue page quadrant 3 L 1 3 L 3 3 L 3 3 R 4 3 R 4 3 L 3 3 R 4 3 R 4 4 R 2 4 R 4 Factor 6 1 CG P 2 CGR P 3 S 4 W/C W 5 G W 6 W/C W 7 F W 8 F W 1 L 3 1 L 4 1 1 L 3 1 R 4 1 L 1 1 R 2 1 - R 4 Factor 7 43 A P 44 L S 45 L S 46 A/A S 47 A/A S 48 L B 49 CGR B 5 R 4 5 R 4 5 L 3 5 R 2 5 R 4 5 R 4 5 L 4 Factor 1 i s c l e a r l y s t i l l dominated by the variables of the Y fac t o r although a number have d r i f t e d out. It has representation of both W and P v a r i a b l e s , includes variables from a l l issues and includes groceries, f u r n i t u r e , cigarettes and wine advertisements among others. Again, t h i s seems to be a content factor possibly of primary i n t e r e s t to female readers. 177 Factor 2 includes representation from P, S and W va r i a b l e s , and a v a r i e t y of content categories. I t also includes a l l var ia bl es from issue 6 and can be thus interpreted. This has been a very stable f a c t o r since the 4 factor space. Factor 3 includes representation from P, S and B variables and a v a r i e t y of content categories. It includes a l l non-W variables from issue 2 and some from issue 3. I t can be loosely interpreted as an X/issue 2 f a c -t o r . Factor 4 s i m i l a r l y represents issue 4 variables not including a l l the possible P va r i a b l e s i n that i s s u e . It can be loosely interpreted as an X/issue 4 f a c t o r . Factor 5, again s i m i l a r l y , represents issue 3 non-W v a r i a b l e s . I t also includes two P variables from issue 4. I t can be loosely interpreted as an X/issue 3 f a c t o r . Factor 6,like f a c t o r 2, includes representation from P, S and W va r i a b l e s , mixed content and, most importantly, a l l variables from issue 1. This again has been a stable f a c t o r and can be interpreted as issue 1. Factor 7 l i k e factors 3,4 and 5 can be interpreted as X/issue 5. Factor 8 cannot be meaningfully interpreted. 178 Table, 7-1(iv): Analysis of Advertising Data (Panel A) - 8 Factor Solution I I I I I I IV V VI VII VIII 1 .05 .16 -.15 .11 .18 P .61 .01 -.02 2 .05 .12 -.18 .10 .23 P .46 .01 -.18 3 -.07 .29 -.28 .24 .20 S .41 .11 -.02 4 .35 .10 -.10 .11 .00 w .61 -.02 .09 5 W .40 .04 -.08 .07 .07 w .61 .13 .12 6 .27 .06 -.01 .13 .06 w .63 -.03 .11 7 .27 .12 -.08 .03 .10 w .64 .26 -.24 8 .29 .11 -.07 .05 .09 W" ~.63 .25 -.25 9 .25 .17 P-.34 .11 .27 .33 .11 .03 10 .03 .18 S-.60 .34 .11 .02 .12 -.01 11 -.02 .14 S-.46 .15 -.13 -.01 -.13 -.29 12 .26 .12 S-.66 .12 .12 .08 .15 -.02 13 .21 .13 S-.66 .14 .14 .08 .17 -.01 14 .15 .15 S-.46 .11 .22 .23 .19 .07 15 .19 .11 B-.59 .15 .17 .22 .04 -.03 16 W .55 .14 -.34 .05 -.00 .30 .05 .10 17 w .65 .03 -.16 .04 .08 .33 .03 -.02 18 w .42 .12 -.40 .10 .04 .21 .04 -.32 19 .36 .14 -.27 .07 P .61 .15 -.00 -.00 20 .36 .19 -.20 .08 P .55 .19 .08 -.18 21 p .49 .09 -.28 .03 P .43 .09 .04 -.06 22 .18 .20 -.29 .23 P .48 .06 .28 .07 23 -.01 .24 S-.48 .08 .36 .19 .33 -.24 24 -.03 .27 S-.43 .10 S .41 .18 .38 -.27 25 .18 .24 -.39 .15 S .63 .03 .16 .06 26 .08 .29 S-.48 .17 s .46 .14 .09 .05 27 .09 .21 B-.31 .25 B .33 .10 .14 .14 28 w .61 .09 -.10 .03 .27 .19 -.00 .18 29 w .55 -.01 -.03 .16 .32 .17 -.04 -.04 30 .34 .22 -.11 P .42 .20 .17 .12 -.00 31 p .48 .21 -.04 P .34 .24 .10 .05 -.17 32 .29 .16 .05 .28 P .62 .19 .13 -.02 33 .29 .22 .05 .28 P .62 .24 .14 -.05 34 .15 .24 -.22 S .59 .36 .08 .06 -.01 35 .06 .21 -.25 S .61 .05 .13 .26 -.32 36 .11 .22 -.22 S .55 .07 .13 .37 -.34 37 .17 .13 -.21 S .69 .17 .13 .23 .00 38 .15 .21 -.20 S .75 .14 .09 .21 -.04 39 .16 .23 -.21 S .62 .19 .18 .06 .19 40 .27 .10 -.15 S .43 .03 .19 .27 -.36 41 w .65 .15 .02 .22 .20 .14 .12 .06 42 p .52 .18 -.24 .18 .08 .29 .26 .14 43 p .42 .18 -.13 .10 .22 .15 P .43 .00 179 Table 7 - l ( i v ) Con't I II I I I IV V VI VII VIII 44 .22 .18 -.38 .27 .07 .08 S .46 .13 45 .16 .32 -.24 .31 .07 .15 S .48 .26 46 .09 .17 -.08 .27 .10 .06 S .79 -.10 47 .10 .19 -.08 .18 .14 .08 S .80 -.11 48 .32 .19 -.27 .24 .05 .13 B .44 .19 49 .29 .31 -.20 .40 .13 .00 B .44 .18 50 W .67 .09 -.06 1 .09 .07 .09 .03 -.01 51 W .70 .17 -.06 .09 .15 .14 .21 -.05 52 w .68 .13 -.14 .06 .04 -.02 .33 -.16 53 w .68 .14 -.14 .09 .04 -.02 .32 -.14 54 p .46 P .43 -.25 .11 .07 .15 .04 -.04 55 p .38 P .58 -.15 .18 .12 .17 .10 -.18 56 p .45 P .52 -.09 .12 .14 .13 .13 -.21 57 -.00 S .65 -.24 .24 .18 .09 .12 .09 58 .11 S .65 -.14 .23 .25 .11 .16 .07 59 .36 S .52 -.11 .06 .21 .16 .02 -.32 60 .38 S .55 -.14 .11 .20 .17 .03 -.30 61 .19 S .58 -.14 .26 .16 .16 .03 .15 62 .14 S .63 -.08 .06 .14 .16 .36 -.23 63 .16 S .66 -.07 .09 .17 .17 .34 -.21 64 .23 S .68 -.32 .20 .03 .04 .16 .19 65 .25 S .64 -.26 .21 .04 -.04 .25 .02 66 w .58 W .34 -.09 .11 .09 .20 -.00 .07 67 w .61 W .40 -.01 .10 .15 .24 -.01 -.07 i n t e r p r e t a t i o n Y issue 6 X/iss2 X/iss3 X/iss4 X / i s s l X/iss5 „ cumulative proportion of variance accounted for = 59%. 180 Right and l e f t pages and page p o s i t i o n i n g do not contribute to the i n t e r p r e t a t i o n of any f a c t o r . The fact that each factor loads on quadrant 3 and 4, i . e . , the bottom h a l f of the page, i s i r r e l e v a n t as almost a l l the quarter pages are from quadrants 3 and 4. I t i s possible that factors might eventually emerge r e f l e c t i n g these influences but these would account for a very small percentage of the variance. summary and discussion The expansion of the f a c t o r space i s halted at t h i s point, p a r t i a l l y because i t i s i n c r e a s i n g l y cumbersome to present and discuss the r e s u l t s . However, fu r t h e r expansion to 9 factors l e f t factors 1 to 7 i n stable r e l a -t i o n s h i p . Factors 8 and 9 were neither stable nor e a s i l y interpreted. Some-what incongruously the "snowtire" factor does not re-emerge which i s d i f f i c u l t , to account f o r . This i s l i k e l y r e l a t e d to the establishment of the issue 5 f a c t o r . A second reason f o r not pursuing the analysis further i s that the amount of variance being accounted f o r by a d d i t i o n a l factors i s i n c r e a s i n g l y small, i . e . , l e s s than 2%. Further, although the r e s u l t s are not unequivocally interpretable,there i s consistency to the expanding factor spaces and the main factors are r e l a t i v e l y s t a b l e . The following general conclusions can be drawn: a) F i r s t , there i s a strong stable f a c t o r , Y,representing content i n the women's section and associated content i n the p u b l i c a f f a i r s and sports sections. A number of v a r i a b l e s , including some W, d r i f t out as the f a c t o r space 181 is increased. This suggests a strong content influence in determining the Y factor, b) Second, there i s a nonstable factor, X, representing primarily S variables. As the factor space is expanded this factor breaks down and issue factors emerge partially reflecting only non-W variables (issues 2, 3, 4 and 5) and partially reflecting a l l variables in the issue (issues 1 and 6). These conclusions await cross-validation on panels B and C. However, some further exploration can be undertaken on the question of why the 'Y' and 'X' factors behave differently as the factor space is expanded. Is this l i k e l y a generalizable result or merely an artifact of the particular variables being analyzed? To investigate this two,sets of variables have been chosen for further analysis: the 17 W variables loading on factor Y and 27 S variables loading on factor X. The object is to eliminate sources of variance due to managerial structure and investigate to what extent content or issue (time) structure influences the extraction of factors. An texpl'o'Ea$oEy»- strategy was again followed, i.e., manipulation of the number of factors in search of an interpretable factor space. The results are presented in Tables 7-2(i) and ( i i ) . In each case only one factor matrix is presented, that thought to be most interpretable while consistent with l a r -ger and smaller factor spaces. Table 7-2(i): This result provides some insight into the sta b i l i t y 182 of the Y fa c t o r . Although there are mixed loadings, a l l variables content analyzed as groceries or wine load on f a c t o r 1. This would i n d i c a t e s e l e c -t i v e exposure to such content over the s i x issues of the study. Hence the strength of Y fac t o r may be p a r t i a l l y l a r g e l y a t t r i b u t a b l e to high i n t e r r e -l a t i o n s h i p s among grocery advertisements overtime. This makes i n t u i t i v e sense as these are frequently purchased products f o r which information would be sought r e g u l a r l y . The remaining factors are generally i n t e r p r e t a b l e on an issue b a s i s . Factors 2, 5 and 6 c l e a r l y r e f l e c t issues 5, 2 and 3. Issue 4 does not emerge - i t i s only represented by a sing l e grocery v a r i a b l e which loads on f a c t o r 1. Factors 3 and 4 r e f l e c t the large issue 1, i d e n t i f y i n g a s p l i t between 3 variables concerning women's clothing and groceries and 2 variables concerning f u r n i t u r e . Table 7-2 ( i i ) : This r e s u l t , s i m i l a r l y , represents the most i n t e r -pretable f i t that can be brought to the sports data. The factor space i s remarkably s i m i l a r to that of the women's data. Factors 1 to 5 can be i n t e r -preted as issues 6, 4, 3, 5 and 2 re s p e c t i v e l y . Likewise, there i s a content "snowtire" f a c t o r but unlike the "grocery" factor above i s of less importance than the issue f a c t o r s . In conclusion, the most basic s p l i t i n the data i s that between the X and Y f a c t o r s . It i s speculated that this may r e s u l t from c e r t a i n a d v e r t i s -ing being of primary i n t e r e s t to women and other advertising being of i n t e r e s t to men. The 2 fa c t o r s p l i t i s supported by the fac t that t h i s s o l u t i o n accounts f o r 41% of the variance i n the o r i g i n a l 67 variables and that a d d i t i o n a l f a c -tors account i n d i v i d u a l l y for less than 3% of the variance. The addition of 183 some 6 variables i n t o t a l adds only 16%. The subsequent s p l i t t i n g of the fa c t o r space c l e a r l y depends both on the content of the va r i a b l e s and issue or time s t r u c t u r e . Whether one overrides the other depends on the products advertised and the repetitiveness of the ad v e r t i s i n g . This v a r i a t i o n probably accounts f or the d i f f e r e n t s t a b i l i t i e s of the X and Y f a c t o r s . It can be expected i n the analysis of panels B and C that the X and Y factors w i l l be stable i n the 2 factor s o l u t i o n and that expanded fa c t o r spaces w i l l p a r t i a l l y r e f l e c t issue s t r u c t u r e . However there i s no reason to expect that expanded fa c t o r spaces generally, nor the r e s u l t s i n Table 7-2(i) and ( i i ) , c a n be cross-validated as both the number and content.of v a r i -ables analyzed w i l l change. Table 7-2(i); Analysis of Advertising Data i n Women's Section - Panel A 184 de s c r i p t i o n of quarter pages I I I I I I IV V VI 1 W/C -.18 .11 W/C-.78 -.15 -.18 .12 2 G G-.35 .12 G-.65 -.25 -.11 .07 . 3 - W/C -.08 .07 W/C-.82 -.14 -.05 .16 4 F -.13 .16 -.22 F-.91 -.14 .11 5 F -.18 .14 -.23 F-.90 -.13 .10 6 WN WN-.42 .14 -.34 -.02 WN-.59 .10 7 G G-.44 .14 -.34 -.10 G-.53 .18 8 F -.12 .20 -.03 -.22 F-.81 .15 9 G G-.38 .08 -.18 -.10 -.25 G .68 10 F -.17 .21 -.17 -.13 -.07 F .83 11 G G-.66 .20 -.18 -.07 -.17 .29 12 G G-.47 G .44 -.16 -.02 -.14 .34 13 G G-.53 G .48 -.20 -.08 -.22 .22 14 F -.20 F .92 -.10 -.15 -.14 .10 15 F -.19 F .92 -.10 -.16 -.14 .13 16 WN WN-.82 .12 -.11 -.15 -.16 .08 17 G -.74 .19 -.19 -.18 -.12 .17 i n t e r p r e t a t i o n groceries issue 5 issue 1: issue 1: issue 2 issue 3 and wine W/C & G F cumulative proportion of variance accounted f o r = 75.7%. 185 Table 7-2(ii) : Analys i s of Advertising Data i n Sports Section - Panel . desc r i p t i o n of fac t o r loadings matrix quarter pages I I I III IV • V VI 1 . L .13 .33 L .37 -.21 L .37 .04 2 A .14 .14 .15 -.09 A .83 .14 3 L .14 .16 .24 -.14 L .75 .08 4 L .19 .12 .25 -.06 L .56 .23 5 CGR .21 .30 CGR .47 -.33 .24 -.00 6 A/A .13 .05 A/A .77 -.16 .20 A/A .36 7 A/A .17 .06 A/A .76 -.24 .14 A/A .39 8 L .26 .27 L .63 -.08 .33 .03 9 L .22 .32 L .66 -?06 .29 -.03 10 CGR .23 CGR .63 .33 -.09 .12 .14 11 A/A .12 A/A .48 .15 -.22 .17 A/A .58 12 A/A .17 A/A .39 .16 -.26 .16 A/A .67 13 A .11 A .71 .15 -.25 .20 .26 14 L .18 L .74 .16 -.23 .12 .30 15 L .31 L .70 .10 -.03 .18 .13 16 A/A .10 A/A .36 .11 -.14 .15 A/A .60 17 L .19 .30 .17 L-.57 .39 -.04 18 L .30 .37 .18 L-.57 .23 -.03 19 A/A .14 .14 .14 A/A-.82 .05 A/A .35 20 A/A .19 .06 .18 A/A-.81 .06 A/A .34 21 L L .62 .29 L .41 -.18 -.00 -.04 22 CGR CGR .65 .31 .39 -.15 -.02 .02 23 L L .64 .37 .09 -.01 .15 .05 24 A/A A/A .68 -.06 .19 -.17 .12 A/A .51 25 A/A A/A .71 -.00 .17 -.18 .13 A/A .49 26 L L .71 .23 .12 -.19 .31 .04 27 A A .69 .16 .07 -.23 .33 .17 i n t e r p r e t a t i o n issue 6 issue 4 issue 2 issue 5 issue 1 A/A cumulative proportion of variance accounted f o r = 69%. 186 Panel B Panel B - working hypothesis H^: Figure 7 indicates that a t o t a l of 90 advertising quarter pages c l a s s i -f i a b l e i n t o 6 content categories were presented to panel B over the s i x issues of the study. However, only 56 of these f e l l within s t r u c t u r a l sections: 8 i n public a f f a i r s , 20 i n sports, 17 i n women's, 2 i n business, and 9 i n entertainment. As a r e s u l t of there being only 2 business v a r i a b l e s , the f a c -t o r space i n the i n v e s t i g a t i o n of H^ was suppressed to only 4 f a c t o r s . The resu l t s are presented i n Table 7-3. Inspection indicates mixed re s u l t s as with panel A. Again, however, fa c t o r 1 could be interpreted as women's, fa c -tor 2 as sports, and f a c t o r 4 as entertainment. The contingency table analysis i s presented i n Figure 18. FIGURE 18: Contingency Table Analysis - Panel B Women' s empirical 17 0 15 X 2 = 35.19 (p <.001) 6 33 41 21 35 56 Sports empirical 9 10 19 x-2 = = 4.72 (p <.05) 6 31 37 15 41 56 187 Table 7-3; Analysis of Advertising Data - Panel B ( a l l issues) a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n sy stem System I I I III IV 1 (Women's/issue #1) (W/l)x-.54 .05 -.31 .10 2 (W/l) (W/l)x-.56 -.04 -.15 .09 3 (W/l) (W/l)x-.55 .06 -.02 .10 4 (W/l) (W/l)x-.55 -.13 -.45 -.00 5 (W/l) (W/l)x-.53 .10 -.43 .04 6 (Entertainment/iss.#1) -.07 ••01 -.13 (E / l ) x .57 7 (Public A f f a i r s / i s s . # 2 ) -.26 .05 (P/2)x-.29 .25 8 (Sports/2) -.08 .10 (S/2)x-.43 .20 9 (S/2) .03 .02 (S/2)x-.42 .24 10 (W/2) (W/2)x-.52 .07 -.12 .18 11 (E/2) -.24 .08 (E/2)x-.38 .30 12 (E/2) -.11 -.00 1 '• -.13 (E/2)x .79 13 (E/2) -.11 .01 -.10 (E/2)x .81 14 (P/issue #3) (P/3)x-.43 .02 -.41 .01 15 (S/3) .01 .20 (S/3)x-.62 .13 16 (S/3) -.06 .22 (S/3)x-.64 .10 17 (S/3) -.25 .24 (S/3)x-.37 .02 18 (W/3) (W/3)x-.59 .05 -.02 .02 19 (S/issue #4) -.08 .29 (S/4)x-.65 .04 20 (S/4) -.11 .31 (S/4)x-.65 .07 21 (S/4) -.21 .33 (S/4)x-.52 .08 22 (B/4) -.12 .26 (B/4)x-.35 .12 23 (B/4) -.24 .37 (B/4)x-.48 .12 24 (W/4) (W/4)x-.32 .26 -.21 .06 25 (W/4) (W/4)x-.49 .17 -.38 .15 26 (W/4) (W/4)x-.52 .14 -.44 .10 27 (E/4) (E/4)x-.47 .17 -.38 .23 28 (E/4) -.19 .22 -.15 (E/4)x .64 29 (E/4) -.19 .27 -.15 (E/4)x .66 30 (P/issue #5) -.13 (P/5)x .65 -.02 .10 31 (P/5) -.13 (P/5)x .66 -.01 .11 32 (S/5) -.25 (S/5)x .52 -.39 -.04 33 (S/5) -.19 (S/5)x .58 -.35 .06 34 (S/5) -.09 (S/5)x .63 -.43 .05 35 (S/5) -.20 (S/5)x .51 -.34 .03 36 (S/5) .03 (S/5)x .55 -.52 -.05 37 (S/5) -.05 (S/5)x .56 -.54 -.06 38 (W/5) (W/5)x-.71 .13 .11 .11 39 (W/5) (W/5)x-.66 .21 -.08 .01 40 (W/5) (W/5)x-.71 .22 -.07 .06 188 Table 7-3 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system system I II III IV 41 (W/5) (W/5)x-.59 .32 -.04 .01 42 (E/5) -.41 (E/5)x .42 -.04 .19 43 (E/5) -.30 .36 .01 (E/5)x .50 44 (P/issue #6) (P/6)x-.60 .40 -.10 .08 45 (P/6) (P/6)x-.68 .33 -.16 .03 46 (P/6) -.38 (P/6)x .52 -.05 .15 47 (P/6) -.35 (P/6)x .52 -.03 .17 48 (S/6) -.01 (P/6)x .50 -.23 .22 49 (S/6) (S/6)x-.52 .30 -.18 .15 50 (S/6) (S/6)x-.57 .35 -.16 .12 51 (S/6) -.09 (S/6)x .61 -.36 .01 52 (S/6) -.09 (S/6)x .66 -.36 -.04 53 (S/6) -.09 (S/6)x .56 -.26 .06 54 (W/6) (W/6)x-.68 .26 -.02 .13 55 (W/6) (W/6)x-.61 .38 -.07 .13 56 (W/6) (W/6)x-.75 .12 .02 .14 i n t e r p r e t a t i o n women's sports p r o b a b i l i t y <.00i <.05 cumulative proportion of variance accounted for = 45%. entertainment <.00i 189 En t e r t ainmen t empirical 6 3 9 JX 2 = 28.47 (p< .001) 0 47 47 6 50 56 exploratory analysis An exploratory analysis of the f a c t o r r e s u l t s f o r Panel B was con-ducted i n the same manner as f o r panel A. The object was to investigate to what extent the expanding f a c t o r space confirms the r e s u l t of panel A. Figure 19 presents the content and s t r u c t u r a l d e s c r i p t i o n of the 56 variables analyzed. The following tables contain selected f a c t o r spaces which adequately describe the r e s u l t s of the exploratory procedure. Table 7 - 3 ( i ) : This table presents the 3 f a c t o r s o l u t i o n . I t was expected, as r e s u l t of the i n c l u s i o n of entertainment v a r i a b l e s , that the 3 f a c t o r r e s u l t would be the most interpretable f a c t o r space. The 2 f a c t o r s o l u t i o n represented the general 'X* and 'Y' factors of panel A but a number of E variables did not load c l e a r l y . In the 3 factor s o l u t i o n these emerged as a separate f a c t o r . The 3 f a c t o r s o l u t i o n b a s i c a l l y forces factors 2 and 3 of Table 7-3 to combine. C l a s s i f y i n g the variables according to t h e i r highest loading the 3 factors have the following description.-'-1. As with Panel A, a l l variables under Factor 1 are from the women's section except where indicated i n brackets; s i m i l a r l y , a l l variables under Factor 2 are from sports and under Factor 3 from entertainment except where indi c a t e d . 190 Factor 1 q/p content q/p content q/p content coding coding coding 1 W/C 25 F 44 M/C (P) 2 W/C 26 F 45 F (P) 3 - 27 F (E) 47 A/A (P) 4 F 38 G 49 M/C (S) 5 •' F 39 G 50 M/C (S) 10 WN 40 G 54 ... WN 14 F (P) 41 F 55 L 18 F 42 WN (E) 56 W/C 24 L Factor 2 15 A/A 23 A/A (B) 36 A/A 16 A/A 30 A/A (P) 37 A/A 17 L 31 A/A (P) 46 A/A (P) 19 A/A 32 A/A 48 L 20 A/A 33 L 51 A/A 21 A 34 A/A 52 A/A 22 F (B) 35 L 53 A Factor 3 6 M 12 M 7 F (P) 13 M 8 'A (S) 28 M 9 L (S) 29 M 11 F 43 M Factor 1 can be interpreted as being of primary i n t e r e s t to women thus confirming the 'Y' factor of panel A. S i m i l a r l y , f a c t o r 2 can be i n t e r -preted as being of primary i n t e r e s t to men confirming the 'X' fa c t o r . How-ever, f a c t o r 2 i s dominated by advertising i n the automobile and accessory cat-egory suggesting a somewhat narrower i n t e r p r e t a t i o n . Factor 3 presents an i n t e r e s t i n g r e s u l t . Although most of the variables are from the entertainment section, these are a l l content c l a s s i f i e d as movies. Entertainment variables 1 9 1 FIGURE 19: Description of Advertising Quarter Pages - Panel B quarter P a 8 e s  1 2 3 4 5 6 content coding W/C W/C F F M section W W w w w E issue 1 1 1 1 1 1 P.age L L L R R R quadrant 3 1 3 2 2 4 9 10 11 12 13 F A L WN F M M P S S W E E E 2 2 2 2 2 2 2 L L L L L R R 3 3 4 3 3 2 4 14 15 16 17 18 F A/A A/A L F P S S S W 3 3 3 3 3 L R R L L 3 2 4 3 3 19 20 21 22 23 24 25 26 27 28 29 A/A A/A A F A/A L F F F M M S S s B B W W W E E E 4 4 4 4 4 4 4 4 4 4 4 R R L R L R R R L R R 2 4 3 4 3 4 2 4 4 2 4 30 31 32 33 34 35 36 37 38 39 40 41 42 43 A/A A/ A A/A L A/A L A/A A/A G G G F WN. M P P S s s s s s w w w w E E 5 5 5 5 5 5 5 5 5 5 4 5 5 5 R R L R R L R R L L L R L R 2 4 3 3 4 3 2 4 3 1 3 4 3 4 192 FIGURE 19 Con't quarter content pages coding section issue page quadrant 44 M/C P 6 R 3 45 F P 6 R 4 46 A/A P 6 R 2 47 A/A P 6 R 4 48 L S 6 R 3 49 M/C S 6 L 1 50 M/C S 6 L 3 51 A/A S 6 L 1 52 A/A S 6 L 3 53 A S 6 R 4 54 WN W 6 L 3 55 L w 6 L 3 56 W/C w 6 R 4 M - movies 193 not c l a s s i f i e d as movies load on factors 1 and 2. Hence f a c t o r 3 does not seem to support the s t r u c t u r a l c l a s s i f i c a t i o n 'hypothesis but rather content s e l e c t i v i t y . Table 7-3 (reconsidered): Again, as with Panel A, the four f a c t o r s o l u t i o n represents considerably more sense within the context of the ex-panding f a c t o r space than with the hypothesis t e s t . Whereas factor 1, the 'Y' f a c t o r , remains stable, the X f a c t o r s p l i t s . Factor 2 can now be i n t e r -preted as X5 s 6 and factor 3 as X2 3 S4» Moreover, fa c t o r 4 now i s very c l e a r l y i n t e r p r e t a b l e as a movies or 'M' f a c t o r . Table 7 - 3 ( i i ) : Thus f a r , the r e s u l t s l a r g e l y confirm those of panel A. There remains only one area of i n t e r e s t . This i s to investigate whether or not the Y f a c t o r continues to remain stable and the X f a c t o r s p l i t as the f a c t o r space i s expanded. Table 7-3(ii) represents the f i v e f a c -t o r s o l u t i o n . Factor 1 continues as a strong Y factor and f a c t o r 4 as a strong M f a c t o r . Factors 2, 3 and 5 represent a continued s p l i t t i n g on an issue b a s i s . However, fa c t o r 5 i s a somewhat curious r e s u l t . I t i s p r i m a r i l y W variables from issue #1 but picks up P and S variables from issue #3. It i s expected that t h i s i s a nonstable occurrence r e s u l t i n g from the l a t t e r var-iables being of some i n t e r e s t to women and thus i s interpreted as Y f a c t o r rather than as an issue f a c t o r . Table 7 - 3 ( i i i ) : The 6 factor s o l u t i o n again generally confirms the expected r e s u l t . However, there are two features which should be noted F i r s t , f a c t o r 3 i s represented across f i v e of the s i x issues. I t loads on S, 194 P and B variables and could be interpreted as a general X factor. Second, although the Y factor remains relatively stable there is considerable d r i f t of W variables both in,issue #1 and #3. Hence there i s an interesting d i f f e r -ence from the panel A result. Although generally the sp l i t t i n g of factors occurs as expected, the difference between the primary X and Y factors is less pronounced. discussion The exploratory analysis of panel B is halted at this point. The three factor solution accounted for almost 41% of the variance and the addi-tional 3 factors have only increased this by a total of 10%. The 6 factor s o l -ution represented only a 3% improvement over the 5 factor solution. However, the major reason is that the panel B results strongly confirm the panel A results. The basic s p l i t between the X and Y factors i s supported. More-over the emergence of a "movie" factor which does not f i t the X, Y pattern is not inconsistent, as one would not expect movies to be of differential i n -terest to male or female readers. The expanded factor space represents some spl i t t i n g on an issue basis. However, the pattern of X, Y sp l i t t i n g found under panel A is not replicated. This was to be expected due to the differing content analyzed under Panel B. The panel A result was found to be an a r t i -fact of the particular quarter page content, i.e., much of the Y factor was dominated by grocery advertising. As this is not the case with panel B, see figure 19, the expanding factor space quite consistently follows a different pattern.. No further analysis of either W or S variables i s undertaken with panel B as any result would reflect the particular types of advertising f a l l i n g 195 into the women's and sports sections. Thus any result would not be general-izable i n terms of the s t r u c t u r a l hypothesis. 196 Table 7 - 3 ( i ) : Analysis of Advertising Data (Panel B) - 3 Factor Solution q/p factor loadings matrix I I I I I I 1 W .51 -.10 -.23 2 W .53 -.01 -.16 3 w .54 .08 -.13 4 w .49 -.14 -.22 5 w .48 -.14 -.24 6 .09 -.01 E-.56 7 .26 -=?17 P-.34 8 .07 -.32 S-.34 9 -.05 -.25 S-.38 10 w .52 -.07 -.21 11 .23 -.25 E-.42 12 .13 .01 E-.76 13 .14 .02 E-.76 14 p .40 -.24 -.19 15 -.02 S-.52 -.33 16 .04 S-.55 -.31 17 .25 S-.39 -.12 18 w .59 -.02 -.03 19 .07 S-.61 -.25 20 .11 .S-.62 i-.ll 21 .22 S-.56 -.22 22 .14 B-.40 -.20 23 .26 B-.55 -.23 24 w .34 -.31 -.09 25 w .49 -.32 -.27 26 w .51 -.33 -.25 27 E .47 -.32 -.34 28 .24 -.19 E-.59 29 .24 -.22 E-.59 30 .22 P-.50 -.03 31 .23 P-.50 -.03 32 .29 S-.63 -.03 33 .24 S-.65 -.08 34 .15 S-.75 -.09 35 .25 S-.60 -.07 36 .01 S-.76 -.06 37 .09 S-.76 -.05 38 W .73 .02 -.04 39 W .67 -.17 -.01 40 w .73 -.16 -.05 41 w .62 -.24 -.03 42 E .47 -.31 -.11 Table 7-3 ( i ) Con't q/p fa c t o r loadings matrix I I I I I I 43 .37 -.21 E-.37 44 P .65 -.33 -.04 45 P .70 -.31 -.04 46 .45 P-.46 -.06 47 P .42 -.39 -.06 48 .07 S-.51 -.20 49 S .54 -.30 -.16 50 S .61 -.32 -.11 51 .15 S-.69 -.04 52 .15 S-.73 .02 53 .16 S-.59 -.05 54 w .71 -.16 -.08 55 w .66 -.29 -.07 56 w .76 -.02 -.10 i n t e r p r e t a t i o n Y X E cumulative proportion of variance accounted for 198 Table 7 - 3 ( i i ) : Analysis of Advertising Data (Panel B) - 5 Factor Solution factor loadings matrix II I I I IV V 1 W .39 -.07 .15 -.11 W-.49 2 W .42 -.05 -.07 -.15 W-.53 3 w .44 -.02 -.14 -.16 W-.45 4 w .34 -.07 .21 -.02 W-.65 5 w .32 -.11 .15 -.07 W-.66 6 .01 -.05 .04 E-.60 -.21 7 .24 -.05 P .45 -.18 -.03 8 .02 -.20 S .42 -.15 -.12 9 -.10 -.14 S .40 -.20 -.12 10 w .48 -.04 .17 -.16 -.18 11 .20 -.11 E .51 -.22 -.06 12 .07 -.00 .18 E-.79 -.11 13 .08 -.00 .15 E-.81 -.10 14 .30 -.16 .26 -.01 P-.44 15 -.13 S-.46 .30 -.15 S-.38 16 -.06 S-.46 .36 -.11 S-.37 17 .18 S-.37 .14 -.04 -.31 18 w .56 -.00 .06 -.02 -.21 19 -.00 S-.49 S .45 -.02 -.26 20 .04 S-.49 S .50 \ -.03 -.22 21 .19 S-.41 S. .53 -.01 -.06 22 .14 -.28 B .47 -.06 .07 23 .22 B-.45 B .43 -.08 -.10 24 w .34 -.23 .29 -.02 .00 25 w .47 -.14 .58 -.06 -.04 26 w .47 -.15 .61 -.01 -.11 27 E .43 -.18 .49 -.16 -.13 28 .23 -.16 .30 E-.60 .07 29 .24 -.19 .31 E-.61 .12 30 .29 P-.53 .05 -.10 .27 31 .30 P-.54 .04 -.11 .27 32 .26 S-.60 .21 .03 -.13 33 .23 S-.64 .18 -.08 -.05 34 .13 S-.73 .21 -.07 -.06 35 - .23 S-.57 .21 -.03 -.06 36 -.03 S-.72 .24 .02 -.11 37 .05 S-.71 .28 .05 -.13 38 w .73 .03 .05 -.10 -.07 39 w .68 -.10 .23 .03 -.05 40 w .72 -.10 .23 . -.02 -.07 41 w .63 -.22 .10 .00 -.05 199 Table 7 - 3 ( i i ) Con't f a c t o r loadings matrix I II I I I IV V 42 E .47 -.35 .01 -.22 -.03 43 E .36 -.28 -.01 E-.53 .01 44 P .64 -.31 .14 -.07 -.06 45 P .68 -.27 .17 -.01 -.16 46 P .43 -.49 -.12 -.22 -.10 47 P .41 -.49 -.14 -.23 -.08 48 .07 S-.54 .08 -.26 .01 49 S .51 -.28 .14 -.15 -.17 50 S .58 -.31 .14 -.12 -.15 51 .12 S-.71 .09 -.06 -.10 52 .13 S-.75 .08 -.00 -.08 53 .17 S-.58 .15 -.07 .06 54 w .70 -.16 .09 -.12 -.11 55 w .64 -.31 .07 -.13 -.11 56 w .74 -.01 .09 -.12 -.16 i n t e r p r e t a t i o n Y x 3,4 5 x 2 4 M cumulative proportion of variance accounted for = Y l , 3 48.2% 200 Table 7 - 3 ( i i i ) : Analysis of Advertising Data (Panel B) - 6 Factor Solution f a c t o r loadings matrix I II I I I IV V VI 1 .23 .08 -.07 .22 W-.61 -.09 2 .24 .12 .01 .02 W-.68 -.11 3 .30 .08 .07 -.05 W-.60 -.13 4 .20 .03 -.21 .24 W-.69 -.01 5 .17 .03 -.20 .18 W-.72 -.05 6 -.02 .04 -.08 .01 -.22 E-.59 7 .20 .01 -.10 ' P .45 -.05 -.20 8 -.01 .07 -.26 S .39 -.08 -.16 9 ' -.06 .05 S-.32 S .31 .00 -.23 10 W .40 .08 -.00 .22 -.28 -.16 11 .13 .06 -.14 E .52 -.09 -.23 12 .06 .00 -.07 .13 -.11 E-.80 13 .09 .00 -.06 .10 -.09 E-.82 14 P .32 .04 P-.36 .18 P-.35 -.03 15 .04 .01 S-.77 .07 -.05 -.20 16 .05 .06 S-.70 .19 -.12 -.14 17 .30 .07 S-.54 .00 -.12 -.07 18 W .60 .04 -.10 .03 -.20 -.04 19 -.01 .21 S-.55 S .38 -.15 -.03 20 -.02 .27 S-.48 S .47 -.17 -.03 21 .14 .27 S-.35 S .53 -.06 -.02 22 .06 .24 -.16 B .49 .01 -.05 23 .17 .31 B-.36 B .43 -.10 -.08 24 .25 .25 -.08 W .36 -.10 -.00 25 W .32 .18 -.05 W .67 -.19 -.05 26 W .33 .15 -.09 W .69 -.23 -.00 27 .34 .15 -.16 E .53 -.20 -.17 28 .19 .18 -.07 E .30 .02 E-.60 29 .20 .23 -.06 E .33 .05 E-.61 30 .15 P .67 -.02 .19 .08 -.05 31 .16 P .67 -.03 .17 .08 -.06 32 .31 S .35 S-.53 .14 -.03 .02 33 .24 S .46 S-.47 .15 -.01 -.07 34 .16 S .48 S-.57 .15 .03 -.06 35 .29 S .35 S-.49 .14 .03 -.04 36 .01 S .42 S-.63 .15 .02 .02 37 .07 S .43 S-.62 .21 -.02 .05 38 W .74 .04 .05 .08 -.15 -.11 39 W .76 .02 -.17 .19 -.02 -.00 40 W .80 .03 -.17 .19 -.05 -.05 41 W .66 .18 -.16 .10 -.08 -.01 42 E .48 E .32 -.20 .00 -.06 -.21 201 Table 7 - 3 ( i l l ) Con't q/p f a c t o r loadings matrix I I I I I I IV V VI 43 E .37 E .30 -.12 -.01 -.03 E-.52 44 P .52 P-.40 -.04 .26 -.24 -.04 45 P .55 P-.33 -.07 .28 -.32 .01 46 .28 S-.58 -.09 -.00 -.28 -.16 47 .29 S-.56 -.12 -.04 -.23 -.18 48 -.01 S-.50 -.27 .11 -.04 -.22 49 S .39 S-.33 -.09 .23 -.30 -.13 50 S .45 S-.37 -.08 .24 -.31 -.09 51 .01 S-.62 S-.38 .14 -.16 -.01 52 .06 S-.62 S-.44 .11 -.10 .03 53 .02 S-.60 -.19 .26 -.07 -.02 54 W .65 -.21 -.04 .15 -.22 -.11 55 W .59 -.33 -.12 .12 -.21 -.11 56 W .68 -.09 .05 .16 -.29 -.12 i n t e r p r e t a t i o n Y x5,6 x X2,4 cumulative proportion of variance accounted f o r Y l = .51.3% M Panel C 202 Panel C - working hypothesis H^: Figure 7 indicates that a t o t a l of 136 advertising quarter pages c l a s s i -f i a b l e into 8 content categories were presented to panel C over the s i x issues of the study. However, only 96 of these f e l l into s t r u c t u r a l sections: 20 i n pub l i c a f f a i r s , 30 i n sports, 17 i n women's, 1 i n business, 11 i n enter-tainment and 17 i n c l a s s i f i e d s . As a r e s u l t of there being only a sing l e business v a r i a b l e the factor space was suppressed to only f i v e f a c t o r s . The r e s u l t s are presented i n Table 7-4. Again the r e s u l t s are mixed; however, fac t o r 1 can be interpreted as sports, factor 2 as women's and factor 4 as c l a s s i f i e d s . The contingency table analysis i s presented i n Figure 20. FIGURE 20: Contingency Table Analysis - Panel C  Sports empirical + a/_p_ Women's a/p 23 7 30 17 49 66 40 56 96 empirical + -12 5 17 10 69 79 22 74 96 X 2 .=• 19.9 (p < .001) X 2 = 23.4 (p < .001) C l a s s i f i e d s a/_p_ empirical + -12 5 17 0 79 79 12 84 96 203 X 2 = 57.44 (p < .001) exploratory analysis An exploratory analysis of the factor r e s u l t s for panel C was con-ducted i n the same manner as for panels A and B. Figure 21 presents the content and s t r u c t u r a l d e s c r i p t i o n of the 96 v a r i a b l e s analyzed. The follow-ing tables contain selected f a c t o r spaces which adequately describe the r e s u l t s of the exploratory procedure. Table 7-4(i): Here again the three factor s o l u t i o n has been selected for presentation. The i n c l u s i o n of quarter pages from the " c l a s s i f i e d s " section and the r e s u l t of H^ suggest a factor beyond the basic X,Y s p l i t found i n panel A and the X,Y,M s p l i t of panel B. However an "M" factor does not emerge i n the 3,4 or 5 factor s o l u t i o n . Hence the most interpretable ture i s a 3 factor s o l u t i o n representing "X", "Y ", and " C" f a c t o r s . 1 Factor 1 a/p content a/p content a/p content coding coding coding 5 L 42 A/A 68 L 6 A/A 43 L 69 A/A 7 L 44 - 70 L 8 A/A 45 WN (E) 71 A/A 14 F (P) 46 F (E) 72 L 15 A/A (P) 50 A/A (P) 73 A/A 16 A/A (P) 51 A/A (P) 83 A/A (P) 1. As above, a l l v a r i a b l e s under Factor 1 are from the sports section except where indicated i n brackets; s i m i l a r l y , a l l v a r i a b l e s under Factor 2 are from women's and under Factor 3 from c l a s s i f i e d s except where indicated. Factor 1 Con't 204 a/p content a/p content a/p content coding coding coding 17 A/A. 52 A/A 84 A/A (P) 18 A/A 53 A/A 85 L 19 A 54 A/A 88 L 20 L 59 M (E) 89 A/A 21 L 60 M (E) 90 A/A 34 A/A (P) 61 A/A (P) 91 L 35 A/A (P) 62 A/A (P) 92 A 41 A/A 67 A/A Factor 2 14 F (P) 56 W/C 77 M (E) 22 WN 57 W/C 78 F 23 L 58 F (E) 79 F 24 F 59 M (E) 80 G 25 F 60 M (E) 81 G 26 F 63 CL (C) 82 F (P) 32 F. (P) 64 CL (C) 86 M/C (S) 33 F (P) 65 CL (C) 87 M/C (S) 36 F (P) 66 CL (C) 93 WN 47 M (E) 74 G 94 L 48 M (E) 75 G 95 W/C 49 F (P) 76 WN (E) 96 M (E) 55 L Factor 3 9 CL 28 CL 37 A 10 CL 29 CL 38 A 11 CL 30 CL 39 CL 12 F 31 CL 40 CL 27 F Factor 1 can be interpreted as of primary i n t e r e s t to men and factor 2 of primary i n t e r e s t to women thus confirming the 'X' and 'Y' factors of panels A and B. Factor 3 can be interpreted as a " c l a s s i f i e d s " or "C" f a c t o r . However notice that v a r i a b l e s 28 to 31, while loading quite highly on factor 3, load most highly on factor 2. This suggests that the C factor i s perhaps otherwise interpretable or at l e a s t not stable as the factor space i s expanded. Generally, there are a number of mixed loadings i n the 3 factor s o l u t i o n including some va r i a b l e s which do not load highly on any p a r t i c u l a r f a c t o r , e.g. #1 to #4. 205 Table 7-4 (reconsidered): In the f i v e factor s o l u t i o n the 'X' factor continues to be strongly represented i n factor 1 except for the issue #1 v a r i a b l e s which have d r i f t e d out. Factor 5 can be e a s i l y interpreted as an issue #1 f a c t o r . Factor 2 can be interpreted generally as a 'Y' factor but again a number of v a r i a b l e s have d r i f t e d out to load on factor 3. Factor 4 loads strongly on the C v a r i a b l e s of issues 1, 2 and 3 i n d i c a t i n g a stable factor from the 3 f a c t o r s o l u t i o n . Table 7 - 4 ( i i ) : C l e a r l y , a further expansion of the factor space i s required. I t i s of i n t e r e s t to investigate the further s p l i t t i n g of the X and Y f a c t o r s , the s t a b i l i t y of the C factor and whether or not an M factor w i l l emerge as with panel B. The 7 factor s o l u t i o n i s the most int e r p r e t a b l e of the expanded fa c t o r r e s u l t s . F i r s t , there i s some s p l i t t i n g according to time structure. Factors 5 and 6 are c l e a r l y i n t e r p r e t a b l e as issue #3 and issue #1 f a c t o r s . Notice, however, that neither of these issues contain any W v a r i a b l e s . Notice that factor 1 has remained quite stable and can s t i l l be interpreted as an 'X' f a c t o r . This i s to be expected because of the large number of A/A content v a r i a b l e s . However i t i s a d i f f e r e n t r e s u l t from panels A and B where i t was the Y f a c t o r which tended to remain stable. The Y factor i s s t i l l i n t e r p r e t a b l e as factor 3 but there i s some d r i f t of W v a r i a b l e s to factor 2 which i s a l a r g e l y uninterpretable mixture of W and C v a r i a b l e s . Second, f a c t o r 4 i s s t i l l i n t e r p r e t a b l e as a C factor but C v a r i a b l e s i n issue #3 have begun to d r i f t i n d i c a t i n g that the C factor over the s i x issues of the study i s not that stable, i . e . the readership of c l a s s i -f i e d s tends to change over time. Third, factor 7 i s c l e a r l y i nterpretable as a "movie" f a c t o r . A l l E v a r i a b l e s content c l a s s i f i e d as movies and only these v a r i a b l e s load on factor 7. discussion 206 The 7 factor solution accounts for 49.4% of the variance and i s an 11.5% improvement over the 3 factor solution. However i t provides only a 2.5% improvement over the 6 factor solution. Again, a l l r e s u l t s from panels A and B are largely confirmed. The basic s p l i t between the X and Y factors i s supported though somewhat less c l e a r l y . Further, the expanded factor space represents a degree of s p l i t t i n g on an issue basis. As expected, however, the pattern of X, Y s p l i t t i n g does not replicate those of panel A or B. Of p a r t i c u l a r interest i s that the 'X' factor remains quite stable unlike the previous exploratory analyses. The strong "movie" factor found with panel B i s replicated under panel C. However, due to the small propor-tion of movie variables i n panel C compared to panel B t h i s factor did not emerge u n t i l the factor space had been substantially expanded. F i n a l l y , the c l a s s i f i e d s section did emerge as a factor but one which does not l i k e l y r e f l e c t selective exposure to this section across issues. Table 7-4: Analysis of Advertising Data - Panel C ( a l l issues) 207 a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system System I I I I I I IV V 1 (P/D .01 .18 .02 -.06 (P/l)x-.42 2 (P/D .13 .16 .21 .02 (P/l)x-.48 3 (P/D .14 .17 .22 , .01 (P/l)x-.48 4 (P/D .14 .04 .11 -.15 (P/l)x-.42 5 (S/l) .21 .05 -.08 -.01 (S/l)x-.61 6 (S/l) .21 -.06 -.01 -.10 (S/l)x-.66 7 (S/l) .21 .04 -.04 .00 (S/l)x-.66 8 (S/l) .33 -.04 .09 -.06 (S/l)x-.68 9 (C/l) .04 .10 .39 (C/l)x-.49 -.36 10 (C/l) .06 .08 .35 (C/l)x-.53 -.40 11 (C/l) .04 . 17 .37 (C/l)x-.53 -.38 12 (C/l) .02 .06 .30 (C/l)x-.57 -.40 13 (E/l) .13 .21 .09 -.07 (E/l)x-.36 14 (P/2) .29 (P/2)x .42 .07 -.25 -.15 15 (P/2) (P/2)x .43 .40 -.07 -.20 -.33 16 (P/2) (P/2)x .42 .39 -.09 -.21 -.33 17 (S/2) (S/2)x .51 .22 -.13 -.18 -.25 18 (S/2) (S/2)x .52 .24 -.11 -.17 -.25 19 (S/2) .35 (S/2)x .35 -.15 -.24 -.31 20 (S/2) (S/2)x .40 .39 -.09 -.16 -.35 21 (S/2) (S/2)x .44 .18 -.00 -.24 -.29 22 (W/2) .04 (W/2)x .63 -. 15 -.12 -.07 23 (W/2) .28 (W/2)x- .54 .04 -.05 -.11 24 (W/2) .15 (W/2)x .56 -.08 -.14 -.12 25 (W/2) .12 (W/2)x .70 .05 -.19 -.06 26 (W/2) .02 (W/2)x .61 .13 -.14 -.05 27 (C/2) .20 .24 .11 (C/2)x-.72 .00 28 (C/2) .16 .15 .13 (C/2)x-.78 -.02 29 (C/2) .23 .14 .09 (C/2)x-.75 -.07 30 (C/2) .17 .19 .13 (C/2)x-.79 -.04 31 (C/2) .22 .06 .13 (C/2)x-.67 -.02 32 (P/3) .25 (P/3)x .53 .19 -. 19 -.04 33 (P/3) .24 (P/3)x .50 .19 -.21 .01 34 (P/3) .43 (P/3)x .35 .06 -.24 -.03 35 (P/3) (P/3)x .44 .36 .09 -.22 .01 36 (P/3) .30 (P/3)x .42 . 15 -.24 -.03 37 (C/3) .40 .09 .18 (C/3)x-.44 -.00 38 (C/3) .37 .10 .16 (C/3)x-.48 -.00 39 (C/3) .05 .14 (C/3)x .45 -.40 -.09 40 (C/3) .06 .17 .41 (C/3)x-.48 -.08 41 (S/3) (S/3)x .58 .09 -.02 -.23 -.18 42 (S/3) (S/3)x .62 .20 -.03 -.21 -.21 43 (S/3) (S/3)x .58 .31 -.00 -.11 -.04 44 (S/3) (S/3)x .49 .31 -.04 -.25 -.05 45 (E/3) (E/3)x .45 .42 .09 -.25 -.05 Table 7-4 Con't 208 a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system System I II III IV V 46 (E/3) (E/3)x .41 .39 .11 -.29 .05 47 (E/3) .22 (E/3)x .40 .13 -.04 -.20 48 (E/3) .23 (E/3)x .41 .12 -.03 -.21 49 (P/4) (P/4)x .44 .30 .31 -.08 -.06 50 (P/4). (P/4)x .59 .26 .21 -.03 -.07 51 (P/4) (P/4)x .57 .29 .20 -.02 -.09 52 (S/4) (S/4)x .52 .08 .04 -.02 -.22 53 (S/4) (S/4)x .65 .17 .03 -.06 -.27 54 (B/4) (B/4)x .41 .06 .17 -.09 -.23 55 (W/4) .26 (W/4)x .40 .31 -.10 -.07 56 (W/4) .11 (W/4)x .55 .39 .05 -.12 57 (W/4) .07 (W/4)x .58 .36 .02 -.14 58 (E/4) (E/4)x .40 .39 .27 -.15 -.15 59 (E/4) (E/4)x .28 .26 .20 .08 -.23 60 (E/4) (E/4)x .30 .28 .20 .07 -.25 61 (P/5) (P/5)x .56 .19 .30 -.02 -.03 62 (P/5) (P/5)x .60 .20 .32 -.05 -.02 63 (C/5) .01 .14 (C/5)x .61 -.19 -.08 64 (C/5) .34 .07 (C/5)x .56 -.28 -.05 65 (C/5) .02 .15 (C/5)x .60 -.18 .10 66 (C/5) .36 .02 (C/5)x .51 -.24 -.02 67 (S/5) (S/5)x .55 .05 .26 -.12 .05 68 (S/5) (S/5)x .70 .17 .15 -.08 -.01 69 (S/5) (S/5)x .76 .08 .16 -.07 -.03 70 (S/5) (S/5)x .61 .14 .10 -.03 -.09 71 (S/5) (S/5)x .71 -.03 .20 -.05 -.16 72 (S/5) (S/5)x .67 .03 .16 -.11 -.08 73 (S/5) (S/5)x .69 -.04 .16 -.06 -. 16 74 (W/5) .02 (W/5)x .51 .50 -.06 -.06 75 (W/5) .28 (W/5)x .45 .38 -.02 -.04 76 (E/5) (E/5)x .40 .33 .28 -.13 -.10 77 (E/5) .28 .29 (E/5)x .41 .00 -.07 78 (W/5) .28 .27 (W/5)x .52 -.16 .04 79 (W/5) .32 .23 (W/5)x .53 -.19 .06 80 (W/5) .19 .13 (W/5)x .63 -. 19 -.09 81 (W/5) .18 .16 (W/5)x .63 -.19 -.09 82 (P. 6.) .32 (P/6)x .39 .28 -.07 -.09 83 (P/6) (P/6)x .52 .36 .20 -.12 -.08 84 (P/6) (P/6)x .54 .32 .17 -.12 -.12 85 (S/6) (S/6)x .57 .17 .03 -.05 -.20 86 (S/6) .38 .25 (S/6)x .43 -.02 -.11 87 (S/6) .39 .23 (S/6)x .47 -.02 -.14 88 (S/6) (S/6)x .61 .29 .15 .00 -.13 89 (S/6) (S/6)x .66 .12 .13 -.17 -.14 209 Table 7-4 Con't a p r i o r i c l a s s i f i c a t i o n empirical c l a s s i f i c a t i o n system System I II III IV V 90 (S/6) (S/6)x .61 .16 .15 -.16 -.17 91 (S/6) (S/6)x .66 .19 .09 -.06 -.13 92 (S/6) (S/6)x .64 .16 .13 -.07 -.18 93 (W/6) .24 (W/6)x .53 .31 .01 -.04 94 (W/6) .38 (W/6)x .48 .28 -.02 .02 95 (W/6) .10 .44 (W/6)x .47 -.11 -.08 96 (E/6) .34 (E/6)x .36 .21 -.00 -.17 cumulative proportion of variance accounted for = 44.2%. FIGURE 21; Description of Advertising Quarter Pages - Panel C 210 quarter pages 1 2 3 4 5 6 7 8 9 10 11 12 13 content coding CG A/A A/A F L A/A L A/A CL CL CL F M section P P P P S S S S C C C C E issue page L R R R L R . R R L L L L R quadrant 3 1 4 4 3 2 3 4 1 2 3 4 4 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 F A/A A/A A/A A/A A L L WN L F F F F CL CL CL CL P P P S S S S S W w w w w c c c c c 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 L R R R R L L R L L R R R L L L L R 3 1 4 2 4 3 4 4 3 3 1 3 3 1 2 3 4 1 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 F F A/A A/A F A A CL CL A/A A/A L WN F P P P P P C C C C S S S S E E 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 R R R R L R R R R R R L L L 1 4 1 4 3 1 2 3 4 2 4 3 3 4 FIGURE 21 Con't 211 quarter content pages coding section 47 M E 48 M E 49 F P 50 A/A P 51 A/A P 52 A/A S 53 A/A S 54 A/A B 55 L W 56 W/C W 57 W/C W 58 F E 59 M E 60 M E 61 A/A P 62 A/A P 63 CL C 64 CL C 65 CL C 66 CL C 67 A/A S 68 L S 69 A/A S 70 L S 71 A/A S 72 L S 73 A/A S 74 G W 75 G W 76 WN E 77 M E 78 F W 79 F W 80 G W 81 G W 82 F P 83 A/A P 84 A/A P 85 L S 86 M/C S 87 M/C S 88 L S 89 A/A S 90 A/A S issue page quadrant 3 R 2 3 R 4 4 R 4 4 R 1 4 R 4 4 R 2 4 R 4 4 L 2 4 L 3 4 R 2 4 R 4 4 L 4 4 R 2 4 R 4 5 R 2 5 R 4 5 R 1 5 R 2 5 R 3 5 R 4 5 L 3 5 R 3 5 R 4 5 L 3 5 R 2 5 R 3 5 R 4 5 L 4 5 R 4 5 L 3 5 R 4 5 L 2 5 L 4 5 R 2 5 R 4 6 R 4 6 R 2 6 R 4 6 R 3 6 L 1 6 L 3 6 R 4 6 L 1 6 L 3 212 FIGURE 21 Con't quarter content pages coding section 91 L S 92 A S 93 WN W 94 L w 95 W/C W 96 M E issue page quadrant 6 R 3 6 R 4 6 L 3 6 L 3 6 R 4 6 R 4 C -CL -c l a s s i f i e d section miscellaneous c l a s s i f i e d content Table 7-4(1): Analysis of Advertising Data (Panel C) - 3 Factor Solution 213 a/p factor loadings matrix I II I I I 1 .17 -.07 -.19 2 .26 -.19 -. 17 3 .27 -.20 -. 18 4 .26 -.04 -.29 5 S .42 .09 -. 17 6 S .42 .14 -.29 7 S .43 .08 . -.18 8 S .52 .04 -.27 9 .04 -.23 C-.66 10 .05 -.17 C-.70 11 .06 -.22 C-.69 12 .10 -.13 C-.72 13 .25 -.16 -.19 14 P .35 P-.35 -.24 15 P .56 -.23 -.20 16 P .56 -.20 -.21 17 S .60 -.08 -. 15 18 S .61 -.12 -. 14 19 S .49 -.13 -.22 20 S .54 -.20 -.18 21 S .53 -.12 -.25 22 .09 W-.53 -.13 23 .35 W-.38 -.02 24 .22 W-.44 -. 15 25 .19 W-.52 -.16 26 .07 W-.51 -.15 27 .22 -.22 C-.63 28 . 18 -.16 C-.71 29 .27 -. 13 C-.69 30 .20 -.18 C-.72 31 .22 -.11 C-.62 32 .26 P-.51 -.18 33 .24 P-.51 -.18 34 P .44 -.33 -.18 35 P .42 -.36 -.15 36 .31 P-.42 -.21 37 .36 -.20 C-.39 38 .34 -.19 C-.42 39 .03 -.36 C-.49 40 .04 -.35 C-.55 41 S .62 -.08 -.19 42 S .68 -. 16 -. 17 43 S .58 -.28 -.03 44 s .51 -.23 -.16 45 E .46 -.39 -.19 46 E .38 -.40 -.21 47 .28 E-.36 -.09 48 .29 E-.37 -.09 Table 7-4(i) Con't a/p factor loadings matrix 214 I II I I I 49 .39 P-.46 -.11 50 P .55 -.39 -:02 51 P .54 -.40 -.02 52 S .55 -.12 -.04 53 S .70 -.18 -.06 54 S .43 -.17 -.15 55 .24 W-.51 -.15 56 .11 W-.65 -.06 57 .09 W-.64 -.08 58 .40 E-.48 -.19 59 E .32 E-.32 -.01 60 E .34 E-.34 -.03 61 P .48 -.41 -.03 62 P .52 -.43 -.06 63 .06 C-.48 -.34 64 .21 C-.46 -.33 65 .07 C-.47 -.34 66 .24 C-.38 -.29 67 S .48 -.27 -.12 68 S .63 -.31 -.02 69 S .69 -.26 -.02 70 S .58 -.23 .00 71 S .66 -.18 -.06 72 S .62 -.20 -.09 73 S .66 -.14 -.07 74 .05 W-.67 -.18 75 .24 W-.61 -.08 76 .38 E-.45 -.16 77 .23 E-.50 -.08 78 .17 W-.57 -.22 79 .20 W-.56 -.24 80 .10 W-.50 -.33 81 .10 W-.52 -.33 82 .31 P-.48 -.11 83 P .51 -.44 -.11 84 P .53 -.39 -.11 85 S .60 -.18 -.04 86 .33 S-.50 -.11 87 .34 S-.50 -.13 88 S .60 -.37 .00 89 S .64 -.23 -.15 90 S .61 -.25 -.16 91 S .65 -.26 -.03 92 S .64 -.25 -.08 93 .23 W-.61 -.03 94 .33 W-.58 -.02 Table 7-4(i) Con't a/p factor loadings matrix  I I I I I I 95 .07 W-.62 -.21 96 1 .36 E-.41 -.06 interpretation X Y C cumulative proportion of variance accounted for = 37.9% Table 7-4(11): Analysis of Advertising. Data.. (Panel C) .- 7 Factor S o l u t i o n 2 ^ a/p I II f a c t o r loadings matrix VI VII I I I IV V 1 .01 .03 -.16 -.06 -.00 P-.39 -.17 2 .17 .21 -.18 -.01 .13 P-.44 -. 14 3 .18 - .22 -.17 -.02 .12 P-.43 -.17 4 .13 .12 -.02 -.12 -.07 P-.42 -.11 5 .20 -.06 -.07 .04 -. 10 S-.64 .01 6 .21 .02 .01 -.05 -.06 S-.71 .08 7 .20 -.01 -.06 .07 -. 11 S-.69 .10 8 .34 .10 .03 -.02 -.05 S-.69 -.04 9 -.07 C .42 -.02 C-.40 -.23 C-.36 -.15 10 -.10 C .38 -.00 C-.43 -.26 C-.41 -.12 11 -.07 C .40 -.03 C-.44 -.25 C-.38 -.16 12 -.05 C .33 .01 C-.49 -.23 C-.40 -.13 13 . 10 .02 -.06 -.11 -.07 -.25 E-.57 14 .29 .10 P-.40 -.26 -.10 -.12 -.15 15 P .44 -.04 P-.43 -.24 -.02 -.30 -.09 16 P .44 -.06 P-.44 -.26 .01 -.30 -.08 17 S .54 -.11 -.29 -.25 .06 -.23 .02 18 S .54 -.09 -.29 -.22 .02 -.24 -.01 19 S .34 -.13 -.36 -.25 -.11 -.29 -.12 20 S .39 -.08 -.37 -.18 -.10 -.31 -.20 21 S .45 .02 -.21 -.24 -.09 -.29 .00 22 .05 .21 W-.62 -.10 -.08 -.05 -.07 23 .28 -vOl W-.53 -.06 -.08 -.08 -.14 24 .17 .12 W-.56 -.17 -.01 -.08 -.12 25 .13 .11 W-.70 -.20 -.08 -.03 -.08 26 .04 .20 W-.63 -.15 -.01 -.03 -.05 27 .20 .16 -.25 C-.73 -.13 .01 -.01 28 .16 .17 -. 15 C-.80 -.09 -.01 -.03 29 .24 . 12 -. 14 C-.79 -.08 -.04 -.06 30 .17 .18 -.19 C-.81 -. 10 -.03 -.04 31 .23 .17 -.08 C-.69 -.08 -.03 .05 32 . 18 .25 -.44 -.02 P-.48 -.08 -.02 33 .18 .25 -.41 -.05 P-.47 -.03 -.03 34 P .36 .09 -.26 -.11 P-.48 -.07 -.06 35 P .37 .11 -.26 -.11 P-.44 -.00 -.11 36 .'24 .21 P-.35 -.08 P-.47 -.08 .03 37 .28 .17 .09 -.27 C-.62 -.03 -.17 38 .25 .16 .08 -.31 C-.61 -.03 -.18 39 .01 C .46 -.01 -.29 C-.32 -.09 -.16 40 .01 C .44 -.06 C-.36 C-.35 -.10 -.11 41 S .50 -.02 -.01 -.10 S-.46 -.22 -.06 42 S .53 -.02 -.11 -.06 S-.52 -.25 -.07 43 S .52 .01 -.23 .02 S-.46 -.07 -.03 44 S .43 -.02 -.24 -.16 S-.40 -.07 -.06 45 E .40 .12 -.35 -.16 E-.36 -.06 -.09 46 E- .35 .13 -.30 -.20 E-.40 .05 -.13 47 .14 .04 -.17 -.04 -.26 -.06 E-.73 Table 7-4(ii) Con't 217 a/p I II factor I I I loadings matrix IV V VI VII 48 . 15 .05 -.18 -.03 -.26 -.07 E-.72 49 P .15 .32 -.27 -.06 -.11 -.05 -.07 50 P .61 .19 -.25 -.06 -.02 -.03 -.12 51 P .60 . 19 -.28 -.05 -.01 -.06 -.10 52 S .51 .01 -.06 -.05 -.04 -.19 -.16 53' S .64 .03 -.16 -.03 -.16 -.27 -.05 54 S .42 .17 -.07 -.09 -.02 -.23 -.03 55 .26 W .34 W-.37 -.05 -.14 -.07 -.07 56 .12 W .41 W-.49 .08 -.07 -.07 -.21 57 .08 w .39 W-.52 .06 -.09 -.10 -.21 58 E .40 .27 -.34 -.15 -.10 -.10 -.21 59 .28 .10 -.12 -.04 .11 -.07 E-.66 60 .30 .11 -.14 -.05 .10 -.08 W-.67 61 P .35 .27 -.12 .00 -.15 -.00 -.17 62 P .60 .30 -.15 -.04 -.13 .00 -. 14 63 .05 c .63 -.11 -.16 .01 -.06 -.04 64 .37 c .55 -.03 -.30 .02 .09 -. 11 65 .02 c .62 -.12 -.15 .00 -.09 -.04 66 .39 c .48 .02 -.28 .03 .07 -.14 67 S .57 .23 -.02 -.16 -.02 -.00 -.14 68 S .70 .12 -.14 -.10 -.08 .02 -.12 69 S .76 .13 -.06 -.08 -.10 -.01 -.08 70 S .60 .07 -.11 -.02 -.12 -.06 -.10 71 S .74 .18 .02 -.06 -.02 -.15 -.01 72 S .66 .14 .01 -.09 -.17 -.07 -.07 73 S .71 .13 .04 -.09 -.02 -.14 -.05 74 .00 w .54 ' W-.47 -.01 -.06 -.05 -.08 75 .28 w .41 W-.39 .03 -.15 -.02 -.12 76 E .40 .27 -.27 -.14 -.08 -.05 -.23 77 .27 .33 -.13 -.04 -.06 .05 E-.54 78 .27 w .54 -.19 -.09 -.20 .05 -.09 79 .32 w .55 -.16 -.13 -.18 .06 -.07 80 .22 w .66 -.10 -.14 -.04 -.09 .02 81 .21 w .66 -.14 -.14 -.05 -.10 .04 82 P .30 p .31 P-.33 .03 -.27 -.11 -.03 83 P .49 .20 -.28 -.07 -.28 -.06 -.17 84 P .50 .17 -.25 -.05 -.29 -.11 -.15 85 S .54 .01 -.10 -.01 -.23 -.18 -.17 86 S .40 s .44 -.21 .01 -.09 -.10 -.09 87 S .41 s .48 -.20 .02 -.08 -.13 -.06 88 S .60 .14 -.24 .02 -.16 -.10 -.16 89 S .64 .11 -.08 -.15 -.18 -.12 -.12 90 S .59 .14 -.12 -.12 -.20 -.16 -.09 91 S .65 .08 -.16 -.04 -.16 -.12 -.09 92 S .63 . 12 -.13 -.04 -.18 -.17 -.08 93 .24 w .35 W-.49 .08 -.18 -.04 -.07 94 .37 .29 W-.41 .02 -.20 .04 -.15 Table 7 - 4 ( i i ) Con't 218 a/p factor loadings matrix I II III IV V VI VII 95 .10 .51 W-.39 -.03 -.16 -.08 . -.06 96 .29 .15 -.20 .02 -.23 -.08 E-.51 i n t e r p r e t a t i o n X - Y C ISS#3 ISS#1 M 1, z , j cumulative proportion of variance accounted f o r = 49.4% External V a l i d a t i o n of Factor Results on Advertising Data 219 The working hypotheses with respect to the advertising data were developed according to the expectation that s e l e c t i v e exposure to a d v e r t i s i n g content i s s t r u c t u r a l l y determined. However much of the evidence indicates factors based on content or content plus s t r u c t u r a l s e l e c t i v i t y . This i s p a r t i c u l a r l y true of the smaller factor spaces where, as has been described, the dominant 'X' and 'Y' factors p e r s i s t across a l l 3 panels and can be interpreted as r e f l e c t -ing male and female ad v e r t i s i n g i n t e r e s t s r e s p e c t i v e l y . In order to v a l i d a t e these r e s u l t s the scores for the 'X' and 'Y' factors were calculated and regressed against the set of independent v a r i a b l e s i n a manner i d e n t i c a l to that described i n Chapter VI for the news r e s u l t s . Tables 7-5 and 7-6 present the r e s u l t s of the stepwise regression proced-ure. Again, for ease of comparison the r e s u l t s are presented for 'X' and 'Y' factors across the d i f f e r e n t panels, i . e . Table 7-5 presents the regression of the 'X' factor scores f o r panels, A, B and C. On the l e f t side of each table the independent v a r i a b l e s are l i s t e d and i n each c e l l can be found the regress-ion c o e f f i c i e n t s where s i g n i f i c a n t . Only v a r i a b l e s with F p r o b a b i l i t i e s l e s s than or equal to .1 are indicated. Table 7-5: The 'X' factors have been interpreted as r e f l e c t i n g advert-i s i n g of primary i n t e r e s t to male readers. The regression analysis tends to confirm t h i s i n t e r p r e t a t i o n . The only consistent v a r i a b l e r e l a t e d to high factor scores i s sex and i t indicates p r i m a r i l y male readership. A l l other independent v a r i a b l e s are n o n - s i g n i f i c a n t or s i g n i f i c a n t i n only one of the three panels. Age i s r e l a t e d i n opposite d i r e c t i o n s probably i n d i c a t i n g a difference i n the content of the advertising across the two panels. Table 7-6: The 'Y' factors have been interpreted as r e f l e c t i n g advert-i s i n g of primary i n t e r e s t to female readers. Again the regression analysis 220 tends to validate t h i s interpretation. Only two variables were consistent across a l l three panels: sex i s p o s i t i v e l y related to factor scores indicating female readership and a favourable attitude towards newspaper advertising i s indicated. There i s also i n two of the panels, a positive relationship to an interest i n hobbies and a negative relationship to a need for understanding, i . e . , a person's wish to understand many areas of knowledge. In summary, these results validate the interpretation of the 'X' and 'Y' factors. This i s especially the case due to the fact that sex was the only consistent discriminator of factor scores. Table 7-5: Regression of .'X' Factor Scores Independent v a r i a b l e s Panel A Panel B Panel C Personality t r a i t v a r i a b l e s Cognitive structure S o c i a l recognition Understanding Manifest anxiety General self-confidence .37 (.08) .55 (.07) Opinion r e l a t e d measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality -.81 (.03) .10 (.01) Leisure i n t e r e s t measures Hobbies Sports .37 (.00) Demographics Age Sex (female = '+') Income Education -.56 (.00) .80 (.02) -.39 (.00) -.10 (.00) -.11 (.00) -.58 (.00) R 2 .13 (.00) .16 (.00) .16 (.00) Table 7-6: Regression of 'Y' Factor Scores 222 Independent va r i a b l e s Panel A Panel B Panel C Personality t r a i t v a r i a b l e s Cognitive structure S o c i a l recognition Und er s t and ing Manifest anxiety General self-confidence -.64 (.05) -.52 (.02) Opinion r e l a t e s measures Newspaper r a t i n g Advertising r a t i n g L i b e r a l i s m Newspaper coverage Newspaper source Newspaper personality 1.00 (.02) .90 (.03) .19 (.00) -.46 (.03) .14 (.00) Leisure i n t e r e s t measures Hobbies Sports .26 (.00) .30 (.00) Demographics Age Sex (female = '+') Income Education .52 (.00) .91 (.01) .72 (.00) .74 (.00) R 2 .19 (.00) .30 (.00) .22 (.00) 223 Chapter VIII SUMMARY AND CONCLUSIONS The object of t h i s study has been to investigate dimensions of aggregate audience exposure to a mass medium. The method employed was factor analysis but, unlike a number of r e l a t e d applications of t h i s tech-nique, p r i o r expectations were introduced into the analysis. This was accomplished through the construction of a model which stated that s e l e c t i v e exposure i s a function of the content and structure of a medium (or media). The model was used to p r e d i c t dimensions of aggregate audience exposure on a content and/or s t r u c t u r a l basis and these predictions were then investigated using factor analysis. The procedure was r e p l i c a t e d across samples and the r e s u l t s were externally v a l i d a t e d through r e l a t i o n to external v a r i a b l e s . Summary of the Results of Analysis on News Data The analysis of the news or nonadvertising data investigated whether dimensions of aggregate audience exposure over time r e f l e c t e d the managerial content and s t r u c t u r a l organization of a d a i l y newspaper. If confirmed, sources of variance known to exist i n the newspaper reading s i t u a t i o n were then eliminated i n order to c l a r i f y the factory r e s u l t s . The va ri ab le s (quarter pages) selected for analysis were c l a s s i f i e d v i a a working hypothesis into an a p r i o r i set of categories l a b e l l e d the a p r i o r i c l a s s i f i c a t i o n sys-224 tern. The same set of variables were then factor analyzed using p r i n c i p a l components analysis where the working hypothesis was used to determine the number of fact o r s . The r e s u l t was l a b e l l e d as the empirical c l a s s i f i c a t i o n system. The working hypothesis was then tested by means of a s t a t i s t i c a l test of independence between the two c l a s s i f i c a t i o n systems. The working hypotheses are reproduced i n Table 8-1 along with r e s u l t s of the t e s t i n g procedure. Working hypothesis investigated whether the p r i n c i p a l dimensions of audience exposure were determined by the a p r i o r i managerial sections over time. As indicated i n Table 8—1 t h i s hypothesis was v e r i f i e d across 3 panels of respondents although i n Panel C there was some question with respect to the Public A f f a i r s section. This conforms to the expectation that audience members are d e l i b e r a t e l y s e l e c t i v e with respect to the news content and supporting structure of the newspaper as s p e c i f i e d i n the model underlying the hypothesis. Subsidiary, working hypotheses B.^ and sought to determine whether a c l e a r e r factor space would r e s u l t from elimination of two confounding sources of variance i n the determination of the aggregate dimensions. With H^, only data from p a r t i c u l a r issues were subjected to analysis thus re-moving the e f f e c t s of d i f f e r e n t issues or time structure. As indicated i n Table 8-1, was strongly v e r i f i e d . Inspection of the fac t o r loadings i n Chapter VI indicates a very c l e a r d i s t r i b u t i o n of loadings compared to those under H^, and a s u b s t a n t i a l increase i n the variance accounted f o r . This was pr e c i s e l y the expected r e s u l t . 225 Table 8-1: Summary of Hypothesis Testing Working Hypothesis H The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper over time are determined by the managerial content/ structure sections. 2 Panel A X Level of Sig n i f i c a n c e H 0 Public A f f a i r s 61.65 .001 r e j , Sports 61.97 .001 Business 61.84 .001 " Women's 65.69 .001 " Panel B Public A f f a i r s Sports Business Women's 52.69 65.38 63.04 78.38 .001 .001 .001 .001 rej, Panel C Public A f f a i r s 7.21 .01 r e j , Sports 22.08 .001 " Women's 17.42 .001 Table 8-1: Continued 226 Working Hypothesis H, The p r i n c i p a l dimensions of audience exposure to the news content of a d a i l y newspaper i n a sin g l e issue are determined by the managerial content categories. Panel A Issue #3 prob. of r e s u l t 0 Panel C II Public A f f a i r s < .005 r e j , Sports < .005 Business < .005 Women's <.005 Issue #6 r e j Public A f f a i r s < .005 Sports < .005 ' " Business < .005 " Women's .005* '" Panel B Issue #3 Public A f f a i r s < .005 r e j , Sports <.005 Business <.005 " Women's <.005 " Issue #5 Public A f f a i r s <.005 r e j , Sports <.005 " Business <.005 ." Women's <.005 " Issue #2 Public A f f a i r s <.005 r e j , Sports <.005 Women's <.005 Issue #6 Public A f f a i r s <.005 r e j , Sports <.005 Women's <.025 II Table 8-1: Continued 227 Working Hypothesis H_ The p r i n c i p a l dimensions of audience exposure to a p a r t i c u l a r category of news content over time are determined by the various issues of the newspaper. Panel A Public A f f a i r s Issue #2 3 4 5 6 prob. of r e s u l t < .005 .0001* .0008* .03 * .014 * H„ re j , Sports Issue #1 2 3 4 5 6 £ .025 < .025 .14 <.025 <.05 rej , II r e j , rej,, Business Issue #1 2 3 5 6 .0003* .002 * .0000* .0001* .0000* rej Women's Issue #L 2 3 4 5 6 . 1 * <L .025 < .01 .1 * < .025 .1 * rej , Table 8-1: Continued 228 Panel B Public A f f a i r s Issue #2 3 4 5 6 Sports Issue #1 2 3 4 5 6 Business Issue #1 3 4 5 Women's Issue #1 2 3 4 5 6 Panel C Public A f f a i r s Issue #1 2 4 5 6 Sports Issue #1 2 3 5 6 Women's Issue #1 2 3 4 prob. of r e s u l t s 0_ < .005 r e j . .0002* " .005 * .005 * < .005 .004 .004 .0007 .003 .25 < .005 rej rej, <.005 <.005 <.005 <.005 .0058* .001 * <.005 .0058* <.005 .0058* < .01 < .005 <.005 .13 * C.05 <.005 <.025 <.005 rej rej <.01 <.01 <.01 <\01 Table.8-1: Continued 229 Working Hypothesis H^ The p r i n c i p a l dimensions of audience exposure to the advertising content of a d a i l y newspaper over time are determined by the managerial sections (as determined by the news content). Panel A X Level of Signi f i c a n c e H„ Public A f f a i r s Sports Business Women's 4.06 34.16 .05 .001 r e j , r e j , Panel B Public A f f a i r s Sports Women's Entertainment 4.72 35.19 28.47 .05 .001 .001 r e j , Panel C Public A f f a i r s Sports Women's Entertainment C l a s s i f i e d s 19.9 23.4 57.4 .001 .001 .001 rej , II rej 230 eliminated content (and managerial section) as a source of var-iance, i . e . , only data from a p a r t i c u l a r content category (and congruent section) was analyzed with the expection that the various issues of the news-paper would s u b s t a n t i a l l y account for the remaining variance. Table 8-1 indicates that was v e r i f i e d across the 3 samples. However, inspection of the factor loadings revealed a less c l e a r r e s u l t than under ^ e s p e c i a l l y i n the case of the P u b l i c A f f a i r s category. The tendency for c e r t a i n v a r i ^ ables (quarter pages) to correlate across issues instead of within issues possibly indicates a degree of content s e l e c t i v i t y within categories. The data i s not s u f f i c i e n t to investigate t h i s . O v e r a l l H^ i s confirmed with the reservation that there may be s e l e c t i v i t y within managerial sections, s p e c i f i c a l l y the Public A f f a i r s section. The aggregate dimensions of audience exposure over time were extern-a l l y v a l i d a t e d through regression of factor scores on a set of external audience descriptor v a r i a b l e s . These r e s u l t s indicated s u b s t a n t i a l d i f f e r -ences among audience members exposed to the various managerial sections. Summary of the Results of Analysis on A d v e r t i s i n g Data The analysis of the advertising data investigated whether aggregate dimensions of audience exposure over time r e f l e c t e d the managerial organiza-t i o n of the news content, that i s , under the assumption that there i s low p o s i t i v e u t i l i t y a r i s i n g from i n d i v i d u a l s e l e c t i v e exposure to a d v e r t i s i n g content, i t was of i n t e r e s t to investigate whether dimensions of exposure were determined by the s t r u c t u r a l organization of the newspaper. If confirm-ed, sources of variance known to exist i n the newspaper readership s i t u a t i o n 231 were again to be eliminated i n order to c l a r i f y the factor r e s u l t s . The set of procedures was s p e c i f i e d through working hypotheses to and were analogous to those f o r the news data. Working hypothesis investigated whether the p r i n c i p a l dimensions of audience exposure to advertising were determined by the a p r i o r i mana-g e r i a l sections over time. As indicated i n Table 8-1, the r e s u l t s for t h i s hypothesis were much weaker than those for working hypothesis concerning the news content. Although there was evidence that dimensions did r e f l e c t a d v ertising i n the Sports and Women's sections and possibly the Entertainment and C l a s s i f i e d s sections, the r e l a t i v e strength of the factor loadings suggested a more fundamental dimensionality to the data. As a r e s u l t , an exploratory analysis was undertaken. The major conclusion drawn from t h i s exploratory analysis, conducted across a l l three panels, was that the most basic dimensions of audience exposure were determined by adve r t i s i n g of d i f f e r e n t i a l i n t e r e s t to male and female readers, the so- c a l l e d "X" and "Y" facto r s . However i t should be added that most of the stronger loadings on these "X" and "Y" factors were on quarter pages placed i n the Sports and Women's sections. Hence, there i s s t i l l some equivocal support f o r s t r u c t -u r a l s e l e c t i v i t y based on managerial section. (It should be r e - i t e r a t e d that the n u l l hypothesis of independence between the a p r i o r i and empirical c l a s s i f i c a t i o n s was i n fact rejected for advertising quarter pages placed i n the Sports and Women's sections.) Further exploratory analysis indicated dimensions which were i n i t i -a l l y interpreted as being determined by the Entertainment and C l a s s i f i e d s sections. However, the dimension thought to be determined by the C l a s s i -f i e d s section did not p e r s i s t across a l l issues of the study. This i n d i -232 cated a f a i l u r e of audience members to be con s i s t e n t l y s e l e c t i v e i n t h e i r exposure to that section over time. Moreover, the dimension thought to be determined by the Entertainment section was found to be more adequately ex-plained by content s e l e c t i v i t y r e s u l t i n g from "Movie" advertising. O v e r a l l , working hypothesis was rejected i n favour of a content i n t e r p r e t a t i o n of the aggregate dimensions based p r i m a r i l y on sex d i f f e r -ences. This was confirmed by expanding the factor space for a l l three panels and v e r i f y i n g that further dimensionality was l a r g e l y accounted f o r by issue or time structure. Dimensions which pe r s i s t e d as the factor space was ex-panded (such as the "Y" factor of Panel A) were found to be a r t i f a c t s of c e r t a i n advertising content and not an i n d i c a t i o n of s t r u c t u r a l s e l e c t i v i t y across time. Subsidiary working hypotheses and Hg were not tested as these are e n t i r e l y dependent upon the v e r i f i c a t i o n of H^. This i s not to say that H^, f o r example, would not have been v e r i f i e d but that such a r e s u l t would have been meaningless i n terms of the s t r u c t u r a l s e l e c t i v i t y model. Any i n t e r e s t i n g r e s u l t would very l i k e l y be unique to the p a r t i c u l a r issue i n -vestigated. The i n t e r p r e t a t i o n of the basic aggregate "X" and "Y" factors was externally v a l i d a t e d through regression of the factor scores on the set of external audience descriptor v a r i a b l e s . This confirmed the i n t e r p r e t a t i o n of these factors as advertising of i n t e r e s t to male and female readers as only 'sex' and 'attitude towards advertising' proved s i g n i f i c a n t l y r e l a t e d to fa c t o r scores across the panels. 233 Conclusions The conclusions to this study can be summarized under three broad areas: the factor analytic methodology, the economic model of selective ex-posure and the results of data analysis. on the use of factor analysis Throughout this study the importance of methodological safeguards in the application of factor analysis has been emphasized. Generally, this has led to certain improvements in the use of factor analysis as a research meth-od when compared to earlier studies. 1) The model of individual selective exposure was developed p r i -marily as a mechanism for introducing prior expectation into the analysis as specifically recommended by Armstrong and Soelberg and indirectly recommended by Ehrenberg. There were two important advantages to this approach. First, i t enabled a careful delimitation of the inputs into the analysis, i.e., data was systematically selected according to c r i t e r i a specified by the model and hypothesized to yield certain results. Second, the specification of a model provided a means with which to evaluate the results, i.e., whether or not they tended to support the a p r i o r i model and the related working hypothesis. This is ;an approach conspicuously absent from the studies cited in Chapter II. Even where the a p r i o r i model was abandoned in favour of a more conven-tional exploratory analysis in Chapter VII, the delimitation of input data imposed s t r i c t constraints on misinterpretation of the results. 234 2) The r e l i a b i l i t y of the factor results was e f f e c t i v e l y demon-strated through cross-validation (replication) across three panels of respondents. This was especially the case with the exploratory analysis where the results are more dependent on the analyst's interpretation. 3) A measure of the v a l i d i t y of the factor results was undertaken through regression of factor scores on a set of external variables. While less conclusive than cross-validation, the relationships found tended to support the interpretation and v a l i d i t y of the factors. Where the appro-priate data are available, such a procedure may provide a more precise method of factor interpretation than the present widely used subjective approach. 4) The results of factor analysis described as the empirical class-i f i c a t i o n system were tested for independence from the a p r i o r i c l a s s i f i c a -tion system using a simple 2 x 2 contingency table analysis of the cross-c l a s s i f i c a t i o n of variables (or quarter pages). The object of t h i s proce-dure was to operationalize the working hypotheses as s t a t i s t i c a l tests. While t h i s approach worked reasonably w e l l , i t must be tentatively concluded that i t has limited usefullness. The simple reason i s that i f a factor can be interpreted i n a certain way i t i s l i k e l y that the n u l l hypothesis of independence w i l l be rejected at a low l e v e l of significance. Accordingly, while the s t a t i s t i c a l test may be sound, i t i s rendered somewhat superfluous by factor interpretation. 5) F i n a l l y , i n more conventional applications of factor analysis the number of factors i s usually determined by arbitrary devices such as the eigenvalue c r i t e r i o n , the scree test or generally some subjective judgement 235 as to s u f f i c i e n t variance accounted for. In t h i s study, the number of factors was a p r i o r i determined by the working hypothesis under consideration. This i s a r e l a t i v e l y common procedure where there i s some a p r i o r i reason to l i m i t the number of f a c t o r s . Under the exploratory analysis, however, there was no such a p r i o r i c r i t e r i o n a v a i l a b l e . As a r e s u l t , the s t a b i l i t y of f a c t o r s , as the factor space was expanded, was used as a means to determining at least the p r i n c i p a l underlying dimensions. This i s a more s a t i s f a c t o r y approach than a r b i t r a r y c r i t e r i a such as the eigenvalue c r i t e r i o n where, unknown to the analyst, the addition or deletion of a s i n g l e factor may cause a r a d i c a l a l t e r i n g i n the d i s t r i b u t i o n of factor loadings. Where the factors are stable as t h e i r number are expanded or larger f a c t o r spaces are i n t e r p r e t a b l e i n terms of smaller ones, the analyst may have more confidence i n the results. One must of course continue to exercise judgement i n l i m i t i n g the extraction of factors as these eventually account for only very small proportions of variance and are equivalent i n number to the o r i g i n a l variables analyzed. on the economic model of s e l e c t i v e exposure While the model was p r i m a r i l y developed as a means for introducing p r i o r expectation i n t o the analysis, i t has proven useful as a d i f f e r e n t approach to s e l e c t i v e exposure i n mass media. 6) The model of s e l e c t i v e exposure used an economic approach where the i n d i v i d u a l was seen as a u t i l i t y maximizer faced with l i m i t e d choices and a time constraint. While consistent with the general d i r e c t i o n of the s e l e c t i v e exposure l i t e r a t u r e , i t provided a new point of departure i n an area dominated by cognitive consistency approaches. 236 7) The model specifically recognized the content and structural aspects of media. Previous studies had sought not to include the structural aspects of media except as a nuisance factor. Here the supporting or con-founding role of structure has been included while acknowledging that i t is only media content which provides positive u t i l i t y to the individual. 8) The object of this study has not been to test the mechanisms of individual selective exposure. However, the sequence of working hypotheses concerning the news data was developed using the model and the results are clearly consistent with that formulation. The working hypotheses concerning the advertising data, however, provided very limited support and restructur-ing of the model is necessary recognizing the importance of advertising con-tent in selective exposure to the medium. on the results of data analysis The actual data analysis sought to confirm the existence of aggre-gate dimensions of audience exposure. 9) In the introduction to this study, i t was argued that a mass medium has a particular configuration of content and structure. The object with respect to the news data has been to validate this configuration as determining aggregate dimensions of audience exposure to a daily newspaper. Subsequent confirmation supports the elements of this configuration, the managerial sections, as vehicles for the delivery of selected news content. Further, i t indicates the,existence of audience segments associated with these vehicles. The results of external validation provide some indication 237 of the character of such audience segments. However, the l a t t e r requires further analysis with a data base designed for that purpose. The establishment of vehicles i n t e r n a l to a mass medium further confirms the need for a reassessment of audience measurement procedures which, as discussed i n Chapter I, currently ignore the p o s s i b i l i t y of s e l e c t -i v i t y within p a r t i c u l a r media (to be discussed below). The use of factor scores for measuring exposure to these vehicles i s also raised as a poss-i b i l i t y but i t i s l i k e l y that, given i n t e r n a l v e h i c l e s have been established, more d i r e c t measures of audience exposure would be appropriate. 10) In previous studies which attempted to discover underlying dimensionality to media exposure, no attempt was made to r e l a t e any of the r e s u l t s to ad v e r t i s i n g exposure. At the same time, the context of the re-search implied that dimensions of audience exposure indicated the existence of a d v e r t i s i n g vehicles and provided a basis f or promotional segmentation. The attempt to impose the content and s t r u c t u r a l configuration upon the advertising data i n t h i s study was generally unsuccessful. This implies that the managerial sections are at least not the primary determinants of s e l e c t i v e exposure to newspaper advertising. However, i t should not be sur-mised that the managerial sections are then r e l a t i v e l y useless as adve r t i s -ing v e h i c l e s . To the contrary, the r e s u l t s of analysis on news data i n d i -cate these are possibly the only means to reach substantial proportions of the audience whether or not ad v e r t i s i n g i s p a r t i c u l a r l y s a l i e n t to these audience segments. A major p r a c t i c a l conclusion following t h i s l i n e of reasoning i s to develop the type of advertisement which i s compatible with and l i k e l y s a l i e n t to whatever audience i s associated with these v e h i c l e s . 238 In the case of the Public Affairs and Business sections, for example, there is substantial selective exposure to news content but l i t t l e selective ex-posure to associated advertising content. These particular sections, at least, could be more effectively exploited. This conclusion is somewhat mitigated by the fact that readers of these sections did not indicate a favourable attitude towards newspaper advertising as did readers of the Women's and Sports sections to varying degrees (nor was a general un-favourable attitude demonstrated). 11) An exploratory analysis, using the conventional factor analytic method, was undertaken and resulted in basic dimensions of advertis-ing exposure reflecting sex differences. The general absence of factors reflecting either content or structural influences as the factor space was expanded tends to confirm an absence of content or structural selectivity except on an issue basis. The principal dimensions were found to be related toapositive attitude towards newspaper advertising and, accordingly, the possibility i s raised that advertising should be structurally organized for convenient audience access as the news i s presently organized. This does not contradict the earlier conclusion that managerial sections are most appropriate for reaching large segments of the audience which may not be favourable disposed towards advertising. Implications of the Study In Chapter I of this dissertation, the section entitled "Importance of the Study" discussed the methodological and managerial importance of the research. This section w i l l discuss the contributions that have been made to these two areas. 239 methodological implications The co n t r i b u t i o n to methodology at t h i s point should be evident. In the conclusions above, i t was demonstrated that factor analysis can be suc-c e s s f u l l y employed i n media research where p r i o r expectations, r e p l i c a t i o n and external v a l i d a t i o n are incorporated into the methodology. The implica-t i o n of t h i s f i n d i n g i s that i n future factor a n a l y t i c . a p p l i c a t i o n s i n media research these methodological safeguards should be introduced. Most im-portantly, as discussed i n Chapter I I , researchers would c l e a r l y benefit from the p r a c t i c e of developing, p r i o r to the analysis, e x p l i c i t hypotheses on the basis of what i s known or thought to be known about the content and structure of the media being analyzed. managerial implications It was stated .in Chapter I that t h i s study, drawing upon predecessor research by S i l l e r , would contribute to the improved "management" of news-papers. S p e c i f i c a l l y , i t would confirm the need f o r a re-assessment of audience measurement procedures which was a major conclusion of S i l l e r ' s work.^ o The basic measures used i n newspaper audience assessment are d u p l i -cated audience or reach, unduplicated or net reach and average frequency of exposure. Duplicated audience or reach i s the aggregate of a l l persons who read any part of a newspaper once for every issue. Net reach simply nets out audience overlap over time. It i s a measure of a l l persons who read any part of the newspaper once, regardless of the number of issues read. 1. Siller,"Newspaper Reading.A Study i n Sel e c t i v e Effects,"p.197-211 240 The average frequency of exposure i s the average number of times a reader i s exposed to the newspaper. These measures are currently determined f or the newspaper as a whole and are based on a l l readers who looked at any part of the newspaper. S i l l e r states, " C l e a r l y , these measures do not represent the actual audience f or any s p e c i f i c part of the newspaper. They represent a conceptual audience composed of the readers of any one part of the newspaper projected as the readers of a l l other parts. I f newspaper readership i s s e l e c t i v e , then readers of one part cannot be con-sidered as readers of a l l other parts and the concept-ual audience becomes an imaginary audience. Calcula-tions on the s i z e and composition of t h i s imaginary audience are of dubious value. The predecessor research to t h i s current study e s s e n t i a l l y dealt with i n d i v i d u a l behaviour with respect to c e r t a i n categories of newspaper content. S i l l e r was able.to demonstrate that most audience members are s e l e c t i v e with respect to these categories on the basis that t h e i r observed readership patterns d i f f e r e d from t h e i r expected readership patterns where 2 the l a t t e r was based on nonselective or random readership. It was con-cluded, as a r e s u l t of t h i s f i n d i n g , that such i n t e r n a l s e l e c t i v i t y of news-paper content could s e r i o u s l y d i s t o r t the above conventional reach and f r e -quency measures used for audience assessment. The inference from i n d i v i d u a l content s e l e c t i v i t y to aggregate 1 .N Ibid. , p. 7 . 2. Ibid ., p. 101-105. 241 audience measurement raises some important issues. F i r s t , how does i n d i -vidual content s e l e c t i v i t y relate to aggregate audience behaviour? Although S i l l e r demonstrated that most individuals are generally selective, where investigating p a r t i c u l a r content categories he was often forced to deal with small subsets of his sample as a result of very few audience members having met the s e l e c t i v i t y c r i t e r i a for these categories. As a r e s u l t , h is con-clusions concerning selective exposure to sp e c i f i c content categories (as opposed to general s e l e c t i v i t y ) were made on the basis of very small sub-sets of readers. There was, i n fact, no measure of aggregate behaviour with respect to these categories. Given that there i s content s e l e c t i v i t y i n newspaper readership, a 1second issue a r i s i n g from S i l l e r ' s conclusions i s to relate such categories to e d i t o r i a l or promotional vehicles which are useful. In other words, does content s e l e c t i v i t y relate to some organizational or structural unit(s) with-i n the newspaper which can be manipulated by newspaper editors of managers? In response to these d i f f i c u l t i e s t h i s current study has dealt with aggregate behaviour of the audience with respect to the newspaper's array or "configuration" of content and structure. The analysis and results concern the entire set of respondents and describes aggregate behaviour resulting from individuals' selective behaviour with respect to the newspaper's content and supporting structural organization. In dealing with the entire sample of respondents, the study avoids the loss of information involved i n using subsets of respondents. Further, by tying exposure to both content and structure, i t focuses on the managerial organization of the newspaper and seeks to confirm whether the ove r a l l dimensionality of aggregate audience 242 exposure (as measured by fa c t o r analysis) i s determined by th i s organization. The broader importance of the findings i s the recognition of a medium's content and structure as a managerial configuration or, more p r e c i s e l y , a constrained set of elements which serves to dimensionalize aggregate audience exposure. This implies that the a l t e r a t i o n of the content and s t r u c t u r a l configuration can e f f e c t i v e l y a l t e r the dimension-a l i t y of audience exposure. Further, i t i s implied by r e l a t i n g exposure dimensions to audience predispositions or biases that the audience f o r a newspaper, or other media, can be e f f e c t i v e l y segmented f o r e d i t o r i a l or promotional purposes. As indicated i n the conclusions, the f a i l u r e of dimensions of advertising exposure to also follow the managerial organization does not mitigate against the use of t h i s organization f o r ad v e r t i s i n g purposes. The negative r e s u l t s suggest the p o s s i b i l i t y that advertising should be designed and placed according to more refined audience measures which recognize the managerial organization of the newspaper. Indeed, advertising i t s e l f might be organized i n t o d i s t i n c t s t r u c t u r a l sections to a greater extent than i s done at ^present. However, the more immediate p r a c t i c a l impact of the findings i s to confirm the present inadequacy of audience assessment procedures. As anticipated by S i l l e r , present measures of duplicated reach, net reach and frequency of exposure should be calculated on the basis of exposure to vehicles i n t e r n a l to the newspaper rather than on the basis of exposure to the e n t i r e medium. S p e c i f i c a l l y , these i n t e r n a l vehicles 243 are the managerial content/structure sections of the newspaper. Areas for Future Study As discussed i n Chapter I, the data used, i n t h i s study were generated as part of a research project c a r r i e d out on a single newspaper i n a medium sized c i t y having only the one major d a i l y . . In order to ensure the generality of the findings the study should be r e p l i c a t e d . Accordingly, one future area of research would be the a p p l i c a t i o n of both the conceptual model and the factor a n a l y t i c methodology to other newspapers. In any future data c o l l e c t i o n , thought should be given to a measuring instrument which does not r e l y on quarter page r e c a l l as t h i s forces the elimination of much data which do not meet the c l a s s i f i c a t i o n c r i t e r i a of Chapter V. Further, i t would be preferable, i n future, to measure exposure with a scale which does not r e l y on a dichotomous score. Where the l e v e l of exposure varies s u b s t a n t i a l l y among quarter pages, the range of the c o r r e l a t i o n c o e f f i c i e n t i s l i m i t e d and can cause problems of i n t e r p r e t a t i o n . The managerial implications a r i s i n g from recognition of i n t e r n a l s e l e c t i v i t y within a mass medium have been elaborated i n d e t a i l elsewhere. S p e c i f i c a l l y , as indicated i n Chapter I, i t has been demonstrated that f a i l u r e to recognize i n t e r n a l s e l e c t i v i t y can s e r i o u s l y d i s t o r t convent-i o n a l reach and frequency measures. For reference, see F.H. S i l l e r and V.J. Jones "Newspaper Campaign Audience Segments," Journal of Advertising  Research, XIII (June, 1973). Further, i t has been demonstrated that i n t e r n a l s e l e c t i v i t y implies a trade off between the net reach and average frequency of exposure that can be expected to r e s u l t from the placement of an advertising campaign wit h i n a newspaper (or other p r i n t medi). That i s , reach and frequency cannot be simultaneously maximized for an advertising campaign of given s i z e i n a s p e c i f i c medium (as i s implied by conventional audience assessment procedures). For reference, see F.H. S i l l e r and V.J^ Jones, "Reach and Frequency Trade-offs within a Vehicle," U.B.C. Working Paper. 244 The factor analytic methodology, where an a p r i o r i model i s specified and cross and external v a l i d a t i o n are performed, could be applied to other mass media. The range of possible applications i s indicated by the l i t e r a t u r e cited i n Chapter I I . This study i s also a f i r s t stage i n the segmentation of newspaper audiences for e d i t o r i a l and promotional purposes. A set of measures should now be developed for d i r e c t l y assessing audience exposure to the int e r n a l vehicles of a daily newspaper. With respect to the data used i n this study, preliminary analysis has indicated that factor scores on the dimensions of audience exposure to news content provide a basis for grouping audience members using a h i e r a r c h i c a l clustering routine. Further, consideration should also be given to the development of a set of audience descriptor variables which may be useful i n predicting exposure to in t e r n a l vehicles and describing associated audience segments. Also, hypotheses and data could be generated concerning questions of sp e c i f i c e d i t o r i a l and promotional interest. 245 BIBLIOGRAPHY Armstrong, H. Scott. "Derivation of Theory by Means of Factor Analysis or Tom Swift and His E l e c t r i c Factor Analysis Machine," American S t a t i s t i c i a n , XXI (December, 1967), 17-21. Armstrong, J.S. and P. Soelberg. "On the Interpretation of Factor Ana l y s i s , " Psychological B u l l e t i n , LXX (1968), 361-364. Bass, F.M., E.A. Pessemier and D.J. T i g e r t . "A Taxonomy of Magazine Readership Applied to Problems i n Marketing Strategy and Media Sel e c t i o n . " Journal of Business, XLII (July, 1969), 357-363. Bruno, A.V. "The Network Factor i n T.V. Viewings," Journal of Advertising Research, XIII (October, 1973), 33-39. C a t t e l l , R.B. Handbook of M u l t i v a r i a t e Experimental Psychology. Chicago: Rand McNally and Co., 1966. C o l l i n s , G. "On Methods: Factor Ana l y s i s , " Journal of Advertising Research, I (September, 1961), 28-32. Cooley, W.W. and P.R. Lohnes. M u l t i v a r i a t e Data Analysis. New York: John Wiley & Sons, 1971. Ehrenberg, A.S.C. "On Methods: The Factor A n a l y t i c Search f o r Program Types," Journal of Advertising Research, VIII (March, 1968), 55-63. Einhorn, H.J. "Alchemy i n the Behavioural Sciences," Public Opinion Quarterly, XXXVI ( F a l l , 1972) 367-378. Ekeblad, F.A. and S.F. Stasch. " C r i t e r i a i n Factor Analysis," M u l t i v a r i a t e Analysis i n Marketing, ed. D.A. Aaker (Belmont, C a l i f . : Wadsworth Publishing, 1971), 228-240. Engel, J.F., D.T. K o l l a t and R.D. Blackwell. Consumer Behaviour. New York: McGraw-Hill, 1967. Eysenck, H.J. "The L o g i c a l Basis of Factor Analysis," Problems i n Human Assessment, ed. D.N. Jackson and S. Messick (New York, McGraw-Hill, 1967), 288-299. Festinger, L. A Theory of Cognitive Dissonance. Evanston, Row, Peterson and Co., 1957. Ferguson, G.A. "The F a c t o r i a l Interpretation of Test D i f f i c u l t y , " Psychometrika, VI (1941), 323-29. Frank, R.E., W.F. Massy and Y. Wind. Market Segmentation. Englewood C l i f f s , N.J.: P r e n t i c e - H a l l , 1972. 246 Green, :P.E. and.D.S. T u l l . . . Research for Marketing Decisions. . Englewood C l i f f s , N.J.: Prentice-Hall, 1970. Gorsuch, R.L. Factor Analysis. Philadelphia: W.B. Saunders and Co., 1974. Harmon, H. Modern Factor Analysis. Chicago: The University of Chicago Press, 1967. Katz, E. "On Reopening the Question of S e l e c t i v i t y i n Exposure to Mass Communications," Theories of Cognitive Consistency, ed. R.P. Abelson e t . a l . (Chicago: Rand McNally and Company, 1968), 788-796. Kirsch, A.D. and S. Banks. "Program Types Defined by Factor Analysis," Journal of Advertising Research, I I (September, 1962), 29-31. Massy, W.F. "Television Ownership i n 1950: Results of a Factor Analytic Study," Quantitative Techniques i n Marketing Analysis, ed. R.E. Frank, A.A. Kuehn and W.F. Massy (Homewood, 111.: Irwin, 1962), 440-460. Massy, W.F. "What i s Factor Analysis?" Multivariate Analysis i n Marketing, ed. D.A. Aaker (Belmont, C a l i f . : Wadsworth Publishing, 1971) 241-245. McGuire, W.J. "Selective Exposure: A Summing Up," Theories of Cognitive Consistency, ed. R.P. Abelson e t . a l . (Chicago, Rand McNally and Company, 1968), 797-800. McNemar, Q. Psychological S t a t i s t i c s . New York: John Wiley & Sons, 1969. M i l l s , J. "Interest i n Supporting and Discrepant Information," Theories of Cognitive Consistency, ed. R.P. Abelson e t . a l . (Chicago: Rand McNally and Company, 1968), 771-776. Nicosia, F.M. Consumer Decision Processes. Englewood C l i f f s , N.J.: Prentice-Hall, 1966. Nunnally, J.C. Psychometric Theory. New York: McGraw-Hill, 1967. Rummel, R.J. Applied Factor Analysis. Evanston: Northwestern University Press, 1970. Rummel, R.J. "Understanding Factor Analysis," Journal of C o n f l i c t Resolution, XI (December, 1967). Sears, D.O. "The Paradox of De Facto Selective Exposure Without Preferences for Supportive Information," Theories of Cognitive Consistency, ed. R.P. Abelson e t . a l . (Chicago: Rand McNally and Company, 1968), 777-787. Sears, D.O. and J.L. Freedman. "Selective Exposure to Information: A C r i t i c a l Review," Public Opinion Quarterly, XXXI (1967), 194-213. 247 Shaw, M.E. and P.S. Costanzo. Theories of Social Psychology. New York: McGraw-Hill, 1970. Sheth, J.N. "Multivariate Analysis i n Marketing," Journal of Advertising Research, X (February, 1970), 29-39. Sheth, J.N. "The Multivariate Revolution i n Marketing Research," Journal of Marketing, XXXV (January, 1971), 13-19. Siegal, S. Nonparametric S t a t i s t i c s . New York: McGraw-Hill, 1956. S i l l e r , F.H. "Newspaper Reading: A Study i n Selective Effects." Unpublished Ph.D. dissertation, University of Western Ontario, 1972. S i l l e r , F.H. and V.J. Jones. "Newspaper Campaign Audience Segments," Journal of Advertising Research, XIII (June, 1973), 27-31. Swanson, CE. "The Frequency Structure of Television and Magazines," Journal of Advertising Research, VII (June, 1967), 8-14. Tatsuoka, M.M. Multivariate Analysis. New York: John Wiley and Sons, 1971. Thurstone, L.L. "The Factor Problem," Problems i n Human Assessment, ed. D.N. Jackson and S. Messick (New York: McGraw-Hill, 1967), 279-287. Tucker, L.R. "The Objective D e f i n i t i o n of Simple Structure i n Linear Factor Analysis," Problems i n Human Assessment, ed. D.N. Jackson and S. Messick (New York: McGraw-Hill, 1967305-308. Twedt, D.W. "A Multiple Factor Analysis of Advertising Readership," Quantitative Techniques i n Marketing Analysis, ed. R.E. Frank, A.A. Kuehn, and W.F. Massy (Homewood, 111., Irwin, 1967), 427-439. Wells, W.D. "The Rise and F a l l of Television Program Types," Journal of Advertising Research, IX (September, 1969), 21-27. Addendum Dunn, S.W. Advertising. New York: Holt, Rinehart and Winston, Inc.,1969. Jones, V.J. and F.H. S i l l e r . "Reach and Frequency Trade-offs within a Vehicle," U.B.C. Working Paper. Lucas, D. and S. B r i t t . Measuring Advertising Effectiveness. New York: McGraw-Hill Co., 1963. Starch, D. Measuring Advertising Readership and Results. New York: McGraw-Hill Co., 1966. 248 APPENDIX I  Description of Data Collection Method'*" The data used i n t h i s study were generated as part of a research project carried out by the London Free Press i n cooperation with the Canadian Daily Newspaper Publishers' Association (CDNPA) and ORC Inter-national Ltd. The research methodology had already been s a t i s f a c t o r i l y tested with a p i l o t project i n Hamilton, Ontario and, accordingly, the same procedures were adopted for the London study. measurement concept The object of measurement i n the data c o l l e c t i o n was the potential reach of the newspaper. A research instrument was designed according to the concept of "opportunity for exposure," a procedure already established i n the broadcast industry. "Opportunity for exposure" i s understood to measure the number of people l i s t e n i n g to or watching a certain radio or t e l e v i s i o n program i n a given time period. Such a measure indicates nothing about what individuals actually hear or see, only that a radio or t e l e v i s i o n was turned on and and "opportunity for exposure" available. This measure has proven acceptable to media buyers and, hence, the concept was adapted to newspapers. The basic measures proposed were as follows: 1) page exposure - reports of a respondent looking at a given newspaper page. 2) space exposure - reports of a respondent looking at a given quarter of a newspaper page. ^"This discussion i s adapted from Chapter IV i n F.H. S i l l e r , "Newspaper Reading: A Study i n Selective Effects," pp. 66-80. 249 Page exposure, i t was argued, was eomparable with the major divisions of a broadcast day, such as hourly segments, while space exposure could be compared to 15 or 30 minute time s l o t s . data c o l l e c t i o n strategy An aided-recall, self-administered questionnaire was used for data c o l l e c t i o n . As any record of exposure necessarily had to be made before r e c a l l faded, i t was decided that the questionnaire should be administered on the day immediately following the publication of the test newspaper. The timing pattern was to measure each Thursday evening newspaper on the following Friday morning for s i x consecutive weeks (October 3 to November 7, 1968). The self-administered, questionnaire method was chosen over personal interviewing as i t was deemed impossible to interview a l l sample members on the day after publication with a f i e l d team of manageable siz e . the newspaper page questionnaire The decision concerning the design of the questionnaire was taken as part of the preceding Hamilton study. This design consisted of a high quality, photo-reduced, offset printed, 8%" x 11" r e p l i c a of a newspaper page. P r i o r to being photographed, each page of the newspaper was divided into quarters by attaching cross-hatched tape. Stickers were then added to the quarter pages asking whether respondents looked at each one (see Figure 5). As a precaution i n the Hamilton study, some respondents were presented with f u l l newspaper pages (rather than photo-reduced pages), 250 while some other respondents were personally interviewed (rather s e l f -administering the questionnaire). Further, some respondents were repeatedly exposed to one of these alternatives and the results compared with those exposed only once. The Hamilton study reported l i t t l e contamination from r e p e t i t i v e measurement and l i t t l e or no differences between using a miniaturized (photo-reduced) newspaper page and a f u l l - s i z e page. However, there was a s l i g h t increase i n the l e v e l of recorded exposure where the self-administered questionnaire was used rather than the personal interview. Nevertheless, the questionnaire technique i s similar to those used i n other media. The entire newspaper was not used to form the questionnaire. Approximately 20 of 60 to 90 pages of the complete newspaper were selected for miniaturization. The minimum number of pages i n any week was 16 and the maximum 22. The pages were selected on the basis of their being representative of a l l sections of the newspaper and th e i r inclusion of test advertisements which were of interest i n the study. communication disguise The sponsorship of the study was disguised from respondents i n order to avoid any possible effects such knowledge might have. Further, respondents were to l d they were p a r t i c i p a t i n g i n a t o t a l communications study and information was collected on both radio and t e l e v i s i o n exposure as w e l l as newspaper exposure. 251 additional interest questions Also included i n each questionnaire were a number of items concerning additional interest topics. These included opinion related measures such as newspaper coverage, newspaper source, newspaper personality, advertising rating and le i s u r e interests (see Chapter IV). follow-up questionnaire In addition to the above, a group of separate attribute scales was administered to the sample i n a separate mail-out questionnaire. The contents of the follow-up questionnaire, investigated newspaper rating, personality t r a i t s and li b e r a l i s m . (Again, see Chapter IV for items used i n this dissertation as part of external v a l i d a t i o n procedure). 1 sampling plan and r e c r u i t i n g procedures The sampled population for the London study was defined as a l l persons within the greater London area who l i v e d i n the home delivery zone for the newspaper and were f i f t e e n years of age and older. The respondent did not have to be a subscriber to the London Free Press. An area sampling technique was used to select households within the e l i g i b l e regions of greater London. Within each household randomization procedures were employed to choose s p e c i f i c respondents.'^ The steps i n the selection procedure were as follows: "1. L i s t i n g of enumeration areas i n the study zone. 2. Random selection of enumeration areas as primary sampling units. ''The study questionnaires are presented i n S i l l e r , pp. 225-270. 252 3. Random s e l e c t i o n of a s t a r t i n g point. 4. Predesignated l i s t i n g of households i n sampling frame. 5. Up to three c a l l s to r e c r u i t within a household. 6. Predesignated respondent s e l e c t i o n procedure with no s u b s t i t u t i o n allowed within the household." 1 The s e l e c t i o n of a s p e c i f i c respondent within a household was determined by information c o l l e c t e d through a recruitment questionnaire. This questionnaire provided demographic data also used i n the d i s s e r t a t i o n for external v a l i d a t i o n . Recruitment was c a r r i e d out by 99 women divided into 9 teams and headed by an experienced captain who reported d i r e c t l y to the head of ORC's Toronto interviewing department.-; Most of the women involved i n recruitment had some p r i o r experience i n research work. A t o t a l of 1,996 people agreed to p a r t i c i p a t e i n the study. Detailed explanations concerning t h e i r involvement were provided but no pressure was applied to p a r t i c i p a t e . Respondents were not paid but were promised a small g i f t at the end of the study. The recruitment rate was 64% ignoring empty houses, language b a r r i e r s and pers i s t e n t not-at-homes. About 11% or 215 people dropped out of the study before i t s completion and an a d d i t i o n a l 531 sample members were eliminated due to various experi-mental treatments over the course of the study. This l e f t a f i n a l sample S i l l e r , p. 79. 253 size of 1,220 respondents of which'5% did not read the London Free Press during the entire study. The sample was partitioned into 3 panels organized according to the newspaper's truck delivery routes. This was done to enable the delivery of somewhat different versions of the newspaper to each panel and thereby f a c i l i t a t e certain experimental manipulations not relevant to this dissertation. The panels were balanced as much as possible according to size and demographic characteristics. Panels A, B, and C contained 402, 404 and 414 respondents respectively. 254 APPENDIX II  The Aided Recognition Measure The measure of newspaper readership i n t h i s study i s an uncontrolled, self-administered, aided-recognition measure based on units of one quarter of a newspaper page. As indicated i n Appendix I, the measure i s based on the concept of providing an "opportunity f o r exposure" and i s analogous to measures used i n the broadcast media which are generally accepted as standard reach information among media buyers. The measure i s also conceptually s i m i l a r to the "noted" type of measurement commonly used i n a d v e r t i s i n g research with respect to p r i n t media. The "noted" measure reports the percentage of readers who state that they have previously seen an advertisement i n a p a r t i c u l a r magazine. The "noted" measure does not imply r e c a l l of message content. A more sophisticated measure, re f e r r e d to as "seen/associated," investigates r e c a l l of the advertised product or service. Another measure, c a l l e d "read most," investigates the a c t u a l amount of written material r e c a l l e d . ^ There are a number of l i m i t a t i o n s associated with .these recognition methods a r i s i n g from the- freedom of the respondent to claim recognition for whatever he wishes and the i n a b i l i t y of investigators to accurately '''Further discussion of these measures and t h e i r widespread a p p l i c -ation i n media and a d v e r t i s i n g research can be found i n the following sources: S.W. Dunn, Advertising (2nd ed., New York: Holt, Rinehart and Winston, Inc., 1969), pp. 603-607. D. Lucas and S. B r i t t , Measuring Advertising Effectiveness (New York: McGraw-Hill, 1963), pp. 50-53. D. Starch, Measuring Advertising Readership and Results (New York: McGraw-Hill Co., 1966). 255 id e n t i f y errors resulting from the measure. Commonly accepted causes of error i n recognition measures can be summarized as follows: "1. A genuine confusion with other material seen. 2. Guessing about recognition under conditions of uncertainty. 3. Making deliberate overstatements or understatements about recognition. 4. Deducing that the material was seen on the basis of recognition of surrounding material. 5. Deducing that the material was seen on the basis of a knowledge of one's own reading habits. 6. A desire to please or disappoint the interviewer. 7. A hesitancy to show ignorance. 8. A misunderstanding of the response i n s t r u c t i o n s . " 1 Other li m i t a t i o n s arise from the long, often t i r i n g , nature of the interviews or questionnaires which may result i n poor performance on the part of both respondent and interviewer. Another i s the d i f f i c u l t y and cost of obtaining representative samples of the population leading very often to the use of judgement samples. F i n a l l y , there i s a need for q u a l i f i e d investigators and high standards i n the execution of the f i e l d work. Despite such l i m i t a t i o n s , recognition measures are widely used i n media and advertising investigations. The measure does have the advantage S i l l e r , p. 282. 256 of assessing exposure under r e l a t i v e l y normal conditions, that i s , i t does not require manipulation i n an a r t i f i c a l environment. Further, as has been discussed, the concept i s widely adaptable to a variety of media. Further, i t remains a r e l a t i v e l y inexpensive means of measuring audience exposure. 257 APPENDIX I I I Relation of Research Findings to Current Audience Measures In t h i s t h e s i s , the primary focus has been on aggregate audience exposure as a function of i n d i v i d u a l s e l e c t i v i t y . As the research method i s not generally used i n day-to-day ad v e r t i s i n g research, i t i s of some i n t e r e s t to consider the findings i n terms of what i s already known about newspaper readership. As should be evident, measures currently used to assess newspaper audiences are based on a l l readers who look at any part of the newspaper. Hence audience measures tend to be based on c i r c u l a t i o n and are aggregated to the extreme. In f a c t , whatever l i t e r a t u r e e x i s t s on s e l e c t i v i t y within newspapers does not recognize that the newspaper may be a highly s e l e c t i v e medium. S.W. Dunn, for example, suggests, "People who read the paper are usually a cross section of the people who l i v e i n the trading area served by each newspaper. In i t s e l f , a newspaper i s not a s e l e c t i v e medium. I t i s read by men and women at a l l educational and income levels."1 Dunn i s , of course, r e f e r r i n g to the medium as a whole and not i t s i n t e r n a l sections. However, th i s i s p r e c i s e l y the point - there i s no: recognition of i n t e r n a l sections. Whatever aggregate measures do e x i s t on i n t e r n a l s e l e c t i v i t y f o r newspapers tend not to deal with s e l e c t i v i t y per se but rather with i d e n t i f y i n g some audience c h a r a c t e r i s t i c s associated with p a r t i c u l a r content. '''Dunn, p. 452. 258 To this extent, both this dissertation and the predecessor research by S i l l e r represent a new departure i n newspaper audience measurement by establishing the s e l e c t i v i t y p r i n c i p l e and associated aggregate measures independently of audience characteristics. I t i s only as a second, step (external v a l i d a t i o n within the context of this, dissertation) that any r e l a t i o n between s e l e c t i v i t y and audience characteristics i s investigated. As a r e s u l t , data are generally unavailable concerning aggregate audience exposure which can act as a meaningful basis of comparison for the results of this thesis. There are, however, two exceptions to t h i s conclusion. F i r s t , there has been substantial concern over some of the minor st r u c t u r a l aspects of aggregate exposure to newspapers. These involve such things as the influence of right versus l e f t page, top versus bottom of the page, etc. Conclusions and rules of thumb concerning such matters are available from Daniel Starch. 1 No corroborating evidence of any major influence a r i s i n g from these lesser s t r u c t u r a l aspects was found i n the course of this study. However, i t should be noted that i d e n t i f i c a t i o n of such sources of variance was not a primary objective of the study. Second, working from the basis of audience characteristics rather than the newspaper i t s e l f , research has demonstrated that sex differences account for substantial variance among interests i n news and promotional information. Starch, p a r t i c u l a r l y , has documented such sex differences. Thus, within the context of t h i s d i ssertation, the Starch, pp. 45-84. 2 I b i d , pp. 85-93. importance of male versus female interests, both i n the interpretation of factors a r i s i n g from the analysis of advertising data and i n the external validation procedure, i s corroborated by more conventional audience assessment. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0100118/manifest

Comment

Related Items