THE 1MPACT OF TIME ON MODELS OF TELEVISION SPOT PRICES. by Benjamin J. Bates

Transcription

1 THE 1MPACT OF TIME ON MODELS OF TELEVISION SPOT PRICES by Benjamin J. Bates presented at the 37th annual Conference of the International Communication Association Montreal, CANADA, May, 1987 Contact Information: Benjamin J. Bates School of Journalism & Electronic Media University of Tennessee Knoxville, TN Copyright, 1987, Benjamin J. Bates. All Rights Reserved ABSTRACT There have been a number of studies in the last twenty years which have derived models of the value, or price, of television broadcast time, in order to determine the influence of various factors. For the most part, the models have been fitted to a single data set collected at a set time. there have been a couple of studies, though, that have suggested that time itself may have an influence on prices and values. This study uses data covering the years to consider several aspects of time's potential impact on prices: the direct effects of time, such as inflation; the influence of factors which only appear at certain times, such as Presidential Election Campaigns; and indirect effects, in which the influence of other factors varies over time. The study finds strong evidence of both the direct and indirect influences of time, although the impact of Presidential Election Campaigns is less clear. The study also found that there are distinctive models for UHF and VHF stations, and that several factors have distinctly different influences for those two classes of stations.

2 THE IMPACT OF TIME ON MODELS OF TELEVISION SPOT PRICES The economics of broadcasting, for the most part, amounts to the study of the supply, demand, and value of that good which finances the broadcasting industry in the United States: the time sold for commercial advertising. The factors which influence or effect the supply or demand of that product (broadcast commercial time), and hence influence its value or price, also influence the behavior of broadcasters as profitseeking organizations. The identification of those factors and the consideration of their patterns of influence are therefore fundamental to the understanding of the American broadcasting system from an economic perspective. There have been a number of different studies which have examined the factors thought to influence the determination the value of broadcasting time, particularly in television (cf Kelley, 1967; Cherington, Hirsch & Brandwein, 1971; Peterman, 1971; Bowman, 1975; Besen, 1976, 1978; Larson, 1976; White, 1977; Park, 1978; French and McBrayer, 1979; Levin, 1980; McFadyen, Hoskins & Gillen, 1980; Bates, 1983; Fournier and Martin, 1983; Wirth and Wollert, 1984). These studies have identified and/or measured a number of different factors influencing the value of television time. The factors which these studies indentified tend to fall into one of three basic types: factors related to the size of the audience for that time; factors related to attributes of the market in which the station operates; and factors related to specific station characteristics, such as network affiliation or operation on the UHF band. These studies have generally done a good job of considering the impact of specific factors, and have resulted in a series of models attempting to measure such impact. With some minor exceptions, though, these studies have sought to consider only general influences of specific factors at specific times. For the most part, they have not specifically considered the influence of time, either directly or indirectly through the possibility that the impact of other factors might well change over time (Bates (1983) is an exception). Nor have most considered the possibility that impacts might differ across stations or markets (here Park (1978) is a notable exception). There is some evidence that the influences of various economic and market factors might not be constant, either over time or across stations. Conditions, and thus influences, are likely to change over time, or might be different for different classes of stations. For example, Park (1971, 1972) argued that the impact of the diffusion of cable would differ for UHF and VHF stations, leading to a diminution of the "UHF handicap" over time. Further, Park (1978) later argued that a study by Besen (1976) had erred in not treating UHF and VHF stations separately in fitting his model of television time rates. A study by Bates (1983) indicated that not only was time, in the form of inflation, itself an influential factor, but that there was some evidence that the size of the impact of several other factors seemed to change over time. Along this vein, Krasnow, Longley & Terry (1982) have argued that changing market conditions and FCC policies have acted to reduce any "UHF handicap" in recent years. These findings suggest, at least, the possibility that the economic influence of some factors may be variable, either over time or across station types, and thus warrants more formal investigation of that possibility. If the earlier studies had utilized similar measures or samples in their models, at least some of the questions regarding variation in the strength of some factor's influence over time could be addressed through an examination of their results. Regrettably, there is little similarity in samples,

3 measures, or models throughout these earlier studies. On the other hand, the samples do not differ enough (or in the appropriate manner) to address the possibility that influence might not be consistent across stations. is a strong need for additional research which specifically addresses the possibility of the variable nature of economic influence in the determination of value and prices in commercial television broadcasting. The goal of this study is to attempt to bypass the limitations of earlier research and meet this need by deriving and considering a general model of television spot prices both over time and across stations and markets. This will permit the analysis of the influence of time as a factor in its own right as well as its indirect impact on the patterns of influence, over time, for other factors which had been identified as influencing the marketplace for television broadcast time. In addition, the consistency of impact across station types can be considered. THEORY This area of research has always been, in the strictest sense, theory-free. What has generally passed for a "theory" of influence has been a hypothetical development as to the direction of the influence of specific factors, usually based upon the underlying assumption that price is directly and positively related to the size of the (expected) audience for the broadcast time sold. Faking this assumption several steps backwards, one can see that it is somewhat based in the fundamental economic theory of supply and demand, under the further assumptions that supply is inelastic, that is, relatively constant, and that the demand for broadcast time is positively linked to the (expected) audience for such time. As the essence of broadcast time as an economic good is the access to an audience which it provides, the assumption of a positive link between the size of the audience for television time and its desirability as a good can be considered reasonable. And it can be argued (Bates, 1985, 1987) that various structural constraints hold the supply of broadcast time relatively constant, There at least in the short term. There is some indirect support for this argument in Bowman's (1975) measurement of supply and demand curves for TV time, which yielded fitted curves consistent the concept that the supply of such time is relatively inelastic. With these assumptions, economic theory predicts that, other things being equal, an increase in demand (an increase in audience size) will result in a higher price for the product, thus lending some credence to the basic assumption that price is positively linked to audience size. This was essentially the scenario utilized for all of the nontime-related factors included in this study. Earlier studies (see above) had used such "theory" to predict that audience and market size had a positive impact upon spot rates, that levels of cable penetration and the number of competing stations had a negative impact (as they increased levels of competition for a given audience within a market), and that operation on the UHF band and/or without a network station affiliation "handicapped" stations (had a negative impact) by reducing their capacity to attract an audience vis-à-vis other stations; predictions which were confirmed by earlier empirical studies. One other factor which had been included in some earlier studies was economic concentration or cross-ownership, although the results of those studies indicated that such factors had limited, or mixed, results (cf. Levin, 1980; McFadyen, Hoskins and Gillen, 1980; Wirth and Wollert, 1984); therefore that factor was not included in this study. However, as time had yet to be fully considered as an influential factor within most of these studies, a "theory" of time's influence on the price of television time needs to be

4 developed. As indicated above, the influence of time may be either direct or indirect. First, time may itself be considered as a factor, or more appropriately, time may be considered as an indicator measure for other, let us say time-based, factors which seem to vary only over time. Inflation can be considered to be the classic example of a time-based factor whose impact can be seen by treating time as a direct factor. Inflation, as a time-based factor, is the aggregation of a number of economic forces which act to increase costs and prices over time. As such, the expectation is that inflation is expected to have a positive impact on prices, as was demonstrated in one previous study by Bates (1983), which found that inflation in television spot rates ran about twice that measured by the Consumer Price Index, even after correcting for possible shifts in other factors of influence. As inflation may seem too abstract a factor, another timebased factor was also examined in this study: the impact of Presidential Election Campaigns. As these campaigns occur only once every four years, their impact can rightfully be considered only by a comparison of influence over time, thus making it eminently suitable for the purposes of this study. These political campaigns create, periodically, a sizable amount of additional demand for broadcast advertising time (Porter, 1963; Abrams and Settle, 1977). Even though FCC policy mandates that such advertising be sold at a station's lowest prices, the fact that such time is sold means that the total supply of advertising time available for other, regular, potential purchasers is reduced. According to the theories of supply and demand, shrinking supply in the face of steady demand will result in increased prices for the remaining commercial spots. The presence of political advertising should, therefore, have a positive impact upon the price paid for television spots. In addition to such direct impacts, it was argued that time might impact models of television spot prices indirectly, through shifts in the patterns of influence for other factors. Conditions do change over time: certain stations might become more competitive as a result of FCC policy; new media and technologies might divert some of the audience from broadcast television; advertiser needs and audience tastes may even change. Park (1971, 1972), for example, argued that the impact of cable penetration upon television station audiences would lessen over time, and would be different for UHF and VHF stations. Or, as argued by Krasnow, Longley and Terry (1982), FCC policy promoting UHF and independent stations, such as the All-Channel Receiver Acts and regulations promoting independent program production and distribution, could have had the impact of reducing the "handicaps" facing such stations over time. Specifically, there would appear to be three potential indirect impacts of time which could be observed by this study. First, it could be argued that the impact of the UHF "handicap" was diminishing over time as a result of FCC policies encouraging UHF stations, such as the All-Channel Receiver Acts. Similarly, other policies promoting the production and distribution of programming outside of the networks has acted to make a wider range of programming available to independent stations, thus tending to make them better able to compete with network affiliates for audience. Thus, one could expect to see the indirect impact of time evidenced in the reduction of the UHF and Independent "handicaps" over time. Park (1971, 1972) considered the impact of cable television upon broadcast stations, and argued that the impact would depend upon the number of alternatives

5 offered by cable systems, and that there would be a difference in impact for UHF and VHF stations. Over the period of this study, the number of viable alternatives on cable systems has increased substantially, and should, therefore, have resulted in a higher level of negative impact over time for VHF stations, and reduce the initially positive impact upon the price of television time on UHF stations. Thus, "theory" would predict that the influence of cable penetration would be increasingly negative for VHF stations, and, while initially positive, increasingly negative for UHF stations. There seem to be, therefore, good reason to predict that time does influence the value of television broadcast time, both directly and indirectly. Theoretical considerations have identified several specific hypotheses which can be tested to demonstrate such influence: that there will be an inflationary increase in the price of television spots; that the additional demand posed by Presidential Election Campaigns will increase spot rates at certain times; that the negative impact of cable penetration will increase over time; and that increasing competitiveness will reduce the size of the so-called UHF and Independent "handicaps." It remains then, to consider such hypotheses. Further, the distinction which Park (1971, 1972) made in the relative impact of cable penetration for UHF and VHF stations provides a ready hypothesis which could be used to test for the consistency of influences across station types. MEASURES AND SOURCES This study expands upon previous research by examining the impact of a set of variables upon a basic measure of the value of television time over an extended period. Data were gathered, covering the period , from various published sources for all commercial television stations operating within the continental U.S. Stations in Alaska and Hawaii were excluded due to the peculiar nature of broadcasting markets, and the unavailability of certain key measures, in those isolated states. The period started in 1974 for two reasons: first, it was felt that the upheaval created by the banning of cigarette advertising would have settled by that time; also, several measures used were not available prior to At the time data were collected, the data for 1984 was the most recent information available. The primary form of broadcast time sold during that period was the 30-second spot, and the measure SPOT$ was defined as the highest price, in dollars, quoted for a 30-second spot in either the Television Factbook or the Broadcasting Yearbook. Such a measure is, regrettably, neither the most accurate or reliable measure of the value of television spots available. It was, however, the most valid measure which was available, and consistently defined, throughout the entire target period and does provide an indication of the level of value for such spots. In order to enhance its reliability and validity as a measure, only those prices which could be reliably determined to be valid at a particular point during a calendar year were used, and were noted in conjunction with that year (YEAR). There was, therefore, a significant amount of missing data for this basic dependent measure, which reduced the original census to a set of 3065 observations of the SPOT$ measure. As the basic measure of audience size, it was decided to use Arbitron's measure of Average Daily Circulation (ADC), measured in thousands of households, as it seemed to reflect the basic level of potential audience for any station. Like the SPOTS measure, it is not the most accurate measure of the actual audience for a specific spot, but it does give a reliable indication of the potential audience for the station. The ADC measures, as well as reports of network affiliation and the

6 broadcast frequency of stations were obtained from the annual TV Factbooks. For unknown reasons, the annual TV Factbooks did not report the 1979 or 1982 ADC measures; these missing values were replaced with the average of the adjacent years' ADC measures. Several market-based measures were included in the study. To maintain consistency in measures, all market definitions were based upon Arbitron's Areas of Dominant Interest (ADI). Yearly estimates of market size were obtained by using ADI ranking (MKTRANK) and the total number of television households, in thousands, in the market (MKTHH), once more as reported in the TV Factbook and the Broadcasting Yearbook. Three basic measures of market economic power, or size, were obtained from the annual "Surveys of Buying Power" published by Sales and Marketing Management. Estimates of total "Effective Buying Income" in terms of millions of dollars (MKTEBI), retail sales (MKTSALES), also in terms of millions of dollars, and an index measure termed "Buying Power Index" (BPI) were obtained for each recognized television market. Two additional measures were derived from these estimates, to provide measures of the relative economic status of markets. Measures of average market income (EBIHH) and average retail sales (SALESHH) in terms of thousands of dollars per household were created from the relevant aggregate measures. The factor of direct competition from other stations was measured by two VHF (VCOMP) commercial stations operating within the market. Indicator variables were created to identify the primary network affiliation of stations (ABC, CBS, NBC, SIN), if any (IND if not affiliated with any network), and whether the station operated on the UHF band (UHF). The "UHF handicap" to be measured by this last variable was said to derive primarily from a UHF station's reduced coverage, and the lack of UHF receivers in the market. These two components were also measured separately; the coverage variable (COVERAGE) created by dividing a station's ADC estimate by its estimated market size, while estimates of UHF receiver penetration (UHF%) were obtained from the various yearbooks. The factor of cable penetration within the market was measured using Arbitron's estimates of basic cable penetration in the market (CABLE%), again obtained from the various yearbooks. As noted earlier, one specific time-based factor was to be considered. Presidential election campaigns provide a significant additional source of demand for television time on a periodic basis. If the impact of this cyclical increase in demand is transitory, then its influence would be confounded with that of the basic time measure (YEAR), and must be looked for in a closer examination of that factor's measured impact. If, on the other hand, the impact on published prices is more permanent, separate measurement is possible through the use of separate indicator variable for the permanent impact of the 1976 (EL76), 1980 (EL80), and 1984 (EL84) Presidential election campaigns. Several caveats must be mentioned with respect to the measures and data used in the study. First, it should be noted that most of the measures used were not necessarily the most accurate which could be obtained. For example, the ADC measure does not measure true audience at any particular time, nor does the highest price quoted necessarily reflect the actual price of any particular spot. However, the necessity of obtaining measures for a large number of stations over an extended historical period forced the use of information which was both consistent and available. Further, as the primary purpose of this study was to consider patterns of influence rather than obtain precise predictions of final price, it was felt

7 that consistent measures of the level of factors would be acceptable for the basic purposes of this study. Second, it should be noted that virtually all of the variables used in the study were highly correlated with one another. Multicollinearity is a problem inherent in virtually all economic models employing more than two variables: economic systems, like social systems, are notoriously complex and interrelated. Similarly, a second potential methodological problem arises with the consideration of many economic models over time: autocorrelation. Again, the nature of the phenomenon under study strongly suggests the existence of a fair degree of autocorrelation may exist. To some extent, problems of multicollinearity and autocorrelation can be corrected and controlled, or at least attenuated, through the use of particular statistical procedures (which will be discussed below). For the most part, however, multicollinearity and autocorrelation are potential problems which must be acknowledged and accepted as a potential limitations rather than resolved. METHODS AND ANALYSIS A linear modeling approach was used to examine the ability of the various measures to explain variations in the price of television time. Consideration of the findings of those studies which reported consideration of multiple models (Bowman, 1975; Bates, 1983), and the models utilized in other studies, suggested that the best models for explaining variations in the price of television time were not strictly linear, but rather involved transformations of some of the basic variables of price and audience size. Examination of various combinations of the original and transformed variables suggested that the best basic models were based upon logarithmic transformations of price (LNSPOT$) as the dependent variable, as well as the logarithmic transformation of ADC (LNADC) as the basic independent variable. This model was also theoretically appealing, as it presumes a constant relationship between price and audience size, and includes most of the indicator variables as proportional influences (that is, impacts exist as a percentage of a base price rather than in constant dollar terms for all stations). The use of a double log model also tends to reduce the problems associated with multicollinearity. Stepwise regression procedures were then used to select from the full set of independent variables those which contributed significantly to the explanation of variation in the dependent variables reflecting the value of television time. One benefit of the use of the stepwise procedure is that, at each step, the procedure selects and fits the variable which best explains the residual variation in the dependent variable, after controlling for the influence of all other entered independent variables. That is, it selects and fits variables on the basis of their independent contribution to the model, thus avoiding the major problem with multicollinearity: the fact that variables are likely to have a degree of common influence which might influence the estimation of coefficients in statistical procedures. [Table 1 about here] The results of the stepwise procedure developing the model for the explanation of variations in spot prices are given in Table 1. As can be seen, audience size is the primary factor influencing the price of 30-second spots, explaining on its own approximately 59% of the variation in spot rates. The variable YEAR, reflecting various timebased factors, explains an additional 17.5% of the variation, while all other included factors, combined, explained about another 6% of the total variation. Several of those "other" factors included in the model were

8 of specific interest. First, it should be noted that of the three Presidential Election Campaign variables, only two were selected for the model, and only one of those (EL76) demonstrated influence in the predicted direction. The contribution of another (EL84) was marginal and in the wrong direction, while the third variable (EL80) appeared to contribute nothing to the explanation of television spot rates. Second, the contribution of cable penetration was statistically significant, but in the wrong direction. Only the contributions of the two "handicap" variables (IND and UHF) were not problematic, being both statistically significant and in the predicted direction. The relative size of the contribution of the time variable (YEAR) in these models indicated the relative importance of time-based factors in determining the value of portions of broadcast signals. The incorporation of time in this manner, however, presumes that the contribution of those factors over the period were uniform, or constant. As there was no basic reason to assume that the influence of time was in fact uniform over that period, a separate procedure was used to fit models which relaxed that assumption. This procedure involved the use of the analysis of covariance statistical procedure with time (YEAR) as the blocking variable, which removed the assumption of constancy in the contribution of the time factor. This procedure also permits specific testing for the influence of the time factor, as well as the assumptions of constancy of impact for the time factor, as well as for the other variables included in the model. The procedure achieves these tests by fitting separate models for each value of the blocking variable, that is, for each time period. In doing so, the procedure also removes any problems with autocorrelation. [Table 2 about here] The basic analysis of covariance model, as well as the more basic linear model (amended slightly to include the full set of election variables) are given in Table 2. The basic linear regression model for spot prices (using LNSPOT$ as the dependent variable) proved to be statistically significant, explaining about 82% of the variation in spot rates among the stations over time (R-Sqr =.8232; F(19, 3012) = ; p <.0001). The basic analysis of covariance models were also highly statistically significant (F(15, 3016) = ; p <.0001). the results of the analysis of covariance model indicated that both the hypothesis of equal adjusted means (F(10, 3006) = 28.55) and the hypothesis of equal slopes (F(140, 2886) = 1.67), that is, the hypotheses that the independent variables would have a constant impact over the period , could be rejected at a significance level of p <.0001 for the spot rates model. This implies that not only was the influence of the time variable YEAR significant, but that the estimated contribution of the various other factors upon the value of television time had not all been constant over time. The differences in the patterns of influence pointed to by this result can be found in Table 3, which reports the values of the fitted coefficients for some of the variables when a separate model is fitted to each year's worth of data separately. there is a fair degree of variation in most of the estimated coefficients; the coefficients for some variables change sign over the period, and many of the estimated coefficients are not significantly different from zero. Thus, the results of this investigation would seem to verify the appropriateness of the need to examine patterns of influence over time. However, it should be recalled and noted that, as the primary independent variables (LNADC) for the years 1979 and 1982 were the results of estimates rather than actual measures, the results for those years are not expected to be as accurate or reliable as

9 those in the other nine individual models. [Table 3 about here] Earlier, it was noted that there was some evidence presented in earlier research that patterns of influence might not be the same for all subsets of stations. With the analysis of covariance procedure, the data was disaggregated on the basis of time. Park (1971, 1972, 1978) had argued that differences also existed between UHF and VHF stations. Thus, the data was disaggregated on the basis of station frequency, and separate stepwise procedures applied to each subsample in order to derive models of television spot prices. The results of those procedures are given in Table 4. [Table 4 about here] While both fitted models are statistically significant (for VHF: F(11, 2069) = 1058, p <.0001; for UHF: F(14, 936) = 168.2, p <.0001) and explain a significant percentage of the variations in the spot rates for the two basic types of television stations, there are some fundamental differences in the direction and size of influence of the various factors. For example, only VHF stations appear to be influenced by the number of competing stations. And, as predicted by Park (1971, 1972), the influence of cable penetration is quite different for VHF (ie, small and negative) and UHF (large and positive) stations. Similarly, there is a difference in the size of the Independent "handicap" experienced by the two classes of stations. ANALYSIS AND CONCLUSIONS What do these models contribute to our understanding of the impact of time and the other forces which influence the value of television time? First, one must conclude that time is an influential factor in the explanation of television spot rates. there were significant variations in the price of television time over the period , not all of which could be explained by other factors. In the basic regression models, the time variable (YEAH) accounted, on its own, for over 15% of the variation in spot prices. The basic regression models indicated that spot prices tended, to increase more than 8% a yearly. Relaxing the assumption of constant effects, the analysis of covariance procedure indicated that time was a significant factor in the model, and estimated yearly (inflationary) increases in spot prices ranging from 4.5% to about 28%. This result, however, should not be construed as invalidating the earlier studies of the factors influencing the value or price of television time, but demonstrating, as many of them did, the need for the consideration of an additional factor in attempts to explain the price of television time. In this case, that additional factor is time. The dominant influence upon the price of television time was still the size of the audience for such time. Variations in the audience size measure (LNADC) accounted for some 70% of the total variation in prices explained by the models. This result is hardly unsuspected, as what purchasers of television time are buying, in essence, is access to an audience. Most of the remaining influences were also linked to the fundamental factor of audience size. Market size, for example, had been used in several earlier studies as indicators of audience (cf. Kelly, 1967; Besen, 1976; French and McBrayer, 1979). A series of five measures of market size were considered in this study, and found to influence spot prices in the predicted direction, although, combined, they accounted for less than 1.5% of the variation in prices after controlling for audience size. Due to the high degree of multicollinearity, it should be noted that the relatively small contribution of the market size measures, after removing the influence of ADC, does not necessarily indicate that these factors would have no influence, when considered singly. Rather, it suggests only

10 that there is no great influence independent of audience size. The actual audience attracted by a station at any time was also said to be heavily influenced by the amount and the nature of the competition which the station faces. The typical measure of competition is the number of competing stations in the market; this study differed slightly in considering VHF and UHF competition separately. The amount of UHF competition faced (UCOMP) contributed significantly to the basic models, although its influence was positive, indicating that the more UHF competitors faced, the greater the value of television time for a station. Although the direction of this influence is opposite to what might be expected from most earlier studies, it might be explained by the fact that since UHF stations are thought to operate at a disadvantage, the more UHF competitors faced by a station in a market, the better that station's competitive posture in the market. On the other hand, the influence of the number of VHF competitors (VCOMP) was in the expected negative direction. These results seemed to indicate that forces of indirect competition, that is, the forces shaping a station's ability to compete on a roughly equal basis with other stations in the market, were more powerful than the forces of direct competition, the simple numbers of other stations in the market. This finding is reinforced by the results of the VHF/UHF disaggregation, in which the VCOMP variable only contributed significantly to the explanation of spot prices for UHF stations, and no measure of direct competition contributed to the explanation of spot rates for VHF stations (see Table 4). This finding is further reinforced by the influence demonstrated by the measures of other forces of indirect competition. For example, this study tended to confirm the findings of earlier efforts that a "UHF handicap" existed; with the basic model indicating that UHF stations had prices from twelve to fifteen percent lower than those of comparable VHF stations. Consideration of the yearly estimates provided by the analysis of covariance model and given in Table 3, however, would appear to indicate that the size of the "handicap" was in fact shrinking, and was not significantly different from zero after 1978, indicating that here was no firm statistical evidence of an impact from that point on. The results thus supported the hypothesis that the size of the UHF "handicap" diminished over time. The influence of network affiliation can also be interpreted as an indication of the effects of indirect competition. Independent stations have largely been unable to attract the audiences of network affiliates because of the differences in the amounts and kinds of programming available to these two types of commercial stations, and may thus be said to operate at a comparative disadvantage. The fitted coefficients for the independent station indicator variable (IND) are uniformly negative and significant for the spot price models. They indicate that, in general, a non-affiliated, independent, station is able to obtain only about half the price for a 30-second spot as are comparable network affiliates. The size of that "handicap," though, also seemed to be decreasing over time from an average discount of about 55% from to an average discount nearer 40% over the last five years of the study (see Table 3), supporting the hypothesis of a reduction in the size of the "handicap" over time. In addition, the results of the model-fitting for the stratified sample given in Table 4 indicated that the size of the handicap was much higher for VHF independents than for UHF independents. The basic influence of cable television upon broadcast television was generally held to be the introduction of additional competition for the broadcast stations in the

11 market (cf. Fisher and Ferrall, 1966; Park, 1971, 1972; Kaplan, 1978; Agostino, 1980; Webster, 1983, 1984; Wirth and Wollert, 1984), which would tend to lower the value of the broadcast time offered for sale. The results of the various models, however, did not support this contention. While cable penetration did contribute significantly to the explanation of spot rates, the overall influence demonstrated in the models was basically positive, indicating that increased cable penetration tended to increase the value of television spots. The VHF/UHF disaggregation presented in Table 4, though, showed that this strong, positive, influence was centered in UHF stations. In addition, the yearly coefficient estimates (see Table 3) indicated that this positive influence was increasing over time. this result, running counter to most theory and research in this area, suggests the need for further examination of the role played by cable in the broadcast marketplace. However, the nature of the results do indicate a change in the pattern of influence over time, thereby giving further evidence in support of the hypothesis that time has an indirect influence on prices through shifts in the influence of other factors over time. The final factor to be considered in the study was the influence of Presidential election campaigns on the price of television time. The results of the study in this regard were decidedly mixed, with the various stepwise regressions providing conflicting indications of influence for the indicator variables reflecting permanent shifts due to the influence of election campaign advertising. [Recall that due to the time-based nature of the variables EL76, EL80, and EL84, it was not possible to include them in the analysis of covariance models for analysis.] No general conclusions about this factor were thus possible, although the lack of general significance might have resulted more from the specific assumptions made about the nature of the influence in the generation of the measures (that is, that any influence might be long-lasting), or the increased capability of stations to anticipate the level of campaign advertising and increase the supply of advertising during such periods, than an absence of any influence resulting from Presidential Election campaigns. In general, the results of this analysis supports both the basic findings of earlier studies and the appropriateness of considering time as an influential factor in the modeling of broadcast spot prices. The variables measuring the basic factors of audience size, market size and characteristics, and station characteristics did prove to influence the prices sought for television time over this extended period ( ), confirming, in general, the results of earlier research. This study also indicated that time, as a separate factor, had a direct and significant impact on prices in the form of inflation, although the results of the consideration of the more specific time-based factor, Presidential Election Campaigns, were inconclusive. This study also demonstrated that time had an indirect influence, as their was strong evidence of changes over time in the strength of two fundamental factors identified by most earlier studies, the "handicaps" said to result from the lack of a network affiliation and operation in the UHF band. That is, this study indicated that UHF stations and independent stations were, other factors being held constant, becoming more competitive over time. Also, the general influence of cable television upon the price of television time, as measured by the degree of cable penetration in a market, was also changing over time, although in a direction opposite of that suggested by other studies. Despite this discrepancy in the direction of impact, however, the presence of a distinctive

12 change the impact of cable penetration, over time, further demonstrates the influence of time as a distinctive factor. Thus, four of the five basic hypotheses derived to demonstrate the influence of time have been supported. The fifth, dealing with the influence of Presidential Election Campaigns, proved inconclusive, although that might be the result of the specific measure used rather than the absence of any effect. In general, though, this study supports the notion that time does influence the value, or price, of television time, and thus should be considered in any attempts to model such prices. This study also indicated that specific patterns of influence varied not only over time, but also varied across station types: that UHF and VHF stations differed in the degree to which certain factors influenced the price of spot time on those stations, and in some cases (e.g. cable penetration) had distinctly different impacts for the two groups of stations. These results, and the findings regarding the impact of cable penetration and the decline in the UHF and Independent station's "handicaps," suggest the need for continued research into the nature and patterns of influence in the market for television time. REFERENCES ABRAMS, B.A, and R.F. SETTLE (1977). "Broadcasting and the political campaign spending 'Arms Race.' Journal of Broadcasting 21: AGOSTINO, D. (1980). "Cable television's impact on the audience of public television." Journal of Broadcasting 24: BATES, B.J. (1983). "Determining television advertising rates," in R.N. Bostrom (ed.) Communication Yearbook 7. Beverly Hills: Sage. -- (1985). "Economic Theory and Broadcasting." Paper presented at the Association for Education in Journalism and Mass Communication conference, Memphis IN, August (1987). "The Role of Theory in Broadcast Economics: A Review and Development," in M.L. McLaughlin (ed.), Communication Yearbook 10. Beverly Hills, CA: Sage BESEN, S.M. (1976). "The value of television time." Southern Economic Journal 42: (1978). "The value of television time: Some problems and attempted solutions: Reply." Southern Economic Journal 44: BOWMAN, G.M. (1975). "Network television advertising: Demand and supply," In B.M. Owen (ed.), Telecommunication Policy Research: Report on the 1975, conference proceedings. Palo Alto: Aspen Institute. CHERINGION, P.W., L.V. HIRSCH, and R. BRANDWEIN (1971). Television Station Ownership: A Case Study of Federal Agency Regulation. New York: Hastings House. FISHER, F.M., and V.E. FERRALL, Jr. (1966). "Community antenna television systems and local station audiences." Quarterly Journal of Economics 80:

13 FOURNIER, U.M., and D.L. MARTIN (1983). "Does government restricted entry produce market power? New evidence from the market for television." Bell Journal of Economics 14: FRENCH, W.A., and J.T. McBRAYER (1979). "Arriving at television advertising rates." Journal of Advertising 8: KAPLAN, S.J. (1978). "The impact of cable television service on the use of competing media." Journal of Broadcasting 22: KELLEY, W.J. (1967). "How television stations price their service." Journal of Broadcasting 11: KRASNOW, E.G., L.D. LONGLEY, and H.A. TERRY (1982). The Politics of Broadcast Regulation. 3rd Edition. New York: St. Martin's Press. LARSON, T.L. (1976). "Aspects of market structure in the U.S. television industry, " PhD. dissertation, University of Wisconsin at Madison. LEVIN, H. J. (1980). Fact and Fancy in television Regulation: An economic study of policy alternatives. New York: Russell Sage Foundation. McFADYEN, S., C. HOSKINS, and D. GILLEN (1980). Canadian Broadcasting: Market Structure and Economic Performance. Montreal: The institute for Research on Public Policy. MELODY, W. (1973). Children's television: The economics of exploitation. New Haven: Yale University Press. PARK, R.E. (1971). "The growth of cable 1V and its probable impact on over-the-air broadcasting." American Economic Review 61: (1972). "Cable television, UHF broadcasting, and FCC regulatory policy." Journal of Law and Economics 15: (1978). "The value of television time: Some problems and attempted solutions." Southern Economic Journal 44: PETERMAN, J.L. (1971). "Concentration of control and the price of television time." American Economic Review 61: PORTER, R.D. (1963). "Some values to the broadcaster of election campaign broadcasting." Journal of Broadcasting 7: WEBSTER, J.G. (1983). "The impact of cable and pay cable television on local station audiences." Journal of Broadcasting 27: (1984). "Cable television's impact on audience for local news." Journalism Quarterly 61: WHITE, K.J. (1977). "Television market shares, station characteristics, and viewer choice." Communication Research 4: WIRTH, M.O., and J.A. WOLLERT (1984). "The effects of market structure on television news pricing." Journal of Broadcasting 28:

14 Table 1: Stepwise Regression Model for Television Spot Prices Step R-Sqr. Std. Error Variable Partial Signif LNADC YEAR SALESHH IND MKTRANK UCOMP UHF EL CBS ABC CABLE% BPI MKTEBI EBIHH MKTSALES VCOMP SIN EL Note: For the final fitted regression, F(18,3013) = , p <.00001

15 Table 2. Estimates from the Basic Models of Television Spot Prices Variable Regression Model ANCOVA Model Coeff. Signif. Coeff. Signif. YEAR LNADC VCOMP UCOMP MKTRANK MKTEBI MKTSALES BPI EBIHH SALESHH UHF ABC CBS SIN IND CABLE% EL EL EL Note: For the basic regression model, Adj. R-Sqr. =.8232, F(19, 3012) = (p <.00001) For the basic ANCOVA model, F(15, 3016) = (p <.00001) and for the test of equal adjusted means, (i.e. the significance of YEAR as a blocking variable) F(10, 3006) = (p <.00001)

16 Table 3. Yearly Estimates from ANCOVA procedure, Selected Variables Variable Adj. Mean UHF * * -.117* -.117* -.091* -.045* -.099* IND CABLE%.017*.406*.272* -.082*.248*.113*.274*.212* *.682 R N Notes: * denotes that estimated coefficient was not significantly different from zero, at a level of p <.05. All regressions were statistically significant at a level of p <.0001.

17 Table 4. Stratified Sample: Stepwise Regression Models for Spot Prices Variable Step VHF Stations Coeff Signif. Partial Step UHF Stations Coeff. Signif. Partial LNADC YEAR VCOMP MKTRANK MKTEBI MKTSALES BPI EBIHH SALESHH UHF% COVERAGE CABLE% NBC IND EL EL N = 2081 N = 951 Adj. R-Square =.8600 Adj. R-Square =.7155 F(12, 2068) = 1058 F(14, 936> = (p <.00001) (p <.00001)