VIEWS: 25 PAGES: 14 POSTED ON: 10/3/2011 Public Domain
DETERMINING TELEVISION ADVERTISING RATES By Benjamin J. Bates Paper presented at the 33rd International Communication ABSTRACT Association conference, Dallas TX, May 1983. Previous work has identified various factors as contributing to the setting of television rates, without giving indication or empirical verification of the A revised version was published as: Bates, Benjamin J. "Determining Television Advertising manner or size of such effects. This study proposes that Rates." In R. N. Bostrom (Ed.), Communication such quantification of effects upon television spot advertising Yearbook 7 (pp. 462-475). Beverly Hills: Sage, 1983. rates can be achieved through the development of a predictive model for such rates over time. Using multiple observations from the period 1973-1981, the Contact Info: development of such a model provided strong evidence that the Benjamin J. Bates identified factors, and the factor of inflation, did affect School of Journalism & Electronic Media the setting of spot rates. In addition, the examination of University of Tennessee rates over time permitted the application of analysis of Knoxville, TN 37996-0333 covariance techniques which indicated that the effects of bjbates@utk.edu certain factors had significantly changed over that period. DETERMINING TELEVISION ADVERTISING RATES the station's broadcast frequency. While providing empirical evidence of effects, the work was limited in that the model did Commercial television stations are, for the most part, not directly consider advertising spot rates and was limited to supported through the sale of a part of their broadcast the examination of a single year. French & McBrayer (1979), schedule to advertisers. This time is sold primarily in the in a qualitative article looking at the factors which determine form of short spots interspersed among the stations' station spot rates, found local market rates to be strongly entertainment programming. As the supply of these spots is influenced by three major factors: demand, competition, and fairly constant, the rates which stations charge for these spots ratings. And demand, they stated, was largely determined by in their schedule are quite dependent upon the demand among the size of the market and local economic conditions. advertisers for those spots. Since the aim of advertising is to This study will involve the empirical verification and convey a message to an audience effectively, and television is measurement of those factors cited in the previous studies, essentially a medium to reach an audience, the demand for namely the size of the market, the quality of the market, local advertising on a station will be dependent upon the audience economic conditions, the direct competition from other that advertising message reaches. commercial stations in the market, differences in station That is, since advertisers are actually purchasing access to coverage, the network affiliation (if any) of the station, and the an audience, it is expected that the price that advertisers would broadcast band in which the station operates. It will refine the be willing to pay for that spot will depend upon that audience. cited factors where appropriate, define reasonable measures, Therefore the price that stations are able to get for their spots and then examine the possible effects of those measured factors will depend on the audience that stations can attract for those upon the rates for television spot advertising. spots. In previous studies over the last fifteen years, a number It is not the goal of this project to provide a deterministic of audience factors have been identified as contributing to the model of the rate setting process and proclaim its validity. The determination of prices for broadcast time on television. actual manner in which rates are set are too indistinct, relying In 1967, W. T. Kelley undertook a survey of broadcast more upon instinct and response to market forces rather than managers in the Philadelphia area, and found that market size specific models or formulae. However, such processes are and quality, station coverage, and competition were all major likely to involve non-concrete consideration of certain key factors which were considered in the rate-setting process. factors. By modeling the apparent relationship between these Kelley did not, however, provide any evidence or proposals as factors and rates, it is possible to provide hard evidence in to the manner in which these factors contributed to the support of the role such factors play in the determination of determination of rates. In an 1976 article, S. M. Besen advertising rates. The goal of this research is to provide such attempted to fill this gap in part by deriving an empirical model evidence. for the value of television time. Using pure, or block, time METHOD rates as the dependent variable, Besen's model included As the factors that this study addresses are considered to be measures of market size, competition, network affiliation, and determinants of spot rates, it was decided to base empirical analysis upon predictive modeling procedures. These would particular point in time were used, and were noted in primarily involve the development of linear or essentially conjunction with that date. The associated date was then used linear models using a combination of real and categorical to match the rate information with other dated measures. (indicator) variables. As the various studies made no proposals A second dependent variable was constructed by adjusting or predictions as to the nature of effects, this modeling process the rate data for the effects of inflation. The adjustment was would not be directed towards the determination of a single made through the use of the Consumer Price Index (as predictive model for advertising spot rates. Rather, through the advertising spots were considered to be finished goods) in an model building process and the consideration of alternative attempt to render the rates collected over time comparable. As models, those factors having a general significant impact can will be pointed out in the analysis section, this attempt was be identified, and general indicators of the particular effects only partially successful. determined. The appearance of effects of similar size and type The measurement of some of the cited factors as across models would, in fact, reinforce the significance and independent variables was quite straightforward. Reports of the validity of a factor in the setting of rates. network affiliation and broadcast frequency of stations were It was also decided to model this process over a period of obtained from the TV Factbook and Broadcasting Yearbook. time. Consideration of the factors and rates over time would A simple nominal measure was constructed to indicate whether allow for the reliable examination of the general significance of a particular station's frequency, or channel, assignment was in the factors without restricting the validity of the procedure and the UHF or the VHF bands. A nominal measure was also the results to a single period. In addition, the presence of constructed to indicate the primary network affiliation of the multiple observations from a single station could allow for the station, if any. direct consideration of a factor's effect as well as reflect the A consideration of the factors cited by the three studies possible presence of additional explanatory factors. Further, reveals the need for the refinement or addition of another the examining of rates over time would allow for consideration factor, as the size of the audience for a message or station is not of possible changes in the size or significance of the various directly considered. The factor of audience size would seem to factors' effects over time, and the possible identification of be the single most important factor in differentiating audiences, trends. and thus the demand for broadcast spots, yet it is not directly The measures used for this analysis were constructed to take cited in those three previous studies. It would appear, advantage of data which are generally available. For the basic however, that indirect consideration was given to audience size dependent factor of rate, or the price of television spots, it was in the inclusion of the market size factor, though reliance on decided to use the highest rate quoted for a thirty second (:30) that measure alone presumes that all stations in the market television spot, as reported in the annual editions of either the cover that market both fully and equally well. Both of TV Factbook or the Broadcasting Yearbook. As these rates these assumptions are generally suspect, leaving an audience were to be collected and analyzed over a period of time, only size measure of questionable reliability and validity. Some of those rates which could be reliably determined to be valid at a the studies attempted to rectify this condition with the added consideration of the factor of station coverage differences, provide competition for broadcast spots, although they did removing major problems, but still leaving the consideration of provide some competition for audience. The station's audience size to indirect means. competition in the market, however, will also affect audience Thus, it would seem appropriate to include the factor of size, as it reflects the number of stations which must share the audience size among the potential factors affecting television potential audience. That is, the actual audience for a station spot rates. This factor's inclusion in the analysis is not only will be a share of potential audience, and that share is theoretically indicated, but it should also provide for a more determined in large part by the competition a station faces. reliable analysis and relieve any confounding of those factors This leaves only the rather ill-defined factors of market which vary with station coverage. quality and local economic conditions which had been cited It was decided to measure audience size through the use of in earlier studies. One aspect of both factors which should the Average Daily Circulation (ADC) for the station for a also be of particular interest to potential advertisers would be given season, as measured by Arbitron and reported in the TV the wealth of the market; the ability of those in the audience Factbook. While not giving a specific measure of absolute to actually purchase the advertised goods or services. One audience at any given time, the ADC is a reasonably valid widely available measure of this ability to buy is the and reliable indicator of the number of households who can, Effective Buying Income (EBI) measure, as developed and and do, watch that station regularly over a period of time. reported by Sales & Marketing Management, Inc., in their Thus, it can be considered a good measure of the potential (or annual Survey of Buying Power. For this study, a scaled, likely) audience for any station. relative measure based upon the median market EBI was The measurement of several other factors depended to a constructed. This market quality measure segmented the certain extent upon the specification of television broadcast median market EBI as to whether it was, in comparison with markets. For the purposes of this study, it was decided to base the national median, over 20% above, between 5-20% above, market definitions upon those used by Arbitron, as reported in within 5%, between 5-20% below, or over 20% below. the Broadcasting Yearbook, only treating Arbitron's A second measure of market quality was included, in the "supplemental" markets as separate markets. From this basis, form of the forecast growth rate for the market. These the factor of market size was measured relatively by the rank of forecasts, in the form of the projected growth in households in the market as reported by Arbitron for a season in the middle of a market, were also obtained from the annual Survey of Buying the data collection period, resulting in an ordinal measure. Power volumes, and the rates scaled and segmented in the Unranked markets were arbitrarily assigned ranks falling below same manner as the EBI measure. It should be noted that the the lowest of the ranked markets. precise market definitions used in the Survey of Buying Power The factor of direct competition from other stations was measures did not always precisely match those of Arbitron, measured by the number of commercial television stations although in all cases the major population center(s) were in licensed and operating in the market at the given date. accord. Noncommercial stations were not included as they did not All measures were obtained on an annual basis where needed to match the dependent variables obtained for the dependent variables, tending to confirm the importance commercial stations within the continental United States over and significance of this factor in the setting of television spot the period 1974-1981. It should be noted, however, that valid rates. High correlations were also found for the initial rates were obtained from these sources for earlier periods, and variables measuring the factors of market size (MKT), station those rates reported for 1973 were also included, along with the competition (COMP), the date of the observed rate (YEAR), corresponding measures of the independent variables. An broadcast band (UHF), and one of the market quality additional variable was then constructed to denote and label the measures (EBI). It should also be noted that there was a stations for which data were collected and included for significant amount of cross-correlation among the analysis. The resulting data set was then restricted to those independent variables as well. data from stations for which at least four observations of rates [Table 1 about here] were reported over the collection period. This yielded a set of The high correlation of audience size with rates 856 observations from a total of 164 stations for preliminary suggested the initial appropriateness of regression analysis. Later, the data set was expanded to include procedures in the modelling process. As there was no observations from other stations in the same markets as those initial reason to assume the inherent linearity of the in the initial set of observations. This second, expanded data relationship between audience size and spot rates, it was set included a total of 1068 observations from 232 stations. decided to examine a variety of essentially linear ANALYSIS regression models. An examination of the plot of RATE The analysis for the significance of the cited factors began vs. ADC indicated the likelihood of an exponential as an exploratory data analysis leading to the development of relationship. To include the consideration of these types predictive models for spot rates. First, the data sets were of relationship in the modeling process, variables for the examined to identify the significant predictive factors for natural logarithms of RATE (LNRATE), ADJRATE television spot rates. Note was taken of the different types of (LNADJ), and ADC (LNADC), and the computation of measures used to derive the various independent variables, and regressions involved the original as well as transformed those initial measures were treated or transformed where variables. appropriate to fit the kinds of effects or relationships revealed These regressions, which are summarized in Table 2, in these exploratory procedures. indicated that the regression procedures "explained" about two- The analysis began with an examination of the cross- thirds of the variation in spot rates between stations and over correlation matrix among the ten variables. Several of the time. These regressions also suggested the relative validity of measures evidenced a correlation with the dependent the adjustment for inflation and the appropriateness of the variables of spot rates (RATE) and the spot rates adjusted for logarithmic transformations, in that an examination of the R2 inflation (ADJRATE). Those correlations greater than 0.1 are statistic indicated that the models employing those adjustments listed by factor in Table 1. A highly significant correlation provided a better fit to the data than the models without the was evidenced between the audience size measure (ADC) and adjusted variables. The validity of these "essentially linear" models was further supported by the analyses of the residuals seemed invalid, particularly in comparison with the resulting from these regressions, as those residuals evidenced multiplicative effects presumed in the other models. less systematic effects. [Table 3 about here] [Table 2 about here] It seemed that a fairly firm basis had been established for While the use of the audience size measure in essentially the determination and development of predictive models for linear models "explained" a great deal of the variation in rates, spot rates in television. At this point, the remaining factor it also left some to be explained by other factors. The measures (such as affiliation, competition, and market size and residuals of these regressions were examined, and all showed quality) were molded, where necessary, into a set of indicator a high correlation with the YEAR variable, even those coming variables which would demonstrate their significance to the from models using the adjusted rates as dependent variables. model, and thus their role in the determination of rates. The This correlation with time, in fact, was in all cases higher than precise derivation of the indicator variables for the nominal the correlation with any other single factor. and ordinal measures was accomplished through an iterative This finding was somewhat of a surprise, as it was procedure of model-building and residual analysis. surmised that correcting for inflation would remove any such This procedure involved the construction of models which effects in models using the adjusted rates. The manner in included the factor in question, then fitting the same model which the YEAR variable appeared to contribute to the without that factor. Examination of the resulting residuals was determination of price indicated, however, that while adjusting then used to indicate the manner in which the factor under for the effects of inflation was appropriate, the Consumer development contributed to the original predictive model. Price Index was not a good measure of inflation in television Separate indicator variables were then constructed to mirror spot advertising rates. that effect. In addition, it should be noted that as factors were Thus, the factor of time, as measured by the YEAR converted to indicator variables, the appropriateness of those variable, was entered into all of the developing models, constructed variables were checked by repeating the above resulting in a series of multiple regressions which are process. summarized in Table 3. All the resulting linear models proved After the iterative procedures were used to develop the to be significant predictors of rate, both when considered in appropriate variables, a series of stepwise regression toto and by variable. The addition of the YEAR variable procedures was used to identify the significant indicator increased in all cases the predictive power of the model, variables within the various model permutations. With the resulting in one case of a model which explained over 80 focus on the transformed, multiplicative, models, the effects of percent of the variation in rates. At this time, it was decided to the individual indicator variables upon the prediction of rates remove the linear, additive, models from further analysis. Not were assumed to be multiplicative, resulting in what could be only was the inappropriateness of these models suggested in considered to be either a bonus or a discount associated with a the analyses of residuals, but the presumption of constant certain state of the basic factor. As the procedure involved the additive effects for the factors which the model implied also consideration of multiple models, both a forward and backwards selection procedure was instituted in the stepwise measure in the models. This was not totally unexpected, as the regressions, resulting in the consideration of eight separate use of transformations placed differing emphases on certain linear models, with multiple R statistics running between ranges of audience size, and the ADC and market size factors .83510 and .92080. A brief summary of these models, fitted to were highly correlated. Still, there was evidence of a separate the data set of 1068 observations, is given in Table 4. effect accruing to the market size factor, although for one [Table 4 about here] treatment that effect was of marginal significance. The effects of station affiliation were apparent in all eight The factor of station competition was also found to have models. In all cases, independent stations were found to have significant effects towards the prediction of spot rates, though significantly lower station rates than other stations. Rates for the size of these effects also varied somewhat with the independent stations were about half of those for network particular treatment of the audience size measure. In affiliated stations with similar audiences. In addition, spot rates particular, a bonus of around 20 percent for all stations in for NBC affiliates were found to be uniformly about 10% markets having at least four stations was indicated, although lower than the rates for affiliates of the other two major no statistically significant difference was noted for markets of networks, when all other factors were taken into account. No four, five, six, or more stations. That is, although rates were significant difference was found between ABC and CBS generally higher in those markets with at least four competing affiliation in any of the models. An indicator variable was also commercial television stations, no additional increase was used to distinguish the affiliates of the Spanish International then noted for the addition of other stations to the market. This Network (SIN), and this variable was found to make a result was fairly uniform across models. Differences in station significant contribution to the prediction of spot rates in all but rates were also noted between stations with one, two, and two of the models. Also, SIN affiliate rates were significantly three competing stations, although the size of these effects different from those for the other network audience (ceteris evidenced greater variation. This, again, was not totally paribus) for half of the models, and only marginally significant unexpected, as the number of stations in a market is highly with opposite effects for the other half. This variability was correlated with market and audience size. possibly due to the small number of observations from SIN The two measures of market quality also proved to make affiliates, and the limitation of the SIN affiliates to the largest significant contributions to the predictive models, which markets. indicated the presence of some effect upon station spot rates, Significant effects for the factor of market size were found although the precise nature of that effect varied from model in all but two of the predictive models. The market size to model. In all cases, however, the basic effect indicated measure was segmented into three groups through the use of was that stations in markets with higher EBI or higher indicator variables: the top twenty five markets, the next growth rates were able to get higher prices for their spots, twenty five markets (ranks 26-50), and all other markets. The with effects in the range of ten percent. size, direction, and significance of these indicator variables Finally, examination of these linear models indicated that differed greatly for the two treatments of the audience size stations in the VHF band were able to charge higher rates for their spots than stations situated in the UHF band, but this indicated that that portion of the model remaining was possibly bonus was uniformly less than ten percent. Thus, the not uniform over the period. That is, the statistics indicated the consideration of the multiple models confirmed the likelihood that some or all of the modeled factors' effects may significance of the cited factors in the determination of have changed over time. The analysis of covariance procedure television spot rates, although the effects did not appear to be used provided for, along with the general model estimation, the uniform across models for all factors. fitting of separate models for each of the covariate values. This For a final estimation of the specific size of the effects, it provided estimates of coefficients for each year. was necessary to pick out a specific predictive model. On the Several interesting developments emerged from a basis of predictive power and other indicators of fit such as consideration of these estimated coefficients. First, it appeared residual analysis, it appeared that the best predictive model was that there were statistically significant differences among the that using LNRATE as the dependent variable and LNADC as fitted coefficients for most of the factors, in the sense that the primary independent variable. The specific estimations of either some coefficients could not be statistically distinguished coefficients, and thus effects, are given in Table 5, along with from zero (indicating the possibility of no effect), or that one or the appropriate statistical measures. more of the estimates greatly varied from other years' [Table 5 about here] coefficients for that variable. While some of this may have As a further indication of the type of effects evidenced by resulted in part from the small sample size of some component these factors, a series of analyses of covariance models was factors, enough regularity in the pattern of estimates remained also derived from the enlarged data set. These models allow in some cases to indicate some interesting relationships in the for the examination of the fitted models on a year by year model. basis, and allow the consideration of the uniformity of effects First, it was noted that the effect of the broadcast band over time. As the relatively small number of observations from factor effectively disappeared after 1976. Uniform, significant certain years made estimation impossible due to problems of coefficients were estimated for the corresponding indicator singularity, the analysis was restricted to the period from 1975 variable for 1975 and 1976, but in all following years the fitted to 1979, inclusive, using the expanded data set of 1068 coefficients were an order of magnitude lower, and could not observations. With the restriction, the data set for analysis was be reliably distinguished from zero. Thus there was no reduced to 812 observations. Further, the consideration of the evidence that a station's frequency had an effect upon the analysis of covariance findings will be restricted to the determination of its rates after 1976. LNRATE-LNADC model which was selected as the "best" In a similar vein, there seemed to be a change in the size of predictive model for television spot rates. the effects attributable to station competition over time. The basic analysis of covariance table yielded several test Specifically, the coefficients for the indicator variable statistics which indicated the validity of this approach. While reflecting those markets large enough to field an independent these statistics indicated the general significance of the station (i.e. four or more stations) evidenced two distinct predictive model and of the covariate YEAR, they also levels. There appeared to be a sizeable effect for that factor until 1978, when the effect was about half that estimated for basis for that verification. While not specifically implying the previous three years. causation in the direction of factors to rates, the unlikelihood of The analysis of covariance also indicated an interesting rates affecting factors such as broadcast frequency, audience elaboration of the effect of station affiliation with NBC which size, and station affiliation infer an asymmetrical relationship had been noted in the earlier models. The results of this in the other direction. That is, the presence of such an effective, procedure indicated that there seemed to be no significant useful, predictive model supports the claim that the factors difference in rates among network affiliated stations until 1977, represented by the variables in the model do have some sort of when the "discounting" effects of NBC affiliation began. In effect upon the setting of television advertising rates. fact, the coefficient for the NBC indicator variable showed In the general models, some effect was indicated for all of steady increase from that point, indicating a worsening of the the factors mentioned in the earlier studies by Kelley(1967), relative position of NBC affiliates. This, it should be noted, Besen(1976), and French & McBrayer(1979). Specifically, the coincided with the dramatic fall of the effectiveness of NBC's empirical study of spot rates indicated the following general prime time programming in attracting audience and the and specific effects. resultant fall of NBC to a distant third in the ratings race. The factor with the greatest effect upon the determination Another regularity which was noted was the overall of spot rates in television was the audience size. This effect consistency of the the market quality measure EBI. While the was strong, and was fairly consistent, explaining on its own individual coefficients showed a fair amount of variability, the roughly two thirds of the variation in rates between stations overall effect noted from high to low values remained fairly over time. The second most important factor found in the consistent. The consistency of the other market quality study, time, was not mentioned specifically in the earlier measure, household growth, was not as evident. research. It is a factor, however, that would only show up in a The coefficients for the other indicator variables, and the consideration over time. The interesting result of this analysis real audience measure LNADC, remained fairly consistent. As was not only the affirmation of the importance of considering no other general pattern in the estimated coefficients emerged, inflation over time, but the models' indication that the rate of they were judged to have had consistent effects over the period inflation in television spot advertising rates was roughly twice 1975-1979. that indicated by the Consumer Price Index over the same period. Inflation in spot rates, the model estimated, ran about CONCLUSIONS 19 percent a year over the period for which data were The goal of this research was to provide quantitative collected. verification of the role of certain specified factors in the The remaining contributions to the predictive models were determination of television spot advertising rates. The made by indicator variables. It should be noted that the development of a predictive model accounting for about 85 possible confounding of market size with a number of the percent of the variation in spot ad rates over a period of eight indicator variables, such as those for the factors of competition years for a group of 232 stations across the U. S. serves as the and market quality, acts to restrict the kind of conclusions which can be drawn. In addition, the differing strengths of the appearances, this is not an unexpected finding. The audience contributions to the predictive model made by the indicator size measure is more a measure of potential audience than variables for separate aspects of factors make any specific actual audience, so the amount of competition will then statements about relative strengths of those factors' impacts indicate how close potential and actual audience is to be for a difficult. station. That is, the ADC measure for a station with no direct Having mentioned the appropriate caveats, the following competition (i.e., located in a single station market) is apt to conclusions and interpretations are offered. The factor of match the actual viewing audience for that station, while the station affiliation proved to be fairly significant in the actual viewing audience for a station with competition is likely determination of station rates, particularly in the distinction to be only a fraction of the ADC (potential audience) measure. between network and non-network affiliates. This may have Thus, the pattern of discounts and bonuses for the set of resulted in part from the relative accuracy of the audience size indicator variables is not unexpected. The effects measure, ADC, as an indicator of prime time audience for demonstrated through these independent variables may reflect stations. Traditionally, and for most markets, independent the differences between a station's potential audience and their stations are not equal competitors with network stations during actual share of prime time audience. that time, as the strength of independent stations in attracting It should be noted that for the final predictive model, the an audience lies in periods outside normal prime-time, discount for the single station market was actually less than whereas prime time spots usually yield the highest rates. Thus, that for the two station market, indicating that there was likely the audience size measure upon which rates are predominantly to be some monopoly benefit accruing to stations without based is apt to be skewed for independent stations as a result competition. However, the nature of the variables and of their apparent inability to compete on an equal basis with measures used did not allow for a precise separation of effects. network affiliates during prime time. It should further be noted that the number of stations in a This may also explain the finding of the growing difference market is confounded with market size, further clouding the in rates for NBC affiliates. The models indicated that, over the precise nature of this effect. sample period, NBC affiliates generally earned less than ABC Lesser effects were noted for the two market quality or CBS affiliates. In fact, the analysis of covariance indicated measures. Stations in markets with higher than average that this was a growing difference over the period from 1975 to Effective Buying Income measures tended to have higher rates 1979, roughly matching that period when NBC was not than stations in markets with lower than average measures. The competing on an equal basis during prime time impact upon bonus accruing to those stations ranged between seven and ten station rates, at least when measured in terms of the number of percent in the various predictive models. As for the growth commercial television stations licensed to a market. In general, measure of market quality, that also showed some effect in the the models indicated that the more competition, the higher the predicted direction, with stations in markets with significantly rates in a market, at least to the point where there were at least higher growth rates (20 percent above the national rate) having four commercial stations in the market. Despite initial about a ten percent higher rates than stations in other markets. It should be noted that these differences showed up only in restrictive measurement of market size. those markets with extreme differences in growth rates from In summation, it appears reasonable to reach the following the national average, indicating that this measure is likely conclusion from this analysis. First, that the factors of considered only subjectively in the determination of rates. audience and market size, station competition, station There did appear to be a UHF/VHF difference in the affiliation, broadcast band, and market quality/local economic rates charged by stations for spot advertising. This effect conditions did have some kind of influence in the was small, however, amounting to only about a seven determination of television spot advertising rates. A listing of percent difference, and did not contribute a great deal to these factors, ranked by their relative importance of their the explanatory power of the models. In fact, the contribution, is given in Table 6. Second, that there has been separate analysis of covariance indicated that this effect inflation in spot rates, and the rate of that inflation has been seems to have disappeared after 1976. The about twice that of the CPI. And last, but not necessarily least, disappearance of effect can possibly be traced to the rise that the effects of factors upon rates need not be, and have not of cable systems in fringe areas, where the signal been, constant; that changing conditions may diminish or even strength differences (resulting in differing picture invalidate the significance of specific factors, or bring forth quality) are greatest, and the resultant diminishing of new factors to consider in the determination of rates. qualitative differences between UHF and VHF stations [Table 6 about here] which might result in differences in audience and thus the rates stations can charge for their advertising spots. Finally, there is the consideration of the market size, which was presumed to have been largely dismissed by the BIBLIOGRAPHY consideration of the audience size. The development of the predictive models did indicate directly (and possibly indirectly) Besen, S. M. "The Value of Television Time." the presence of a market size factor above and beyond audience Southern Economic Journal, 42:435-441 (1976) size. The precise nature of the market size effect indicated by Broadcasting. Yearbook. (annual) the model, however, was dependent upon the nature of the French, W. A., and McBrayer, J. T. "How Television Stations relationship assumed between audience size and rates in the Price Their Service." Journal of Advertising, 8:15-18 particular model which was fitted (that relationship being (1979) determined by the transformations used). Such differences in Kelley, W. J. "How Television Stations Price Their effects, while reaffirming the link between audience and Service." Journal of Broadcasting, 11:313-323 market size, also indicates a differential effect at various levels (1967) of market/audience size. Should this area be studied further, Sales and Marketing Management, Inc. Survey of Buying this is one aspect which should be further investigated, as this Power. (annual) procedure was hampered in that regard by a possibly too Television Digest, Inc. Television Factbook. (annual) Table 1. Correlations Correlation with Variable RATE ADJRATE ADC .7832 .8502 MKT -.4950 -.5530 COMP .4558 .4909 YEAR .2984 .1936 EBI -.2184 -.2421 UHF .1387 .1574 Note: all the above correlations were significant at the p=.01 level Table 2. Initial Regressions Variables R-square Rate Audience RATE ADC .61352 ADJRATE ADC .72288 LNRATE ADC .63199 LNADJ ADC .69935 LNRATE LNADC .66762 LNADJ LNADC .73723 Note: the F-test statistic for all the regressions indicated their significance at a level of p<.001 Table 3. Multiple Regressions Variables R-square Dependent Independent RATE ADC, YEAR .67428 ADJRATE ADC, YEAR .74164 LNRATE ADC, YEAR .78342 LNADJ ADC, YEAR .74717 LNRATE LNADC, YEAR .80957 LNADJ LNADC, YEAR .77947 Note: the F-test statistics for all regressions indicated their significance at a level of p<.001 Table 4. Predictive Models Variablesa Forward Selection Backwards Selection 2 Rate Audience Mult R R Mult R R2 LNRATE LNADC .92049 .84730 .92078 .84783 LNRATE ADC .90951 .82720 .91041 .82884 LNADJ LNADC .90871 .82576 .90910 .82646 LNADJ ADC .89588 .80200 .89689 .80441 Note: the F-test statistics for all models indicated their significance at a level of p<.001 a The stepwise procedure also selected from the full set of indicator variables and the variable YEAR in all cases Table 5. Estimates of Effectsa Factor Variable Partial Coefficient T-Stat Constant -.49926 -1.420 Audience LNADC .70111 .72056 31.906+++ Inflation YEAR .68332 .17264 30.370+++ Bdcst Band UHF' .07210 .07094 2.346+ Network INDb .37406 .79780 13.089+++ Affiliation SINb .10242 .45169 3.341+++ NBCb .12431 .10662 4.0654++ Competition ONESTNb .11455 .16134 3.742+" TWOSTN .19065 .24874 6.302+++ FOUR+- -.14904 -.19519 -4.891+++ Mkt Size TOP25b -.29171 -.63604 -9.900+++ NEXT4 -.12483 -.18446 -4.023+++ Mkt Quality HIEBI' -.07108 -.072008 -2.312+ LOEBI .05862 .05337 1.905 HIHHGROb -.06938 -.054387 -2.2574 a + p<.05 " p<.01 +++ p<.001 With LNRATE as dependent variable, using the backwards selection procedure b Due to the requirements of the particular statistical package used, these indicator variables were given a value of 1 if true, and 2 if false (i.e. not having the indicated value) Table 6.Factors Affecting the Determination of Rates (ranked in order of relative importance) Factor Variables 1 Audience Size ADC, LNADC 2 Time (Inflation) YEAR 3 Network Affiliation IND, SIN, NBC 4 Market Size TOP25, NEXT25 5 Competition ONESTN, TWOSTN, FOUR+ 6 Market Quality HIEBI, LOEBI, HIHHGRO 7 Broadcast Band UHF