.n w .1 '3“:- 4' 'p . r 4”. “,.v4, A u. x A .. \ . .c. .u‘ r....‘.-~M flit-uh... .. A q‘l‘1 us . ,. z "' - -.~\‘~" “n," u a - ‘ “ ~ '34 v‘l‘a'x'ti“ -~._‘ \ ' 3/03“ '1! ‘vvv I ‘- . 0 ‘ »\ , u ‘ 34;" - - vV . . ‘ J b i‘I.~- . 1 ”guxk'u in ' x- ? ‘4“ l‘ l ‘) .c ,...~ '4 4.“: ” , «If u 1. u. - _ g. . )VHJJy. « ‘, . - ~ :9 ', A - ‘ wggazyum via-n3 H '. Nu" -" Vt . ' ' ~ I ' 5“ ' ’ " :“‘J’;‘. 1‘(‘ Ho” ‘ . ' ‘J ; ,1 ‘ s .:<‘( ' l Add . 1“ 1 ~ ‘3 5‘2“". (I ‘ . . . , . 4;. .7 31:13! , («J ‘ \. w ‘ L .. ‘1 ' fl ' ; f , . -! vhf; ’v Jh'x‘“ ~. 0‘. 3 'yJJi 1". 14 . . ‘ ' I - I ‘ «49!: m .‘AJ/‘f i Lb '; ' 1 v ’ "\ r'v. L423 3"; » a‘nA . - Iv - 31‘.‘ c. .x “‘3“ W < "I! z ‘ ) fi‘Vrivx‘ AU ‘1“. (r- ‘ n $1.1 . n .l 3',"¢;u‘i‘c‘i~"- l. a . w ‘1 ’ 2‘1, 5 5'54 3 ’32 nu “I“ \- . an 0.1, ,5! . 1“?” ' X J ~. A “V f ‘ ‘r‘i'fi-fi‘jhifi V J “~“’;§‘,')" ' * n . l u. ‘V . )‘ .. 1‘. 4» : ,-_ d r 3, . . e 1,. ‘ “if? 2"." ." ‘f a (‘.‘.‘5‘4"‘~3“‘ I #315973} ‘ '1 in . I“ f '7‘ ‘ 4“?! ”cf; , 1 n , , - ( . . .. ;. -J l-(".‘€‘zfl"”rr, . . 1:. . . ”My .. 130,-" 3}:‘-"*’¢s‘3"'>‘-3" I .l 4 \ 4 , -' A W H ‘ . ‘ " h ‘ d «H‘K‘scf‘f ‘ ' N" '( ,, ‘ , . - _ . 3'. , . ~ a {53.{‘J:5;1§‘.gfm,“ 1 9-33-3325? 1 “in. I' h 4 #1.," ‘ aw ~ « "‘ )‘J ’ .- "l‘ Pi { "' 'v’ k ‘ ' ‘ #7 L‘Tia.£'a‘.’"'u’ ‘ ‘ ‘ ’ ‘1 «Jam? c M‘. H ‘ S" L- '- - .4."‘§'(’3l" ( ~‘r ‘Jf‘f'f ‘ was" .xfl‘a’. {‘x‘r‘fl ‘ o. n, . .v ' .‘ ,1 1“ .;-m€‘.x . 4' At"" 1-3;”th 3?" ”‘0: ‘1 in v H . ’7 ‘mwQ " ,w. ,. h f‘.“”“-’.- 4' ,{ihv'P‘TQ‘I , ,‘r‘ «in. . ‘, r m. - 4. ,2” 32"}.7'43?" 91:?" o ' ‘ ‘ F“- 1“ ‘11:] ‘ v , , :. . . ’ n t I, . A ”I {if Ivj‘x‘var n f“: 4‘ n , I a ’r‘Hfia n)...“ ,t _ 7J1?) 3 1293 00538 614 « LIBRARY Michigan State University This is to certify that the dissertation entitled A MODEL OF RADIO LISTENER CHOICE presented by Edward E. Cohen has been accepted towards fulfillment of the requirements for Ph.D. . Mass Media degree m Date “‘1’83 MSU is an Affirmative Action/Equal Opportunity Institution 042771 MSU RETURNING MATERIALS: Place in book drop to ngaARIES remove this checkout from w. your record. FINES will be charged if book is returned after the date stamped below. WNW!” ~.r;. ‘ ,y ~ ~ 00 , oec¢*"?ft 'jj4“‘ 335¢ 4 060 .- .1 I. ”i“. ‘ ~35 13.. Q 1" MAR 2 3 2013 {3113 13 A MODEL OF RADIO LISTENER CHOICE By Edward Ellis Cohen A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Ph.D. Program in Mass Media Department of Telecommunication 5454451 ABSTRACT A MODEL OF RADIO LISTENER CHOICE By Edward Ellis Cohen The nature of how radio listeners choose the stations they listen to has never been researched in the context of the post-television radio industry. In addition, what little research does exist about radio is often proprietary to radio stations that commission the work. This study proposes a two stage model of radio listener choice, one for initial choice and another for day to day use once a station repertoire has been chosen. As part of the development of the model, the construct of listener loyalty is tested through use of a new scale, and the question of how listeners make changes to their station set is examined. The model proposed for initial listener choice is based on Tversky’s elimination by aspects (EBA) model, a noncompensatory model that states decision makers will select choices based on an order of elements involved in the choice. Listeners will choose an element that is most important. All choices that meet this criterion will then be part of the set for the next element. The process continues until one choice is left. A modification of the model allows for more than one choice to be left at the end of the process. The second stage of the model proposes a "first acceptable choice“ model for day to day changing. The research involved a sample of 904 radio listeners 18 years of age or older in a large, medium, and small market. The telephone study was conducted during May and June, 1987 and involved an approximately fifteen minute interview. The data show strong support for the EBA model as a model Of initial listener choice, however there was a lack Of support for the first acceptable choice model. The loyalty scale proved tO be reliable and showed gOOd validity. Additionally, numerous valuable findings for broadcasters based on demographic data were discovered. Copyright by EDWARD ELLIS COHEN 1988 ACKNOWLEDGEMENTS A great number of people helped out in bringing this dissertation to fruition. I am grateful to the people at the National Association of Broadcasters for their financial assistance through the NAB Research Grant program and for hiring me when it came time to leave Michigan State. My thanks to Dr. Rick Ducey and everyone else in Research and Planning who have been so encouraging in helping me to finish this monstrosity. My appreciation also goes to my committee members. Drs. Tom Baldwin, Brad Greenberg, and Linda Kohl have all been very helpful with comments and advice. Certainly, my greatest debt is to Dr. Carrie Heeter, a good friend as well as dissertation advisor; her guidance and wisdom were always welcomed. She always believed that this project would be f inished, even when I didn’t. To my mother, Phyllis Cohen, thank you for everything. Although you probably still don’t know why I went after a Ph.D. (sometimes I don’t either), you’ve always supported me and that is important. There are numerous others who deserve some specific acknowledgement, but then this section would be longer than the rest of the dissertation. All I can say is a heartfelt but insufficient "thank you.” Finally, I must thank the medium of radio. It has been a part of me for some time and has treated me well, both professionally and personally. Despite attempts to find other areas to work in or study, radio is the first love, always drawing me. This dissertation is only one more manifestation of that attraction. TABLE OF CONTENTS Introduction and Literature Review .................................. 1 Introduction ............................................... 1 Literature Review ........................................... 3 A Choice Model for Radio .................................... 12 Research Questions, Propositions, and Design ........................... l9 Pretesting .................................................. 19 Research Questions .......................................... 21 Design .................................................... 27 Results ......................................................... 30 Demographics .............................................. 30 Time Spent Listening ........................................ 33 Station Repertoire ........................................... 36 RQl: Initial Selection of Repertoire ............................. 38 Importance of Format Elements ........................... 38 The Structure of Listener Choice .......................... 54 RQ2: Stability of Station Repertoire ............................. 58 Listener Satisfaction ................................... 58 Listener Loyalty ....................................... 62 RQ3: Patterns of Listening Change .............................. 74 Effect of Location ..................................... 74 What Causes the Change ................................. 81 The Process of Finding Another Station .................... 92 Conclusions ...................................................... 100 Implications for Broadcasters .................................. 100 Implications for Researchers ................................... 109 Strengths and Weaknesses of the Study ........................... 113 Suggestions for Future Research ................................ 115 Appendix: Questionnaire ........................................... 118 Bibliography ..................................................... 126 vi LIST OF TABLES Table 1 Demographics .......................... 31 Table 2 Time Spent Listening ..................... 34 Table 3 Station Repertoire ....................... 36 Table 4 Importance Scores of Radio Format Elements .......... 39 Table 5 Element Importance Scores by Gender .............. 41 Table 6 Correlations Between Time Spent Listening and Importance. . . . 42 Table 7 Element Importance Scores by Income .............. 43 Table 8 Element Importance Scores by Education Level ......... 44 Table 9 Element Importance Scores by Age ............... 46 Table 10 T-Test Results for FM Versus Stereo Importance ......... 48 Table 11 Element Scores by Market Size ................. 49 Table 12 Regressions on Importance of Elements ............. 51 Table 13 "Victories" for Tested Radio Elements and Distribution of Hierarchy Scores .......................... 55 Table 14 Hierarchy Scores by Demographic Groups ............ 56 Table 15 Correlation with Hierarchy Score ................ 57 Table 16 Listener Satisfaction by Demographic Groups .......... 59 Table 1? Correlation of Satisfaction with Other Variables ......... 60 Table 18 Regression on Satisfaction ................... 62 Table 19 Factor Analysis and Means for Loyalty Scale .......... 63 Table 20 Loyalty Scores by Demographic Groups ............. 64 Table 21 Correlations with Listener Loyalty Scores ............ 66 vii Table 22 Regression t, Table 23 How Likely Would You Be To Try A New Station... Table 24 Trying a New Station You Didn’t Know by Demographic Groups . Table 25 Try a New Station That Sounded Like Your Favorite by Demographics ........................... Table 26 Correlations with Trying a New Station You Didn’t Know . . . . Table 27 Correlations with Trying a New Station That Sounds Similar to Your Favorite ........................... Table 28 Table 29 Table 30 Table 31 Table 32 Table 33 Table 34 Table 35 Table 36 Table 37 Table 38 Table 39 Table 40 Table 41 Table 42 Table 43 Table 44 Table 45 Table 46 Regression on At Home Change Behavior ............ Regression on In Car Change Behavior ............. How Often Do You Change Stations .............. At Home Change Behavior by Demographic Groups ....... In Car Change Behavior by Demographic Groups Correlations With At Home Change Behavior .......... Correlations With In Car Change Behavior ............ Regression on Frequency of Changing At Home ......... Regression on Frequency of Changing in the Car ....... How Often Do Items Cause Change ............... Correlations With Causes of Change ............... Causes of Change by Age ................... Causes of Change by Education ................. Causes of Change by Income .................. Causes of Change by Gender .................. Causes of Change by Market Size .............. Inter-Item Correlations for Change Behavior ......... Regressions on Possible Causes of Station Changes ....... Method of Change ...................... viii .Iyalty ...................... 67 68 69 7O 72 72 73 73 74 75 76 78 79 8O 80 82 83 86 86 87 87 87 89 90 92 Table 47 Chi-Squares for Change Strategies at Home ...... 94 Table 48 Chi—Squares for Change Strategies in the Car... 97 ix Figure 1 LIST OF FIGURES Example of Preference Tree ..................... Figure 2 Program Element Scores. . . . . ....... . ......... 18 40 Chapter 1 Introduction and Literature Review In r i The purpose of this dissertation is to explore the process of how listeners choose the radio stations to which they listen. The key word is ”explore" because in the thousands of pages of radio research that are issued every year, none of it looks at the actual choice process. We have a great deal of information about what people are listening to. Arbitron Ratings and Birch/Scarborough Research release quantitative syndicated radio ratings reports on a continuous basis. In fact, Birch/Scarborough will also let a subscriber know what products radio station listeners buy. On a national basis, Statistical Research Inc. conducts studies for the radio networks known as RADAR. Many other companies produce information for sale, such as Simmons, IRI, and others. While much of the above information makes its way into the hands of those in the radio industry, there is another large segment of information that remains totally secret. This is the output of the custom research companies that conduct studies for individual radio stations. The competitive nature of the radio industry means studies probing the attitudes of listeners in local markets are not made available to anyone outside the station that pays for the work. The needs of the market and a lack of interest on the part of the academic community have combined to produce a very limited amount of public 2 research on how listeners interact with radio. This study will go beyond the simple "who is listening to what station at what time" to give a look at the methods listeners use to choose the radio stations they listen to. There are good reasons for singling out United States radio for study. The medium is unique, both among the mass media and within the greater context Of how choice theory is applied.CRadio stands out from most other forms of mass media in that different radio stations are programmed to attract different segments of the total radio audience. In all but the smallest markets, commercial radio is ”formatted'tland even most public radio stations, while offering occasional block programming, do tend to cater to specific audience segments, usually those interested in classical music.¢ln broadcast television, most programs are designed to maximize audience and to cut across large population segments. Further, programs are presented for different audiences on the same broadcast outlet. Even a television program aimed at a very specific audience segment, for example teenagers, may be followed by one designed for middle-aged adults (the tenets of audience flow notwithstanding).3 . A case can be made for the similarity of radio and most magazines. With only a few exceptions, magazines cater to a small segment offering articles and advertising aimed specifically at those who share a particular interest, such as a specific hobby or trade. Even with the similarities, two major differences remain. One is cost; very few consumer magazines are free. The other is the delivery system; consumers must make an effort to get the magazine. Radio is literally everywhere. Some cable television networks do present continuous programming for specific audience segments. MTV is an obvious example along with ESPN, CNN, and the Nashville Network. Again, radio differs from cable in the same way 3 that it differs from magazines; cost and delivery. The cost here is direct; the fee the cable operator charges the consumer for the service. The delivery difference is the necessity of the cable outlet or satellite dish for reception of the television signal. One of the factors affecting choice in radio may very well be where the listener uses the medium. While cable offers only a couple of locations (home or public places such as a restaurant or bar), radio remains ubiquitous, going anywhere and everywhere. The key difference between radio and nearly all other products or services studied by choice researchers is obvious to any beginning marketing student: two of the "four P’s" are missing. Product and promotion remain, but price and place do not. Place is not a problem if you are within range of the signal of the station you wish to hear and you own a radio (it is unlikely any listener would travel just to hear a radio station). Price competition does not exist because all radio stations are free to the listener. Many of the factors that researchers consider when examining consumer choice simply are not relevant with respect to the choice decision in radio. This is what makes radio unique both among the mass media and the greater array of products and services. r v' w There is no literature that directly deals with the radio choice process. Instead, the research has centered on the use of the medium (Troldahl and Skolnik 1968, Lull, Johnson, and Sweeny 1978, Wober 1984b) or who is listening and how much time the individual spends with the medium (Teel, Bearden, and Durand 1979, Schlinger 1981, Lull, Johnson and Edmond 1981, Hagerty 1983). Another research area has been the relationship of music and radio listening (Wober 1984a, Baldwin and Mizerski 1985), and measurement has occasionally been studied (Beville 1983, Beville 1985, Cohen, Baldwin, and Samuels, 1987). 4 Without questioning the quality of the writings cited here, the quantity of research on radio since the advent of television has been skimpy. The general choice literature contains a large amount of research that can be best reviewed by breaking down the choice process. Choice consists of different elements although not all researchers agree on what those elements are. A common thread is that choice occurs only in situations where a subject must decide between alternatives in order to best achieve some goal or outcome. This implies that a choice involves more than one alternative. Conversely, a situation with only one possible outcome is not a choice. The first element of choice is recognition by the decision maker of a situation where a choice must be made. There must be needs, wants, or desires to be fulfilled to allow someone to enter a choice situation. The need may be something major such as buying a home or choosing a college to attend. It may be very minor, for example, the choice of what brand of laundry detergent to buy at the store. Later, the question‘of habitual choice will be discussed, that is, whether or not habit can eliminate what would appear to be a choice situation. In the meantime, it is assumed that humans recognize choice situations. After recognition of a choice situation, search behavior may take place. The amount of search should correlate positively with the level of involvement of the consumer in the decision. High involvement choices such as the choice of a college to attend may involve greater search than a low involvement decision (Chapman 1986). Stemming from Krugman’s work in the ’60’s, involvement with a product is thought to determine how much of a decision process takes place. Stone (1984) defines involvement as the time and/or intensity of effort expended in 5 the undertaking of behaviors. Involvement is important in describing the radio choice process. Is radio a high or low involvement product? Does it differ for different users? A recent research note (Bolton 1986) suggests radio is a low involvement product and that it is a low priority for most listeners. Bettman (1979) with his information processing approach to choice behavior suggested there is both internal and external search. Internal search would involve consumers reviewing stored information while external search is action the consumer takes to gain information about products. Schneider and Shif f rin (1977) reviewed how memory may be searched for information. Their psychological work has applications for consumer behavior. Search may not necessarily involve examining all possible alternatives. Heeter (1985) discussed exhaustive and restricted orienting search in the realm of cable viewing. This is directly applicable to radio (cable and radio being similar in the number of choices available to the consumer) as it would suggest that listeners do not search out and sample every radio station available to them prior to making a choice. Sheluga, Jaccard, and Jacoby (1979) agree and suggest that consumers may make better decisions when they do not process all the information available to them. ”Information overload" may exist for consumers in the search process. Numerous factors can affect external search. Writing in the consumer behavior (product) context, Moore and Lehmann (1980) reviewed six areas that can be determinants of how extensive the search for information may be prior to purchase. These were market environment, situational variables, potential payoff and product importance, knowledge and experience, individual differences, and conflict and conflict-resolution strategies. These categories can be applied to radio choice if one allows for minor interpretation in the payoff /importance category. While search is the physical action of attempting to change the informational environment to make a better choice, attention involves the narrowing 70f the range of stimuli to which the consumer responds (Howard 1977). Howard states that attention comes from arousal and arousal is specific only to a product class (i.e. radio) and not a particular brand (i.e. WXXX). Howard also believes that human behavior prior to "purchase" can be divided into three categories: extensive problem solving, limited problem solving, and routinized response behavior. This brings up the question of whether or not a complete choice process occurs for every choice situation (i.e. search, choice of a strategy, etc.). Howard offers the viewpoint that consumers move along a continuum from extensive problem solving to routinized response behavior as they become more familiar with products. Other researchers question whether a choice is actually made each time. Olshavsky and Granbois (1979) questioned the conventional wisdom that because two or more alternative actions exist, choice must occur. Their study led them to conclude that ...f or many purchases a decision process never occurs, not even on the first purchase. How could someone buy something with no decision process? The researchers suggested culturally mandated reasons, interlocked purchases, conformity to group norms, imitations, or reliance on the recommendations of others. As an example, just observe the number of men who bring along wives, girlfriends, or others to a clothing store and allow them to make the purchases (the author included). In addition, Olshavsky and Granbois remarked that even when purchase behavior is preceded by a choice process, it is likely to be limited. Consumer 7 behavior can be viewed as ritual behavior (Rook 1985), meaning activities that occur in a fixed, episodic sequence and tend to be repeated over time. Having reviewed the pre-choice actions, the choice process itself is next. Starting with economics, choice theory can be viewed from a base of utility theory. In the economic view of the rational man, choices are made on the basis of how much will be gained from them (subjective expected utility) (Wright 1984). It is possible though, to define rationality in other ways than simply maximization of utility (Einhorn and Hogarth 1981). Nonetheless, for each possible outcome in a choice, there is some associated utility (although this utility can be zero). It is assumed that the perceived utility can be measured in some fashion (whether or not it is accurate, hence the subjective part of SEU theory). Additionally, probabilities can be assigned to possible outcomes. Multiplying the utility value for an outcome by the subjective probability will yield the expected utility, leaving the decision maker with clearly valued options to which various strategies of how to choose can be applied. For example, the decision maker may choose to minimize loss or to maximize gain. Once a strategy is chosen, a decision can be made. SEU theory is easy to apply to decisions that lend themselves to direct measurement of costs and benefits, such as a choice between two investments. Expected payoffs and potential risk factors can be easily quantified. When dollars and cents choices are not involved, SEU may have little to offer students of choice processes. SEU theory does allow for the discussion of two concepts that arise often in the study of choice theory and some time should be spent on the applicability of each to this topic. The first is uncertainty, the concern each of us has that we may not have all the information needed to make the best 8 decision on a matter. Without that information, we may regret the choice we make (Janis and Mann, 1977). In the case of radio listening, uncertainty may be operationalized as the chance that a listener will regret the listening choice made. However, the existence of uncertainty may be questionable due to the absence of place and price. While uncertainty exists for a product that one must travel to purchase and then pay for, a radio choice that proves to be poor can be remedied quickly with the touch of a button or the turn of a dial. All that has been lost is a short amount of time. Due to the lack of penalty and the fact that much listening is done individually, uncertainty may be of minimal importance. The other concept is risk, much like the risk involved in visiting Las Vegas or Atlantic City. Kozielecki (1981) defines a risky situation as "one in which we are not quite certain of the outcome and when making a decision, we cannot predict with any degree of certainty if it will lead to success or failure, whether we win or lose." Much of the work done on risk has involved the use of experimental monetary gambles. The research has shown most people to be risk-averse (Kahneman and Tversky 1984). Given two choices that are the same yet are worded differently, most people will choose the one that appears to offer less risk. Risk in the consumer behavior sense will generally mean the potential for loss of money or social loss, such as embarrassment. Again, the lack of a cost for radio negates the monetary risk and the personal listening eliminates most of the social risk involved. Much like uncertainty, risk may not be germane to the discussion of radio choice, because radio is a low risk choice situation. If one eliminates uncertainty and risk, a good portion of the choice research is eliminated, too. However, there remains the question of strategy. 9 A major strategy distinction is whether people use compensatory or noncompensatory methods to make their choices (Harrell 1986). A compensatory strategy is one where a decision maker will add up scores (mathematically or otherwise) for products on various criteria that are important and the product that scores the highest overall will be chosen. High scores on some criteria can compensate for low scores on other criteria. A number of noncompensatory strategies have been advanced. In lexicographic models, consumers rank choice criteria from highest to lowest importance. The choice alternative that scores highest on the most important criterion will be chosen. If there is a tie, the person will then choose the choice alternative that scores highest on the second most important criterion and so on, until a choice is made. In sequential elimination models, criteria are again ranked, but now the choice alternatives which lack the most important criterion (or score below a preset minimum) will be eliminated. This elimination process continues until only one choice alternative is left. In conjunctive models, consumers set minimum acceptance levels on a number of criteria.’ The decision maker chooses the choice alternative that meets acceptable levels on all of the criteria. If more than one alternative exists, then more criteria are added until only one choice remains. Finally, a disjunctive model finds the consumer setting minimum levels for a few crucial criteria. Alternatives that meet any of the criteria are considered acceptable. Depending on the type of decision to be made, an individual may use any of these methods. One area of controversy in decision theory is the question of how well a model can describe what happens in the average person’s mind. Previously in 10 this paper, SEU theory was rejected for just that reason as it is assumed that individuals simply cannot and do not assign mathematical weights to alternatives in an effort to achieve an optimal decision in the choice of a radio station. Involvement was described as an important factor because the less interest in the decision on the part of the decision maker, the less likely he/she is to follow a model that requires a large amount of thought. Compensatory models assume an extensive information processing capability on the part of the decision-maker (Park 1978) and many individual decisions faced by a person in today’s society may simply produce too much of a burden if these processes are followed. Park criticizes lexicographic models based on the potential for not only a suboptimal choice, but a ridiculous one at that. Second, the more alternatives involved, the more complex the decision and therefore, the greater the possibility of suboptimal choice. Finally, he criticizes conjunctive models for their rigidity and their inability to always generate a specific choice. All of his criticism comes in the context of presenting a sequential conflict resolution model that will be discussed later. When comparing the various choice models advanced in the literature, the problem is to decide which should be the most likely to be used in a situation. Involvement has been cited as an influence in choice models. The problem of search was mentioned. The problem of the amount of cognitive activity on the part of the decision-maker is important, regardless of the involvement level. Further, there is a problem with the number of alternatives to be presented in the choice situation. Research has shown that a decision maker (DM) may choose differently depending on the number of alternatives available (Chakravarti and Lynch 1983, Huber and Puto 1983). The DM may also choose ll differently depending on the range of appeal of the alternatives, the ”range effect” (Hutchinson 1983). For example, if three alternatives are given, the choice outcome may be different from that made with ten alternatives. If one alternative is changed to an outlier, one of the other choices may appear to be a ”better” choice to the DM. Beyond that, the process used to make the choice may vary with the number of alternatives available (Johnson 1984). C; All of the above considerations are important to radio. In most markets, the number of choice alternatives is large. Listeners may not be aware of all alternatives. They may not sample every station possibly due to taste considerations. Also, context effects may differ greatly over time, for example, if a station changes format from easy listening to become the third adult contemporary station in the market, there may be more attention paid to the adult contemporary format, at least temporarily. That may increase overall listening to the format, even if the new station is not successful. The cause may be the greater amount of attention caused by the change. I In his discussion of psychological theories of consumer choice, Hansen (1976) divided choice into situational variables, predispositional variables, and interaction rules. Situational variables involve a two by two matrix between actual physical stimuli, perceived aspects of the situation, specific aspects, and general aspects. In other words, how the environment acts upon the consumer and how the consumer perceives the environment may not only affect choice but may affect how the choice is made. Little research has been done in this area. The predispositional variables include personality, general attitudes, values and interests, specific attitudes, beliefs and images, and choice-specific predispositions. The interaction rules signal a return to the debate between 12 compensatory and non-compensatory rules. Hansen concludes by suggesting that a model of consumer choice should be flexible enough to explain what choice principle is applied and what predispositional variables have become important in making the choice. Einhorn and Hogarth (1981) mention the possibility of "meta-strategies" of choice. Kassarjian (1982) points out the lack of examination of cultural factors in choice. A consumer’s mood may also play a role such as the satisfaction or dissatisfaction with a product (Gardner and Vandersteel 1984). The educational background of the consumer changes his/her choice habits, too (Gronhaug 1974). As an example, a positive correlation has been found between education and the consumption of mass media as well as between education and the reading of advertisements. A ' M 1F i The question of what choice model best applies to radio is the goal of this dissertation. Thus, for the same reasons that subjective expected utility theory was removed from consideration, it is necessary to eliminate all compensatory (additive) models. Again, it is simply too hard for the average consumer to use some form of mathematics, however crude, to make a decision on a product as unique as radio. Radio does not have a price and no ready numbers to use. Radio also involves taste; in fact, taste should be a major part of a listener’s choice of a station. A listener will not generally choose something he/she does not enjoy. Without offering an answer yet to the question of high or low involvement, it appears that radio is a product that does not fit well into a compensatory choice strategy, where the decision-maker assigns scores on individual criteria and adds the scores to make a final choice. Many strategies still remain on the non-compensatory side. One common feature of these is however unlikely some models may seem at times (in terms 13 of yielding a "rational” choice), they all are relatively simple to use. They also handle larger numbers of alternatives more easily on a cognitive basis, allowing the DM a simpler method to make the choice. The noncompensatory choice model that appears most reasonable to use as a starting point is the Tversky model (1972). This is the "elimination by 'aspects" model, a relatively simple noncompensatory probabilistic method. It operates with the DM selecting criteria that matter to him/her in the choice situation and ranking the criteria in order of importance. The choice alternatives are then judged on the most important aspect. Those that fail to meet the criterion established by the DM for that aspect are eliminated from further consideration. If only one alternative remains, the choice is made. If more than one remains, the DM proceeds to the second most important aspect and the process continues until only one choice alternative is left. To transform EBA into a proposed theory of radio listening choice involves two stages. It is proposed that when making initial choices, listeners do view radio as a high involvement product, using the elimination by aspects model to make a choice. This initial choice will occur when a listener has moved into a new city, has become completely dissatisfied with his/her current choice set, or at any time that a major change in listening patterns is necessary. Use of this model may yield only one station or there may be several that meet all of the relevant criteria, though these may have some sort of preference order. It is further proposed that as one station may not satisfy the listener continuously, he/she will make changes. If these changes are made, the listener will use a simpler conjunctive or disjunctive strategy, treating radio as a low involvement product and only considering a limited set of altenatives. Much listening, especially in-car listening, involves a great deal of dial [4 switching, the radio equivalent of cable zapping (Heeter and Cohen, 1988). This can be explained through a change in mood, a conflict in taste, or some other factor usually brought on by something the radio station is currently programming. It may be news content to a person that uses radio for escape or it may be a song that is distasteful to the listener. The reason for the change is not important here. The key component is the triggering once again of the choice process. Technology also plays a role in the application of choice models in radio. When radio first came on the scene, a listener had to tune in a station by hand, a process that still exists for most radios today. However, technology gave us the pushbutton radio for cars many years ago and now, many stereo receivers for the home make use of microprocessor technology. The listener may store anywhere from four to twenty four stations in the memory of the receiver and listen to any of them at the touch of a button. Further, many home units now have remote controls that operate much like those for television sets, giving the listener the chance to "zap" radio stations. In television, the advent of remote control boxes for cable television and videocassette recorders led to profound changes in viewing habits (Heeter, 1985). As many car radios have the pushbutton feature and more home units now have memories and remote controls, there is a need to consider the ramifications of technology for the choice process. In the case of quick changes in listening, such as the car or home zapping, the conjunctive or disjunctive noncompensatory method may also be used. If a listener simply punches buttons upon hearing something he/she doesn’t wish to hear, it is proposed he/she will listen to the first acceptable alternative. Conjunctive/disjunctive methods may be used quite often depending 15 on the listener’s taste and the convenience of pushing buttons. The reasoning behind this high/low involvement proposal is the belief that most listeners are at least initially involved with what they listen to. Radio is important to them and at some point, they will take the time to evaluate what is available to them and identify a ”repertoire" of acceptable stations. They may do this once, after moving to a new city, upon reaching an age where they make the judgement, or even when they buy a new car and must set the pushbuttons on the new radio. The process may be triggered again by external or internal cues. External cues from the environment may be advertising for a new station that appears interesting to the listener, a change in programming of the listener’s current favorite station, comments from friends about another station they listen to, or even some sort of promotion a rival station is conducting. Internal cues may be changes in taste or changes in maturity levels. Any of these cues can trigger the elimination by aspects process again. In the absence of cues to trigger reevaluation of a listener’s station repertoire, there is no need for the listener to go through the relatively long (when compared to the conjunctive/disjunctive model) EBA process. Once the listener is satisfied with his/her set of choices, radio can become a low involvement product and the listener enters habitual choice. He/she either stays with one station or switches between a small set of stations (similar to the Heeter (1985) channel repertoire for cable) that have survived the EBA test, ignoring all others. Most listeners cannot tell you what station is on each button on their car radio, nor can most people name more than a few stations in a market (Heeter and Cohen, 1988). Reviewing Arbitron reports will show that the average listener samples between two and three stations a week depending on market size. No matter how many stations are available, only a 16 select few will survive the EBA process for most listeners. Another unique aspect of the proposed radio choice model beyond the two levels of involvement is the presumption that one single final choice does not have to be made. The EBA process can yield two, three, or more preferred choices due to radio’s lack of price and place. With unlimited switching allowed at no penalty, there is no reason for the listener to commit to a single, permanent choice as is the case with an automobile or even toothpaste. Proposing two levels of choice is not entirely unique. Park (1978) put forth the sequential conflict resolution model which uses a process involving one stage similar to both lexicographic and elimination by aspects models followed by a possible second stage that utilizes a satisf icing plus process. Going beyond the EBA model, a modification of the model may be made, the preference tree (Tversky and Sattath 1979). The use of a tree model makes the EBA model easier to handle cognitively. In fact, it now becomes EBT or elimination by tree. Visually, a preference tree appears to be upside down (see figure 1). This means that the DM makes his/her choice on the most important aspect, then follows that particular link to the next decision point. As with EBA, the process ideally continues until only one choice is left. As stated earlier, there is no necessity to have to eliminate down to one choice. The radio station repertoire can be the final choice through the EBA/preference tree model. In practice, the model should work as follows: Suppose a listener in the Detroit market has between thirty and forty stations from which to choose. The listener may choose music as the most important aspect to him/her, which would start the process on the music side of the preference tree, eliminating the all talk and all news stations. Next, the choice may be for country music, 17 another branch of the tree. This action will eliminate all but the country music stations. As there is still more than one choice available, another decision must be made. Let us assume the next most important aspect is personality, that is, a station that has a larger amount of talk on the air. The listener will now eliminate any country stations that emphasize a ”more music” approach. This should leave only one station in a market like Detroit, yet if it does not, there is no requirement to proceed further in the elimination by aspects process. A set of two or three stations is quite manageable for the average listener and he/she will then choose from among those few. If no choice remains after the EBA process, the listener must move back up the tree to find the point where the last stations were eliminated. At that point, a different criterion must be chosen to eliminate some of the options, or all must be kept in the set and a choice made from among those. 18 Figure 1 Example of a Preference Tree m3mz HH< xana :OfiunauomcH mmog mBOfim xama muuaaoo hufiamaomuom coaumaH0mcH one: hhuasou wawaouqu know can: . z . muucsoo :oamsz duos: mo< mmo xuom.unwwg. umsuo Chapter 2 Research Questions, Propositions, and Design The problem to be studied is listener choice. Commercial and academic research previously cited indicated that listeners have a small set of radio stations they listen to regularly. Three basic research questions relating to listener choice and the set of regularly listened to stations will be addressed: RQl: How do listeners choose their initial set of stations? RQ2: What causes listeners to add or remove stations from their set once the initial choices have been made? RQ3: What process do listeners use to make changes within that set when actually listening to radio? Each research question will be described in detail, including a brief overview of the issue, elaboration of the questionnaire items intended to address each question, and discussion of the analyses to be applied to the question. Study design and pretesting will be discussed first. r in Four focus group sessions were held during April, 1987, with nineteen MSU students to get a better idea of what factors are involved in listener choice. In addition to open-ended questions, sample questionnaires were administered to the groups to test questions, wording, scales, and indexes. The focus groups showed that, at least among the student population, more emphasis should be placed on social factors in choice than was previously thought. While many of the students did not claim to listen to the same 19 20 stations as their friends, asking about the initial choice process often elicited comments such as "I tried what other people around me were listening to" or "I heard a song I liked on the radio in a friend’s room, so I tried that station.” Students in a residence hall are a more homogeneous group when compared to social groups in other situations such as work, but the statements are important. Further questioning found some focus group members did use an EBA process when first arriving in East Lansing, reporting they listened to all the stations available shortly after they settled in. Further probing discovered that when a focus group member said he or she checked ”all" the stations, this often meant "all the FM stations.” Many had never tried AM and felt there was little reason to do so. This suggests that listeners may not be able to articulate their true hierarchy. For example, when asked what elements would be most important in an ideal radio station, focus group members invariably talked about music and disc jockies. If they were then asked whether they would listen to their ideal station if it were on AM, some said they wouldn’t. In that case, the first element in the hierarchy would be the method of transmission, not one of the programming elements. This issue was addressed in the questionnaire. Another noteworthy result of the focus groups was the determination of which elements are most important to listeners. Through talking and the pretest questionnaire, six key elements (out of fifteen presented to the groups) were identified. These were music, personalities (disc jockies), the number of commercials, news, weather, and one larger category that can be called "sound quality." This sound quality element consists of the band, the reception, the presence of stereo, and general sound quality. Due to the similarity of the elements, all four can be combined as one for measurement 21 purposes, although the groups tended to consider stereo slightly less important than the other criteria. When the group members were asked about trigger mechanisms for changing stations, just about anything seemed able to start them changing. As expected, programming elements could start a change, usually a disliked song, obnoxious commercial, or too much talking on the part of the disc jockey. Other obvious causes given were elements that were undesired at the time they were aired, for example, news when news wasn’t wanted. In some cases, mood played a role. Some members said they occasionally changed ”just to change" with no good reason for doing it. Others reported that as their moods changed, their listening choices changed. The questionnaire corroborated the spoken data regarding triggers for the change process. The conjunctive/disjunctive model appeared to be accurate with many in the groups. They often spoke specifically of changing to the next station that met a minimum criterion (usually a song the listener liked), however a few said they check all stations and then select the best from the group. Finally, station loyalty was discussed and tested. The students did not sound as if they were very loyal to their favorites. A scale and index were administered to attempt to assess if the level of station loyalty could be measured. Both were included in revised form in the questionnaire. W RQl: How do listeners choose their initial set of stations? It is proposed that listeners use a hierarchical approach to initially determine the stations they wish to listen to, specifically Tversky’s "elimination by aspects" (EBA) method. EBA suggests that decision makers choose by selecting first what they believe to be the most important aspect important 22 aspect of the ”product." All choices not having this aspect will be eliminated. If a number of choices remain, the process is repeated using the next most important aspect. Finally, one choice remains or the process must be repeated. A change is proposed in the process that would allow more than one choice to be available. Recognizing the unique status of radio as a "product” (no price or place), a listener may very likely have a set of chosen stations rather than a single option. The proposition that listeners choose their set of regularly listened to stations using a hierarchical EBA method was operationalized in several ways in the survey. First, ratings for the importance of fifteen radio elements on a seven point scale were requested to give interval measurements, permitting identification of which attributes are considered most important. The results were mean levels of importance for each of the elements that can then be ranked. Different profiles can be constructed for different subgroups, i.e. heavy versus light listeners, large repertoire (number of stations) listeners versus small repertoire, and various demographic comparisons. The results should give an idea of the specific elements that listeners prefer, both in total and in the groups. While those analyses will yield some specifies, that alone will not give full support to the proposition of a listener hierarchy. For that reason, a second analysis involved pair comparisons to test the hierarchical EBA model. The six items that made up the fifteen pairs were identified in the focus groups, pretests, and previous research as the most important elements and allow a reasonable number of comparisons for use in a telephone survey. The analysis is very simple. An algorithm for use in SPSS was devised that determines the structure of each listener’s responses. A perfect hierarchy 23 would find one element would be paramount, that is, chosen over the other five elements it is compared against, while another element would be chosen over four other elements, yet another would beat three, and so on. If the number of ”victories” for each element is squared and then summed across each respondent, an index of "degree of hierarchy" results. The perfect hierarchy gives a sum of 55. If an element is chosen in each comparison (rather than choosing ”not sure"), the least hierarchical score would be 39, based on three elements ”beating" three others and the other three "beating” only two. Without a doubt, some respondents will not choose between a pair of elements in some cases. This is not missing data, rather it should be considered a ”non-choice." Lack of choice between two elements signifies a lack of a hierarchy on the respondent’s part for that particular comparison. Therefore, using these ”non-choices" will result in lower ”degree of hierarchy” scores such that a respondent making no choices at all will have a hierarchy score of zero. Support for the proposition will come from higher "degree of hierarchy" scores from respondents. Lower scores will suggest that listeners use some other model for their listening choices. However, further subgroup comparisons will be run to determine if particular groups are more likely to use the hierarchical model. T-tests and analyses of variance will be used to provide the comparisons. RQ2: What causes listeners to add or remove stations from their set once the initial choice has been made? The research question asks how listeners make changes in their ”permanent" sets of stations. What stimuli will cause an addition or deletion from the set? Programming factors are certainly involved, but other 24 possibilities must be investigated. The roles of listener satisfaction with and loyalty to favorite stations probably play a role. Research question 2 will be analyzed in a few different ways. First, a ten item scale has been designed to be a measurement of ”loyalty" to the listener’s favorite station. The scale will be used as an index with a top score of 70 indicating someone extremely loyal to their favorite. It is assumed that an extremely loyal listener is less likely to change to a new station if a new one that is similar to the favorite were to come on the air. While this is an exploratory test of such a scale, it would be a valuable tool for station operators planning to change formats or buy stations. In this research, the degree of loyalty must be ascertained to determine whether listeners will change their permanent set of stations. In the pretest, this scale had a reliability of .76 (alpha) with the eleven items used. The scale measures the loyalty of listeners to their favorite station for use in answering research question two. Higher loyalty scores should mean a lower likelihood of a listener changing any part of his/her set of stations. Other questions established which station is the respondent’s favorite and how long the listener has claimed it as a favorite, other ways of measuring how loyal listeners are to their favorite station. Another question asked for a rating on an interval scale of how happy respondents were with their favorite station. This establishes a potential measure of the strength of the listener’s ties to the station as well as a dependent variable for use in regression analysis to be discussed next. Another assumed indicator of listener propensity to change stations permanently must be the degree of satisfaction with the current favorite(s). Two measures will be used here. 25 First, a direct measurement of overall satisfaction is requested using a seven point interval scale. Next, another set of questions directly measures the likelihood of the respondent trying a new station under two different conditions. One condition is simply "a new station" and the second is "a new station that sounds similar to your favorite.” In each case, the results should also show the likelihood of the respondent trying something new. If they are unlikely to try a new station, they are therefore unlikely to be considering adding stations to their listening set. RQ3: How do users make changes within their set of stations when they actually listen to the radio? Once listeners have chosen an initial set of stations, they can use this set for day to day listening. If it is accepted that the typical listener has more than one station in his/her set, then there must be some process for changing within this set. It was proposed earlier that listeners use a conjunctive or disjunctive strategy to choose the next station to be listened to. That means choosing the first station available that meets a minimum set of criteria or only one criterion (e.g., a song the listener enjoys is currently playing). It is proposed that listeners do not go through the drawn out EBA process proposed for initial choice. While the strategy is important, there must be a reason for the choice process to be triggered. While no hypotheses are specified, the research delves into three areas: RQ3a. What triggers the proposed "short version" choice process during day-to-day listening? RQ3b. What is the propensity of different listeners to change? RQ3c. What is the effect of the listening location on change behavior, 26 specifically in-car and in-home listening? The first question covers the cues that trigger the process, both from the radio station’s programming (music, commercials, personalities, etc.) and those from outside (mood). Listeners were asked about specific items that started the change process. Another set of questions addressed the change in daily listening situations, specifically the propensity for changing and the possible causes. Four of the questions concerned the change patterns and the test of the conjunctive/disjunctive strategy proposition. Specific format elements can cause a listener to change stations and various questions deal with this. Another question asks about whether the listener’s mood causes changes, something not controlled by the radio station, yet potentially equally important in terms of change. Beyond simple frequencies, the results can be broken down by subgroups using t-tests and analyses of variance, again by radio usage and demographic groups. The second research sub-question asks if different groups change more often than others (i.e., young versus old). It has been reported that younger listeners, specifically teens, are noted button pushers and dial twirlers, while older listeners may stick with one station much longer. While empirical evidence of this exists for television, the lack of evidence for radio dictates that this should be a question rather than a hypothesis. Individual listener characteristics to be used for comparison include: radio station channel repertoire, radio usage, age, sex, income, education, and intermarket comparisons. Sub-question three is proposed because of the radical differences in listening environments. In-car listening is often solitary, may generate more 27 attention to the programming (depending on where and when one is driving), and has different technical characteristics, both in reception (FM stereo can be especially difficult to receive clearly in a moving vehicle) and receiver (car radios are much more likely to have pushbuttons and seek/scan mechanisms than home receivers). The expectation based on Heeter and Cohen (1988) is that there will be more station switching in cars. T-tests and one-way analyses of variance will be used for comparisons of means will be used to compare various groups’ responses on these questions. Design This study was conducted by telephone in three markets: Seattle-Tacoma, Washington, Greenville-Spartanburg, South Carolina, and Fargo-Moorhead, North Dakota-Minnesota. Telephone was chosen because it enables a researcher to use a large sample size generating results that can be generalized easily to larger populations, making the results more useful to the broadcast industry. Drawing the sample was easy and both interviewers and facilities were readily available. Co-operation rates are generally high with telephone surveys and the questions asked here were readily adaptable to the telephone methodology. The main drawbacks are the lack of control and short length of the interview. The reasoning behind the markets chosen was relatively simple. It was desirable to be able to compare large, medium, and small markets. Seattle-Tacoma is a top twenty market, Greenville-Spartanburg is in the mid-sixties, and Fargo is considered a small market, as ranked by Arbitron Ratings Company, the larger of the two syndicated radio ratings services. The geographic separation was for efficiency in conducting the survey. Using a West Coast market allowed phone calls to be made beyond midnight Eastern time. Each market contained no more than three counties in the metro area as 28 defined by Arbitron Radio, making the markets manageable for drawing the sample. Other factors involved in the choices included the author’s familiarity with the Greenville-Spartanburg market and the desire to use a small market that did not receive outside stations from a major market. Such markets may exhibit listening characteristics similar to a large market due to the large number of choices. In the case of Fargo, the market is extremely isolated from any larger market. Approximately three hundred completed interviews were conducted in each market. That number was chosen to allow comparisons between markets with a reasonably low sampling error (approximately 5.7 percent at the 95 percent confidence level for each market when the full sample is analyzed). The sample list was drawn from the local phone books using a systematic sample with random digit dialing. Each phone book within each market was weighted for the approximate number of residential listings in that book. Screener questions were asked of each potential respondent, first to eliminate radio station employees who may give biased answers, believing the survey to be on behalf of a particular radio station and then to eliminate those potential respondents that did not spend a minimum amount of time listening to radio, at least two hours a week. A total of 178 potential respondents were eliminated on the listening question and 11 radio employees were found, removing some of the potential for misleading or uninformed answers. The number of ”less than two hour” listeners was 12.6 percent of all numbers where a qualified respondent was found (completes plus refusals plus non-listeners). The telephone study was conducted from May 27 through June 8, 1987 from East Lansing, Michigan. Both students and temporary help were used and nearly all interviewers had previous experience at telephone interviewing. All 29 were trained prior to the calling and were supervised by the author. A total of 904 completed interviews were conducted, 306 in the Greenville-Spartanburg metro, 294 in the Fargo-Moorhead metro, and 304 in the Seattle-Tacoma metro area. Sampling error for the overall sample is 3.25 percent at the 95 percent confidence level. Each metro area matched the metro survey area for the market as defined by Arbitron Radio. All respondents were 18 years of age or older and said they listened to radio for at least two hours a week. The response rate was 48.1 percent after eliminating business and disconnected numbers, while 17.6 percent refused, 34.3 percent were no answers, and 5.3 percent were busy or answering machines where no resolution of the number was made. All phone numbers were called a minimum of three times. Chapter 3 Results Presentation of results will parallel the research questions. First, background characteristics of the sample will be reported: demographics, time spent listening to radio, and station repertoire. Next, findings related to research question 1, how listeners choose their initial station repertoire, will be examined. These include the importance of various format elements and whether a hierarchical choice process is evident. Research question 2 considers the stability of listeners’ station repertoire exploring their satisfaction with and loyalty to favorite stations and their likelihood of trying new stations. Research question 3 addresses the issue of switching stations within listeners’ regular repertoire: frequency of changing stations, factors which cause them to change, and their approach to changing stations. For each research tapic, the impact of individual differences across listeners will be assessed. Interrelationships between major study constructs will also be tested. Thus, each facet of radio listening that was investigated here will be presented individually, followed by correlations with previously presented data, one way analyses of variance with each of the demographic elements and finally multiple regressions to bring together all of the data. mummies The following tables tell the story of the sample. Universe estimates come from Arbitron Ratings’ population estimates for the three markets, at the time of the survey. 30 Men Women White Black Hispanic Other Refused Less than H.S. High School Some College College Degree Graduate Work Refused 31 TABLE 1 Demographics Gender of Sample N Pct Universe Estimate 433 47.9% 48.7% 471 52.1 51.3 Race of Sample N Pct Universe Estimate 820 90.7% 93.4% 52 5.8 6.0 2 0.2 1.6 19 2.1 N / A l l 1.2 N/ A Educational Level of Sample N Pct Universe Estimate‘ 61 6.7% 25.1% 262 29.0 34.0 276 30.5 20.4 188 20.8 20.4” 1 14 12.6 3 0.3 ’Universe estimate for education based on population 25 and over. The sample is based on 18 and over. ”This percentage is for all college graduates (including graduate work). Under $10,000 310,000-820,000 820,000-830,000 530,000-540,000 840,000-350,000 >850,000 Refused Household Income Adjusted N Pct Pct Universe Estimate 78 8.6% 10.2% 16.6% 156 17.3 20.4 20.8 203 22.5 26.5 17.2 161 17.8 21.0 15.5 83 9.2 10.8 10.5 85 9 4 11.1 18.5 138 15:3 N/A N/A 32 TABLE I (CONT’D.) Age of Respondents N Pct Universe Estimates 18-24 196 21.7% 16.3% 25-34 253 28.0 25.9 35-44 203 22.5 - 19.9 45-54 86 9.5 12.5 55-64 84 9.3 1 1.4 65+ 55 6.1 14.0 Ref used/ Unclear 27 3.0 N / A The biggest skew from reality is in education, where the sample shows a much higher percentage of college educated individuals than the universe estimates. In terms of age, the sample is skewed to the younger ages and the income levels show a greater percentage of middle income households, which may be due to refusals. Gender and race are not far off from universe estimates with the exception of Hispanics. Only two were interviewed, or less than one quarter of one percent of the sample versus over one and a half percent in the population. Despite the skews, no weighting will be used for the parts of the sample that are off from the universe. Instead, the reader is advised to be aware of the potential for bias in some circumstances. The method of presentation of results will be to build one layer of results upon the previous ones. Each facet of radio listening that was investigated here will be presented individually followed by correlations with previously presented data, one way analyses of variance and t-tests as appropriate with each of the demographic elements and finally multiple regressions to bring together all of the data. Because non-white respondents represent only eight percent of the sample (52 respondents), race will not be used in the subsequent analyses. 33 Ti e Spent Listening In commercial radio, time spent listening (TSL) is a measure commonly used by programmers to determine the staying power of their stations with audiences. Usually, these people are concerned with how long the average listener is spending with their station. In this study, TSL is a measure of how much time each listener spends with radio in general. The measure is valuable as heavy listeners may have different expectations of radio and different uses for the medium than light listeners. Time spent listening was calculated by combining the results of two questions. First, respondents were asked how many hours they listened to radio on an average weekday. The same question was repeated for weekends. Those measures were multiplied by five and two days respectively, and summed to represent TSL in an average week. For those few respondents (less than ten) who reported that they did not work a normal work week (e.g. Wednesday through Sunday), interviewers were instructed to have them adjust the definitions of weekday and weekend to their work week. The reader should keep in mind that TSL means presented here are somewhat higher than those given by other sources, such as the commercial ratings services, Arbitron, Birch, and RADAR. This difference is accounted for by the screener question at the start of the survey. Only respondents who listened to radio at least two hours a week were surveyed. This would be expected to raise the overall average somewhat when compared to other measures. 34 TABLE 2 Time Spent Listening (in hours per week) Overall Mean 29.1 Hours/Week Age Group TSL 18-24 32.3 F=2.86 25-34 29.9 df-762 35-44 24.6 D=-014 45-54 26.4 55-64 31.7 65+ 29.0 Educational Level Less than H.S. 36.5 F=6.63 High School Graduate 33.5 df-896 Some College 28.1 p=.001 College Graduate 24.5 Post Graduate Work 25.6 Income Under $10,000 33.9 F=2.06 810,000-820,000 31.8 df =7 62 820,001 -830,000 26.4 p=n.s. 830,001-840,000 28.7 840,001-850,000 27.5 More than 850,000 26.1 Gender Male 26.9 t=2.77 Female 31.1 p=.006 Market Size Large 27.9 F-l.50 Medium 28.4 df=899 Small 31.0 p=n.s. The mean amount of listening time by respondents in an average week was 29.1 hours or just over four hours a day with a standard deviation of 23.2 hours. Three respondents who reported listening 168 hours per week (24 hours 35 per day) were coded as missing data for TSL. Following the lead of the ratings services which regularly delete respondents who report listening levels that are considered "too large," these "continuous" listeners were eliminated. It is unlikely that anyone can "listen" 24 hours per day, whether or not they choose to sleep at some time. Bivariate analyses were performed to assess the relationship between TSL and demographics. Every demo except income and market size showed statistically signficant results. For age, the overall ANOVA yielded an F of 2.86 (p=.014), indicating a significant relationship between age and TSL. Scheffe comparisons were used to pinpoint significant differences between categories. Scheffe post hoc comparisons were run on all ANOVAs in this dissertation. Scheffe offers the advantages of a more conservative test and eliminates the use of a large number of t-tests, which would have certainly resulted in some significant results regardless of the true outcomes. The 35-44 group listened to radio the least, only 24.6 hours per week. On the other end, the 18-24 cell listened over 32 hours per week and the 55-64 group spent approximately 31 2/3 hours per week listening to radio. A Scheffe comparison showed a significant difference between the 18-24 year old group and the 35-44 group. No other significant differences between age groups were found. Education had a significant relationship to TSL (F=6.63, p=.001). The less than high school education group spent over 36.5 hours per week with radio, while the college degree group listened fewer than 24.5. The post-graduate work cell listened only 25.6 hours per week. Scheffe comparisons showed significant differences at the .05 level between the college degree cell and both the less than high school and high school degree groups. In this sample, those with less education spend more time listening to radio. 36 Gender showed significant differences in listening levels, based on the t-test statistic. There was a significant difference betweeen men and women (ta2.77, p=.01). In this sample, women spent over four hours a week more with radio than men did. Women listened an average of 31.1 hours. Men spent fewer than 27 hours with the medium. Statign Repertgirg The number of stations each respondent listened to was determined by two questions. First, each respondent was asked what stations they listened to on a regular basis. Next, each was asked to name if there were any other stations they listened to once in a while. The sum of the two figures is a number that can be referred to as station repertoire. The number from the first question is also useful on its own, as a set of regularly listened to stations. Information for both will be presented in this section. TABLE 3 Station Repertoire Number of Stations Regular Stations 2.0 Other Stations 1.0 Total Repertoire 3.0 Distribution Number of Stations Frequency Percentage l 141 15.6% 2 252 27.9 3 243 26.9 4 128 14.2 5 67 7.4 6 38 4.2 7 20 2.2 More than 7 14 1.5 Don’t Know 2 0.2 37 TABLE 3 (Com) Age Group 18-24 3.3 F=4.80 25-34 3.2 df=874 35-44 3.0 p=.001 45-54 2.7 55-64 2.7 65+ 2.3 Education Less than High School 2.5 F=3.55 High School Graduate 2.8 df=898 Some College 3.3 p=.007 College Graduate 3.0 Post Graduate Work 3.2 Income Less than 810,000 2.8 F=l.l6 810,000-820,000 3.3 df =7 63 820,001-830,000 3.0 p=n.s. 830,001-840,000 3.0 840,001-850,000 3.1 More than 850,000 3.3 Gender Male 3.2 t=2.90 Female 2.9 p=.004 Market Size Large 3.3 F-6.l 1 Medium 3.0 df=897 Small 2.8 p=.002 In this study, the average listener used three stations, two of which were listened to on a regular basis. There were no correlations above .01 between how many stations respondents listened to and how much time they spent with radio. Education, age, market size, and gender showed differences between groups for station repertoire. Greater education meant more stations (ANOVA, F=3.55, p=.01) and younger listeners tended to listen to more stations (ANOVA, F=4.80, 38 p=.001), although individual comparisons were not significant. For gender, the t of 2.90 was significant at a probability level below .01. Male respondents listened to more stations than female respondents. Market size results confirmed what would seem logical to most readers and has been shown in nearly every syndicated ratings report: given more options, listeners will listen to more stations. In this case, not only is the overall station repertoire ANOVA significant (F=6.37, p=.002), but the Scheffe comparisons show the large market respondents listened to more stations than either the medium market respondents or the small market respondents. Finally, age was statistically significant (F=4.80, p=.001). Scheffe comparisons showed differences at the .05 proability level between the 18-24 cell and the 65+ cell and between the 25-34 cell and the 65+ group. In this case, it appears that younger listeners will shift around between more stations than older listeners. RQl; Initial Selectign 9f Repgrtgire The study examined the importance of radio format elements in selecting an initial repertoire and the hierarchical nature of these attributes. Importance of Format Elements Respondents were questioned as to how important fifteen different elements of radio formats were to them. A one to seven scale was used with seven meaning "extremely important" and one equalling ”not important at all." While some format element may have been left out, the list was an attempt to be exhaustive (see table 4). 39 TABLE 4 Importance Scores of Radio Elements Mean Reception 6.5 Music 6.3 Sound Quality 6.3 Stereo Sound 5.6 News 5.2 Weather 5.2 PM 5.2 Limited Commercials 5.1 Community Involvement 4.9 Disc Jockies/Personalities 4.7 Traffic Reports 3.6 Sports Reports 3.4 Contests 3.2 Phone-In Talk Shows 3.1 Play by Play Sports 2.8 Reception, music, and sound quality were the most valued elements among the respondents. Stereo sound placed in between other elements. A group of secondary importance consisted of news, weather, FM, limited commercials, community involvement, and disc jockies. Finally, traffic reports, sports reports, contests, phone-in talk shows, and play by play sports were not considered to be as important by this group. Graphically, the differences become more clear, as shown in figure 2. 7 If any one independent variable shows major differences in the importance of the dependent variables, it is gender. Of the fifteen t-tests run, fourteen showed statistically significant differences at the .05 level of probability or better. Limited commercials was the only format element that was not different for men and women. The results for each element are shown in table 5. 40 Figure 2 PRISM ELBAENTS PROGRAM ELEMENT SCORES PLAY BY PLAY MEAN seem TALK SONS O 1 2 3 4 5 6 7 7=VERY INPOHTANT IN IDEAL RADIO STATION 41 TABLE 5 Element Importance Scores by Gender Element Male Female t p Sports Reports 4.1 2.7 10.93 .001 Play by Play Sports 3.2 2.4 5.84 .001 Weather 4.8 5.5 5.36 .001 Music 6.2 6.5 4.28 .001 Phone-in Talk Shows 2.9 3.4 3.57 .001 Disc Jockies/Personalities 4.5 4.9 3.05 .002 Contests 3.0 3.4 2.79 .005 Stereo Sound 5.7 5.4 2.66 .008 News 5.1 5.4 2.65 .008 FM 5.3 5.0 2.53 .012 Reception 6.4 6.5 2.47 .038 Community Involvement 4.8 5.0 1.99 .047 Traffic Reports 3.5 3.8 1.97 .050 Sound Quality 6.2 6.4 1.96 .050 Limited Commercials 5.2 5.0 1.43 n.s. Of the significant results, women rated ten of the elements as more important. The men in the sample rated only four elements higher, two of which, sports reports and play by play, could have easily been hypothesized in advance of the study. The other two are sound quality related, FM and stereo; however, the women in the sample rated both reception and sound quality higher than the men. Another independent variable that was analyzed was time spent listening, described earlier in this section. Correlations were run between TSL and each of the importance variables. Table 6 shows the matrix. 42 TABLE 6 Correlations Between Time Spent Listening and Importance 1' D Contests .19 .001 Phone-in Talk Shows .15 .001 Disc Jockies/Personalities .12 .001 Community Involvement .09 .01 Stereo .08 n.s. Music .07 n.s. Play by Play Sports .06 n.s. Sound Quality .06 n.s. Weather .04 n.s. Limited Commercials .04 n.s. Traffic Reports .03 n.s. Sports Reports .02 n.s. FM .00 n.s. News -.01 n.s. Reception -.02 n.s. It appears that light radio listeners are a somewhat different group than heavy listeners. The differences in importance scores could be interpreted as differences of involvement, that is, for light listeners, not much of anything is that important to them with radio being similar to toothpaste. For heavy listeners, by virtue of the amount of time they spend with the medium on a weekly basis, everything is more important. Another comparison was made with the independent variable of how many stations a listener used. Again, Pearson product-moment correlation was used and in this case, not a single correlation was statistically significant. Thus, the number of stations in a listener’s repertoire has no relationship to the importance of the various format elements. 43 TABLE 7 Element Importance Scores by Income Element <810K$10K- 820K- $30K- 840K- 850K+ F df p 820K 830K 840K 850K Contests 3.5 3.7 3.2 3.1 2.7 2.6 4.22 765 .001 Community Involve 4.7 5.3 4.7 5.0 4.4 4.7 3.75 765 .003 Phone-in Talk 3.4 3.2 2.9 3.4 2.8 2.6 2.89 764 .014 Disc Jockies 5.1 5.0 4.7 4.6 4.6 4.3 2.55 765 .027 Ltd. Commercials 5.1 4.9 5.1 5.4 5.0 5.4 - 1.73 765 n.s. News 5.3 5.2 5.1 5.2 5.2 5.6 1.40 765 ms Reception 6.4 6.5 6.5 6.6 6.2 6.5 1.36 765 n.s. Stereo Sound 5.5 5.8 5.7 5.4 5.3 5.4 1.27 764 n.s. Sports Reports 3.4 3.2 3.4 3.5 3.6 3.8 1.26 765 n.s. FM 5.1 5.4 5.2 5.1 4.8 4.9 1.23 765 n.s. Music 6.5 6.3 6.3 6.4 6.1 6.3 1.01 765 n.s. Play by Play 2.9 2.6 2.8 2.6 2.8 3.0 0.86 764 n.s. Weather 5.2 5.3 5.0 5.1 5.2 5.3 0.49 765 n.s. Traffic Reports 3.8 3.8 3.6 3.5 3.5 3.5 0.48 758 n.s. Sound Quality 6.3 6.3 6.3 6.3 6.3 6.2 0.31 765 n.s. Income level proved to have little explanatory power regarding importance of format elements. Four of the fifteen ANOVAs showed statistically significant results, but two of these, disc jockies (F=2.55, p=.027) and talk shows (F=2.89, p=.014) showed no significant differences in the Scheffe comparisons. On the other hand, contests and community involvement did yield some significant comparisons. Contests (F-4.22, p-.001) showed a significant difference between the 810,000-820,000 cell and both the 840,000-850,000 cell and the 850,000+ cell. A likely explanation is that someone making less than 820,000 a year needs whatever is being given away more than someone making over 840,000 a year, but this does not explain the lack of a significant difference between the upper income groups and the under 810,000 category. Community involvement (F=3.75, p=.003) showed one statistically significant comparison. The 810,000-820,000 cell rated this element significantly more important than did the 840,000-850,000 cell. 44 TABLE 8 Element Importance Scores by Education Level Element $50K 45 16 14 (7.0) (2.5) (2.2) X3-4.8 df-10 p=n.s. Number of Stations 1 Station 2 Stations 3 Stations 4+ Stations 97 TABLE 48 Chi-Squares for Change Strategies in the Car (Percentages in Parentheses) By Station Repertoire Know Ahead of Time 36 (4.8) 109 (14.7) 131 (17.6) 1 16 (15.6) X’s-18.9 df=6 par-.004 Gender Male Female Know Ahead of Time 212 (28.4) 181 (24.3) X2866 df-2 p=.037 Age Group 18-24 25-34 35-44 45-54 55-64 65+ Know Ahead of Time 77 (10.6) 108 (14.9) 98 (13.5) 42 (5.8) 36 (5.0) 20 (2.8) X3-15.5 df=10 p=n.s. Listen to First One 27 (3.6) 44 (5.9) By Gender Listen to First One 74 (9.9) 99 (13.3) By Age Listen to First One 45 (6.2) 58 (8.0) 38 (5.2) 7 (1.0) 14 (1.9) 7 (1.0) Check Before Choosing 11 (1.5) 48 (6.5) 47 (6.3) 74 (9.9) Check Before Choosing 97 (13.0) 83 (11.1) Check Before Choosing 51 (7.0) 55 (7.6) 37 (5.1) 13 (1.8) 14 (1.9) 6 (0.8) Hierarchy Know Ahead Group Of Time 40 or less 45 (6.0) 41-42 7 (0.9) 43-44 21 (2.8) 45-46 16 (2.1) 47-48 29 (3.9) 49-50 27 (3.6) 51-52 56 (9.4) 53-54 48 (6.4) 55 144 (19.3) X3-2.5 df=16 p-n.s. Income Know Ahead Group of Time <310K 25 (3.9) $10K-$20K 59 (9.2) $20K-$30K 92 (14.4) $30K-$40K 75 (l 1.7) $40K-$50K 36 (5.6) >$5OK 47 (7 .4) X3-11.0 df=10 p=n.s. 98 TABLE 48 (Cont’d.) By Hierarchy Score Listen to First One 19 (2.5) 2 (0.3) 14 (1.9) 10 (1.3) 9 (1.2) 11 (1.5) 9 (1.2) 31 By Income Listen to First One 16 (2.5) Check Before Choosing 21 (2.8) 4 (0.5) 12 (1.6) 2 (0.3) 11 (1.5) 14 (1.9) 25 Check Before Choosing 18 (2.8) 33 (5.2) 38 (5.9) 33 (5.2) 15 (2.3) 18 (2.8) 99 TABLE 48 (Cont’d.) By Education Educational Know Ahead Listen to Check Before Level of Time First One Choosing Less than HS. 27 10 9 (3.6) (1.3) (1.2) High School 102 54 52 (13.7) (7.3) (7.0) Some College 120 55 63 (16.2) (7.4) (8.5) College Grad 90 38 30 (12.1) (5.1) (4.0) Post Grad 54 15 24 (7.3) (2.0) (3.2) X3-7.8 df =8 p=n.s. By Time Spent Listening Hours Know Ahead Listen to Check Before of Time First One Choosing 10 or less 78 35 35 (10.5) (4.7) (4.7) 11-20 107 54 48 (14.4) (7.3) (6.5) 21-30 78 29 37 (10.5) (3.9) (5.0) 31-40 43 20 21 (5.8) (2.7) (2.8) 41+ 85 34 38 (1.1.5) (4.6) (5.1) X3=1.9 df=8 p-n.s. Chapter 4 Conclusions This final chapter will consider what has been found and how the results of this study can be used. In reality, the present research serves two masters: the research and academic community in the attempt to model radio choice procedures by individuals, and the radio industry in the attempt to find out more information to help stations more successfully serve their audiences. For that reason, this chapter will be divided into two segments: implications for research that will deal with the theoretical aspects of the study and implications for broadcasters for programming and strategic purposes. Im 1i i n r Br r Theoretically, the broadcaster has one primary goal: maximizing profit within the parameters of his/her license. Deregulation aside, the broadcaster continues to be "controlled” by an agency of the federal government, specifically in technical areas (power, antenna height, frequency, etc.) and in other realms including commercials (sponsor identification rules, lotteries) and equal employment (FCC equal employment rules). The broadcaster also does not have the ability to move to another city in order to improve his/her market position. For example, broadcasters cannot move their stations to the Sunbelt in order to take advantage of lower wage rates or growing economies. The broadcaster cannot "outsource" his/her station overseas to cut costs (although satellite delivered programming may be considered a similar action), and though he/she may sell the station and buy another in a more desirable location, the station and the license remain in the same location. The station licensed to 100 101 Buffalo remains in Buffalo, ownership notwithstanding. This points to the proposition that broadcasters attempt to maximize profit with some limitations. Pouring some money and effort back into the community or producing programs that serve limited audiences but generate goodwill are not part of a short term profit maximization strategy, but may help in the long-run.