Noel D. Uri and Keith Brown
[WJMCR 7:1 December 2003]
The objective of this study is to provide an assessment of the extent to which direct broadcast satellite service competes with cable television service for subscribers in the multichannel video programming subscription service market. After recounting some of the institutional characteristics of the market, the theoretical underpinnings of the modeling approach are presented. The model being pursued is known as the dominant firm-competitive fringe model. In this model, as is characteristic of the multichannel video programming subscription service market, the dominant firm (cable systems) sets a price that fringe firms (DBS service providers) take as given in deciding how much to supply. The empirical results presented endeavor to determine (1) whether de facto the competitive fringe constrains the dominant firm�s pricing behavior and (2) if so, to what extent. A simultaneous equations econometric model is rigorously developed with attention focused on many of the important modeling issues frequently overlooked in studies of the market for cable service. The model estimated consists of five equations – demand equations for analog cable service, digital cable service, and DBS service and supply equations for analog cable service and digital cable service. With regard to the issue of interest, the competitive fringe does serve to constrain the dominant firm. The estimates, however, do not indicate that changes in the relative price of analog or digital cable service have a quantifiable impact on the demand for DBS service.
Currently in the United States, there are three primary technologies that deliver television service to individual households: over-the-air broadcasting, cable television, and DBS or Direct Broadcast Satellite. Each of these technologies is covered under a unique regulatory framework. Additionally, the channels that viewers watch on television fall into two primary categories: broadcast channels which include the broadcast networks and independent local channels and subscription channels. Over-the-air broadcasting is free to consumers with a television set and a suitable antenna. Cable television and DBS are multichannel video programming subscription services. A model of the market for multichannel video programming subscription service is presented, including both the demand and supply of multichannel video programming. The model estimates empirically the extent to which cable television service and DBS service compete for subscribers. The model is based on data collected by the Federal Communications Commission on its 2002 FCC Annual Cable Price Survey. Before delving into the empirical issues, however, it is critical to understand the market for multichannel video programming subscription service.
(a) Demand Considerations
The broadcast industry has two key components. The first is composed of local television stations. All television stations in the United States must be licensed by the Federal Communications Commission (FCC). The FCC license gives a station the right to use a specified portion of the radio spectrum to transmit video programming in a specific geographic region. A group of local television stations serving the same geographic region make up a television market.1 Because the video signal from a local television station is broadcast through radio waves or �over-the-airwaves,� this method of providing television is called over-the-air broadcast television.
The broadcast television industry is funded primarily through national, regional, and local advertisements that are aired along with other programming on local television stations. Households that receive only broadcast television pay no subscription fee for access to the signals of the television stations in their geographic region. They only need to have a television set with an adequate antenna. The availability of broadcast stations has been shown to impact the market for television service. For example, a greater number of free broadcast stations will tend to reduce cable subscriptions and cable rates.2 Based on the results of the 2002 FCC Annual Cable Price Survey,3 33.6 percent4 of households in the United States did not subscribe to cable or any other subscription television service but relied on over-the-air broadcast technology for their multichannel video programming.
The second component of the broadcast industry is cable television. Cable television developed as a way of providing the signals of local television stations to rural and mountainous areas that could not get adequate reception of those signals through conventional antennas. Cable systems obtain a franchise authorization under agreed upon terms and conditions from a local authority such as a city, county, or a township that grants them the right to operate in a specified area (known as the franchise area) and run cables along public rights-of-way.5
During the 1970s developments in satellite technology enabled video signals to be transmitted economically via satellites permitting the development of new networks (e.g., CNN and HBO) designed primarily for the distribution of programming via satellite to cable systems throughout the United States. Unlike the broadcast networks which earn revenue mainly through advertising, these subscription networks are supported through advertising revenue and/or fees paid by cable systems. The cable operator generally receives two kinds of signals: signals broadcast by local television stations from television towers in the geographic area and signals via satellite from subscription networks. These signals are all provided to subscribers through the cable system�s wires.
Once subscription networks were added to cable systems� channel offerings, cable systems became more than an antenna service. New cable systems were built in previously unserved areas. In a relatively short period of time, the growth in the number of cable systems was quite impressive. In 1970, for example, there were 2490 cable systems. In 2002 there were 9947 cable systems.6 This represents a 4.2 percent annual growth rate in the number of cable systems over the period. Where available, consumers were confronted with a choice between having access to only the over-the-air broadcast channels for free or paying to obtain the broadcast channels as well as several new subscription network channels over a cable system. Based on the results of the 2002 FCC Annual Cable Price Survey, about 95.6 percent7 of the households in the United States have access to a cable system and 56.2 percent8 of these households subscribe to a cable service.9
The most significant competitor to cable today is the direct-to-home satellite television industry. Satellite multichannel video programming subscription service emerged in the early 1980s as an alternative to cable service in rural areas not passed by cable systems. Program packagers bundled the subscription networks offered by cable systems and sold them to consumers who received the signals on large satellite dishes. They did not distribute broadcast signals. In 1994 Direct Broadcast Satellite (DBS) service was introduced.10 Subscribers could receive the video signals using small reception dishes that could be mounted on rooftops or window sills. Satellite multichannel video programming subscription service is available nationwide and each DBS company11 typically offers the same programming packages and prices throughout the United States. Monthly service charges are comparable to monthly cable rates for comparable packages of programming services.
The penetration of DBS service was impeded in its early years because there were legal limitations on DBS providers� ability to transmit local broadcast signals. This put DBS service at a disadvantage relative to cable service. DBS providers were governed by the 1988 Satellite Home Viewers Act which was passed at a time when satellite providers did not possess the technology to transmit local broadcast signals to many markets throughout the country. In general DBS firms had no license to provide broadcast signals to households in urban or suburban areas that generally could receive over-the-air local broadcast signals. In late 1999 Congress enacted the Satellite Home Viewers Improvement Act to, among other things, allow DBS companies to provide local broadcast signals to subscribers.
Because DBS service was developed using digital technology, these systems have a greater channel capacity and transmit clearer video and audio signals with less degradation of the signals than analog cable systems. Additionally, digital technology uses radio spectrum more efficiently. Only recently have cable systems started to offer digitally transmitted service along with their analog service.12
Based on the 2002 FCC Annual Cable Price Survey, 82.0 percent13 of households in the United States that subscribe to a multichannel video programming subscription service use cable television although DBS subscriptions are growing.14 In 2001, for example, based on data collected on the 2002 FCC Annual Cable Price Survey, DBS accounted for 14.4 percent15 of the multichannel video programming subscription service while in 2002, it accounted for 18.0 percent.16
There are a number of aspects of DBS service that make it attractive to some subscribers. First, through the use of digital technology, DBS�s video and audio quality is state-of-the-art.17, 18 Second, DBS providers typically offer a larger number of channels than most analog cable systems. Moreover, while most cable systems now offer a digital tier, their channel lineup is typically less complete than that offered by DBS. For subscribers who, for example, are interested in movies or sports, DBS may offer a better selection of channels than the competing cable system(s).
(b) Supply Considerations
Although there are 10,146 cable systems19 in the United States, the ownership of cable operations is relatively concentrated. Additionally, there are ownership ties between cable systems and related firms. For example, there are vertical relationships between cable companies and program suppliers, horizontal concentration among cable companies, and clustered cable systems whereby cable companies consolidate ownership within a geographic area. Each of these relationships has implications for the cost of supplying cable service.
As cable system deployment and the number of subscribers grew, the ownership structure of the industry changed. Early on, the cable industry was characterized by privately owned, small systems scattered throughout the United States. The expanding market for cable service in the 1980s, however, attracted the interest of large media companies.20 These companies purchased cable systems throughout the United States, emerging as large national cable owners known as multiple system operators (MSOs).21 Several of these MSOs also invested in the development of cable programming and thereby established ownership ties between two vertically integrated markets – companies producing and supplying cable programming and companies purchasing that programming and delivering it to subscribers.22 Nearly 98.7 percent23 of the cable systems responding to the 2002 FCC Annual Cable Price Survey indicated that they were owned by a MSO.
In recent years cable systems have engaged in a clustering strategy in order to consolidate their systems in and around specific cities or regions. Cable companies can obtain increased economies of scale from clustering as compared to having noncontiguous cable systems that are more geographically diffuse.24 The clustering approach enables firms to consolidate facilities for receiving and transmitting programming, reduce the number of repair crews, have regional service centers, and reduce the size of management.25 Additionally, clustering can serve to enhance advertising revenues.26 Slightly more than 81.1 percent27 of the cable systems surveyed on the 2002 FCC Annual Cable Price Survey were part of a cluster.
The various ownership interrelationships that exist in the cable industry ostensibly provide efficiencies to cable companies that result in reduced costs of providing cable service. For example, relatively larger cable providers may realize reduced programming costs and also have cost savings in management and related overhead functions.28
As noted previously, the most significant competitor to cable is DBS. Because DBS uses a digital technology, these systems have a greater channel capacity and transmit clearer video and audio signals with less degradation of the signals than many analog cable systems although these issues do not necessarily impact the cost of providing service. This technical advantage of DBS has been short lived as cable operators have moved to upgrade their systems to include digital technology. Results from the 2002 FCC Annual Cable Price Survey indicate that digital cable service is available to 92.9 percent29 of subscribers and that 28.3 percent30 of subscribers choose digital cable service.
This is the basic institutional structure of the market for multichannel video programming for cable service and DBS service in the United States. This structure will be used in developing an econometric model of the market. Given the nature of the data that are available and the inherent interest in measuring the potential for competition between cable television service and DBS service, only these two multichannel video programming subscription services are considered.
In 2002 cable systems had a 82.0 percent share of the multichannel video programming service market.31 In the context of this market, cable systems can be thought of as the dominant distributors of multichannel video programming market, while the other providers – DBS firms – can be thought of as fringe suppliers. Consequently, to examine the competitive influence of DBS on the demand for cable service, it is necessary to consider a model based on the multichannel video programming service market defined to include both cable television service and DBS service rather than on a more narrowly focused market for just cable television service.
The theoretical underpinning of the model is known as the dominant firm-competitive fringe model.32 This model assumes that there is a single firm or jointly acting group of firms holding a dominant position33 and several smaller price-taking firms, typically referred to as the competitive fringe. Because of its position, the dominant firm is modeled as selecting a price that the fringe firms take as given in deciding how much to supply.34 An important insight provided by the dominant firm model is in showing how the existence of a competitive fringe restrains the dominant firm�s pricing behavior. Suppose the fringe was absent so that the dominant firm was instead a monopolist. In this case its demand would equal market demand and price would be set at the monopoly level. The existence of a competitive fringe results in the dominant firm charging a lower price.35
The dominant firm sets a lower price because its demand is weaker as a result of the competitive fringe. Additionally, it takes into account how the fringe will respond to its price. Knowing that the fringe supply increases as it increases its price, the dominant firm sets a lower price in order to reduce fringe supply.36
Cable providers (considered as the dominant firm) and DBS operators (considered to be the competitive fringe) can be regarded as differentiated, not so much because they use different technologies to provide services, but because the two types of technologies confront different laws and regulations that influence their cost structures as well as the type of services they provide. For example, cable systems generally pay a local franchise fee for the use of public rights-of-way and may be required to provide capacity for public, educational, and government channels depending on their franchise agreement. Satellite providers have only a limited compulsory copyright license to provide broadcast channels. They are, however, required to reserve a portion of their channel capacity37 for noncommercial programming of an educational or informational nature. The capacity must be made available under reasonable prices, terms, and conditions and the DBS provider cannot exercise any editorial control over the contents of this video programming.38
Under a generalized dominant firm-competitive fringe model, the demand for cable service will depend on cable rates and cost conditions affecting both the cable and noncable providers of multichannel subscription video service. A desirable attribute of this model is that it is possible to incorporate the competitive influence of noncable providers – i.e., DBS providers – on the demand for cable. As noted, DBS represents the single largest competitor to cable service. To measure the competitive influence of the noncable providers (i.e., the competitive fringe), the DBS share of video subscribers in each franchise area is used in modeling the market for multichannel video programming subscription service.39
Given the nature of the dominant firm- competitive fringe relationship between cable service and satellite (DBS) service, the appropriate approach to modeling the market for multichannel video programming service is with a simultaneous equations model that captures both demand and supply considerations. The structure of the model is necessarily contingent on the data that are available. That is, while the model will be consistent with economic theory, it will not include all of the factors that impact the demand and supply of multichannel video programming service because the data are simply not available. Thus, for example, data on production costs for analog and digital cable service are not available40and hence, variables reflecting these costs on supply are not included in the simultaneous equations model specification. The primary data source is discussed in the next section.
2002 FCC Annual Cable Price Survey
Section 623(k) of the Communications Act, as amended by the Cable Television Consumer Protection and Competition Act of 1992,41 requires the FCC to publish annually a statistical report on cable prices, or more specifically, average rates for the delivery of basic cable service, cable programming service, and equipment.42The requirement to compare the price of cable service for systems where effective competitive has been found and the price of cable service where effective competition has not been found is important given the objectives of the 1992 Cable Act. The primary data used in this study rely on the results of the survey conducted as a result of this requirement.
The 2002 FCC Annual Cable Price Survey requested data from a sample of cable systems as of July 1, 2002 and July 1, 2001. The 2002 FCC Annual Cable Price Survey was structured to allow the FCC to compare prices charged by two groups of cable systems: (1) systems that are deemed to face effective competition (nominally referred to as the competitive group); and (2) systems that do not face effective competition (the noncompetitive group).43 Cable systems in the competitive group are limited to geographic areas where a cable operator has sought and obtained a FCC finding of effective competition. For these purposes the FCC�s formal legal decisions were used as a basis to determine whether effective competition exists based on the statutory definition of that term.
The 2002 FCC Annual Cable Price Survey collected information about average monthly rates for the basic service tier (BST) and major cable programming service tier (CPST).44 The BST typically consists of local stations (e.g., broadcast channels) plus a few satellite channels and public, educational, and government access (PEG) channels if they are carried.45 The major CPST typically consists of satellite-delivered channels. About 90 percent of cable subscribers take both the BST and major CPST while the remaining 10 percent take BST only. In addition data were collected for the most highly subscribed digital tier of service. Information was also collected on the average monthly charge for equipment, consisting of an analog addressable converter and remote control and digital converter plus remote control. The 2002 FCC Annual Cable Price Survey further sought information needed to determine average rates per channel and to explain changes in rates. Finally, information was gathered on other factors that affect cable prices and competition in the multichannel video programming service market such as the cable system�s best estimate of the number of subscribers to DBS service in the franchise area,46 as well as the availability of advanced services such as Internet access, cable telephony, and interactive programming. A summary of the major findings of the 2002 FCC Annual Cable Price Survey for all cable systems surveyed, and separately for the competitive and noncompetitive groups, is available elsewhere and not reproduced here.47
To compare average monthly rates of competitive and noncompetitive cable systems,48 a separate sample was selected from each group. These samples included 282 of the 356 systems in the competitive group and 473 of the 9,790 systems in the noncompetitive group. To ensure that the samples were representative and to gain more precise estimates, both groups were stratified into subgroups (or strata) and a portion of the sample from each stratum was selected. The competitive group was divided according to the test by which effective competition was determined and the noncompetitive group according to the number of subscribers in each operator�s cable system.
Competitive systems were divided into five strata based upon the test for which effective competition was determined. Cable systems meeting the overbuild test were further subdivided into two strata: (1) wireline overbuild and (2) DBS overbuild.49 The remaining systems in the competitive group were divided into three strata as follows: (1) low penetration, (2) municipal, and (3) LEC. The LEC stratum consists of both the incumbent cable systems who competed with an affiliate of a LEC at the time that a finding of effective competition was made and the LEC affiliates. The other strata, except for the DBS stratum, similarly consist of the incumbent cable systems as well as the relevant competitors. The DBS stratum includes only the incumbent because monthly rates of DBS operators are not part of the 2002 FCC Annual Cable Price Survey.
For the LEC, wireline overbuild, DBS overbuild, and municipal strata, all 95, 37, 42, and 14 systems, respectively, were included because of the relatively small number of systems in each of these four strata. A random sample was chosen for the low penetration stratum because that subgroup had a large number of systems, 168 in all, from which 94 were randomly selected for that stratum. This resulted in a total of 282 systems for the competitive group.
Noncompetitive systems also were divided into five strata. The number of cable systems selected from each stratum depended on the number of subscribers nationwide in that stratum. A sample of systems not stratified by size would have placed a disproportionately greater emphasis on smaller systems relative to the number of subscribers the smaller systems serve. The very large stratum includes systems serving more than 50,000 subscribers in a single community. The large stratum contains systems serving more than 50,000 subscribers, but with no individual community exceeding 50,000 subscribers. The medium stratum is comprised of systems serving from 10,001 through 50,000 subscribers. The small stratum includes systems serving from 1,001 through 10,000 subscribers, and the very small stratum includes systems with 1,000 or fewer subscribers.
The high proportion of subscribers nationwide represented by the very large stratum resulted in the selection of all 99 cable systems in that stratum. Other selections include 109 of the 169 large systems; 153 of the 888 medium-sized systems; 72 of the 2,717 small systems; and 40 of the 5,917 very small systems. Because of the low proportion of subscribers nationwide represented by very small systems, the formula for calculating sample size initially produced fewer selections from that stratum. The number of selections was adjusted upward to 40 systems, however, to ensure that there were a sufficient number of observations from that stratum for valid statistical inference.
Of the 755 Survey questionnaires mailed to cable systems from both groups, respondents completed 694 questionnaires. Of the 61 incomplete questionnaires, cable systems explained that the requisite data were unavailable for 22 systems that had recently been sold or combined with other systems, and 12 questionnaires were undeliverable. Competitive cable systems submitted 262 of the completed questionnaires. Of these, 163 systems had direct competition in their geographic area, with 91 meeting the LEC test, 72 meeting the overbuild test (with 31 in the wireline overbuild subgroup and 41 in the DBS overbuild subgroup), and 13 served a community in which the municipality owned one of the systems (thereby meeting the municipal test). Of the remaining respondents in the competitive group, 86 served less than 30 percent of households in their service area (thereby meeting the low penetration test). Noncompetitive cable systems submitted the remaining 432 responses.
The 2002 FCC Annual Cable Price Survey data are keyed to the five digit Zip code associated with the greatest number of subscribers in the franchise area. This allows for merging the data from the 2002 FCC Annual Cable Price Survey with social, economic, housing, and geographic information from auxiliary sources. These supplementary data potentially serve to influence both the demand and supply of multichannel video programming service.
Data on the social, economic, and housing characteristics of subscribers by five digit Zip code were obtained from the 2000 Census of Population Summary Files provided by the U.S. Census Bureau. These data are quite comprehensive consisting of information on school enrollment, educational attainment, marital status, disability, language spoken, country of birth, employment status, commuting patterns, occupation, income, type of housing unit, number of occupants, house heating fuel, rent versus own, mortgage status, tenure of occupancy, and so on.
The final set of data merged with the data from the 2002 FCC Annual Cable Price Survey consisted of information on the size of the Zip code area, its latitude and longitude and the recommended azimuth, elevation, and skewness of the satellite dish.50 The area, latitude, and longitude information was obtained from the U.S. Postal Service. The azimuth, elevation, and skewness data were obtained from the DirecTV web site. These data are important because physical features of the landscape can limit the demand if, e.g., the angle of elevation is too low, or impact costs if subscribers are relatively spatially diffuse.
The use of supplementary data is essential in modeling the demand and supply of multichannel video programming service in a simultaneous equations framework. Both the demand and supply of analog and digital cable service and the demand for DBS service are influenced by the economic, social, and demographic characteristics of consumers and the geographic area in which they live. To ignore these factors would result in a model that is significantly misspecified.
The model specification is, of necessity, dependent on the data that are available. While there is a desire to have the specification as consistent with the previously discussed theoretical considerations as possible, empirical relationships frequently do not conform precisely to economic theory. Hence, the model specification is tempered by the data that are available. Additionally, the model specification is a straightforward extension of previous studies on the demand for cable service, including their strengths and, to the extent possible, mitigating their weaknesses.
For the demand side of the model presented here, Mayo and Otsuka,51Rubinovitz,52 Biel et al.,53 Ford and Jackson,54 and Beard et al.55 provide the foundation.56 These papers model the demand for cable service as a system of equations where the price and demand for multichannel video programming service defined to include cable service and DBS service are jointly determined. None of these papers explicitly looks at the demand for DBS service nor do they separate the demand for cable service into its analog and digital components of cable service. The primary reason for this is that none of these studies compiled or had access to data on digital cable service. The majority of the studies were done during a period when digital cable service was in its infancy or simply was not available.
Rubinovitz uses the number of channels delivered over the cable system as a measure of service quality and estimates a system of equations including a demand function, a �quasi-supply� or price function, and a quality function where quality is proxied by the number of channels using two-stage least squares. Ford and Jackson closely follow Rubinovitz but add a programming cost function to the system in order to account for the cost of the quality level chosen by the cable operator. More recently, Beard et al. present a simultaneous equations model where the price of the basic service tier (BST), the price of the major cable programming service tier (CPST), the number of satellite channels (i.e., non-broadcast channels) offered, and the number of subscribers are endogenously determined. The number of satellite channels is used in this study as a proxy for service quality.
There are a number of interesting observations to be made about these previous studies. First, each uses a log-linear specification57 asserting that it is the preferred specification. This is an empirical issue that should be tested but never is. Second, the number of channels offered by a cable operator is treated as being endogenous in the more recent studies. There are different sets of channel offerings provided by cable systems. There is the basic service tier and a major programming service tier. Within the basic service tier, there are over-the-air broadcast channels that subscribers could, in most instances, otherwise get free.58 There are also premium and pay-per-view channels although these are virtually never offered as part of the basic service tier or the major programming service tier.59 This mix of channel offerings in the model should be considered explicitly. It has the potential for providing some insight into subscriber behavior. Moreover, there is some disagreement as to whether the total number of channels or a subset of channels is important to subscribers and, in fact, truly endogenous. For example, Anstine60finds that program guide channels have a negative marginal value. Jayarantne61finds that the number of broadcast channels is unimportant to subscribers. Given these results, the endogeneity of the number of channels in a simultaneous equations model of the market for multichannel video programming service is another issue that should be empirically investigated. Third, there is scant attention in previous studies to the possible presence of outliers in the data that serve to unduly influence the coefficient estimates.
Based on the preceding discussion, in order to portray accurately the demand for multichannel video programming service, three separate demand equations are considered including the demand for analog cable service, the demand for digital cable service, and the demand for DBS service. The demand for cable service is disaggregated into its analog and digital components for a number of reasons. First, the primary data used in the estimation from the 2002 FCC Annual Cable Price Survey allow for this level of detail. Second, with digital cable, cable systems provide a higher quality signal and hence a better video image than is possible with analog cable service. Cable systems use digital technology to compress video signals, allowing more than one program service to be carried in the bandwidth space normally required for one analog program service. Typically, the signal is sent to the home and decompressed in the set-top box for display on the television.62 Finally, digital cable service offers channels not available via analog cable service. Based on results from the 2002 FCC Annual Cable Price Survey, digital cable channel offerings not available through analog cable service include, for example, Discovery Espanol, EuroNews, Rio de la Plata, Saigon Broadcasting Network, Sundance Movie Channel, TV Asia, TV5 (French), TRIO, the Word Network.
Disaggregating the demand for cable service into analog cable demand and digital cable demand will enable an assessment of whether analog cable subscribers respond differently than do digital cable subscribers to changes in various economic, demographic, and social factors. This is something that has not previously been done.
The demand for analog service is a function of the price of analog cable service (which includes the price of BST plus CPST service), the price of digital cable service, and the price of DBS service. With regard to the price of basic DBS service, it is uniform nationally with no inter-subscriber variability.63 Thus, its effects are captured in the constant terms of the demand relationships. DBS price is not explicitly included in the specification. The relevant prices are the monthly subscription prices for analog and digital cable service as well as the price of DBS service. These are the marginal prices upon which subscribers make their decisions.64 Also included in the demand equation specifications are variables reflecting the size of the market measured by the number of households passed by the cable operator in the franchise area for analog and digital cable demand and the total number of households in the franchise area in the case of DBS demand, the number of channels, and a set of economic, demographic, and social variables that impact potential subscriber�s choice of whether to subscribe to a multichannel video programming service and, if so, what type of service to subscribe to. The demand for DBS service has an added variable, the angle of elevation of the satellite dish. This is potentially important because if the angle of elevation is too low then satellite reception will be hampered by the presence of obstacles such as multi-story buildings, mountains, and ground clutter (e.g., trees) thereby reducing demand. The demand relationships are straightforward and closely follow conventional demand theory.65
The interesting additional variable included in this study in the analog and digital cable demand equations is the share of DBS service subscribers in the franchise area. This variable captures the market penetration of DBS service. It is used as the measure of the impact of the competitive fringe on the demand for the service supplied by the dominant firm. The extent to which DBS service impacts the demand for analog and/or digital cable service can be empirically determined by specifying in the demand equations for analog and digital cable service the number of DBS subscribers relative to the total number of subscribers in a franchise area. That is, ceteris paribus, the greater the relative number of DBS subscribers in an area (that is, the greater the DBS penetration in a franchise area), the fewer the relative number of potential analog and/or digital cable subscribers and hence the fewer the actual number of analog and/or digital cable subscribers. If DBS service is a viable competitor to cable service and operates to constrain price increases by the dominant firm, this variable should have a statistically significant impact on the analog and/or digital demand for cable service. Moreover, to the extent that DBS penetration impacts both analog and digital demand, the degree of this impact can be objectively determined. That is, for example, it is possible to determine whether DBS has a relatively larger impact on analog or digital demand for cable service.
The simultaneous equations model component of the supply of multichannel video programming service will be limited to just the supply of analog and digital cable service. Data simply are not available to model the supply of DBS service.66 With regard to the supply of analog cable service, data on costs were not collected on the 2002 FCC Annual Cable Price Survey. Hence, proxy variables are needed to adequately reflect cost differences between cable systems. To this end, qualitative (dummy) variables are used to differentiate cable systems by type (e.g., small versus large, municipal systems versus overbuilders) with the implication being that in the presence of economies of scale or other cost efficiencies, costs will vary by size of operator. Also, qualitative variables are used to differentiate between whether a cable operator is part of a MSO or not and between whether it is a member of a cluster or not. As noted previously, in each instance cable systems which are part of a MSO or who are members of a cluster ostensibly have lower production costs.67 Other factors that potentially serve to impact supply include whether there exists effective competition in the franchise area and whether basic service is subject to local regulation and these factors are also considered. Finally, the potential subscriber density of the franchise area supplied will directly impact costs. That is, if potential subscribers are relatively more spatially dispersed, costs of supplying these subscribers will be higher due to higher cabling costs, increased costs of getting the required construction permits and rights-of-way, and so on, all other things equal, than if subscribers are relatively concentrated.
With regard to digital supply, there is one additional factor. To supply a digital signal to a subscriber, the cable subscriber needs to purchase or rent a set-top digital converter from the cable system. Larger cable systems can purchase these converters at a lower cost than can relatively smaller cable systems due to substantial volume discounts. Hence, costs of supplying digital cable service would be relatively less for larger cable systems.
Some Preliminary Empirical Issues
Before turning to the actual estimates of the simultaneous equations econometric model of the demand and supply of multichannel video programming service, some preliminary issues need to be dealt with. First, consider the issue of the appropriate functional specification. Previous studies have asserted that the appropriate functional form for each of the equations in a simultaneous equations model of the demand for cable television service is linear in logarithms. That is, both the dependent and explanatory variables are transformed by loge (i.e., Napierian logarithms) before empirically estimating the relationships. To test the credibility of this assertion, a straightforward nonnested test is used. The test is applied to each of the demand and supply equations individually. (This is simply the nature of these tests.) The test chosen is the J-test developed by Davidson and MacKinnon.68 The basic idea of the test is to embed both of two competing regression functions in a more general one and then test one or both of the original models against it. Three functional specifications are considered – a linear specification, a log-linear specification where all of the explanatory variables are transformed by loge, and a semi-log-linear specification where just the dependent variables are transformed. Each of the specifications is considered in pairwise fashion with each of the functional specifications alternately serving as the correct specification (i.e., the null hypothesis). The results, presented in Table 1, are insightful. There is no clearly preferred functional specification for any of the equations except the demand for DBS. In this instance a linear specification is clearly preferable. For the other equations, there is no empirical basis for selecting one functional specification over another.
|Table 1. J-Test of Alternative Functional Specifications*EquationComputed Value of the Test Statistic1. Analog Cable Demand Equationa.H0: Linear, Ha: Log-linear9.85b.H0: Linear, Ha: Semi-log-linear21.16c.H0: Log-linear, Ha: Linear9.79d.H0: Log-linear, Ha: Semi-log-linear10.41e.H0: Semi-log-linear, Ha: Linear3.65f.H0: Semi-log-linear, Ha: Log-linear2.292. Digital Cable Demand Equationa. H0: Linear, Ha: Log-linear2.28b. H0: Linear, Ha: Semi-log-linear16.76c. H0: Log-linear, Ha: Linear2.11d. H0: Log-linear, Ha: Semi-log-linear9.70e. H0: Semi-log-linear, Ha: Linear8.69f. H0: Semi-log-linear, Ha: Log-linear9.223. DBS Demand Equationa. H0: Linear, Ha: Log-linear1.50b. H0: Linear, Ha: Semi-log-linear1.72c. H0: Log-linear, Ha: Linear3.28d. H0: Log-linear, Ha: Semi-log-linear6.53e. H0: Semi-log-linear, Ha: Linear3.67f. H0: Semi-log-linear, Ha: Log-linear5.634. Analog Cable Supply Equationa. H0: Linear, Ha: Log-linear3.47b. H0: Linear, Ha: Semi-log-linear20.79c. H0: Log-linear, Ha: Linear3.02d. H0: Log-linear, Ha: Semi-log-linear7.71e. H0: Semi-log-linear, Ha: Linear6.63f. H0: Semi-log-linear, Ha: Log-linear5.005. Digital Cable Supply Equation*a. H0: Linear, Ha: Log-linear2.61b. H0: Linear, Ha: Semi-log-linear21.97c. H0: Log-linear, Ha: Linear2.26d. H0: Log-linear, Ha: Semi-log-linear3.14e. H0: Semi-log-linear, Ha: Linear4.29f. H0: Semi-log-linear, Ha: Log-linear4.83* The critical value at the 5 percent level is 1.96 given a sample size of 582. That is, if the computed test statistic is less than the critical value, the null hypothesis, H0, is accepted. The alternative hypothesis is denoted as Ha.|
This suggests that the functional specification for each of the analog and digital demand and supply equations should be selected using another criterion. For the current model, a linear specification is used for each of the demand and supply equations. This makes the analog and digital demand and supply equations functionally consistent with the DBS demand equation. A linear specification has the additional desirable property that it does not imply that subscribers or suppliers (cable systems) respond in an invariant way at every point along the demand or supply curve to, for example, a given percentage change in the price of analog service or a change in one of the explanatory variables (e.g., the number of households passed in the franchise area). This is a characteristic of a log-linear specification. Nor does the linear specification imply that subscribers or suppliers respond in a constant fashion at ever point along the demand or supply curve to an equal incremental change in one of the explanatory variables. This is a characteristic of a semi-log-linear specification.
Next, consider the issue of the exogeneity of the number of channels provided by cable systems. Are the number of analog and digital channels provided in a cable system�s franchise area simultaneously determined along with the price and the number of subscribers? To examine this issue, a Wu-Hausman test for exogeneity is used.69 First, recall that there are different measures of the number of channels available. For analog cable channels these include the number of BST channels, the number of CPST channels, the sum of the two, the sum of the two less the number of broadcast channels, and the total number of channels available including premium and pay-per-view. For digital cable channels the measures of the number of available channels include the number of digital channels of the most highly subscribed tier and the total number of channels including premium and pay-per-view. The data that are available for the number of DBS channels that are coincident with the 2002 FCC Annual Cable Price Survey are constant across subscribers since DBS service is marketed nationally.70
For analog cable channels, the number of BST plus CPST channels less the number of broadcast channels is used in the analysis while for digital cable the number of digital channels on the highest subscribed digital tier is used.71 This measure for the number of analog channels seems appropriate since, at the margin, most consumers subscribe to cable service because it provides them with access to channels not otherwise available for free. This measure is consistent with the measure used by Beard et al.72 and the results of Crawford.73
The motivation for the test for exogeneity rests on the observation that if the variables assumed to be exogenous in a simultaneous equations model are not, in fact, uncorrelated with the structural disturbances, significant inferential problems arise. Since all of the asymptotic properties claimed for estimators rest on the assumption of uncorrelated structural disturbances, this specification error can have quite serious consequences.74
A single equation version of the Wu-Hausman test has been devised by Spencer and Berk.75 It is used here. The null hypothesis is that the variable of interest is exogenous. The test is based on the existence of two alternative estimators – one that is consistent and asymptotically efficient under the null hypothesis and one that is not asymptotically efficient under the null hypothesis but is consistent under both the null and alternative hypotheses. By comparing the estimates from both estimators and noting that their difference is uncorrelated with the efficient estimator when the null hypothesis is true, a test is derived based on the asymptotic distribution of the difference in the two estimators. The test statistic is distributed as a chi-squared with the number of degrees of freedom equal to the number of variables of interest. Since the equations for the number of analog cable channels and the number of digital cable channels are considered separately, the number of degrees of freedom is one.
The Wu-Hausman test is conducted twice. In the first instance, the demand relationships serve as the reference and in the second the supply equations serve as basis of the test. For the number of analog cable channels using the demand equations as the reference and the number of channels measured as the number of BST plus CPST channels less the number of broadcast channels, the computed value of the test statistic is 0.071. This is less than the critical chi-squared value of 3.841. Hence, the null hypothesis is accepted implying that the number of analog cable channels is exogenous. This same result also holds when the other measures of the number of analog channels are used. When the supply equations are used as the reference, the computed value of the test statistic is 1.170. Again the null hypothesis is accepted. Also, while the computed values of the test statistic are not reported for the other measures of the number of analog channels used, they produced analogous results. Reporting exhaustive details of these results in this instance provides no insights.
For the number of digital cable channels equation, just the number of digital cable channels on the highest subscribed tier is used as the number of channels measure. Most subscribers of digital cable service purchase just the major digital tier although there are frequently additional digital tiers offered. For the number of digital channels using the demand equations as the reference, the computed value of the test statistic is 2.186. This is less than the critical chi-squared value of 3.841. The null hypothesis is accepted and the number of digital cable channels is concluded to be exogenous. When the supply equations are used as the reference, the computed value of the test statistic is 0.626. Again, the null hypothesis is accepted.
There is one final issue to be considered before turning to the estimation of the simultaneous equations model. This involves the presence of outliers in the data. Regression diagnostics is a tool for assessing the quality and reliability of regression estimates. It is used here to test for the presence of outliers. It is especially useful for cross sectional data (the type being used here) in helping to determine in a systematic way the location of data points that are either unusual or inordinately influential.76
Regression diagnostic tests were developed by Belsley et al.77 The tests systematically search for unusual or influential data. That is, regression diagnostics look for observations that lie outside patterns set by other data, or those that strongly influence the regression results. The impact of such data points is rarely apparent from even a close inspection of the basic data series. The basis of regression diagnostics is an analysis of the response of coefficient estimates to controlled perturbations of the inputs including the parameters to be estimated, error and model specification, and the ordering of the data. Four separate regression diagnostic tests are available including RSTUDENT, HATDIAG, COVRAT, and DFFITS. The interested reader is referred to Belsley et al.78 for information on the distribution of the test statistics.
Each of the demand and supply equations is subjected to the regression diagnostic tests individually. An observation is judged to be an outlier if it fails two or more of the tests. For the analog cable demand equation, seventeen observations are judged to be outliers, for the digital cable demand equation, twenty five observations are judged to be outliers, and for DBS demand, seventeen observations are judged to be outliers. There are no overlaps in the observations determined to be outliers. That is, there is, for example, no observation that is judged to be an outlier for both analog cable demand and digital cable demand. Moreover, a closer inspection of the observations judged to be outliers do not reveal any obvious problems such as a misplaced decimal point in recording the data or a data entry error. For the supply equations, analog cable has twenty two outliers and digital cable has twenty five outliers. Again there are no overlaps in the observations judged to be outliers. There also are no overlaps with the demand equations outliers.
For each demand and supply equation, a qualitative (dummy) variable is introduced to mark the outlier. The variable is defined to equal zero if two or more of the regression diagnostics tests are passed and one if two or more are failed. Table A in the Appendix provides specific definitions of the regression diagnostics dummy variables.
Estimation Results for the Simultaneous Equations Model
A considerable amount of preliminary analysis went into determining the final specification of the simultaneous equations model for multichannel video programming service. There is little to be gained from recounting all of the details this preliminary analysis.79 Simply note that the objective has been to develop a model that is theoretically consistent and one that yields credible and robust coefficient estimates80 but also one that can give some insight into the nature and extent of the substitutability and, consequently, the potential for competition, between cable television service and DBS service. The model is also data dependent. That is, it has been developed based on the data collected on the 2002 FCC Annual Cable Price Survey.
For the final estimation, not all 694 observations could be used. For nine of the observations, the Zip code information was either missing or inaccurate and could not be corrected. Additionally, there were 103 cable systems that did not provide digital cable service in 2002. It was necessary to drop these observations from the data set. The majority (i.e., 70.1 percent) were either very small or small noncompetitive systems or LEC systems in the competitive group. This gave a total of 582 observations used in the estimation. By dropping these observations, any inferences with regard to the behavior of subscribers and cable systems and the impact of the competitive fringe (DBS service) on the dominant firm (cable service) must be qualified accordingly. This understanding is implicit and the point is not pursued in the subsequent discussion.
The preferred estimation technique is the full information maximum likelihood approach. Other studies presenting a simultaneous equations model of the market for cable service have generally used three-stage least squares estimation. With the full information maximum likelihood approach, the likelihood function for the entire system is maximized by the choice of all system parameters, subject to all a priori identifying restrictions. The resulting estimators are consistent and asymptotically efficient. They also have the same asymptotic properties as three stage least squares estimators, including the same asymptotic covariance matrix. A major advantage of full information maximum likelihood estimation approach over three stage least squares, however, is that with this technique it is possible to use in the estimation process a wide range of a priori information, pertaining not only to each equation individually but also to several equations simultaneously.81
The coefficient estimates of the demand portion of the simultaneous equations model of the demand and supply of multiprogramming video subscription service is presented in Table 2.82 The variables are defined and their sources are given in Table A in the Appendix.83 Each equation in the model is overidentified which allows for the estimation of all of the coefficients. That is, for each equation the number of excluded predetermined variables exceeds the number of included endogenous variables. The rank condition for each of the equations is also satisfied. The rank condition specifies that the total number of variables excluded from the equation must be at least as great as the total number of endogenous variables of the system less one.84 There are five endogenous variables including the number of analog cable subscribers, the number of digital cable subscribers, the number of DBS subscribers, the analog price of cable service, and the digital price of cable service.85 The number of analog and digital cable subscribers includes the number of subscribers reported for the franchise area (i.e., estimated by the cable system responding to the survey questionnaire). The number of DBS subscribers is the estimated number of DBS subscribers in the franchise area. The analog price represents the price for the combined BST plus CPST service and the digital price represents the price for digital service for the tier with the highest number of subscribers plus the price for BST and CPST analog service.86
|Table 2. Simultaneous Equations Model Coefficient Estimates|
1. Analog Cable Demand Equation
Constant-2456.495879.38-0.41781[.676]Analog Cable Price-24.3872156.451-0.15587[.876]Digital Cable Price59.8857161.9330.36981[.712]Households Passed0.204300.0154013.2662[.000]Number of Analog Channels8.5965062.76080.13697[.891]Unemployment Rate (%)881.840211.2904.17360[.000]Fuel Type (Electricity)62.068320.87692.97307[.003]House Value0.030940.003648.48803[.000]Proportion of DBS Subscribers-17598.83146.19-5.59370[.000]RD Analog Demand65871.32259.8929.1480[.000]
2. Digital Cable Demand Equation
Constant706.5823059.570.23094[.817]Analog Cable Price98.271181.06351.21227[.225]Digital Cable Price-77.221075.7934-1.01884[.308]Households Passed0.085890.012676.77900[.000]Number of Digital Cable Channels51.835420.66782.50802[.012]House Value0.003800.002351.61782[.106]Proportion DBS Subscribers-6480.731786.18-3.62827[.000]RD Digital Demand26863.6669.52140.1236[.000]
3. DBS Demand Equation
Constant-8050.253891.62-2.06861[.039]Analog Cable Price128.530119.2931.07743[.281]Digital Cable Price41.381884.34920.49060[.624]Households in the Franchise Area0.052300.0024621.2070[.000]Number of Digital Cable Channels-26.470221.0587-1.25697[.209]Number of Analog Channels-15.260741.6833-0.36611[.714]Language Spoken (Spanish)-6.6290618.7264-0.35399[.723]Retirement Income0.135630.054252.49998[.012]Elevation57.449343.05101.33445[.182]House>1050.318312.90913.89788[.000]RD DBS Demand47577.7978.07248.6443[.000]
The results are rather mixed from a statistical significance point of view. For none of the equations in the model is the price of analog cable service or the price of digital cable service statistically significantly different from zero. This is different from several previous studies where, for example, Ford and Jackson87 find that a one percent increase in price of cable service results in a 2.4 percent reduction in the number of cable subscribers and the General Accounting Office88 finds that a one percent increase in price results in a 2.1 percent reduction in the number of subscribers. The important difference in the studies finding statistically significant price effects and this one is that just a single service – analog cable service – is considered in these other studies. The demands for the three unique types of multichannel video programming service are considered here. Additionally, the data sets employed in the estimation in each of the studies are different as are the estimation techniques.
The most significant factor explaining both analog and digital cable demand is the number of households passed in the franchise area. This variable is introduced in the demand equations to provide for a control of the market size. Larger systems, for example, have a larger number of potential customers and, mutatis mutandis, have a larger number of subscribers. The simple correlation between the number of analog cable subscribers and the number of households passed is 0.84. A similar result holds for digital cable subscribers where the simple correlation is 0.83. The estimated coefficients suggest that for each additional 100 households passed by a cable operator, 20 will subscribe to analog cable service and 8 will subscribe to digital cable service. This result is consistent with the findings of Ford and Jackson89 and Beard et al.90 That is, they find a comparable analog cable demand response to a change in the number of households passed to that estimated here.91
Since the simultaneous equations model assumes equilibrium, the supply of analog and digital cable service should increase by amounts comparable to the increase in the demand for analog and digital cable service, respectively, when the number of households passed increases. This is precisely what the estimation results indicate.92 As the cable system expands to provide service to previously unserved areas or as the number of households increase because of in-filling construction,93both the demand and supply (see Table 3) of analog cable service increase by comparable amounts.
For DBS the simple correlation between the number of subscribers and the number of households in the community is 0.67. The estimated coefficient suggests that for each additional 100 households in the franchise area, the number of DBS subscribers will increase by 5.
The number of analog channels is not statistically significantly different from zero at the 5 percent level or better in the analog demand equation. The implication of this is simply that subscribers do not value the number of channels offered for BST plus CPST cable service less the number of broadcast channels and that it is not a suitable variable by which to measure cable service quality. Several different measures were also considered including the number of BST channels, the number of CPST channels, the sum of the two, and the total number of channels available including premium and pay-per-view. None of these alternative measures proved to be statistically significantly different from zero. This result is not inconsistent with the finding of others. For example, Ford and Jackson94 do not find the number of channels offered by a cable operator to influence the number of subscribers. Mayo and Otsuka95 likewise do not find the number of channels to be a statistically significant factor in explaining the demand for analog cable service.
It is interesting to speculate on why the number of channels offered by a cable system does not prove to be a statistically significant factor in explaining the demand for analog cable service. The incumbent nature of analog cable service coupled with consumer habit formation or inertia as subscribers become accustomed to a certain number of channels in the analog cable offering is most likely a factor in this. One thing that appeared in the preliminary analysis was that when a Houthakker-Taylor stock-habit type specification was used to characterize the demand for analog cable service, the number of subscribers in the previous period is so highly correlated with the number of subscribers in the current period that the impact of any independent variable in the specification disappears (in a statistical significance sense).96 Thus, the data are not able to discern any impact of the addition or subtraction of a channel to an analog cable systems offering nor are they able to identify any effect of other independent variable in this type of specification. Habit formation is simply too pronounced in the demand for analog cable service. Additionally, a hedonic price study using data from the 2002 FCC Annual Cable Price Survey indicates that only 24 of 105 channels are valued by subscribers.97 Hence, it is not possible to differentiate between the analog demand for cable service for a system with 45 channels versus one with 50 channels if many or most of these channels are not valued by subscribers.
With regard to digital cable demand on the other hand, there is a positive and a statistically significant at the five percent level relationship between the number of digital cable subscribers and the number of channels offered on the major digital tier. For each additional digital cable channel offered on the major digital tier, the cable operator attracts about 52 additional digital cable subscribes in a franchise area.
The number of analog cable subscribers is positively related to the unemployment rate. Moreover, this result is statistically significant at the 5 percent level. This has interesting implications suggesting that as the unemployment rate increases, the number of cable subscribers increases because those who are unemployed have a higher propensity for cable subscription than those who are employed, all other things equal. Obviously such a conclusion needs additional empirical verification. None of the other cable demand studies, however, explicitly consider this factor.
As the number of households in a franchise area who have electricity as their primary type of energy increases, there are more analog cable subscribers. This result is statistically significant at the five percent level. House value is used as a proxy for purchasing power or affluence in a franchise area. The more affluent an area or the relatively greater purchasing power it has, the larger will be the number of analog cable subscribers, all other things being equal. This suggests that analog cable service is a normal good.98
Defining an appropriate measure to capture accurately the impact of purchasing power on the demand for analog and digital cable service proved to be problematic. Use of more conventional measures of purchasing power such as median income and per capita income or a variant of these did not yield statistically significant estimates for the analog demand for cable service. Most other studies either do not find a straightforward income measure to be significant (e.g., Beard et al.99 or if they do, they generally have the wrong sign (e.g., General Accounting Office100).101 The proxy for purchasing power or affluence for digital cable demand is only marginally significant (at about the 10 percent level) suggesting that other factors besides relative purchasing power are more important in explaining the demand for digital cable service.
The penetration of DBS in a franchise area has a quite significant and negative effect on both the number of analog cable subscribers and the number of digital cable subscribers. Recall that this measure is being used to capture the degree of competition that cable systems as the dominant firm face from noncable providers, the fringe firm.102 A one percent increase in the number of DBS subscribers will result in a reduction of about 176 analog cable subscribers in a franchise area and a reduction of about 65 digital cable subscribers. These results are statistically significant at the 5 percent level. This result is consistent with that found by the General Accounting Office,103 the only other study explicitly looking at DBS penetration in the context of its impact on the demand for cable television service. In percentage terms, a one percent increase in DBS penetration results in about a 0.9 percent reduction in the number of analog cable subscribers in a franchise area and about a 0.8 percent reduction in the number of digital subscribers. While the loss of subscribers by cable systems to an increase in DBS penetration is not relatively large, it is statistically significant.
The regression diagnostics variables in the analog and digital cable demand equations are introduced to account for anomalous behavior in the data for specific franchise areas. These variables are defined in such a way as to not completely negate the impact that the outlier data might have on the estimation results.104 As such they do not have any readily interpretable meaning. They just note that the observations of interest (i.e., those judged to be outliers in preliminary analysis) have an identifiable relationship with the number of analog and digital cable subscribers. This observation holds for each of the regression diagnostics variables in each of the demand and supply equations.
The estimation results do not indicate that DBS demand is influenced by the number of channels offered on the major digital tier by a cable system, the number of analog cable channels,105 or the language spoken. There has been the suggestion that subscribers who live in households that are fluent in languages other than English have a higher propensity for subscribing to DBS service because of its expanded foreign language offerings. Thus, the expectation was that franchise areas with a relatively larger proportion of households where, e.g., Spanish is spoken fluently would have more DBS service subscribers. The results reported here are for the variable for Spanish.106 Other foreign languages were also considered singly and jointly with other language variables in preliminary analysis. None of the variables proved to be statistically significantly different from zero at the 5 percent level.
Retirement income is positively related to the demand for DBS. This result is statistically significant at the 5 percent level. Thus, the greater the retirement income of households in a franchise area, the larger the number of subscribers, other factors remaining unchanged. This is consistent with what one would expect. A one percent increase in retirement income leads to a 0.21 percent increase in DBS demand. There are no other studies that have explicitly considered this factor in modeling the demand for multichannel video programming service.
The angle of elevation does not seem to be a significant factor influencing the demand for DBS service. The factor is considered because if the angle of elevation is low enough, surrounding ground clutter, buildings, or other factors might serve as an impediment to satellite signal reception. The General Accounting Office107finds the angle of elevation to be a significant factor impacting the penetration of DBS. The results here do not lend support to that conclusion.
The final factor considered with regard to the demand for DBS service is the percentage of multi-dwelling housing units in a franchise area that have ten or more units. This variable is introduced to capture the effect of market size on the demand for DBS service and, consequently, serves to reinforce the number of households in the franchise area variable which was also included in the DBS demand equation. The relationship is positive and statistically significant at the 5 percent level. The estimate indicates that a one percent increase in the number of multi-dwelling housing units in a franchise area that have ten or more units leads to a 0.6 percent increase in the number of subscribers.
The coefficient estimates of the supply portion of the simultaneous equations model of the demand and supply of multiprogramming video subscription service is presented in Table 3. As before, the variables are defined and their sources are given in Table A in the Appendix.
|Table 3. Simultaneous Equations Model Coefficient Estimates|
1. Analog Cable Supply Equation
Constant6224.7311622.60.53557[.592]Analog Cable Price-61.0659171.929-0.35518[.722]Digital Cable Price-55.2595171.659-0.32191[.748]Households Passed0.162950.0159710.2035[.000]Number of Analog Channels-36.544872.7487-0.50234[.615]Size – Medium4059.288249.620.49205[.623]Size – Large7057.418242.740.85619[.392]Size – Very Large17013.58202.562.07417[.038]Type – LEC2087.204172.190.50026[.617]Type – Overbuild-418.1824565.06-0.09160[.927]Type – Municipal-1335.5515953.9-0.08371[.933]Type – LP-2764.584844.45-0.57067[.568]Regulated1551.991385.301.12033[.263]Competitive-1101.849080.23-0.12134[.903]Cluster1093.011292.530.84563[.398]MSO-1510.355132.58-0.29426[.769]Population Density0.198990.079312.50884[.012]Capacity4.207427.549370.55732[.577]RD Analog Supply60177.71830.8532.8687[.000]
2. Digital Cable Supply Equation
Constant2345.103436.150.68247[.495]Analog Cable Price3.2044186.21050.03717[.970]Digital Cable Price-86.378076.3291-1.13165[.258]Households Passed0.0789900.012686.22949[.000]Number of Digital Cable Channels66.691621.19883.14601[.002]Regulated238.504730.0660.32668[.744]Competitive-2993.75798.690-3.74832[.000]Population Density0.123570.044912.75133[.006]Capacity1.440003.284430.43843[.661]Subscribers Nationally0.000110.000061.67730[.093]RD Digital Supply25395.8850.22429.8695[.000]
There are a few factors clearly impacting the analog supply of cable service at the 5 percent level of significance. Besides those discussed previously, these factors include whether the cable operator is very large and the population density of the service area. Very large cable systems have a larger supply. Also, as population density increases, analog cable supply increases. Both of these results are consist with what was expected ex ante. Thus, larger cable systems realize economies of scale in their operations and are able to supply the marginal (most recent) subscriber at a lower cost. Additionally, an increase in population density reduces such things as trenching and cabling costs resulting in an increase in supply. For example, a one percent increase in population density leads to a 0.2 percent increase in the quantity of analog cable supplied. Note that there is nothing comparable in other studies to support the integrity of this estimate. Other studies of the cable industry explicitly ignore supply side considerations. Studies that introduce some sort of quasi-supply equation in the form of an ad hoc price equation find mixed results with regard to a measure of potential subscriber density. For example, neither Beil et al.108 nor Rubinovitz109 find it to be a significant factor while Emmons and Prager110 find it to be statistically significant at the 5 percent level.
With regard to the statistically insignificant factors that might potentially impact the supply of analog cable service, there are a couple of issues to note. First, most cable systems in the sample are both part of a cluster and are affiliated with a MSO.111 Hence, the lack of finding of any statistical significance for these variables is most likely an artifact of there being not enough variation in these explanatory variables across the entire sample. Also, system capacity measured in terms of megahertz does not appear to be significant in explaining the supply of analog service. A larger capacity should lead to a higher cost and reduced supply. This is not indicated in the data from the 2002 FCC Annual Cable Price Survey. This result, however, is also probably a function of there not being enough variation across cable systems in the sample.
Finally, digital cable supply increases in a statistically significant way as population density increases. This is the same result observed for analog cable supply. For each one percent increase in population density, the quantity of digital cable supplied increases by 0.14 percent, all other things remaining unchanged. The number of subscribers nationally, a proxy for the size of the cable system nationally, is only marginally statistically significantly different from zero (at the ten percent level). It was expected that large national companies with a greater number of subscribers could demonstrate a relatively greater degree of market power in purchasing components to supply digital cable service so that their costs would be lower. With lower costs supply would be greater. The results do indicate that the larger the cable system, the larger the supply but only marginally.
One interesting result is the positive relationship between the number of cable channels in the digital tier with the highest number of subscribers and digital cable supply. This result is statistically significant at the 5 percent level. As the number of cable channels increase, costs such as programming costs, increase. This should lead to a reduction in the supply. This is not, however, what the estimation results indicate. This is obviously an anomalous result.
This study has developed a simultaneous equations model of the market for multichannel video programming service consisting of cable service and DBS service. The objective has been to measure the extent to which DBS service competes with cable television for subscribers in this market. After recounting the institutional characteristics of the market, the theoretical underpinnings of the modeling approach are presented. The model being pursued is known as the dominant firm-competitive fringe model. In this model, as is characteristic of the multichannel video programming service market, the dominant firm (cable operator) sets a price that fringe firms (DBS service providers) take as given in deciding how much to supply. The empirical results presented endeavor to determine (1) whether de facto the competitive fringe constrains the dominant firm�s pricing behavior and (2) if so, to what extent.
The simultaneous equations econometric model is rigorously developed with attention focused on many of the important modeling issues frequently overlooked in studies of the market for cable service. These issues include the appropriate functional specification, endogeneity of the number of channels offered, and the presence of outliers in the data.
The model estimated consists of five equations – demand equations for analog cable service, digital cable service, and DBS service and supply equations for analog cable service and digital cable service. The model is estimated using the full information maximum likelihood technique. The estimation results are mixed. With regard to the issue of interest, the competitive fringe does serve to constrain the dominant firm�s pricing behavior. As the penetration of DBS increases, both the number of analog and digital subscribers is reduced. The estimates, however, do not indicate that changes in the relative price of analog or digital cable service have a quantifiable impact on the demand for DBS service.
|Table A. Variables and Their Sources1. Analog Cable Demand is the total number of cable subscribers to just analog cable service for a franchise area. The data come from the 2002 FCC Annual Cable Price Survey. It is computed as the total number of cable subscribers less the number of subscribers taking digital cable service.2. Constant is just intercept term.3. The Analog Cable Price is the sum of the BST and CPST cable price. The data come from the 2002 FCC Annual Cable Price Survey.4. Digital Cable Price is the price of the most highly subscribed digital tier. The data come from the 2002 FCC Annual Cable Price Survey.5. The Number of Households Passed is the number of households passed by a cable operator in a franchise area. The data come from the 2002 FCC Annual Cable Price Survey.6. The Number of Analog Channels represents the number of BST plus CPST channels less the number of broadcast channels offered by the cable operator. The data come from the 2002 FCC Annual Cable Price Survey.7. The Unemployment Rate is the rate of unemployment corresponding to a specific cable franchise area keyed by Zip code. The data come from the 2000 Census of Population.8. The Fuel Type (Electricity) represents the percent of houses in a cable franchise area whose primary energy type is electricity. The data come from the 2000 Census of Population.9. House Value is the median value in dollars of a house in the cable franchise area. The data come from the 2000 Census of Population.10. Proportion of DBS Subscribers is the ratio of the total number of multichannel video programming subscribers who subscribe to DBS to the number of subscribers to all multichannel video subscription service. The data come from the 2002 FCC Annual Cable Price Survey.11. RD Analog Demand is a qualitative (dummy) variable introduced to mark those observations that are judged to be outliers for the analog cable demand equation as observations of interest.12. Digital Cable Demand is the number of cable subscribers who also subscribe to the most hightly subscribed digital tier for a franchise area. The data come from the 2002 FCC Annual Cable Price Survey. It is computed by multiplying the percent of subscribers taking digital cable service by the total number of subscribers.13. The Number of Digital Cable Channels is the number of channels offered on the digital tier. The data come from the 2002 FCC Annual Cable Price Survey.14. RD Digital Demand is a qualitative (dummy) variable introduced to mark those observations that are judged to be outliers for the digital cable demand equation.15. DBS Demand is the number of DBS subscribers in a cable system�s franchise area. The data come from the 2002 FCC Annual Cable Price Survey.16. Language Spoken (Spanish) is the percent of households in the franchise area in which Spanish is spoken fluently. The data come from the 2000 Census of Population.17. Retirement Income is the monthly amount of retirement income received by residents in the franchise area. The data come from the 2000 Census of Population.18. Elevation is the angle of elevation of the satellite dish required for a DBS Subscriber to receive a satellite signal. The data are taken from the DirecTV web site.19. House>10 is the percentage of multi-dwelling housing units in the franchise area that have ten or more units. The data come from the 2000 Census of Population.20. RD DBS Demand is a qualitative (dummy) variable introduced to mark those observations that are judged to be outliers for the DBS demand equation.21. Analog Cable Supply is the total number of subscribers to which analog cable service is being supplied in a franchise area (i.e., subscribers to BST service). The data come from the 2002 FCC Annual Cable Price Survey. It is computed as the total number of cable subscribers less the number of subscribers taking digital cable service.22. Size – Medium, Size – Large, and Size – Very large correspond to the definitions of medium, large, and very large noncompetitive cable systems found in the text. The value of the variable is defined to equal one if the cable operator meets the definition criterion and zero otherwise.23. Type – LEC, Type – Overbuild, Type – Municipal, and Type -LP correspond to the definitions of LEC, Overbuild, Municipal, and LP competitive cable systems found in the text. The value of the variable is defined to equal one if the cable operator meets the definition criterion and zero otherwise.24. Regulated denotes whether the basic service tier price is subject to local regulation for the franchise area. It is defined to equal one if there is local regulation and zero otherwise. The data come from the 2002 FCC Annual Cable Price Survey.25. Competitive denotes whether the FCC has made a finding of effective competition within the community. It is defined to equal one if there is effective competition and zero otherwise. The data come from the 2002 FCC Annual Cable Price Survey.26. Cluster denotes whether the cable system is part of a MSO cluster of two or more systems. It is defined to equal one if the cable system is part of a cluster and zero otherwise. The data come from the 2002 FCC Annual Cable Price Survey.27. MSO denotes whether the cable system is affiliated with a multiple system operator (MSO). It is defined to equal one if the cable system is affiliated with a MSO and zero otherwise. The data come from the 2002 FCC Annual Cable Price Survey.28. Population Density is defined to equal the total population of the franchise area divided by the geographic extent of the area (i.e., the number of square miles). The population data come from the 2002 Census of Population and the geographic extent data come from the U.S. Postal Service.29. Capacity is the cable system capacity as of July 1, 2002 measured in terms of megahertz. The data come from the 2002 FCC Annual Cable Price Survey.30. Digital Cable Supply is the total number of subscribers to which digital cable service is being supplied in a franchise area. The data come from the 2002 FCC Annual Cable Price Survey. It is computed by multiplying the percent of subscribers taking digital cable service by the total number of subscribers.31. RD Analog Supply is a qualitative (dummy) variable introduced to mark those observations that are judged to be outliers for the analog cable supply equation.32. Subscribers Nationally is the total number of subscribers that the parent company of the individual cable system has in the United States. The data come from the 2002 FCC Annual Cable Price Survey.33. RD Digital Supply is a qualitative (dummy) variable introduced to mark those observations that are judged to be outliers for the digital cable supply equation.|
The authors are with the Industry Analysis Division, Media Bureau, Federal Communications Commission, Washington, DC 20554. The views expressed are those of the authors and do not necessarily represent the policies of the Federal Communications Commission or the views of other Federal Communications Commission staff members.