h1 . '3; B: :td‘L F 13313313 ‘31 M 3:33 3 33' 3333 3 3; 3 u 111313 3311111 313 .3513“ 3'. 3 1 5 33333333 3.13333 1333133333 33333333 2333:3333 3'3 . : buzz.“ ' '33 1 .3 31:3: 33333'M.?1 13 1111333353433 ;.:i‘. .‘h‘ ’1 .‘j “§:.‘ at“ i? t“t‘l!" . .. E; :i:.::i' .' ‘3 1 1 1 '33 1.3313301" v 3333: 3 31313313333131.11 3331 .3313 33333 331113333 1.3. 1.1, 33 “3 3333! 3.3333 3 3 33313 .3 3333331 133313 333333 33 3" 3m 3 3 3333: 1313113111 . 1 ‘;=‘- "'3 13“”; 113.41 | . I}: V ‘ ' o .i‘ 5”“ .""’__'.:.: 0‘ ‘ 1.]‘-..=. “a '33,? '. ' W- n ' ‘ o _"‘ .— a O 1 :O— H ._ ,7“- ‘ O a. n“ .0 m“ 3 111.1» 1, 33 313 1:111} _—-—- fl ‘ ‘ - C 2' .. u": u- m, C' ' ' ’ 1 w . ‘ "o ' ' “f. 1' _' t ‘35 ;. gar.» 9:11}. Ex ”'3‘ “V I (- ‘ T v :3. S" ‘.fi.e.t.?' “vu-:; :5 2. ~ . II ..V uL h.» .‘J ..- J ‘y ‘ g“. ’5 ,. .E. L--- c 7?? "s?".“"‘ ‘1'.- . .'(X 4‘ an“ ‘i~ «Swab w; it v ‘1 4 . . 'r .\. w Fro This is to certify that the dissertation entitled A MODEL OF ENTRY INTO SELECT U.S. RADIO MARKETS BETWEEN 1973 AND 1978 presented by Robert Earl Yadon has been accepted towards fulfillment of the requirements for Ph 0 D 0 degree in MaSS Media Major professor Dme January 26, 1983 MSU is an Affirmative Action/Equal Opportunity Institution 0— 12771 MSU LlBRARlES v RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. ‘1— irrvlffclq ID a: /‘%’;.’f§’.. 91,35 y A MODEL OF ENTRY INTO SELECT U.S. RADIO MARKETS BETWEEN 1973 AND 1978 BY Robert Earl Yadon A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY College of Communication Arts and Sciences Mass Media 1983 ABSTRACT A MODEL OF ENTRY INTO SELECT U.S. RADIO MARKETS BETWEEN 1973 AND 1978 BY Robert Earl Yadon The Federal Communications Commission has held for over 40 years that competition between radio stations is desir- able, and that the public interest is best served by com- peting broadcast stations. Over the same period, however, the FCC allowed existing stations to protest, and thereby delay, the entrance of new stations into a market based on a claim of "economic injury." It wasn't until 1964 that the Commission adopted formal pleadings requirements and developed criteria for the eval- uation of the economic impact of new station entry on an existing station. Unfortunately, the Commission failed to indicate the precise standards by which a determination of economic injury might be made, leaving it up to the courts to decide when sufficient "potential revenue" existed to warrant the addition of a new station without detriment to the public interest. This study addresses some of the station and market characteristics that are determinants of new station entry, and recognizes the need for management, Robert Earl Yadon entrepreneurs and students alike to accurately predict growth opportunities where they exist. Select market and station variables were measured over a six-year period, 1973-1978, in 218 radio markets across the United States. Relationships between market and station variables were analyzed, covariations noted, and a factor analysis employed to eliminate as much collinearity as possible. The reduced variable list was then used in both static and dynamic regression models to predict the number of radio stations a market was capable of holding. A new list of dynamic variables, those dealing with the relation- ships between variables or among variables over time, were derived from the original variable set, and were employed in a discriminant analysis in an attempt to predict the point (year) of entry. The results of this study demonstrate and confirm the strong relationship between station and market variables. This study also disclosed that most variations between these variables were coincident over time, and therefore impractical for use in time series applications. Use of dynamic variables, however, proved more satisfactory and demonstrated that entry was dependent upon the degree of change in market and station variables during preceding years. Copyright by ROBERT EARL YADON 1983 ACKNOWLEDGMENTS The author is indebted to the many people who made a significant contribution to this study. First, special thanks is expressed to Dr. Thomas Baldwin, chairman of my doctoral committee, for his guidance, encouragement, and constructive criticism. This writer also extends special thanks to Dr. John Abel, Dr. Martin Block, and Dr. Donald Taylor, for their valuable assistance throughout this study. Special thanks goes to Dr. Robert Schlater, former chairman of the Department of Telecommunication, for his understanding, patience, and encouragement throughout this study. His continued support was a major factor in the completion of this research. In addition, the writer also extends his appreciation to the National Association of Broadcasters (NAB) for their financial support. The findings and conclusions contained ‘herein, however, are those of the author alone and are not necessarily endorsed by the NAB. Finally, special appreciation is expressed to my wife, Linda, for her understanding, encouragement and patience during the completion of this study. iii TABLE OF CONTENTS LIST OF TABLES . . . . . . . . . . . . . . KEY TO SYMBOLS . . . . . . . . . . . . . . CHAPTER I. INTRODUCTION . . . . . . . . . . . Scope of the Study . . . . . . . Limitations . . . . . . . . . . . II. REVIEW OF LITERATURE . . . . . . . I I I O METHODOLOGY 0 O O O O O O O O O O O The Variables . . . . . . . . . . Specification of Variables . . . Hypotheses . . . . . . . . . . . Analysis of Data . . . . . . . . IV 0 RESULTS 0 O O O O C O O O O O O O 0 Introduction of Market Indices . Individual Market Index . . . Normality Index . . . . . . . . Potential Revenue . . . . . . . Radio-Dollar Index . . . . . . Relationship Between Station and Market Characteristics . . . . Reduction of the R—Matrix . . . Regression Coefficients . . . . Lagged Regression Model . . . . Entry as a Dependent Variable . . Discriminating Variables . . . Discriminant Analysis . . . . . iv Page vi viii 10 12 16 36 36 37 47 50 57 57 58 60 65 65 69 73 76 80 81 85 TABLE OF CONTENTS (continued) CHAPTER V. SUMMARY, CONCLUSIONS, AND Summary . . . . . . . . Conclusions . . . . . . Recommendations . . . . APPENDIX A . . . . . . . . . . . APPENDIX B O O O O O O O O O O 0 APPENDIX C O O O O O O O O O O O SELECTED BIBLIOGRAPHY . . . . . RECOMMENDATIONS Page 92 92 101 112 121 123 128 129 Table II. III. IV. VI. VII. VIII. IX. XI. XII. XIII. XIV. LIST OF TABLES Correlation of New AM Station Entry and Per Station Income in Communities Over 50,000 P0pu1ation Between 1939 and 1950 . Correlates of Financial Behavior of AM Radio Stations in 1964 . . . . . . . . Correlates of the Financial Behavior of Select Oklahoma AM Radio Stations in 1973 Total Revenues by Station Type . . . . . . Correlates of the Financial Behavior of Sampled Radio Markets in 1971 . . . . . . Correlates of the Financial Behavior of Sampled Radio Markets in 1973 . . . . . . One-Way Analysis of Variance Between IMI and Market ID . . . . . . . . . . . . . . One-Way Analysis of Variance Between IMI and Year 0 O 0 O O O O O I C O O O O O 0 Mean IMI and Range of Normality, by Year . Analysis of Variance Between NINDX and ENTRY O O O O O O O O O O O O C O O O O 0 Correlation Matrix of Station and Market Characteristics 1973 to 1978 . . . . . . Relationship Between Distance (DIST) and Select Market Characteristics . . . . . . Rotated Factor Matrix . . . . . . . . . . . Correlation Sub-Matrix for Factor One . . . vi Page 17 20 26 29 31 31 6O 61 62 63 66 68 7O 71 Table XV. XVI. XVII. XVIII. XIX. XX. XXI. XXII. XXIII. XXIV. XXV. XXVI. XXVII. Page Reduced R-Matrix After Factor Analysis . . . . Analysis of Variance of Regression Model Predicting the Dependent NUMSTA78 . . . . . Regression Coefficients of Market Variables Predicting the Dependent NUMSTA78 . . . . . Relationships Between NUMSTA78 and Lagged Market Characteristics . . . . . . . . . . . Analysis of Variance of Lagged Regression Model Predicting the Dependent NUMSTA78 . . Regression Coefficients of Lagged Market Variables Predicting the Dependent NUMSTA78. Comparison Between Lagged and Unlagged Models. Nonparametric Correlation Coefficients for Entry in Combined and Lagged Models. . . . . Stepwise Discriminant Selection Procedure . . Canonical Discriminant Functions . . . . . . . Canonical Discriminant Function Coefficients . Classification Results . . . . . . . . . . . . Classification Function Coefficients for ENTRY78 O O O O O O O O O O I O O O O O O 0 vii 72 74 75 77 78 79 80 82 86 88 88 89 90 AMREV AVGPOP AVGREV BPI COMPIN DIST DIVER ENTRY FMREV ID IMI NINDX NUMSTA PERCHG POTREV PRFREV RANK KEY TO SYMBOLS Total revenue for AM and AM/FM combination stations in the market Average total radio revenue per capita (see Equation (9)) Average total radio revenue per market Buying power index Competition index Distance between test market and next closest market of equal or greater population Diversity index Entry of a new station into a test market Total revenue for independent FM stations in the market Unique market identification code Individual market index (see Equation (2)) Normality index Total number of radio stations in a test market Percentage change in total radio revenue from one year to the next (see Equation (7)) Potential radio revenue in a test market Percentage of total radio revenue captured by FM stations in a test market (see Equation (12)) Population rank of test market viii RDINDX REVPOP SMPOP STAPOP STATRS TCOMP TOTREV TRS Radio-dollar index Total radio revenue per capita (see Equation (8)) SMSA population of test market Number of radio stations per 100,000 population (see Equation (10)) Number of radio stations per $1,000,000 in total retail sales for test market (see Equation (11)) Total number of competing media units in the test market (radio, television, newspaper) Total radio revenue in test market Total retail sales in test market ix CHAPTER I INTRODUCTION In the broadcast industry, as in many industries, the proper evaluation of market conditions is necessary for future economic growth. This dissertation reports the results of a study into the growth of radio broadcasting between the years of 1973 and 1978, and attempts to identify those market variables which, when combined, constitute favorable conditions for new station entry. As this analysis demonstrates, economic success in the marketplace may be increasingly dependent upon positive action based on the proper evaluation of select market conditions. Further, the correct analysis of market con- ditions is likely to have the additional benefit of promoting the efficient utilization of limited spectrum space, and maximization of local service to each respective community. This may especially be true with expansion of services to markets, where existing stations may provide a barrier to entry, and thus new competition, by raising the question of economic injury. While the Federal Communications Commission (FCC) reviewed a proposal to reduce the standard AM channel width from ten kilohertz to nine kilohertz in order to allow 1 additional stations into the AM band,1 ironically it was another FCC decision in September 1969 that changed the status of FM radio. The 1969 proposal called for nearly all future radio expansion to be put into FM. Major reasons cited for the change included the fact that FM was a full-time service whereas AM was in- creasingly daytime-only service (especially for most new stations), and that FM stations could still be assigned without the interference or problems which plagued the addition of almost any AM outlet. Regardless of whether future radio expansion is in the AM or FM band, two factors dominate the case for economic analysis. Initially, each station owner is concerned with the overall performance of his or her station as weighed against competition and market conditions. Second, for the existing station owner, the possibility of additional competition through new station entry compounds all factors related to the existing competitive equilibrium,and calls for increasing levels of expertise to maintain a profitable market share. ~Likewise, any outside individual interested in constructing a new station in the market must consider these same performance characteristics. The prOper evaluation of current market conditions is necessary to determine the validity of "economic injury." A claim of "economic injury" by the existing station owner makes it possible for him or her to protest, and thereby delay, the entrance of a future station on the basis that the market will not economically support the additional station, and that entrance of the new station will only lead to the overall reduction of public service to the community. Action by the FCC in the matter of economic injury goes back to the 1930's, when "the Commission regularly took into account the nature and extent of economic injury which would be caused to existing stations by the grant of a new applicant."3 In fact, the position of the FCC at that point in time was strongly in favor of competition, and it proclaimed on more than one occasion, "Where the economic situation of a community permits, the public interest is best served by competing broadcast facilities."4 Where the Commission clearly indicated film; preference for competition, it failed to indicate the standards by which determinations of economic injury were made. Even under similar market conditions, uniform court decisions did not always result.5 In 1938, the Commission denied an application for a construction permit for a second station in Fall River, Massachusetts, on the grounds that no showing of need and market advertising support were apparent, even though the existing station was enjoying some degree of financial success.6 It is also the only case where the FCC refused to grant a construction permit based on economic injury. The first major court test of the issue of economic injury, and precedent for future cases, is based on the Supreme Court's 1940 landmark decision in FCC v. Sanders' Brothers Radio Station.7 Here the court held that the Communication Act was neither intended nor designed to protect licensees against competition. The court noted: ...resu1ting economic injury to a rival station is not, in and of itself, ...an element the partitioner must weigh, and as to which it must make findings, in passing on an application for a broadcast license.8 At the same time, the court added that competition should not be disregarded as it could cause not only financial hardship to the existing complaining station, but also an overall reduction of "public service" to the community. Expanding upon this position, the U.S. Court of Appeals (D.C.) held in the 1958 case of Carroll Broadcasting Co. v. 9 FCC, that economic injury to an existing station is not in and of itself a matter of public concern; however, it becomes vital when it spells diminution or destruction of overall service. The court stated: When an existing licensee offers to prove that the economic effects of another station would be detrimental to the public interest, the Commission should afford an opportunity for presentation of such proof and, if the evidence is substantial (i.e., if the protestant does not fail entirely to meet his burden), should make a finding or findings.10 Where Carroll established the relevance of economic injury to the public interest, it did not determine how detailed, or what specific evidence the protesting licensee would need to offer as "proof" to entitle him or her to a hearing. The initial Commission response was to set for hearing any case where the protestants alleged economic injury, and offered to prove that the public interest would be adversely affected. The Commission established no plead- ing guidelines but gave the protestant the benefit of the doubt. Due to the ease with which these hearings were then granted, this policy provided an inefficient use of the Commission's time, and energy. As one might expect, the FCC's policy opened the floodgates and nearly every case was set for hearing. Congress, in 1960, amended the Communication Act of 1934 to create the "petition to deny." The statute11 provides a petition to deny must: ...contain specific allegation of fact sufficient to show ... that a grant of application would be prima facie inconsistent with (the public interest, convenience and necessity)... The statute also states that if: ...there are no substantial and material questions of fact and a grant of the application would be consistent with the public interest the petition shall be denied and the grant made without hearing. 13 The 1960 amendments were viewed as a congressional response to the lax methods by which the Commission granted hearings. Thus, Congress provided that a "petition to deny" filed under Section 309(d) must make a "substantially stronger showing of greater probative value than (was previously) necessary."14 The Commission finally reacted to the 1960 amendments in late 1962 when it began requiring petitioners to plead a prima facie case of injury to the public "before" it would set a hearing.15 In implementing this policy, how- ever, the Commission failed to specify, in reasonably precise terms, the type of information necessary to entitle a petitioner to an evidentiary hearing on the Carroll issue. Problems surrounding the vague nature of the Commis- sion's policy were first encountered by the courts in KGMO Radio-Television, Inc., v. FCC,16 in which the court reversed a Commission order denying a Carroll hearing because the petitioner failed to plead sufficient factual data to make out a prima facie case. While the court noted that it was within the Commission's authority to require more informa- tion, it held that since the petitioners had "no notice... that more would be required, the petition should not have been denied on the ground that more was not furnished."17 On remand of KGMO, the Commission established precise pleadings requirements calling for analysis of existing station's economic loss based on specific criteria, and as set forth in Missouri-Illinois Broadcasting Company.18 See Appendix A. Rigid application of the Commission's criteria for pleadings, as established in Missouri-Illinois, apparently went too far and led to a near extinction of the right of a petitioner to a hearing on the Carroll issue. During this period, the U.S. Court of Appeals (D.C. Cir.) held that the Commission abused its discretion in denying virtually every petition for a hearing, which made judicial inter- vention necessary in Southwestern Operating Co. v. FCC.19 Here the Court ruled that although the Commission can insist on more than general and conclusory allegations, it may not use its pleadings requirements as a means to justify complete disregard of evidence before it. The Court, in Southwestern, did not intend to overrule the Commission's pleadings rules. The Court recognized that since competition is, in fact, a favored element in broad- casting, then: ...the temptation to an existing licensee to postpone as long as possible the advent of competition warrants special care by the Commis- sion in the scrutiny of requests for hearing in Carroll circumstances. The effect of the Southwestern decision was that it opened decisions to the particular circumstances of each individual case, and it placed a burden upon the petitioner to plead specific factual data sufficient to establish a prima facie case of injury to the public interest in order to meet the requirements for a hearing. Another limitation on the Commission's pleadings requirements came out of the court's decision in Folkways 21 Broadcasting Co. v. FCC, where the petitioner's request for a hearing was denied, in part, because he failed to allege the particular advertisers that would be lost and the particular public service programming that would have to be taken off the air as a result of increased competi- tion. The court in Folkways held: ...a Carroll hearing may not be limited to a case in which preknowledge of the exact economics of the situation is necessarily available. Requiring such precision would eliminate the doctrine as a practical matter.22 Thus, the Commission modified its Missouri—Illinois requirements to eliminate the need for "preknowledge," but at the same time continued to insist upon specific factual data rather than generalized and conclusory allegations of public injury. In early 1972, the court reaffirmed the Commission's requirements as set in Missouri-Illinois, when it ruled in WLVA v. FCC.23 The court stated: Specifically, the petitioner must raise substantial and material questions of fact as to whether (1) the revenue potential of the market is such that a grant will cause the petitioner to suffer a significant loss of income; (2) the effect of this loss will be to compel the petitioner to eliminate some or all of its public service programming; and (3) this loss of programming will not be offset by the increased non-network programming proposed to be offered by the applicant. Since these three aspects of the Carroll issue are essentially inter- dependent, a failure to satisfy any one of the three is likely to be dispositive of the petitioner's claim to a hearing. While the WLVA case supported the Commission's require- ments as outlined in Missouri-Illinois, it also added the concept of "future economic injury" to the list. The court said: For even in the typical Carroll situation, the Commission must consider not only existing, but also future competitive conditions. What may appear in the short run to be a substantial finan- cial loss may often pale to insignificance when the overall revenue potential of the market is taken into account.2 Thus, the court required the petitioner to demonstrate sound reasoning to indicate that the market would not improve its economic standing to the point where the second station "might" become viable in the near future. Finally, in 1974, another restriction was placed on the pleadings requirements, this time indicating that the petitioner must not only provide necessary data, but also establish the accuracy of that data. The Commission, in 26 Lenawee Broadcasting Co., found: Lenawee's estimate of $581,000 requirement in advertising revenue to support this station has not been challenged. Nevertheless, to success- fully raise the Carroll issue, the petitioner must establish some basis for determining the accuracy of its projected figure of $550,000 in anticipated market revenue potential.27 This would seem to indicate that not only must the infor- mation presented be complete, but it also suggests that projections made by the petitioner with regard to "potential" revenue in the marketplace must be verified in some manner if they are to be a part of the petition. In summary, a petitioner must provide the basic statistical evidence originally set forth in Missouri- Illinois, but later modified and restated in 3325 and Folkways, to exclude the conditions of "preknowledge" and "exact calculations." This data must be linked to specific arguments concerning public service programming, while projections and findings concerning the "potential" revenue available in the marketplace must be verified as to accuracy as indicated in Lenawee. 10 Scope of the Study This dissertation reports the results of an investiga- tion into the performance and characteristics of over 200 select radio markets throughout the United States between 1973 and 1978, the existing radio stations therein, and the possible introduction of additional stations into these markets. It is anticipated that this study will provide a base upon which an additional body of knowledge about the future growth and performance of radio broadcasting through- out the United States can be formulated. Several related, but fundamental, measurement and analysis tools will be utilized, with the following objectives in mind: 1. To establish the amount of average "potential" radio revenue available in each market as a function of total retail sales and other select market characteristics. 2. To determine if select market variables play a significant role in the economic performance of radio stations therein. 3. To examine the interrelationship between radio revenue and the proximity of the select market to other markets of similar or larger populations. 4. To project these findings in order to generate a model which will ascertain the probable economic viability of new station entry into each market. 11 5. To identify individual variables and groups of variables which merit more detailed analysis. The need for research of this type is centered over three major areas of discussion. First, the term "economic injury" as presented in Carroll and later defined in Missouri—Illinois is sufficiently vague as to allow the determination of new station entry based on existing station income, perceived injury, and some measure of "total adver- tising revenue potential" in the marketplace. Second, unlike television where, in most markets, oligopoly con- ditions insure product (programming) standardization, radio stations are forced through competition to address program~ ming to segmented markets. The pleadings requirements in Missouri-Illinois concentrate on "public service" programming which generates no direct revenue for the station and, in turn, probably has little bearing on the ultimate "financial success" of either the existing or proposed station. In reality, public service programming goes with the territory (license), and the ultimate question must be whether there is economic room for survival of both the new and existing broadcasting stations? Finally, the study of the viability of new station entry to markets across the United States is germane to recent concerns over increased minority ownership in the broadcast industry. As a general rule, the cost of entry is much lower for new stations where there are no premiums 12 to be paid in a purchase price for intangibles such as goodwill. A major criticism of a study of this type might center on the appropriateness of economic research which deals with new station entry. It has been suggested that there are fewer and fewer channels or frequencies available for new stations to occupy, and that this type of analysis will not be relevant for very long. In response, the current table of frequency assignments28 was developed by the FCC in the late 1940's. Given that it is technically possible for a community to have up to 50 FM stations through the modifica- tion of that table on a station-by—station basis, coupled with current discussion at the Commission over decreasing the channel width in the AM band to allow additional stations on-the-air,the argument that we are reaching a saturation level in radio station growth is technically incorrect. Limitations This dissertation has some specific limitations which should be brought out at this time. First, this study does not attempt to define the overall "success" of any given individual station. This research is solely concerned with the relationship of select market and aggregate station variables, on a market-bydmarket basis. Second, this study is primarily quantitative in nature, and in no way accounts for all the qualitative aspects of 13 some of the variables included. For example, levels of management expertise in some markets will exceed that in other markets. While this will have a bearing on the economic performance of the individual stations therein, it is not a part of this study. Finally, this study is limited by the sometime inaccurate and incomplete nature of some of the data. In particular, the reporting methods of the FCC are highly suspect and will be addressed later in this paper. In response, if a particular variable was in question, it was coded as missing data. FOOTNOTES 1"Where Things Stand," BroadcaStinflj XCVII (October 1' 1979)] p. 10. 2Christopher H. Sterling, "Decade of Development: FM Radio in the 19603," Journalism Quarterly, XLIIX (1971). pp. 229-230. 3Frederick W. Ford, "Economic Considerations in Licensing of Radio Broadcast Stations," Federal Communications Bar Journal, XVII (1961), p. 192. 4In the Matter of Arthur Lucas, 5 FCC 464 (1938). See also, In the Matter of Wichita FalIs Broadcasting Compapy, 3 FCC 386 (1936); In the Matter of James E. Davidson, 4 FCC 594 (1937); and In the Matter of Dorrance D. Roderick, 5 FCC 563 (1938). 5Frank J. Kahn, "Regulation of Intramedium 'Economic Injury' by the FCC," Journal of Broadcasting, XIII, No. 3 (Summer, 1969), p. 224. 61n the Matter of Fall River Herald News Publishing Company, et a1., 5 FCC 483 (1939). 7309 U.S. 470 (1940). 81bid. 9103 U.S. App. D.C. 246 (1953), loIbid. 11 74 Stat. 889. 1247 U.S.C. 309(d) (1). 14 15 1347 U.S.C. 309(d) (2). 14S. Rep. No. 690, 86th Cong., lst Sess., 3 (1959). See also, H.R. Rep. No. 1800, 86th Cong., 2d Sess. 11 (1960). 15Tri-Cities Broadcasting Co., 24 R.R. 691 (1962). See also, Rhinelander Television Cable Corp., 37 FCC 1071 (1964). l6119 U.S. App. D.C. 1 (1964). l'7Ibid, at 3. 18 3 R.R. 2d. 232 (1964). 19122 U.S. App. D.C. 137 (1965). 20Ibid, at 138 n. 2. 21126 U.S. App. D.C. 123 (1967). 22Ibid, at 127. 23459 F 2d. 1286 (1972). 24Ibid, at 1297. zsIbid, at 1298. 2648 FCC 2d. 1181 (1974). 27Ibid, at 1182. 2847 CFR 73.202; Sec. 316, 66 Stat. 717; 47 U.S.C. 316. CHAPTER II REVIEW OF LITERATURE Previous studies in the area of broadcast economics have either been inconclusive, eroding imivalidity with the passage of time, exclusive of FM station growth, or in need of further testing with larger samples. In 1971, Levin1 published a study which examined the pattern of new station entry during the period 1939 to 1950. The results are listed in Table I on page 17. Analysis of the Levin study suggests that construction of new stations during the period 1945-48 correlates posi- tively with the changes in per-station income from the previous time period of 1939-45. Likewise, the pattern of new construction during the period 1948-50 correlates positively with the change in per-station income during the preceding time period, 1945—48. Rank-order techniques used by Levin suggest that AM station entry during a current time period (To) depends somewhat on the rate of change in per—station income in the marketplace during the preceding time period (T_l), which in turn decreases per-station income during the current time 16 17 TABLE I CORRELATION OF NEW AM STATION ENTRY AND PER STATION INCOME IN COMMUNITIES OVER 50,000 POPULATION BETWEEN 1939 AND 1950 Stations On—The-Air 1939—45 1945-48 1948-50 Per Station Income 1939-45 .10 .38 1945-48 -.66 .49 1948-50 -.10 Source: Harvey J. Levin, The Invisible Resource: Use and Regulation of the Radio Spectrum (Baltimore, 1971). period. Based on a revised profit picture, a new pattern of station entry would emerge during the next time period (Tn) . Unfortunately, the Levin study was conducted in communities of 50,000 people and over, and dealt solely with AM stations. Since the 1940's, the regulatory policies of the FCC have changed to limit the growth of AM while allowing FM to become an independent growth medium.2 Today, while there may still be a positive correlation between the change in per-station income during the current time period and entry of new stations in some future time period, changes in FCC policy have a significant impact on the present applicability of the Levin research. In addition, 18 attempts to explain entry variance with only one financial variable, per-station income, and new station construction under current FCC policy would seem a waste of time. The selection of "income" as a variable also provides cause for concern. The employment of different accounting methods with regard to depreciation, amortization, and capitalization, make this criterion variable meaningless without the removal of extraneous expense items (e.g., corporate country club membership, private airplanes, cars, etc.). A much better predictor variable would be gross revenues where little variance can be expected station-to- station. Another problem area in the Levin research seems to be the significance of the findings. While all correlations are significant at the .05 level of reliability, the highest correlation (rho = .49) accounts for only a quarter of the variance (rho2 = .24). This would indicate that there are a number of unexplained factors that contribute to the pattern of new station entry within any given market. The unexplained factors are not discussed in the Levin study. In 1966, Wagner3 published a study which attempted to find a relationship between the characteristics and growth of the broadcast industry. The Wagner study briefly touched on some factors, which for 30 years (1936-66), have influenced the growth patterns of broadcasting. In conclu- sion, Wagner stated: 19 Competition between media has always been intense, but the growth of broadcasting has made it more so in the past 30 years. The data reviewed in this study suggest that this competition is now in a period of intensification, and as further growth becomes more difficult to achieve, other factors, such as skill of management and efficiency in operation, more and more will provide the formula for success.4 Late in 1966, Saunders and Till5 took the first quantitative look at some of the unexplained correlations of the financial behavior of broadcasting stations. Their results, sampled from 2,082 AM stations in 1964, are listed in Table II on page 20. While the Saunders-Till study was perhaps the most exhaustive national research at that time, the authors were cognizant of the limitations of their research. For example, the authors state in the last chapter of their study, "As is typical in preliminary studies of this type, the conclusions are necessarily limited, and the recommendations are rather extensive."6 One area where the Saunders—Till research was valuable was to eliminate a number of variables that are unrelated or reflections of more meaningful measures of the financial behavior of a radio station. Recommendations in the Saunders and Till study included: 1. With the exception of the four variables mentioned above (term of ownership, per capital effective buying power, per household effective buying power, and market quality index), each independent variable employed in this study should be examined in detail using more sophis- ticated methodology to specify the direct effects of each variable. 20 .mHuvH .dm .xppma .mamauaC m20wpmum mcflummopmoum mo uofi>msom Hafiocmcfim on» mo mmumHmHHou wanwmmom mo cowummflumm>cH cm .HHwB .m Hanund cam mumpcnmm .w mmEmn "mouaom so.) mH. am. pm. om. Hm. ma. mason mcflssm .mmm paonmmsom ppm Ho. 8m. Np. Am. mm. om. mm. meoocH Hmcomumm co. «m. Np. nm. em. om. mm. mmem Hemumm oo. am. mm. mm. mm. Hm. mm. mpaosmmsom oo. am. mm. mm. mm. Hm. mm. coflupandom moHemHmmeoamamo ammmaz OH. mm. as. we. 58. em. mm. “A4 no pump» .02 moHemHmmaoamemo oneaam 113.8 I 3.1 8.1 H”; uuN Hum u 1.. u u m m m mmw m n m n Wnuw m a e a 9 any a I a n mmm N Z ”OHQEmw xou m UT. uT. uT. no no 1.3 a S n n nu n1 3.... a a a a? an". U? T. ..MI voma ZH mZOHedfim OHQ¢mmm AdHUZdZHh ho mmfiflqmmmou HH mamfle 21 2. The correlation matrices resulting from this analysis should be used as raw data for studies using factor analytic and multiple regression techniques. From such analysis it might be possible to develop formulas to predict the profitability of broadcasting stations from their known characteristics. None of the previously mentioned studies on the finan- cial indices of radio broadcasting has attempted to explain the variance contributed by correlates of financial behavior of broadcasting stations, nor were any data applied to the future growth of radio broadcasting. With no scientifically based data generated to explain the future growth of radio, a new study was undertaken in 1975 by this author.8 In 1966 Saunders and Till maintained that a majority of the independent variables included in their study (e.g., population, households, retail sales, and personal income) correlate meaningfully with the dependent variables (e.g., network revenues, national revenues, local revenues, total revenues, total expense, and income) and warrant further investigation. The Saunders study was repeated in 1975 by this author using 1973 data from ten single-station small markets in Oklahoma. The reasons for reducing the sample to single-station markets was twofold. First there were extreme administrative problems in obtaining per-station financial data from the FCC. Second, by studying single- station markets, any problem associated with separation of the effects of competition on the data set were eliminated. The purpose of the 1975 exploratory study was to help determine which dependent and independent variables were the 22 best possible indices of financial and market character- istics, and which variables account for a majority of the variance between the two groups in single-station markets. The results were then compared with the Saunders study of 1966. In addition, it was possible to develop a formula through multiple regression techniques, to predict the total radio revenue assumed by an existing station within a single-station market, when only market characteristics are known. Finally, using the Levin and Saunders studies as a model, it was possible through the introduction of new variables to account for a majority of variance when a new station was introduced to an existing, single-station market. Prior to the initial construction of a correlation matrix in his 1975 study, this author felt that one or more additional variables needed to be generated and tested. In order to weigh the economic performance of an existing station within a particular market, parameters must be established. In this case, it became necessary to examine the economic variability of the station against some measure of potential revenue. To generate this new variable, it was first necessary to select an index of relative market strength, and then examine its relationship to existing station revenue data. One well-known barometer of market activity is the variable "total retail sales." The other criterion variable, revenue data from the top 300 radio markets in 1972, was published 23 by the FCC. Therefore, it was possible to establish a relationship between total retail sales and the broadcast financial variable, total radio revenue, within each respective market sampled. The concept of developing a constant to describe the average potential radio revenue available in the market- place was not new. In 1960, Chapman9 first suggested that total retail sales be used as a predictor of average potential radio revenue in the marketplace. Chapman arbitrarily divided all markets into four distinct groups based on market size (e.g., major market, metropolitan, medium, and small), and calculated the ratio of retail sales to total radio broadcast revenue as reported each year by the FCC for each group. While Chapman repeated his study in 1972,10 in neither case was there an attempt made to analyze or statistically treat the data beyond the mere reporting of ratios. In the Yadon study of 1975, a random sample (i.e., N = 30) of the top 300 radio markets was studied because of internal properties of the larger markets. Factors such as the number of stations per market, and therefore the corresponding level of competition, suggest that total radio revenue in these markets should be nearly identical to the total potential radio revenue available. In select cases, total radio revenue in some markets will exceed the average potential revenue predicted, probably at the expense of advertising revenue from some other medium. Likewise, in 24 some cases, the inverse would be true. Finally, there was always the possibility that some unexplained market charac- teristic influenced the ability of radio stations to capture the average potential radio revenue predicted for their market. The correlation coefficient between total retail sales and total radio revenue in the sample (r = .95), indicated a relationship that was positive and very strong. In addition, the coefficient was significant at the .001 level, with over 90 percent of the common variance accounted for (r2 = .91). Thus, total retail sales and total radio revenue showed a greater than chance relationship 999 times in 1,000. Given the nearly one-to-one linear relationship of the two criterion variables, total retail sales and total radio revenue, it was possible to project these findings to other markets with the generation of a constant. By testing for the standard error of the mean (SEM), where M = .3963, one could be 95 percent confident that if all 300 markets were included in the study, the new variable, average (mean) potential revenue, should fall between .3476 and .4450 percent of the other variable, total retail sales. By using the new constant (Kpr = .0039) in a single station market where N = 1, it is possible to generate an estimate of the average potential radio revenue available per market. Using essentially the same variables considered by Saunders in 1966, a correlation matrix was generated to show the relationship of market variables to the dependent or 25 financial variables. Table III on page 26, shows how the coefficients generated by this author in 1973 correspond to the Saunders study of 1966 (see Table II, page 20). Based on the national probability sample of Saunders, and the results reported by this author in 1975, it would be safe to say that the four independent variables common to both studies (i.e., population, total retail sales or potential revenue, households, and personal income) would continue to show a positive relationship with a majority of the financial variables if the Yadon study were repeated in other markets. Differences in the strength of relationships between the two studies may be attributed to the data set utilized by each researcher. Saunders utilized individual station data, while Yadon used aggregate station data per market, and thus reported higher correlations. Only those variables indicating a station's gross profit margin, and efficiency of assumption of potential revenue (i.e., financial efficiency index, and radio-dollar index) showed a consistent negative or negligible relation- ship with the four independent market variables. The neg- ative relationship between the independent market variables and the dependent radio-dollar index was expected. This may suggest that the larger the single-station market gets, the less likely the existing station is of capturing the potential radio revenue available. Or, it could also suggest that the constant (Kpr) decreases as market size and level of competition increase. Regardless of market size, one 26 .mbmfl .muwmum>fl:9 oumum meoapaxo .mflmmnu .m.z pmnmgansdas ca ..mnma :A mpmxnmz oflppm :oflpmum mamcwm mEoanxo mo How>mswm Hmwocmcflmg .cowm» .m puonom "monsom .moamm Hamumn Hmuou mo cofluocsm m we moambmu Hafipcmuoms 58.- so. pm. pm. mm. mm. as. meoqu Hmcomumm mn.u mo. Nb. mm. mm. om. up. kmscw>mm Hpfiucmuom me.) so. so. me. me. up. we. mpaozmmsom mn.n mo. pm. as. mm. om. HA. aofiumasdom mUHBmHmMBU¢M¢mU Emmm¢mmm AdHUZfiZHh HEB m0 mMEdqmmmOU HHH mqmdfi 27 can expect that once the distance between the total radio revenue assumed by the existing station and the potential radio revenue available becomes great enough, the market would become viable for new station entry. This would warrant further study in markets of different size over time. The negligible relationship between the independent variables and the dependent financial efficiency index is consistent with the Saunders study. While profits tend to increase along with relative market size, the relationship between total radio revenue and total net income (profit) remains fairly static. This may be due, in large part, to the fact that expenses tend to increase at a near linear rate with relative market size. Beyond this, the measure- ment of income (profit) and expense, and their ultimate inclusion in any study is not recommended.11 Covariations between dependent variables and the inde- pendent variables alike suggest that near linear relation- ships exist. For example, the four independent variables were so highly related that no rotation was possible with factor analytic techniques. Due to collinearity, any one market variable (i.e., population, total retail sales, house- holds, or personal income) may be selected to describe the overall covariation of market characteristics to financial dependent variables. Likewise, the number of dependent var- iables may be reduced in a similar fashion. The linear re- lationship that exists between national/regional revenue, local revenue, total revenue, and total expense, suggest 28 that any one variable could explain adequately the var- iance of this group. The strength of the factor loadings in the 1973 study by this author would tend to support the contention by Saunders that radio is primarily a local and/or regional advertising medium. Another quantitative look at the relationship between radio revenues and total retail sales was made in 1975 by 12 Paul Kagan. While examining AM stations only, Kagan came to the following conclusion concerning the application of the potential revenue constant: In may be that some station buyer decides to enter a market precisely because it is under the national average, and thus seems to have more potential for increasing its share-of—market. On the other hand, radio is a mature business in most locations, and one must raise serious questions about markets that fall far below the average. It cannot all be accounted for in the management column. It may well have to do with the characteristics of the stations in the area (power, frequency, etc.), the competing media, and the media habits of the populace.1 While Kagan is correct in his warning concerning the application of national average, across the board, to markets of differing characteristics, there are a number of problems with his research. First, the study only examined AM, and AM/FM combination revenues. While AM, AM/FM revenues clearly account for a majority of total radio revenue in the United States in 1975, the increasing impact of indepen- dent FM radio cannot be discounted as indicated in Table IV below. Second, it may well be true that differences between the potential of a market and the existing station revenues 29 cannot all be accounted for in the management column; however, some of the factors which Kagan cites as alter- native explanations are concerns indirectly related to station management. For example, factors such as competi— tion and media habits of the populace are items which the manager of a radio station must consider in the day-to-day operation of the station, and may even control, to some degree. Other factors, such as station power and frequency, while meaningful and related to potential of the station, are technical in nature and are beyond the direct control of management.14 TABLE IV TOTAL REVENUES BY STATION TYPE (In millions of dollars) 1 Year AM, AM/FM % of Ind. FM % of Total Revenue Total Revenue Total Revenue 1973 1,316.1 .90 153.6 .10 1,469.7 1974 1,369.3 .88 193.4 .12 1,562.7 1975 1,430.2 .85 245.3 .15 1,675.5 1976 1,622.6 .83 332.5 .17 1,955.1 1977 1,761.4 .80 428.7 .20 2,190.1 1978 1,974.9 .78 570.5 .22 2,545.4 1Includes revenues from FM stations associated with AM stations, but reporting separately. Source: FCC Annual Report, 1978. Based on the conclusions and recommendations of both Saunders in 1966, and this author in 1975, Yadon continued 30 his investigation of market and financial characteristics in March 1976.15 This time, however, the author drew his sample (N = 63) from large markets, representing data from over 1,000 AM and FM stations over each of the two sample years, 1971 and 1973. Using essentially the same variables considered by Saunders in 1966, and by this author in 1975, a correlation matrix for each test period (e.g., 1971 and 1973) was generated to show the relationship of market variables to the dependent or financial variables. In addition, new factors included the distance of the individual test market from the next largest SMSA, and a measurement of internal test market efficiency (e.g., marginal propensity to consume). Table V and Table VI on page 31, show relationships for 1971 and 1973 which range from positive, with a definite to very dependable correlation, to negative with a definite relationship. As the two tables would seem to indicate, the variables total retail sales and consumer spendable income were judged the best possible individual predictors of total radio revenue. However, due to the vast amount of collinearity between the independent variables, either one would adequately represent market variance when using regression equations. Other independent variables in the study correlate meaningfully with the dependent total radio revenue. Our new independent variable, distance index, supports the con- tention of this author that the distance between the test 31 TABLE V CORRELATES OF THE FINANCIAL BEHAVIOR OF SAMPLED RADIO MARKETS IN 1971 “303 U) U) U) a w 8 8 Sample: N = 63 FIB-S 8 c H : (0,044 0) OJ €00) 88:3 8 8 ‘68 8.28. Em End [—.m MARKET CHARACTERISTICS Distance Index .40 .42 .29 .41 Marginal Propensity to Consume -.32 -.38 -.38 -.38 Population .80 .75 .68 .74 Total Retail Sales .89 .98 .96 .98 Consumer Income .88 .98 .96 .98 TABLE VI CORRELATES OF THE FINANCIAL BEHAVIOR OF SAMPLED RADIO MARKETS IN.1973 LH U) U) 38 8 8 Sample: N = 63 73.213 5 5 '35 888 8 8 88 aazsm 5:: 5.2 84m MARKET CHARACTERISTICS Distance Index .54 .49 .36 .47 Marginal Propensity to Consume -.32 -.35 -.31 -.35 P0pulation .87 .78 .70 .77 Total Retail Sales .85 .96 .96 .96 Consumer Income .86 .98 .96 .98 Source: Robert E. Yadon, "Economic Behavior of Differen- tiated U.S. Radio Markets, 1971 & 1973." Unpub. Report, Dept. of Telecomm., MSU, March 1976. 32 market and the next larger SMSA varies directly with total radio revenue. In other words, the proximity to competition in larger markets also seems to be important. The other new variable, marginal propensity to consume, shows an inverse relationship with the dependent variable. This internal market indicator, which is an index of the relationship between annual growth of total retail sales and consumer spendable income, suggests that as the marginal propensity to consume index increases, the amount of total radio revenue tends to go down. Differences between retail sales and consumer spendable income may be partially accountable to capital investments of various forms, or indicative of spending outside the respective marketplace. Since retail sales and consumer spendable income show a near linear relationship with total radio revenue, the efficiency index may well be a better predictor of market activity and the effect it has on total broadcast revenues. Because of differences in the economic structure of radio and television, much of the published work in tele- vision economics is not directly applicable to this study. At the same time, some of the previous work supports some of the positions of this author, and therefore justifica- tion for this dissertation, and deserves mention. Park, for example, claims that beyond the study of gross revenues of stations, as reported to the FCC 33 annually, other financial variables such as station expenses and net income (profit) are meaningless and inherently weak: The large variation observed in the profits of apparently equally situated stations suggests that financial data filed by individual stations have little usefulness for policymaking purposes...comparisons of individual station performance are questionable because of differ— ences in station operating modes and other factors that cannot be systematically taken into account.16 Others, such as Webbink,l7 support the use of discrete variables to predict entry, while Besen and Hanley18 use a regression model to predict the number of households needed to support a given number of television stations. In summary, it should be apparent that past research into the relationship of market characteristics and the economic performance of radio stations therein is limited. This dissertation will attempt to resolve some of the issues and lay the groundwork for predicting the viability of market entry. FOOTNOTES 1Harvey J. Levin, The Invisible Resource: Use and Regulation of the Radio Spectrum (Baltimore, 1971), pp. 358—368. 2Lawrence D. Longley, "The FM Shift in 1945," Journal of Broadcasting, XII (1968), p. 360. 3Paul H. Wagner, "Changing Growth Patterns in Broad- casting," Journal of Broadcasting, X (1966), p. 337. 4Ibid. 5James G. Saunders and Arthur R. Till, An Investiga- tion of Possible Correlates of the Financial Behavior of Broadcasting Stations, Report No. 1, Ohio University Center for Research Broadcast Management and Economics (Athens, 1966). 6Ibid., p. 31. 71bid, pp. 32-33. 8Robert E. Yadon, Financial Behavior of Oklahoma Single Station Radio Markets in 1973. An unpublished M.S. thesis, Oklahoma State UniverSity, 1975. 9"How Much Volume Should You Do?," Broadcasting, LIX (July 11, 1960), PP. 82-84. 10"Finding Par for Business at Radio Stations," Broadcasting, XXCII (June 19, 1972), pp. 39-40. 11"How Much of Radio's 'Spot' Leaks into 'Local?'" Television/Radio Age (February 4, 1974), p. 32. 34 35 12Paul Kagan, Communications Investor, No. 89 (February 4, 1975). 13Ibid, p. 3. 14David E. Schutz, "Is That Station Really a Bargain?," BM/E (July 1978): PP. 82-87. 15Robert E. Yadon, "Economic Behavior of Differentiated U.S. Radio Markets, 1971 & 1973." An unpublished report, Department of Telecommunication, Michigan State University, March 1976. 16Rolla Edward Park, Leland L. Johnson and Barry Fishman, Projecting the Growth of Television Broadcasting: Implications for Spectrum Use (Santa Monica: Rand Corp., 1976)! p. ix. 17Douglas W. Webbink, "Regulations, Profits and Entry in the Television Broadcasting Industry," Journal of Industrial Economics, XXII (1973), p. 167. 18Stanley M. Besen and Paul J. Hanley, "Market Size VHF Allocation, and the Viability of Television Stations," Journal of Industrial Economics, XXIV (1975), pp. 41-54. CHAPTER III METHODOLOGY This study employs a relatively large number of variables, based on recommendations and findings of 1 in 1966, Levin2 in 1971 and this Saunders and Till author3 in 1975 and 1976. New variables, not previously considered were also introduced. The Variables Previous research regarding the station and market characteristics of radio suggest that there are at least two distinct groups of variables for examination.4 First, there are the aggregate station variables of those radio stations currently in each test market as reported annually by the FCC. These station characteristics would include total radio revenue, AM and AM/FM combination revenue, independent FM revenue, and a new variable, average station revenue, generated internally by the computer. Items such as station expense and net income (profit) are excluded from this study for reasons previously cited.5 36 37 The second group of variables, characteristics of the individual sampled markets over the test period, include total retail sales, population (rank), buying power index, total number of commercial radio broadcast outlets in the market, distance to the next largest SMSA, a discrete variable indicating entry of a new station in the market, and an index for competition. Specification of Variables The following are some of the variables and terms used in this dissertation that may require additional definition: 1. Sample Market For purposes of this dissertation, a sample or test market will consist of those radio markets, be they an SMSA or county, that also qualify for inclusion in the FCC's Annual Report for each year between 1973 and 1978, inclusive. As such, between 1973 and 1977, only those markets with three or more AM, AM/FM combinations, and/or independent FM stations, meet this criteria. As of January 1, 1973, there were only 218 markets with commercial radio broadcast properties that qualified for inclusion in this study.6 See Appendix B. In 1978, the FCC changed its reporting procedures to combine revenue data on all stations 38 (e.g., AM, AM/FM combinations, and independent FM) under one heading, and list the total, aggregate number of stations per market. While this caused some confusion in comparisons of markets over time, it also eliminated the need to code revenue data on some markets as "missing data" because there was less than three stations in one or more of the reporting categories (e.g., AM, AM/FM combination, or independent FM). Total Broadcast Revenue (TOTREV) Total broadcast revenue consists of total radio time sales plus talent and program sales, plus other incidental broadcast revenues, less commissions, as reported annually by the FCC. This figure will include revenues reported for both AM and AM/FM combinations, plus inde- pendentETIrevenue data where appr0priate. In cases where either AM, AM/FM combination revenue or independent FM revenue is missing due to less than three stations reporting, or where not all stations in the market reported on time, total broadcast revenue and that specific variable will be coded as missing values. The variable "Total Broadcast Revenue" is the sum of the following two variables, "AM and AM/FM Revenue," and "Independent FM Revenue." 3. 39 AM and AM/FM Revenue (AMREV) AM and AM/FM revenue represents data from individual AM stations, and AM/FM combinations that report financial information as a single station, and consists of total time sales plus talent and program sales, plus other incidental broadcast revenues, less commissions, and as reported annually by the FCC. In the aggregate, this figure will exclude nearly 700 independent FM stations and over 400 FM stations associated with AM stations but reporting separately each year. Independent FM Revenue (FMREV) Independent FM revenues consist of data reported by nearly 700 independent FM stations plus over 400 FM stations that are associated with AM stations by reporting separately each year. This figure represents total time sales plus talent and program sales, plus other incidental broadcast revenue, less commissions, as reported to the FCC on an annual basis. Average Station Revenue (AVGREV)' A term used to identify the average or mean total revenue generated in a sample market per station. This figure will be generated in the 40 computer based on the actual number of stations per market. If total broadcast revenue is coded as a missing value, or where the precise total number of radio stations on-the-air is impossible to ascertain, this variable will be coded as a missing value. Total Retail Sales (TRS) This variable includes all sales or cash receipts of businesses within the Standard Metropolitan Statistical Area (SMSA) of the sampled markets, and as estimated annually in Spot Radio Rates and Data. BuyinggPower Index (BPI) Annually, Sales and Marketing Management magazine reports the relative buying power of each SMSA across the United States. This index is comprised of a weighted measure of a market's population, effective buying income, and total retail sales, and is expressed as a percentage of the nation's total buying power. Total Number of Stations (NUMSTA) In most markets, the total number of radio stations on-the-air will not equal the total number of stations reporting revenue to the FCC for inclusion in their annual report. 41 For example, in 1977 the FCC reported seven AM and AM/FM combinations, and one independent FM station, in the Lansing, Michigan market (SMSA). In truth, this represented a total of 14 individual stations, comprised of one individual AM station, six AM/FM combinations, and one independent FM station. Thus, eight stations reporting revenues for Lansing, actually amounted to 14 individual stations in the aggregate. Further, because the FCC only reported the number of AM and AM/FM combinations in a market, versus the number of similar combinations reporting revenues, it was impossible to calculate exactly how many stations, in the aggregate, were actually in the market or the actual number reporting. In 1978, the FCC changed reporting procedures and combined the separate independent FM report with the AM and AM/FM combination data. Further, they began reporting the total aggregate number of stations per market. They forgot, however, to report the aggregate number reporting revenues as in previous reports. Thus, the problem shifts from trying to calculate the aggregate number of stations per market, to one of trying to deter- mine if all stations reported revenue figures on time for inclusion in the annual report. Because 1978 was the first year for the FCC to report under 42 the new procedures, and the 1978 report included 1977 data, it was relatively easy to compare reports to determine if all stations in the market reported revenue figures for 1978. In 1982, all the above reporting problems were eliminated when the FCC decided to no longer collect financial data from stations. For purposes of this study, the number of radio stations in a market will be the aggregate number of individual stations on-the-air in each sampled market during the test period, where such information may be accurately determined, regard- less of whether part of an AM/FM combination. Individual Market Index (IMI) The individual market index may be defined as the percentage of total retail sales in a market captured by the total aggregate number of existing radio stations on an annual basis. Thus, the IMI is simply the total radio revenue of a market divided by total retail sales, with the result multiplied by 100. Individual indices over the six-year test period may then be compared and contrasted to develop Operational parameters for expected or potential radio revenue in markets of different size. 10. ll. 43 Potential Revenue (POTREV) It is anticipated that the performance of existing radio stations, and in turn the entry of new stations into each market, will be based to some degree on the efficiency of those exist- ing stations in capturing the potential radio revenue available. In previous studies, only one constant (Kpr) was developed using the mean IMI of all markets under study, and this constant was used to project potential revenue as a percentage or function of total retail sales. Chapman,7 however, recognized the markets of different size, due to competitive differences, would perform in different ways. Thus, he suggests using different constants (Kpr's) for each group or subset, based on market population or rank. Potential revenue, therefore, is a variable which is generated internally by the computer using the mean IMI of a subset of markets as a constant, in order to project the average po- tential revenue available asaafunction of total retail sales. Normality Index (NINDX) If the average potential revenue available in a market is calculated, then parameters may 12. 13. 44 be developed to describe a "range of normality" in which markets within each group or subset can be expected to operate. This range can be arbitrarily established as one standard deviation above and below the mean IMI, or Kpr' for that group of markets. For purposes of discrimination, markets falling below the normative range will be coded "-1", while those within the range will be coded "0", and those above the range will be coded "+1". The performance of each market, in terms of total radio revenue, will be measured against the above parameters and the proper value encoded for each year of this study. Radio-Dollar Index (RDINDX) A market's total radio revenue reported as a percentage of the average potential radio revenue projected for the sample market. Distance Index (DIST) This variable will be a measurement, in miles, of the distance from the city of license for a majority of stations in one market, to the next largest city of license. 14. 15. 45 Entgy of a New Station (ENTRY) Entry of one or more new radio stations into a market during one or more of the test years will be indicated with a discrete variable during the year of entry where "0" indicated no entry during the year under study, and "1" indicates entry of a new radio broadcast station. Competition Index (COMPIN) Radio is primarily a local or regional advertising medium. It should therefore follow that a good deal of the competition for adver- tising revenues in the marketplace will come from the local daily newspaper(s), and the local tele- vision station(s). Therefore, some method must be developed to express the relative strength of newspaper and television advertising, and, in turn, the level of competition in each select market. Television revenues for most markets are published by the FCC annually. In turn, Dessart8 recommends using the total advertising linage per year of each daily paper, as reported by Editor and Publisher, to project the annual revenues of newspapers in each market. The problem with using dollar (revenue) estimates, the actual revenue figures, or for 46 that matter an aggregate count of the number of papers and television stations in the market, is that the larger markets will always generate larger revenues due to a greater number of daily newspapers and television stations. Thus, there will be a problem with collinearity with the other predictor variables unless this measure can be expressed in some other fashion. One suggestion might be to express the total number of radio stations in the market as a weighted percentage of the total number of com- peting units (e.g., television, radio and news- paper). In effect, this will generate a variable which will indicate the level of competition between the three media. The higher the ratio, in theory, the better the chance radio has of fractionalizing the market and, in turn, captur- ing a larger percentage of total advertising revenue available. Some thought has been given to extending this concept to include a weighted value for one additional radio station. The problem here, however, is that this research centers on aggre- gate variables and ascertaining whether there is room for one additional station. The issue of "impact" of an additional radio outlet is too closely related to the respective management 47 expertise of that outlet, and is therefore outside the scope of this research. For purposes of this study, newspapers were counted only if published on a daily basis from within the respective radio market as defined earlier in this section. Television stations, on the other hand, were counted only if the main studio and transmitter were located in the radio market under study. Hypotheses A number of major questions are left unanswered by previous research. For example, it is known that as markets become larger, there is normally a proportional increase in the number of radio outlets. This does not, however, suggest that large market stations, in the aggregate, perform the same as those in a smaller market. Thus, the research question becomes whether the relationship between a select market variable such as total retail sales, and the aggregate performance of radio stations in a given market as indicated by total radio revenue, is consistent regardless of market size or number of individual radio outlets? Hypothesis 1: There is a difference in the relationship of total retail sales to total radio revenues, represented by the individual market index (IMI), in markets of different size. 48 All markets are not alike in media buying habits, or levels of radio station management expertise. In turn, these differences could be expected to explain a portion of the level of competition in these markets. In short, the level of competition goes beyond the aggregate number of stations per market, and includes the sophistication of market media buyers, and the aggressiveness of station management and local sales personnel. These differences might be demonstrated through the examination of the proximity of aggregate radio revenues in the market to some measure of "potential" market radio revenues based on comparison of the test markets to similar markets through- out the United States. Using the normality index described earlier in this section, it is possible to segregate stations into three discrete classifications; those that fall above or below one standard deviation of the mean IMI for a market in a given year, and those that fall within one standard deviation above and below the mean IMI. Hypothesis 2: New station entry will occur more often in markets that fall one standard deviation or more below the mean IMI, as measured by the normality index (NINDX), than in markets that are within or above one standard deviation. As explained earlier, as market size increases the level of competition, or aggregate number of stations, also increases. Yet, competition can be both "inter" and and "intra" market for available advertising dollars. Thus, 49 by comparison of the relative statutory distance of the test market to markets of similar or greater population, the impact of cross-market competition on market revenues should be evident. Hypothesis 3: The closer a test market is to another market of equal or greater population, the less likely the specific market will be of capturing available radio advertising revenue. The arbitrary classification of test markets by the normality index may not be sufficient to discriminate among predictor variables for purposes of identifying those markets ripe for entry by a new station. If actual radio revenue for a test market, as reported by the FCC, were taken as a percentage of the predicted radio revenue avail- able, then this new variable, radio-dollar index, might be employed to demonstrate if a relationship exists. Hypothesis 4: The greater the difference between potential radio revenue (POTREV) and actual radio revenue (TOTREV), ceteris paribus, the greater the likelihood of new station entry (ENTRY). While there is no doubt that larger markets have more radio outlets than smaller markets, there remains a question as to when a market demonstrates sufficient "economic" room for entry. Certain conditions must exist within the marketplace to insure the new station an adequate chance for survival. At the very least, it would not be prudent for an entrepreneur to enter a market with the preknowledge 50 that a new radio station had little chance for success. Therefore, there must be certain "signals" a market exhibits to indicatei1:is ripe for entry. In other words, there are select market variables which, in the right combination, indicate the potential for entry by a new radio outlet. It is the identification of these variables which is paramount. Hypothesis 5: Entry of a new radio station into a market will be dependent upon changes in relationships within and among select market and station variables over time. Analysis of Data The variables for all the 218 markets reported will be punched on data processing cards. The cards will then be read into the Michigan State University's Cyber 750 computer system, and the data tabulated and treated according to various Statistical Package for the Social Sciences (SPSS) computer programs.9 Coefficients of correlation will be computed to describe the magnitude and direction of relationships between each of the station (dependent) variables and each of the market (independent) variables. With pairs of continuous variables, the Pearson product-moment coefficient of correlation (rp) will be used. For comparisons involving one continuous variable and one discrete variable, such as "entry of a new station," a rank-order coefficient of correlation will be used. 51 A one-way analysis of variance will be performed on the variable "individual market index" by market identifica- tion (ID) to test for significant differences in the amount of total retail sales assumed in markets of different size. A a posteriori contrast test, such as Least Significant Difference (LSD) will be used in an attempt to define groups of markets that are homogeneous. The result may be that groups of markets will be divided into subsets where differ- ences between their mean IMI's will not be significant at the .05 level or greater. The alternative is, that no differences exist in the aggregate performance of stations across time. The 218 markets will be examined individually. Therefore, it is possible that over 100 subsets could develop. While it is expected that significant differences exist be- tween markets, fewer than 50 subsets should materialize. It is from this a posteriori analysis that parameters defining the normality range of operation for each subset can be developed should differences exist. If subsets among the 218 markets are identified, a new mean IMI would be calculated for each group or subset and the standard deviation above and below the mean IMI calculated for each subset. Classification of a market which operates within, or outside the range will be beneficial in the later deter- mination of criteria for new station entry. 52 Multiple Regression Analysis A method of multivariate analysis, multiple regression allows a researcher to study the collective and separate contributions of two or more independent variables to the variation of a dependent variable. As such, the major purposes of regression analysis are prediction and explana- tion.10 The output of the SPSS regression package used in this study is designed to supply regression weights (b) which serve as a means to identify the relative contribution of independent (exogenous) variables to a dependent (endogen- ous) variable. Therefore, the sample regression coefficient, 11 or weight, becomes the estimate of the population. The general format can be expressed as: Yi = f(X1, ..., Xn) (l) where the Yi are the station (endogenous) variables, and Xn are the market (exogenous) variables. Regression analysis will be valuable for a number of reasons. First, it allows the researcher to ascertain the contribution of the independent market variables to the dependent total radio revenue in markets of different size. Second, it allows one to examine the relative role of these indepen- dent variables to the aggregate performance of stations as differentiated by the normality and radio-dollar indices. 53 Discriminant Analysis In order to study the entry of a new station into a market, discriminant analysis will be employed to identify those variables which determine when a market is ripe for new competition. Through the classification of markets that added stations, or did not add stations, over the six- year test period, it is possible to identify a group of variables which dictate whether entry is probable. By coding each of the 218 markets with a discrete variable where "1" indicates entry of a new station occurred, and "0" where no new entry took place, it is possible to examine markets in terms of the potential for future entry based on the past performance of these same markets. The discriminant program in the SPSS package will form a linear combination of those market and financial variables with respect to the sample market that added a station during the six-year test period. The computer will then establish a set of classification functions which will be used to produce the probability of membership of each sample market in the "entry" or "no entry" groups. Thus, the output contains regression coefficients which become the estimates of the probability of entry. The final stage of analysis is the review of the original data to deter- mine how many markets were correctly classified based on the variables selected for discrimination. In this manner, it is possible to determine the relative accuracy of the formula for prediction in other markets over time. 54 In the case of this study, Rao's V will be used as the stepwise criterion for selection and inclusion of independent variables in the discriminant analysis. Rao's V is a generalized distance measure, which means it is an indication of which independent variables contribute the largest change in the value of V when added to the variables already selected. A decrease in the value of V would indicate that a variable contains a large amount of informa- tion already explained by variables previously selected. Independent variables are tested for retention in the model at each step, based on their partial multivariate F ratio. The minimum F ratio to avoid removal in stepwise analysis is 1.0. Evaluation of the discriminating power of the discriminant function will be based on the canonical correla- tion, or ability of the function to separate the groups, ENTRY (0,1). Wilks' Lambda, and its associated chi-square test of statistical significance, indicates the discriminat- ing power in the variables being used. FOOTNOTES 1James G. Saunders and Arthur R. Till, An Investiga- tion of Possible Correlates of the Financial Behavior of Broadcasting Stations, Report No. 1, Ohio University Center for Research Broadcast Management and Economics (Athens, 1966). 2Harvey J. Levin, The Invisible Resource: Use and Regulation of the Radio Spectrum (Baltimore: The Johns Hopkins Press, 1971). 3Robert E. Yadon, Financial Behavior of Oklahoma Single Station Radio Markets in 1973, an unpublished M.S. thesis, Oklahoma State UniVersity, 1975. Robert E. Yadon, "Economic Behavior of Differentiated U.S. Radio Markets, 1971 & 1973," an unpublished report, Department of Telecommunication, Michigan State University, March 1976. 4Saunders, op. cit. 5"How Much of Radio's 'Spot' Leaks into 'Local'?" Television/Radio Age (February 4, 1974), p. 32. 6FCC Annual Report, 1973. 7"Finding Par for Business at Radio Stations," Broad- casting, XXCII (June 19, 1972), pp. 39-40. 8George Dessart, ed., Television in the Real World: A Case Study in Broadcast Management (New York: Hastings House, 1978), P. 151. 9Norman H. Nie, et al., Statistical Package for the Social Sciences, 2nd ed. (New York: McGraw-Hill, 1975). 10Fred N. Kerlinger and Elazar J. Pedhazur, Multiple jRegression in Behavioral Research (New York: Holt, Rine- hart and Winston, 1973) , P. 4 55 56 11W. J. Dixon, BMD: Biomedial Computer Programs (Berkeley: University of California Press, 1970), pp. 15-21. CHAPTER IV RESULTS The discussion which follows first examines bivariate relationships among aggregate station and market variables measured in 218 test markets over a six-year period. Next, specific hypotheses are addressed and discussed. Finally, models are presented and discussed with respect to entry of new outlets in radio markets of varying market character- istics. Introduction of Market Indices ’Prior to the initial construction of a correlation matrix, a number of variables must be generated and tested. First, in order to weigh the strength of a market, para- meters must be established. In the case of this study, it is necessary to examine the aggregate relationship of station viability in a test market to the marketplace itself. To generate such a variable, it is first necessary to select an index of relative market strength and then examine its relationship to existing radio revenue data. One well-known barometer of market activity is the variable 57 58 total retail sales. The other criterion variable, total radio revenue in each test market, is published by the Federal Communications Commission (FCC) on an annual basis. Therefore, it is possible to establish a relationship be- tween total retail sales (TRS) and total radio revenue (TOTREV). As is suggested in earlier studies, a strong relation- ship exists between the criterion variables. In fact, when all 218 markets in this research were examined over the six-year test period, a total of 1308 individual observations (cases), the positive relationship between total retail sales and total radio revenue (r = .9726) indicates that this combination is nearly linear, signifi- cant at the .001 level, with nearly 95 percent of the common variance explained (r2 = .9460). Therefore, total retail sales and total radio revenue show a greater than chance relationship 999 times in 1,000. Individual Market Index While a strong relationship exists between the market variable total retail sales (TRS) and the aggregate station variable total radio revenue (TOTREV), differences may manifest themselves between markets over time. In order to test this proposition, a new variable must be created that exhibits the relationship of these criterion variables on a market-by-market basis. 59 Individual Market Index (IMI) = TUTREV x 100 (2) The new Individual Market Index (IMI), was calculated for each market over the six-year test period. This index simply represents total radio revenue for a market as a percentage of total retail sales. A one-way analysis of variance was performed on the data between IMI and the discrete market identification variable (ID). As Table VII indicates, using the Least Significant Difference (LSD) test for a posteriori contrasts, with alpha set to .001, significant differences did appear between markets. Unfortunately, while 24 homogeneous subsets were identified, those subsets were not unique and extensive overlap of markets between subsets preclude meaningful analysis at this time. It should be sufficient at this point to state that differences do exist in the aggregate performance of individual radio markets as measured by the IMI index, and that in the future each market should be treated as a sepa- rate entity. While unique subsets were not apparent, the IMI index will be retained in this study as a meaningful variable. The results of this analysis support the first research hypothesis, that a difference exists in the relationships between total retail sales (TRS) and total radio revenue (TOTREV), represented by the individual market index (IMI), in markets of different population. 60 TABLE VII ONE-WAY ANALYSIS OF VARIANCE BETWEEN IMI AND MARKET ID Source df s.s. m.s. F (Prob.) Between Groups 206 5.3357 .0259 17.877 (.001) Within Groups 671 .9722 .0014 TOTAL 877 6.3079 Normality Index If the new variable IMI is helpful in measuring the aggregate performance of stations with respect to the marketplace itself, then this new index may also be useful in establishing parameters upon which to gauge this per- formance. Should the parameters prove meaningful, they may be used later in this research to discriminate between markets that add a new station and those that do not add a new station. To generate these parameters, a new variable was created which establishes an artificial range of normality as a measure of market performance. To generate this range, the first step is to examine the data on a year-by- year basis to determine if differences exist between mean IMI's. Table VIII below indicates that differences do exist between markets over time. Thus, it is relatively 61 easy to use the mean IMI for each year to establish artificial performance standards. TABLE VIII ONE-WAY ANALYSIS OF VARIANCE BETWEEN IMI AND YEAR Source df s.s. m.s. F (Prob.) Between Groups 5 .3650 .0730 10.771 (.0001) Within Groups 898 6.0865 .0068 TOTAL 903 6.4515 Table IX below indicates the mean IMI for each year under study, plus the arbitrary range of one standard deviation above and below the mean used to establish the upper and lower limits to the new variable, normality index (NINDX). TABLE IX MEAN IMI AND RANGE OF NORMALITY, BY YEAR Year Mean IMI S.D. Upper Limit Lower Limit 1973 .3072 .0858 .3930 .2214 1974 .3024 .0785 .3809 .2239 1975 .2836 .0788 .3624 .2048 1976 .2948 .0721 .3669 .2227 1977 .3257 .0858 .4115 .2399 1978 .3439 .0908 .4347 .2531 For example, if the IMI for a specific market fell within plus or minus one standard deviation of the mean IMI for that year, the new variable (NINDX) would be coded "0"; and if the IMI was above one standard deviation the market was coded "+1"; and finally should the IMI fall one standard deviation below the mean IMI, the market was coded "-1". In order to ascertain whether the new variable, NINDX, exhibits a significant relationship with the dependent ENTRY, two statistical tests were performed. First, the nonparametric Spearman rank-order coefficient of correla- tion (rs) and Kendall's tau (I) were computed (r8 = -.0326; T = -.0314). In each of the 962 cases analyzed, however 63 the relationship was small, negative and not significant at the .05 level of confidence. The second test was to determine if differences exist between NINDX groups (-1, 0, +1), and the dependent ENTRY. Distribution of the 692 cases over the NINDX groups was as follows: -1 = 207; 0 = 407; and +1 = 78. Table X below indicates that significant differences do not exist between the three NINDX groups based on the ENTRY index (0,1). In that these are two nonparametric variables, the Kruskal-Wallis Test was also used to confirm the rela- tionship. Again, the results indicate that there is not a significant relationship between the NINDX coding scheme and entry of new stations into a market. Obviously, the results of these tests do not support the second research hypothesis that ENTRY will occur more often in markets that fall one standard deviation below the mean IMI as measured by the normality index (NINDX). TABLE X ANALYSIS OF VARIANCE BETWEEN NINDX AND ENTRY Source df s.s. m.s. F (Prob.) Explained 2 .067 .034 .551 (.577) Residual 689 42.007 .061 TOTAL 691 42.074 .061 64 Potential Revenue Given the nearly one-to-one linear relationship of the two criterion variables, total retail sales and total radio revenue, it is possible to develop another arbitrary means of evaluating radio markets. By using the mean IMI figures presented in Table IX, it is possible to generate a dollar-figure for the "estimated" average potential revenue available per market, per year. Simply stated, the mean IMI per year can be multiplied by total retail sales in each market for that same year to create an index upon which to evaluate the assumption of potential revenue available by the aggregate stations per market. Potential Revenue (POTREV) = TRS x Mean IMI (3) Because the mean IMI per year changes over time, the new variable (POTREV) can be applied to the changing revenue climate per market to estimate the efficiency of each market in capturing, vis-a-vis other markets, potential revenue available. In its raw form, the new POTREV variable would be little more than a mirror of the relation- ship between total retail sales and total radio revenue. Thus, additional treatment is required to make this new variable meaningful. 65 Radio-Dollar Index The application of the potential revenue variable (POTREV) per market allows a new index of revenue assumption to be generated. The new variable, radio- dollar index (RDINDX), is the percentage of average potential revenue assumed by the aggregate stations in a given market. _ TOTREV " POTREV (4) Radio-Dollar Index (RDINDX) In retrospect, it is possible to generate a dollar estimate for average potential revenue in each test market, and at the same time calculate the percentage of assump- tion by the existing stations. Relationship Between Station and Market Characteristics Using essentially the same variable considered in earlier studies, plus those generated herein, a correla- tion matrix was generated to demonstrate the relationship of market variables and station variables. Table XI indicates relationships which range from positive with a very dependable correlation, to negative with a moderate to marked relation. Where the correlation coefficient was not significant at the .05 level of confidence, the letters "N.S." were entered into the matrix. Even a casual review of the coefficients presented in Table XI suggest the immediate problem of 66 . .flmuuwfio mucfiom HmEHown .moamcflmcoo mo H0>OH no. on» an u:mo«macmfim #02 n .m.z O! x .m.2 .m.z and soon .m.z mom who .m.z ooo .m.z .m.z .m.z and .m.z .n.z Naanu .m.z x and won «me: on» .m.z vno vow mmn com «um aoo poms poo mud >uuhoa .m.z and x own howl ode .m.z ooo van duo and mna ooo hnwu onH can KH>HO om~ woo -~ x oonl woo aoN own man hnm moo coo omo «om: «mo .ooo >mmo>£ Foo: anal howl ooNs x «oat bro: coal .m.z mud; maul maul .m.z .m.z qua: and Zamtou .m.z ano oafi Now Hon: x woo mno NHH nun ado vvo boa one: «do un~ A2009 moo .m.z .m.z aom who) «no x goo on" ooo .m.z .m.z oho ova .m.z .m.z HIM who. vhm omo mom «out nno aoo x oho can nvo arm has «om: vvo .m.z >N¢POP .m.z 1:. ..2 o: .m.z SA on” «S x .m... 6.2 So . 2: .m... . .m... 3... H523: ooo mmm doc nmm mad: men omo vac .m.z x ago new vnn Nod: ano wad BmHn .m.z moo ~n~ mmo NNAI mam .m.2 nvo .m.z ago X ooo nos ooml boo moo Hon .m.z boa mm~ new owe) vvm .m.z nho moo non ooo x how new: aha boa mm? .m.z goo one ono .m.z poo who has oou vnn man now x own: ”oh oma mmeoo Hoo. omm. moo >mmeoo .mom. ooo. mo oomm>Ho moo. omm. moo mm>Ha «Hoo. ooH.) moH oh>mmo>a ooo. ooo. moo >mmo>m «moo. ooH. mo ooszzoo Hoo. oom. moo szzoo .ooo. ooo. on moozooa moo. Hmm. moo ozooe .moo. ooH.u mod oonH .ono. moo. moo HzH .moH. moo. omm mo>mmeoe .oHH. ooo. moo >mmaoa Aezaemzooo nu Hoo. oom. moo Hszmmz ooo. oam. on moaoHo moo. «Hm. moo emHo «mom. moo. mam ooHom moo. ooo. moo Hom «mom. ooo. mam oomme moo. ooH. moo mme .ooo. ooH. mam oomeozoz Hoo. mom. ooo «emzoz .mom. moo.) omm oomzam omo. ooo.) moo xzmm .moa. moo. on homomzm omo. woo. moo oomzm x.omoo u mu 2 mflnmmpp> 1.omoo u mu 2 mabmmup> omoomq omszzoo mAmQOZ mewdfl 92¢ QWZHmEOU ZH Nmazm mom mBZWHUHmmmOU ZOHBdAmMMOU UHMBflZflMfimZOZ HHxx mqmdfi 83 functional form. New stations will enter markets of different size, differing economic characteristics, and at different points over time. Thus, variables like SMPOP, RANK, TRS, BPI, etc., Offer little to identify those points at which markets are ripe for entry. In Table XXII, market variables show a decline in correlation strength between the combined and lagged examples. On the other hand, indices like IMI and RDINDX, regardless of significance levels, show a marked improvement in variance explained. Therefore, it is not the individual market var- iables that are important, but the relationships between variables that seem to be significant. Using essentially the same variable set, it is possible to generate six new, additional variables to assist in the discrimination analysis. 1. Percentage Change (PERCHG) The percentage change in total radio revenue, year-to-year. PERCHG = (TOTREVT/TOTREV ) -1 (7) T T-l 2. Revenue Per Capita (REVPOP) Total radio revenue in the marketplace, allocated to total population. REVPOPT = TOTREVT/SMPOPT (8) 84 3. Average Revenue Per Capita (AVGPOP) The average radio revenue per station, allocated to total population. AVGPOPT = AVGREVT/SMPOPT (9) 4. Number of Stations Per 100,000 Population (STAPOP) Number of radio stations in the marketplace, allocated per 100,000 population. = * STAPOPT (NUMSTAT/SMPOPT) 100,000 (10) 5. Number of Stations Per $1,000,000 in Total Retail Sales (STATRS) = * STATRST (NUMSTAT/TRST) 1,000,000 (11) 6. Percentage FM Revenue (PRFREV) Percentage of total radio revenue in market- place allocated to FM stations . PRFREV = FMREVT/TOTREV T (12) T Rather than combining this new variable list with the group of independent variables previously generated, only those market and station characteristics related to ENTRY were selected as discriminating variables. Thus, in addition to the six new variables generated above, IMI, COMPIN, AVGREV and RDINDX will be retained for analysis. 85 Discriminant Analysis Stepwise discriminant analysis was selected in order to eliminate the "less" useful variables in the list prior to performing the actual analysis. The criterion selected for entry into the discriminant model was Rao's V, a generalized distance measurement which allows the greatest separation between the dependent groups--ENTRY (0,1). Because each market does not have an equal prior probability of entry, adjustments must be made before analysis. In this case, the author elected to allow the data set to be used to estimate the population distribution. The prior probability for ENTRY was 16 percent, and for NO ENTRY it was 84 percent. Again, ENTRY78 was used as the dependent variable in the model, and measures Of all ten independent discriminat- ing variables were collected between 1975 and 1977. Prior to 1975, the number of missing values reduced the valid observations to a level that makes analysis difficult. In addition, previous research presented indicates that a clear majority of variance is accommodated in these three years. Table XXIII indicates that only eight of the original 30 independent variables entered were actually selected, and only seven variables were retained for analysis. At Step 9, the independent variable REVPOP76 was removed from the model when the change in Rao's V became negative and 86 Hmoa. mmmm.HI Hooo. mmvm.mm Homo. oohmv. ohmom>mm o momo. vam.m Hooo. mmow.Hm mavo. oomme. momom>mm m Homo. nmon.o mooo. vmmm.om ommo. vowmw. ohmmeeam o baoo. oamv.m mooo. hmmm.~m mmmo. omaom. mh>mmmmm o voaa. oowm.m oHoo. ommm.ma ommo. mmmvm. mommemam m ommo. mmmo.m Hmoo. omvm.oa mHNo. omhom. nowmummm v Haao. ohmv.o «woo. ohHH.mH vamc. omomo. ohmom>mm m mono. mamo.m mmmo. Homo.o oHoo. mvmhh. mhomummm N mmmo. omom.m mmmo. omwm.m mane. mmmom. oo>mmmmm H .mflm >< .mflm > m.oom .mwm mpnfimq po>OEom commucm moum mxaoa mHQMHHm> MMDDmoomm ZOHBUMAmm BZflZHZHMUmHQ mmHZQWBm HHHxx mqmdfi 87 nonsignificant. In the end, seven variables were analyzed. In Table XXIV, canonical discriminant functions are analyzed. With 100 percent of the variance accommodated in the single function for the model, as specified, a canonical correlation of .7498 for the single function, and a Wilks' Lambda of .43769, this list of discriminating variables provides a very good degree of separation between the dependent groups, ENTRY (0,1). With two groups, only a single function is possible. The Wilks' Lambda of .43769 corresponds to a chi-square of 16.112 with a significance of .0241. A lambda of this size or smaller has a .0241 prob- ability Of occurring by chance. If the canonical correlation (RC = .74987) were squared (R: = .56231), it would indicate the amount of variance in the single function explained by the two discriminant groups, ENTRY (0,1). Table XXV displays the unstandardized and standardized canonical discriminant function coefficients. The unstan- dardized coefficients are used to obtain a discriminant score for the single function by multiplying each coeffic- ient by the respective variable value, and summing the products plus the constant, market by market. Standardized coefficients represent the number of standard deviations that variable is away from the mean for all variables in the single function. Further, the standardized coefficients represent the relative contribution of the variable to the function. In the end, markets are classified into the group with a pattern most like its own. 88 TABLE XXIV CANONICAL DISCRIMINANT FUNCTIONS Percent of Variance Canonical Wilks Chi- Correlation Lambda Squared Sig. 100.0 .7498715 .4376927 16.112 7 .0241 TABLE XXV CANONICAL DISCRIMINANT FUNCTION COEFFICIENTS Variable Standardized Unstandardized Coefficients Coefficients PERCHG75 1.51339 16.10863 PERCHG77 .58397 7.51070 REVPOP75 - .82899 - .26936 STATRS75 - 4.12087 - 800.06640 STATRS76 3.55897 753.24180 PRFREV75 4.76402 37.75887 PRFREV76 - 5.97474 - 42.23632 (CONSTANT) 3.16520 89 Of the 218 cases (markets) processed for ENTRY78, unfortunately 193 had at least one missing discriminating variable, leaving a total of 25 cases used in the analysis. Appendix C identifies the 25 markets and their respective discriminant scores. Table XXVI displays the output of the discriminant analysis, and shows that all markets were correctly identified in this model. TABLE XXVI CLASSIFICATION RESULTS Predicted Group Membership Actual Group N 0 1 NO ENTRY (O) 21 21 0 (100.0) (0.0) ENTRY (l) 4 0 4 (0.0) (100.0) Percent of Grouped Cases Correctly Classified -- 100% Table XXVII shows the classification function coeffi- cients required to apply the discriminant model on a market- by-market basis. Measures from the seven discriminating variables, plus the constant, are used to generate a classification score for each of the ENTRY and NO ENTRY groups. The classification equation for either group can be expressed as 90 C. = c. PERCHG75 + c. PERCHG77 + c. REVPOP75 + 1 11 12 13 C. STATRS75 + c. STATRS76 + c. PRFREV75 + 14 15 16 C. PRFREV76 + ki 17 (13) where C1 is the classification score for either the ENTRY or NO ENTRY group, ci are the coefficients listed in Table XXVII, and ki is the constant. There will always be a separate classification score calculated for each group, ENTRY/NO ENTRY, with the market assigned to the group with the highest classification score. TABLE XXVII CLASSIFICATION FUNCTION COEFFICIENTS FOR ENTRY78 Variable 0 l PERCHG75 -44.l305 3.6394 PERCHG77 -.3092 21.9637 REVPOP75 1.6322 .8334 STATRS75 2433.4080 60.8185 STATRS76 -l456.7040 777.0270 PRFREV75 -104.2404 7.7331 PRFREV76 131.4837 6.2322 (CONSTANT) -l6.9292 -12.1911 91 Based on the results of the regression analysis Of NUMSTA78 and this analysis of ENTRY78, the entry of new radio station outlets in a market is directly proportional to the activity Of that market. The regression and dis- criminant analysis support the fifth research hypothesis, that entry of a new radio station into a market is dependent upon changes in relationships within and among select market and station variables over time. CHAPTER V SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS Summary The purpose of this exploratory study was to determine which market and station characteristics were the best predictors of the number of stations per market (NUMSTA), and the entry of new stations into the marketplace (ENTRY). In addition, another purpose of this study was to test specific relationships that exist in radio markets, and in turn what bearing these relationships have on the dependent variables. Earlier studies into the behavior of broadcast stations maintain that market variables play a significant role in determining the financial behavior of radio properties, and the pattern of new station entry. For example, Saunders examined the relationships between market characteristics and station financial variables in 1964.1 Unfortunately, this study was limited to a single year, precluding all but a simplistic look at the correlations between indi- vidual station and market variables. Thus, the problems of collinearity, or benefits Of developing a lagged model were never addressed. 92 93 In 1971, Levin published a study of new station entry in markets of over 50,000 population between 1939 and 1950.2 For his study, entry was not a discrete variable but simply the aggregate number of stations on-the-air. The other criterion variable, per-station income, was used to demon- strateéi"lagged" relationship between the two variables. The conclusion was, that the rate of change in per-station income during a previous time period would have a positive effect on AM station entry during the current time period. In turn, the entry of a station in the current time period would tend to decrease per-station income. The problems of using "income" have been covered by this author and others, and as with the Saunders study, Levin only examined AM stations. Since 1970, a clear majority Of the growth in this industry has been in the FM service. The exclusion of FM data in previous research is only one justification for renewing this investigation. Another is the overall simplicity of these earlier efforts. For example, Saunders attempts to explain the performance of radio stations using simple correlations between station and market variables which display an extremely high degree of multicollinearity. Levin, on the other hand, uses a single lagged variable in an attempt to explain new station entry. As his own research demonstrates, the largest amount of variance explained within his model comes from a rank-order correlation of r = .49 (r2 = .24). 94 The door was essentially Open to a more exhaustive treatment of market-station relationships, and ultimately the investigation of new station entry. Using most of the variables employed in earlier studies, this research examines 218 radio markes over a six-year period, 1973 to 1978. In turn, a number of new variables not introduced in previous research were covered. The relationships between station and market variables in this study were, overall, as expected. In a majority of cases, correlations were strong, very dependable, and consistent with earlier research. However, because this study incorporates variables not previously considered, with more extensive treatment, additional discussion is called for. First, previous studies document the close relationship between market variables like population, households, total retail sales and personal income, and the financial perform- ance of broadcast properties. Yet these earlier studies did not address whether these relationships were consistent market-to-market, regardless of size. Because of the collinearity of these market variables, it is unnecessary to test every relationship. For this study, only total retail sales (TRS) and total radio revenue (TOTREV) were selected for initial study. The Pearson correlation coefficient between TRS and TOTREV is strong and highly dependable (rp = .973). It would be reasonable to assume that a single variable, total 95 retail sales (TRS), is an adequate predictor of total radio revenue (TOTREV) in the market. Yet, this bivariate correla- tion says nothing about the relationship between these two variables, market-to-market. If a new index were generated by dividing total retail sales by total radio revenue, and multiplying the resultant by 100, the product would be the percentage of retail sales captured by the aggregate stations in each market, or individual market index (IMI). If total retail sales were a good predictor of total radio revenue, one would expect the percentage to vary little, market-to- market. If, on the other hand, one expected differences in levels of radio station management expertise between markets, differences in levels of cross-media competition, or differences in market economics, then another conclusion might be warranted. As was demonstrated in this study by the one-way ANOVA between individual market index (IMI) and the market identification variable (ID), significant differences do exist in these relationships between markets over time. Second, if significant differences do exist between markets based on the relationship of retail sales to radio revenue, then this index might be used as a relative measure of new station entry. If the variable IMI were calculated for-all 218 markets over the six-year period (1973-1978), a total of 1308 Observations, then some assumptions could be made concerning the impact of this index. For example, relative to the mean IMI per year, do those markets that 96 fall below theIMfinihave a higher probability of new station entry? In order to test this proposition, markets with IMI's falling one standard deviation below the mean IMI for that year were coded "-1." Those markets between one standard deviation below and one standard deviation above the mean were coded "0", and those with an IMI above one standard deviation were coded "+1." This classification scheme developed a new variable, the normality index (NINDX). Both rank order correlation coefficients and an analysis of variance demonstrate that no significant difference exists between market IMI's with respect to the NINDX classification and new station entry (ENTRY). Third, it is recognized that no radio market operates in a vacuum. That is, the relationship of one radio market to the next closest radio market could have a bearing on the financial behavior Of the market under study, especially if the market next door is larger. If the number Of radio stations a market can support is based, to some degree, on the ability of the market to support those stations, then any factor which would tend to diminish or siphon Off radio revenues would also tend to have a negative impact on new station entry. To test this proposition, the distance between each Of the 218 test markets and the next closest market with equal or greater population was noted. This became the new variable for distance (DIST). It then became a 97 relatively easy matter to correlate the relationships between DIST and the various market and station character- istics. As was demonstrated in this study, DIST had a positive effect on both the total radio revenue available in a market, and the overall efficiency of stations in capturing available revenue. Fourth, as indicated earlier in Table IX, the mean IMI fluctuates year-to-year, probably in response to the aggregate pressures of the national and local economies. At the same time, this figure provides a reasonable annual index Of the aggregate performance of radio stations across the country with respect to their individual, local economic environment (e.g., total retail sales). Thus, if the mean IMI is used, year-by-year, to develop a "potential" radio revenue figure for each market, this new variable (POTREV) could be measured against the actual revenue figure (TOTREV) to document the performance of each station in the 218 test markets. In the end, the potential radio revenue figure (POTREV) would be compared to the actual radio revenue reported (TOTREV), and a new index indicating the assumption Of potential radio revenue generated (RDINDX). In theory, one might expect that those markets that show the greatest distance between the actual radio revenue gen- erated, and the expected or potential revenue calculated, tO be the "best" candidates for new station entry. Unfortunately, the relationship between ENTRY and this new variable RDINDX was extremely small, negative and not 98 significant at the .05 level of confidence. Thus, the level of potential revenue assumption, using a national index, was demonstrated to be unrelated to new station entry. Finally, the question of new station entry was approach- ed in a direct manner. Given a substantial list of market and station variables, it should be possible to identify a group or subset of variables which influence the aggregate number of radio stations in a specific market. This can be accomplished in two ways. First, it may be possible to estimate the number of stations a market is capable of hold- ing. When compared to the actual number of stations on-the- air, some assumptions might be made concerning the intro- duction of a new radio outlet. Second, it may be possible to identify the specific point at which a market becomes ripe for entry, regardless of the current number of stations on-the-air. In order to accommodate the large number of variables employed in this study, a method was needed to reduce the variable set and thereby eliminate as much multicollinearity as possible. Factor analysis was used to examine the entire variable set, with the application of the principal components method and varimax rotation. In the end, three select factors appeared. A single variable was selected, one from each factor, to be representative of all loadings within each factor. Those three variables were the popula- tion of the SMSA (SMPOP), market rank (RANK), and the individual market index (IMI). 99 The importance of the three select variables deserves some discussion. First, market SMSA population (SMPOP) provides the model with a continuous measure of market size, an obvious factor in the number of stations per market. The second variable, market rank (RANK), is almost the inverse of population, but more correctly identifies the importance of a specific market to media buyers at the national level, since some advertisers make buys according to rank (e.g. buy stations in the tOp 50 markets). The final variable, the individual market index (IMI), is the ratio of retail sales to total radio revenue. This variable reflects the interaction of station and market revenues. With 1978 as a typical year, the three variables were used in a regression equation as independent variables to examine their collective impact on the variance of the dependent number of stations (NUMSTA78). Analysis indicated that these three variables account for more than 78 percent Of the variance of the dependent NUMSTA78 (r2 = .78921) for 1978. Thus, this three—variable model would provide a reasonable approximation of the aggregate number of radio outlets per market. The obvious problem is in using the independent vari- ables from a single year to predict a dependent variable from that same year. If the amount of variance accounted for in a lagged model of these same three variables could not improve on the single year regression equation, then application of this model, as specified, would be relatively useless. 100 The three independent variables were tested over a three-year period prior to 1978, and entered into the step- wise procedure Of multiple regression with the dependent NUMSTA78. The results indicated that three variables, SMPOP77, RANK77 and IMI75, accounted for more than 80 percent of the common variance with the dependent NUMSTA78 (r2 = .80106). Thus, the lagged model provided a better fit, although marginal difference, with the dependent variable. Finally, it may be possible to identify the point Of entry for new stations within a test market. To do so, it requires a new variable set to be generated. As previous correlation matrices demonstrated, the discrete variable ENTRY (0,1) was not highly correlated with most market var- iables (e.g., SMPOP, TRS, BPI, RANK, etc.). If the original variable set offers little to identify those points in time where entry is likely to occur, new variables must be generated. In this case, the new independent variables are actually indices or relationships among the variables already discussed. For example, percentage change in total radio revenue (PERCHG), radio revenue per capita (REVPOP), average radio revenue per station (AVGREV), the number of radio outlets per 100,000 population (STAPOP), the number of stations per $1,000,000 in total retail sales (STATRS), and the percentage of total radio revenue going to FM radio stations (PRFREV). Rather than using this new list of independent variables along with all those previously generated, only IMI, COMPIN, AVGREV and RDINDX were 101 retained from the original list for further analysis. Using discriminant analysis, it is possible to test this new set of ten independent variables and select those which allow for the greatest separation between the depen- dent groups ENTRY (0,l). Again, 1978 was used as the base year for the dependent ENTRY, with the independent variables examined over the previous three years (1975-1977). The discriminant analysis produced seven variables that provided a good degree Of separation between the dependent groups, with over 50 percent of variance of these variables explained by the two discriminant groups (R: = .56231). The discriminating variables were PERCHG75, PERCHG77, REVPOP75, STATRS75, STATRS76, PRFREV75, and PRFREV76. In the end, the discriminant model was tested against the original data set. Of the 25 cases where there was no missing values in any of the discriminating variables, the model correctly identified the point (year) of entry, with 100 percent of the grouped cases correctly classified. Conclusions This study found that most independent variables cor- related meaningfully with the dependent NUMSTA. Based on previous research, and the findings of this study, it is safe to assume that most market characteristics would continue to show a positive, very dependable relationship with the number of stations in a market (NUMSTA) in the 102 future. Only those variables indicating the relationships between total retail sales and total radio revenues on a national and local scale (MEANIMI, IMI), the level of media competition (COMPIN), the diversity of radio station owner- ship in a test market (DIVER), and the level of assumption of potential radio revenue (RDINDX), demonstrated a negligible or insignificant relationship with the dependent variable. The negative relationship between number of stations in a market (NUMSTA) and market rank (RANK) was expected. Covariations between the dependent variable NUMSTA and independent variables suggest that linear relationships exist. The high degree of interrelationship between the independent variables was evident in the r-matrix (see Table XI). This multicollinearity was demonstrated when a clear majority of the market variables loaded on a single factor during factor analysis. For example, six independent variables were so highly related that rotation would have been impossible with factor analytic techniques had these same variables been examined exclusively. Any one of the six variables (SMPOP, TRS, BPI, TOTREV, AVGREV, POTREV) could be selected to describe the over-all covariation of market characteristics. Three variables previously thought to be indicative of station activity in the marketplace (e.g., TOTREV, AVGREV and POTREV), are actually more de- scriptive of the over-all market behavior and therefore were dropped from further consideration. 103 A general conclusion from this analysis might be that aggregate station financial performance is directly related to the economic performance Of the marketplace. A linear relationship does exist; however,this conclusion is too broad and does not recognize unique differences that exist between markets. This concern was addressed when the relationships between two highly correlated variables, total retail sales (TRS) and total radio revenue (TOTREV), were examined in 218 markets over the six-year test period. The results of the a posteriori contrasts test revealed that unique subsets of markets, based on the IMI, do not appear. It is therefore impossible to say with any degree of cer- tainty, as was suggested by Chapman,3 that markets can be identified and grouped together. Markets previously recog- nized as outliers,or those markets with extremely low IMI's relative to the national average, may perform this way for many reasons, some of which are outside the scope of this study. For example, levels of radio station management expertise, market unemployment trends, etc. Therefore, while the IMI variable may be a good index of aggregate station-to-market performance, it can only be applied on a market-to-market basis. If an index like the IMI variable is helpful in describing the relationship of station and market variables, it may also be of assistance in identifying the year of entry for a new radio outlet. In order to use the IMI as a measurement tool, a point or points of reference must be 104 established. In this study, the mean IMI for each year was used for this function, and boundaries established that correspond to one standard deviation above and below the mean set to identify outliers. This established a range, or index of normality (NINDX). Markets that fell one standard deviation below the mean IMI for a given year were coded "-1", those within one standard deviation below and above the mean were coded "0", and those markets with an IMI above one standard deviation were coded "+1". The research question was whether markets classified or coded as "-1", demonstrated a greater probability for new station entry? In other words, in markets that might be described as "less efficient" in generating advertising revenue, is there a greater propensity for entry? The conclusion, based on analysis of variance, is that no significant difference between markets exists using the NINDX coding scheme (-l,0,+l), when evaluating these same markets for entry of a new station (ENTRY). It would seem that markets which consistently generate lower IMI's do so because there are, in fact, fewer advertising dollars available in the marketplace. Examination of these markets over the six-year test period (1973-1978), indicates that poorer performance relative to the national mean IMI is consistent, not an anomaly, and probably not due to the aggregate incompetence of station management in these markets. 105 One reason for the poor performance of stations in these markets may be the proximity of the test market to other radio markets of similar or greater population. In other words, there may be a shift in advertising revenue to the larger radio market if it is too close. Media buyers may tend to allocate advertising budgets in a dis- proportionate manner to stations in these close, but larger markets. To test this proposition, the distance between test markets and the next closest market of equal or greater population was noted. This new variable, called the dis- tance index (DIST), was correlated with select market characteristics. The results, listed in Table XII, demon- strate the distance is related positively to the individual market index (IMI), and to the estimated efficiency of the market in capturing projected advertising revenue (RDINDX). The conclusion is, therefore, that distance between markets has a small, positive impact on the aggregate ability of radio stations to generate advertising revenue. This does not suggest that radio stations in these outlier markets are not profitable, only that their ability to generate advertising revenue is capped, to a certain extent, by their proximity to a larger market. In the end, the greater the distance between radio markets, the greater the relative earning potential for stations in the smaller market. Finally, there is the question of how to determine the number of stations a market is capable of holding, or the point (year) at which a new station can enter a market? 106 There are two ways to approach this question. In the first, it may be possible to "predict" the number of outlets a market is capable of supporting through examination of select existing station and market characteristics. The second approach is to estimate the point of entry through the study of changes in relationships between variables, or changes in the variables themselves over time. The results Of this study indicate that the number of stations in a market is determined, to a large extent, on the economic characteristics of the marketplace. The only question being what form these characteristics or variables take? Referring to Table XI, the dependent variable number of stations (NUMSTA) is highly correlated with most of the market variables (e.g., total retail sales, buying power index, SMSA population, etc.). Therefore, the link between market Characteristics and the number Of radio outlets is established. Unfortunately, a majority of the independent variables are also highly correlated among themselves. The immediate problem in dealing with this variable set was one Of multi- collinearity. To deal with this problem, the r-matrix was subjected to factor analysis with varimax rotation. The results were that the original variable set loaded on only three factors. One variable was selected to represent each factor, and those three variables were used in a regression equation to estimate the dependent NUMSTA. 107 The results of the initial regression analysis, using measures from 1978 only, demonstrated that the linear combination of the three variables identified in the factor analysis (e.g., SMPOP78, RANK78, and IMI78) were an improve- ment over the bivariate coefficients. The largest single correlation with the dependent NUMSTA reported in the r-matrix was with the variable TRS (r = .807), where slightly over 65 percent of the common variance was explained (r2 = .65125).* Looking at the first regression equation, over 78 percent of the variance in the dependent NUMSTA78 was accounted for (r2 = .78921), for a net increase in variance explained of nearly 14 percent. Thus, in a static model, where both the dependent and independent variables are from the same year, the linear model provides a very good estimation of the number of stations for that same year. The application of a static regression model provides little information about the future. It is only a tool to evaluate what has already happened. In this case, the analysis would be ex-post-facto, up to a year later when measures of market variables from the previous year were finally published and available. Therefore, to address the question of predictability, the three independent variables (SMPOP, RANK and IMI) were lagged up to three years in an attempt to improve upon the amount of variance explained. *Variables like TCOMP were excluded from comparison with NUMSTA because they are functions of that variable. 108 All measures of the three independent variables were entered into the stepwise selection process of multiple regression for the years 1975 through 1977. Again, the dependent variable was the number of stations in a test market in 1978 (NUMSTA78). The results indicated that SMPOP77, RANK77 and IMI75 met the requirements for entry into the regression model, and these three variables accounted for slightly more than 80 percent of the variance in the dependent NUMSTA78 (r2 = .80106). In regression analysis, it is possible to test for autocorrelation (e.g., that the residuals are not inde- pendent from one Observation to the next). For each model, the Durbin-Watson statistic was calculated. With 132 Observations, it is safe to say that statistics as high as those calculated for both the static (D.W. = 2.0749) and dynamic (D.W. = 2.0258) regression models demonstrate, at the .01 level of significance, that positive autocorrelation is not a problem. The results Of the lagged model were not the improve- ment hoped for over the static equation. They did, however, provide valuable information concerning the interaction of market variables over time. The results indicate that market variables in their static, functional form are not good predictor variables. While the amount of common var- iance explained in both the standard and lagged models was very good and significant at the .05 level of confidence, the lagged model suggested that market variables were, at 109 best, coincident indicators not leading indicators of NUMSTA. This conclusion is based on Table XVIII, where both SMPOP and RANK coefficients demonstrate a progressive correlation over time with the dependent variable NUMSTA78. In other words, no single lagged measure provided an ideal or "best fit" solution with the dependent variable. This progression indicates that had the measures for 1978 for these two variables, SMPOP and RANK, been entered into the model as in the first equation, they would have been selected by the stepwise procedure. The small improvement in variance explained in the lagged equation Obviously comes from the independent variable IMI75, a ratio of total retail sales to total radio revenue. The only conclusion warranted is that market variables cannot be used in lagged models. The reason rests with the fact that they are not leading indicators of market activity, but at best are coincident indicators, or those whose movement coincides roughly with the current performance of economic activity.4 Therefore, unless the model is changed, the best the current list of variables can accomplish is to provide the ex-post-facto system of checks and balances on the number of stations in a market. Where a dynamic process like "entry" is examined, it must be done with dynamic variables (e.g., those that measure relationships between variables, or among variables over time). To explain differences that might exist between markets where entry will and will not occur, the existing 110 variable list was modified to generate new indices of eco- nomic activity. In addition to the four indices already incorporated in the study (e.g., IMI, COMPIN, AVGREV and RDINDX), six new variables were generated. These were measures of change in total radio revenue year-to-year (PERCHG), total radio revenue per capita (REVPOP), average radio revenue per capita (AVGPOP), radio stations per 100,000 pOpulation (STAPOP), radio stations per $1,000,000 in total retail sales (STATRS), and percentage of total radio revenue captured by FM stations (PRFREV). The four existing variables, plus the six new indices, were measured over three years between 1975 and 1977, and entered into stepwise discriminant analysis. Of the original 30 variables entered, only seven variables were retained for analysis and used to discriminate between the two groups, ENTRY78 (0,1). The seven variables retained were PRFREV76, REVPOP75, PERCHG75, PERCHG77, STATRS75, PRFREV75 and STATRS76. The results of the discriminant analysis indicated that these seven variables provided a very good degree of separa- tion between the dependent groups, with over 50 percent of the variance explained (R: = .56231). Of the 218 cases processed, only 25 were used in the discrimination, but all 25 were correctly identified later during the classification phase of the analysis. The value of discriminant analysis rests with the ability of the researcher to test the contribution Of select variables in explaining the difference between two groups. 111 In the case of this analysis, the seven discriminating variables did a very good job in distinguishing between the ENTRY and NO ENTRY markets. There are two methods available to the researcher in applying a discriminant model. First, it is possible to utilize the canonical discriminant function coefficients to generate discriminant scores. By comparing the discriminant scores to the group centroids, it may be possible to classify markets into the ENTRY or NO ENTRY groups. The second, and preferred method is to utilize the classification function coefficients for each group (ENTRY and NO ENTRY), whereby the market is assigned to the group with the highest classification score. Upon examination, the seven discriminating variables employed in this research fall into two categories. First, there is the variable PERCHG, indicating a percentage change in a single variable, total radio revenue, year-to-year. Second, there are the variables like REVPOP, STATRS and PRFREV, that demonstrate relationships between two variables within a given year. Further, three variables, PRFREV, PERCHG and STATRS, entered the discriminant model twice, at different points in time, which may indicate that they too should be measured year-to-year. With only 25 out of 218 radio markets entering the discriminant model, a major concern must be systematic bias in the selection process. The number of markets analyzed was restricted due to the FCC reporting techniques and does not appear to be systematic in any way. Further, the 25 112 markets would seem to cover a wide cross-section of the country, ranging in market rank from Indianapolis, Indiana (33) down to Sherman, Texas (308). Regardless, the discriminant model presented lends itself to the conclusion that entry of radio stations in markets is predicted on a dynamic process, and can only be determined through the examination of changes in total radio revenue, radio revenue per capita, number of stations per $1,000,000 in total retail sales, and the percentage of total radio revenue captured by FM stations. In conclusion, entry into radio markets by new stations is dependent upon the relationships between station and market variables, as well as the dynamic economic activity of the marketplace. Recommendations The recommendations which can be drawn from this study fall into two general categories. The first category deals with application of the various results and their impact on telecommunication policy. The second deals with further research in the area of radio markets, and the entry Of new radio station outlets in those markets. The policy recom- mendations are considered first. Since the late 1950's, the Federal Communications Commission (FCC) has attempted to define the term "economic injury.‘ The pleadings requirements to raise the Carroll issue are set forth in Missouri-Illinois, but were later 113 modified and restated in WLXA and Folkways to exclude the conditions of "preknowledge" and "exact calculations." These criteria are presented in Appendix A. What should be evident after reading the pleadings requirements as developed in Missouri-Illinois, is that the thrust of these guidelines surround a determination of the impact of new station entry based on (1) existing radio station income, (2) perceived injury, and (3) some measure Of total advertising revenue potential in the marketplace. This last point seems to be paramount. If there is suffi- cient potential revenue in the marketplace for the addition of a new radio outlet, then can there be a sustained case for economic injury? The answer, based on previous court rulings, would be an unqualified no. The court in §QEEB‘ western placed the burden on the petitioner to plead specific factual data sufficient to establish a prima facie case of injury to the public interest in order to meet the require- ments for a hearing. Thus, if it were demonstrated that sufficient "economic" room existed for a new station, claims of injury to the public interest would be speculative at best. The models developed in this study directly address the problem of economic injury. Rather than trying to predict the amount of potential revenue available, they examine market characteristics to ascertain when sufficient room exists for increased competition. In support Of this approach, the FCC has stated on numerous occasions that, 114 economic conditions permitting, the public interest is best served by competing broadcast stations.5 Of course the application of both the non-lagged regression and the discriminant models are predicted on the availability of both market and station data in the future. While market data are available from a number of sources, station financial information was disseminated on a controlled basis by the FCC. In 1982, as part of an over- all move toward de-regulation, the FCC stopped requiring radio stations to file annual financial reports (FCC Form 324). The loss of this valuable source of information was quickly noted by a number of professional broadcast organi- zations, and a movement is currently underway to indepen- dently poll stations for financial data. This effort is being sponsored by a consortium comprised of the National Association of Broadcasters (NAB), the Television Bureau Of Advertising (TvB), the Broadcast Financial Management Assoc- iation (BFM), and others. Their ultimate success in this effort, and the validity Of their results, remain a question. Regardless of which group collects the information, there are a number of lessons to be learned. Certainly station reporting techniques can be improved. First, each station should be counted as a separate entity. The pre- vious FCC policy of allowing AM/FM combinations to be reported as a single entity rendered much of the financial information useless. Second, the aggregate count of stations in the market should reflect the exact number Of outlets in 115 the SMSA, regardless of simulcasting. Finally, markets With missing data (e-g-r some stations failed to file data) should be identified in the report. Only through accurate reporting techniques can this data be utilized in the future. With de-regulation also comes the question as to whether economic injury remains a salient issue. Research presented in this study would tend to support the free market concept, and thereby endorse the de-regulation move- ment. Radio stations, in the aggregate, direct their efforts in an attempt to satisfy their clients, at a profit. Their performance is related to the economic activity of the marketplace, and therefore subject to positive and negative shifts in demand. Their ability to react accordingly will dictate, to a large degree, their success and ultimate sur- vival. A major argument used by the FCC in promoting de- regulation was that the marketplace was capable of control- ling the "business" of radio broadcasting. In a growth economy, those radio stations not attentive to the needs of their various publics would suffer the competitive con- sequences. In turn, if the public interest is best served by competition, then under de-regulation the issue of "entry" is that much more compelling. While the Carroll Doctrine may no longer be politically important, the economic realities of competition keep the issue alive for current and future broadcast station owners. 116 Adoption of the models presented in this study could help to insure that radio markets remain in equilibrium, expand- ing only as fast as market conditions dictate. Existing station owners could anticipate the arrival of new competi- tion, while entrepreneurs could gauge the filing of construction permits on dynamic market conditions. Radio broadcasting exists in a monopolistic competition market structure. Positive profit by existing stations will trigger entrance of new broadcast outlets which, in turn, will drive down price to the zero profit level. Conversely, if the existing stations are already at a negative profit level, some stations may be forced off the air, leaving more business for the survivors. The result is that demand and prices go up. Traditionally, if entrance to the broad— cast industry were free, the long-run equilibrium would be in a zero profit position. Of course in broadcasting, entry is not free. At some point the market will no longer respond due to technical considerations, that is there are no available frequencies left. Excess profits in this case will likely encourage new technologies to enter the marketplace as close substitutes. The second category involves the continuation of research into radio markets and the aggregate performance of stations therein. The results of this study are obviously not exhaustive, and as with most exploratory research are plagued by a few problems. One of the major concerns of the author was the volume of missing aggregate station 117 financial data, in large part created by questionable FCC reporting techniques. Until 1978, the FCC allowed FM stations associated with AM stations (e.g., AM/FM combina- tions) to report revenues as a single entity. Further, independent FM stations were reported separately. Because the FCC required a minimum of three stations to report revenues in order to publish financial data, this requirement often provided an inaccurate picture of radio economic activity in small to medium markets. In that a number of new variables were created within this study as a function of aggregate station financial data, the problem was compounded. Where pairwise treatment of data were possible, the problem was accommodated within the SPSS computer package. However, where listwise treat- ment was required, as in discriminant analysis, the impact was obvious. To remedy this problem in the future, it is recommended that the researcher (1) gain access to per- station data on FCC computer tapes, or (2) plan an extended visit to the Washington, D.C., area in order to ascertain the actual aggregate count of stations per market and related financial data. Sufficient information is contained in this study to support continued research. A detailed analysis of the impact of the individual market index (IMI) is recommended. While the relationship between retail sales and radio revenue has been documented, differences in the aggregate performance of stations as measured by the IMI index were 118 not addressed. These differences could help explain why some markets perform better than others, and to establish potential revenue per market, absent future FCC reporting of financial information. Where this study did disclose the impact of distance (DIST) on the aggregate performance of broadcast stations, other factors were not discussed. Factors which may explain differences in the performance between markets include the specific demographics of the markets (e.g., age, sex, education, minority populations, etc.). Furthermore, if competition (TCOMP and COMPIN) were indexed more meaningfully than was possible in this study it might become a factor in the differences. A new lagged regression model (linear or log/linear) to estimate the number of radio stations in a market at some future time is required. Recognizing the inappro- priate use of static market variables in lagged models, it is recommended that future research investigate the possi- bility of employing a lagged dependent variable (NUMSTA) within the new model. Another possibility would be to employ the dynamic variables utilized in the discriminant analysis in future regression models. The success Levin enjoyed in using a single dynamic variable, per station income, lends support to this proposal. Finally, additional research is called for to identify new variables that might explain additional variance in the discriminant model presented in this study. While the model presented here was extremely successful in identifying 119 the ENTRY and NO ENTRY groups, the canonical correlation and Wilks' Lambda suggest that a good deal of unexplained variance remains. The functional form of the independent variables employed in this study might be changed. For example, rather than examine relationships between two variables from a given year, it may be more helpful to examine these same relationships over time. The change in a single variable, or relationship between variables, year- to-year may provide additional information as to which dynamic characteristics of a market contribute to growth. Until such time that new research in this area is undertaken, the formulae presented in this study may be utilized to examine radio markets, expand existing service where appropriate, and initiate new service to communi- ties throughout the United States. FOOTNOTES 1James G. Saunders and Arthur R. Till, An Investiga- tion of Possible Correlates of the Financial Behavior of Broadcasting StatiOns, Report No.]q.Ohio University Center for Research Broadcast Management and Economics (Athens, 1966). 2Harvey J. Levin, The Invisible Resource: Use and Regulation of the Radio Spectrum (Baltimore, 1971). 3"How Much Volume Should You Do?," Broadcasting, LIX (July 11, 1960), pp. 82-84. 4G. H. Moore and J. Shiskin, Indicators of Business Expansions and Contractions, (New York: National Bureau of Economic Research, 1967), pp. 8-28. 55 FCC 464 (1938); 3 FCC 386 (1936); 4 FCC 594 (1937); and 5 FCC 563 (1938). 120 APPENDIX APPENDIX A CRITERIA FOR ECONOMIC INJURY The following presents the pleadings requirements as pre— sented in Missouri-Illinois, 3 R.R. 2d 232 (1964). (1). (2). (3). (4). (5). (6). (7). (8). What is the total amount of retail sales in the community and in the area for the preceding three years? If sales are in a decreasing pattern, set forth the reasons which should include information as to any unusual economic conditions in the area. What is the total number of businesses in the community and the area? What is the total advertising revenue potential in the community and the area? What is the amount of local advertising revenue actually earned by your station in the community and the area? Set forth, for at least the three preceding years, your total revenues, total expenses, net profit or loss and the average number of employees. How many of the existing businesses in the community and the area do not now advertise on the radio? State, in detail, the specific advertisers that would shift their advertising to the proposed station. How many advertisers will split their advertising time between the existing station or stations and the proposed station? What are the other competing advertising media in the community and the area? 121 (9). (10). (ll). (12). 122 Appendix A State, in detail, how a grant of this proposal would cause a net loss or degradation of program service to the area. (a). What public service spot announcements do you now broadcast? (b). How many public service spot announcements do you broadcast each week? (c). What public service programs would have to be discontinued? Spot announcements? (d). What public service programs will have to be shifted to other time segments? Spot announcements? (e). As to (c) and (d) what percentage of the total broadcast time is represented by the public service programming? (f). What is the cost of carrying these programs and what saving will be effected in dropping or shifting this programming? What is the cost of carrying the public service spot announcements and what saving will be effected in dropping or shifting spot announcements? What programming personnel changes will be required? What information, if any, do you have that some or all of the public service programming will not be carried by the proposed station? Will a grant of the proposal require you to make substantial changes in your total present program format and policies? State full details. Set forth any other information which is sufficiently related to the economics of broadcasting, including the specific relationship between any assumed losses in revenue to the withdrawal of particular programs or program services in support of the question raised in petition to deny concerning the inability of the area to support another broadcast station without loss or degradation of program service to the area. APPENDIX B RADIO MARKETS INCLUDED IN STUDY Abilene, TX Akron, OH Albany, GA Albany, NY Albuquerque, NM Alexandria, LA Allentown, PA Amarillo, TX Anchorage, AK Ann Arbor, MI Anniston, AL Appletown-Oshkosh, WI Asheville, NC Atlanta, GA Atlantic City, NJ Augusta, GA Austin, TX Bakersfield, CA Baltimore, MD Baton Rouge, LA 123 Battle Creek, MI Beaumont, TX Billings, MT Biloxi, MS Binghamton, NY Birmingham, AL Bloomington-Normal, IL Boise, ID Boston, MA Bridgeport, CT Buffalo, NY Burlington, NC Canton, OH Cedar Rapids, IA Champaign, IL Charleston, SC Charleston, WV Charlotte, NC Chattanooga, TN Chicago, IL Cincinnati, OH Cleveland, OH Colorado Springs, CO Columbia, SC Columbus, OH Corpus Christi, TX Dallas/Ft. Worth, TX Davenport/Rock Island, IA/IL Dayton, OH Daytona Beach, FL Denver/Boulder, CO Des Moines, IA Detroit, MI Duluth/Superior, MN/WI El Paso, TX Elmira, NY Erie, PA Eugene/Springfield, OR Evansville, IN/KY Fargo/Moorhead, MN/ND Fayetteville/Springdale, AR Fitchburg-Leominster, MA Flint, MI Florance, AL Fort Smith, AR/OK 124 Appendix B Fort Wayne, IN Fresno, CA Ft. Lauderdale, FL Ft. Myers, FL Gadsden, AL Gainesville, FL Galveston, TX Gary/Hammond/E. Chicago, IN Grand Rapids, MI Great Falls, MT Green Bay, WI Greensboro/Winston Salem, NC Greenville/Spartanburg, SC Hamilton/Middleton, OH Harrisburg, PA Hartford, CT Honolulu, HI Houston, TX Huntington/Ashland, WV/KY Huntsville, AL Indianapolis, IN Jackson, MI Jackson, MS Jacksonville, FL Johnstown, PA Kalamazoo, MI Kansas City, MO/KS Kileen/Temple, TX Kingsport/Bristol, TN/VA Knoxville, TN LaCrosse, WI Lafayette, LA Lake Charles, LA Lakeland/Winter Haven, FL Lancaster, PA Lansing/E. Lansing, MI Las Vegas, NV Lewiston/Auburn, ME Lexington, KY Lima, OH Lincoln, NE Little Rock, AR Lorain/Elyria, OH Los Angeles, CA Louisville, KY/IN Lubbock, TX Lynchburg, VA Macon, GA Madison, WI Manchester, NH 125 Appendix B McAllen, TX Melbourne, FL Memphis, TN/AR/MS Miami, FL Midland, TX Milwaukee, WI Minneapolis, MN/WI Mobile, AL Modesto, CA Monroe, LA Montgomery, AL Muskegon, MI Nashville, TN Nassau-Suffolk, NY New Britain, CT New Haven, CT New London, CT New Orleans, LA New York, NY Newark, NJ Newport News, VA Norfolk, VA Northeast Penn, PA Odessa, TX Oklahoma City, OK 126 Appendix B Omaha, NE Rochester, MN Orlando, FL Rochester, NY Oxnard, CA Rockford, IL Parkersburg, WV Sacramento, CA Pensacola, FL Saginaw, MI Peoria, IL Salem, OR Petersburg, VA Salinas, CA Philadelphia, PA Salt Lake City, UT Phoenix, AZ San Angelo, TX Pine Bluff, AR San Antonio, TX Pittsburgh, PA San Diego, CA Pittsfield, MA San Francisco, CA Portland, ME San Jose, CA Portland, OR Santa Barbara, CA Poughkeepsie, NY Santa Rosa, CA Providence, RI Sarasota, FL Provo, UT Savannah, GA Pueblo, CO Seattle, WA Raleigh, NC Sherman, TX Reading, PA Shreveport, LA Reno, NV Sioux City, IA Richland, WA Sioux Falls, SD Richmond, VA South Bend, IN Riverside, CA Spokane, WA Roanoke, VA Springfield, IL Springfield, MO Springfield, MA St. Cloud, MN St. Joseph, MO St. Louis, MO Stockton, CA Syracuse, NY Tacoma, WA Tallahassee, FL Tampa, FL Terre Haute, IN Texarkana, TX Toledo, OH Topeka, KS Trenton, NJ Tucson, AZ Tulsa, OK Tuscaloosa, AL 127 Appendix B Tyler, TX Utica-Rome, NY Vineland, NJ Waco, TX Washington, D.C. Waterbury, CT Waterloo, IA W. Palm Beach, FL Wheeling, WV Wichita, KS Wichita Falls, TX Williamsport, PA Wilmington, DE Wilmington, NC Worcester, MA Yakima, WA York, PA Youngstown, OH APPENDIX C RADIO MARKETS INCLUDED IN THE DISCRIMINANT ANALYSIS BETWEEN ENTRY AND NO ENTRY IN 1978 Market Actual Predicted Discriminant Entry Entry Scores 1. Alexandria, LA 1 1 2.0385 2. Allentown, PA 0 0 .0366 3. Anniston, AL 0 0 .5998 4. Appletown-Oshkosh, WI 0 0 -.3459 5. Baton Rouge, LA 0 0 -.2925 6. Beaumont, TX l 1 1.6850 7. Cedar Rapids, IA 0 0 -.7894 8. Champaign, IL 0 0 -1.7775 9. Charlotte, NC 0 0 -.3489 10. Des Moines, IA 0 0 -2.2364 11. Elmira, NY 0 0 -.2428 12. Evansville, IN/KY 0 0 -.6615 13. Fayetteville, AR 0 0 .2287 14. Ft. Lauderdale, FL 0 0 .3633 15. Gadsden, AL 0 0 -l.5553 16. Indianapolis, IN 0 0 -.2024 17. Johnstown, PA 0 0 .4601 18. Kileen/Temple, TX 1 1 3.6947 19. Provo, UT 1 1 2.5459 20. Salem, OR 0 0 .9658 21. Sherman, TX 0 0 -.5024 22. Southbend, IN 0 0 -2.6849 23. St. Joseph, MO 0 0 .4737 24. Trenton, NJ 0 0 —1.9353 25. Wichita Falls, TX 0 0 .4831 Group Centroids: E = 0 -.47448; E = 2.49101 1 128 SELECTED BIBLIOGRAPHY SELECTED BIBLIOGRAPHY Books, Periodicals and Reports Besen, Stanley M., and Paul J. Hanley, "Market Size VHF Allocation, and the Viability of Television Stations." Journal of Industrial Economics, XXIV (1975), pp. 41-54. Dessart, George. Ed. Television in the Real World. New York: Hastings House, 1978. Dixon, W. J. BMD: Biomedical Computer Programs. Berkeley: University of California Press, 1970. "Finding Par for Business at Radio Stations." Broadcasting, XXCII (June 19, 1972), pp. 39-40. Ford, Frederick W. "Economic Considerations in Licensing of Radio Broadcast Stations." Federal Communications Bar Journal, XVII (1961), p. 192. "How Much of Radio's 'Spot' Leaks into 'Local'?" Television/ Radio Age, (February 4, 1974), p. 32. "How Much Volume Should You Do?" Broadcasting, LIX (July 11, 1960), pp. 82-84. Kagan, Paul. Communications Investor, No. 89 (February 4, 1975). Kahn, Frank J. "Regulation of Intramedium 'Economic Injury' by the FCC." Journal of Broadcasting, XIII, No. 3 (Summer, 1969), p. 224. Kerlinger, Fred N., and Elazar J. Pedhazur. Multiple Regression in Behavioral Research. New York: Holt, Rinehart and Winston, 1973. Levin, Harvey J. The Invisible Resource. Baltimore: The Johns Hopkins Press, 1971. 129 130 Longley, Lawrence D. "The PM Shift in 1945." Journal of Broadcasting, XII (1968), p. 360. Moore, G. H., and J. Shiskin. Indicators of Business Expansions and Contractions. New York: National Bureau of Economic Research, 1967. Nie, Norman H., et. a1. Statistical Package for the Social Sciences, 2nd ed. New York: McGraw-Hill, 1975. Park, Rolla Edward, Leland L. Johnson and Barry Fishman. Projecting the Growth of Television Broadcasting: Implications for Spectrum Use. Santa Monica, Calif.: Rand Corporation, 1976. Saunders, James G., and Arthur R. Till. _An Investigation of Possible Correlates of the Financial Behavior of Broadcasting Stations. Athens, Ohio: Ohio Univer- sity Center for Research Broadcast Management and Economics, 1966. Schutz, David E. "Is That Station Really a Bargain?" BM/E, (July, 1978), pp. 82-87. Sterling, Christopher H. "Decade of Development: FM Radio in the 19605." Journalism Quarterly, XLIIX (1971), PP. 222-230. Wagner, Paul H. "Changing Growth Patterns in Broadcasting." Journal of Broadcasting, X (1966), p. 337. "Where Things Stand." Broadcasting, XCVII (October 1, 1979), p. 10. Yadon, Robert E. "Economic Behavior of Differentiated U.S. Radio Markets, 1971 & 1973." Unpublished report. East Lansing, MI: Department of Telecom- munication, Michigan State University, 1976. Yadon, Robert E. Financial Behavior of Oklahoma Single Station Radio Markets in 1973. M.S. thesis. Stillwater, OK: Oklahoma State University, 1975. 131 Court and FCC Decisions Carroll Broadcasting Co. v. FCC, 103 U.S. App. D.C. 246 FCC v. Saunders' Brothers Radio Station, 309 U.S. 470 (1940). Folkways Broadcasting Co. v. FCC, 126 U.S. App. D.C. 123 (1967). In the Matter of Arthur Lucas, 5 FCC 464 (1938). In the Matter of Dorrance D. Roderick, 5 FCC 563 (1938). In the Matter of Fall River Herald News Publishing Company, et. al., 5 FCC 483 (1939). In the Matter of James E. Davidson, 4 FCC 594 (1937). In the Matter of Wichita Falls Broadcasting Company, 3 FCC 386 (1936). KGMO Radio-Television, Inc. v. FCC, 119 U.S. App. D.C. 1 (1964). Lenawee Broadcasting Co., 48 FCC 2d.1181 (1974). Missouri-Illinois Broadcasting Co., 3 R.R. 2d.232 (1964). Rhinelander Television Cable Corp., 37 FCC 1071 (1964). Southwestern Operating Co. v. FCC, 122 U.S. App. D.C. 137 (1965). Tri-Cities Broadcasting Co., 24 R.R. 691 (1962). WLVA V. FCC, 459 F. 2d. 1286 (1972).