New methods for mapping media concentration: network analysis of joint ventures among firms

Measuring media concentration

Debates and discussions about how to measure media concentration have been one of the major concerns of media scholars and policy media makers around the world. Since the Hutchins Commission report, which examined the consequences of concentrated ownership in the American journalism industry as well as how the emergence of new communication technologies would affect the journalism industry, there has been no consensus on how to address media concentration (Downing, 2011; Iosifidis, 1997; Murdock and Golding, 1973; Noam, 2018). Nevertheless, the different regulatory bodies around the world, including the Federal Communications Commission (FCC) in the United States, the Office of Communications (Ofcom) in the United Kingdom, the Canadian Radio-television and Telecommunications Commission (CRTC), the Instituto Federal de Telecomunicaciones (IFT) in Mexico, and the Korea Communications Commission (KCC), among others, evaluate the concentration of their media markets by using the same parameters as other industries. The most commonly used indices are the HHI, which is calculated by squaring the market share of each company competing in the market and then summing the resulting numbers, and the concentration ratios CR4 and CR8, which represent the market share of the four and the eight largest firms, respectively (Albarran, 2002).

However, the convergence between the media, telecommunication, and technological industries is exacerbating the difficulty in adequately evaluating the concentration in these respective industries, as transindustrial conglomerates hold properties in multiple industries. Furthermore, a particularly unique challenge related to media industries is the methodological challenge of how to measure whether media industries are ensuring a plurality of voices and a diversity of cultural representations. As Noam (2016) remarks, ‘media is a powerful instrument of influence over hearts, minds, wallets, and votes’ (p. 4). This is important to emphasize particularly because other industrial sectors do not have these same capabilities. Indeed, in noting the public good and other unique characteristics intrinsic to media industries, Victor Pickard (2014) frames this particular shortcoming of media as a market failure.

To compensate for the shortcomings of existing indices, Noam (2009) proposed the Media Ownership and Concentration Diversity Index (MOCDI), also known as Noam Media Concentration index (NI), which takes into account the number of voices available. According to Noam (2016), market power indices like the HHI do not accurately reflect media diversity. Thus, the HHI measure misrepresents the actual degree of pluralism in media markets. Noam’s model incorporated the idea that each media outlet has to be counted as a ‘voice’. The intention of Noam’s index is to try to reflect more accurately the health of the marketplace of ideas rather than simply measuring market power. He explains,

as a radio listener, I am better off with 20 stations on the dial or another newspaper sold at the news kiosk, even if fewer people listen to or read them. Their availability provides an option that carries value even if it is unexercised by most readers and authors. (Noam, 2016: 22)

To incorporate this logic into his index, Noam’s MOCDI takes a commonly accepted measure of market power (HHI) and divides it by the root of the number of voices in the market. In doing so, the MOCDI index is an effort to combine a measure of market power with a measure of the potential plurality of voices (Noam, 2009). However, this index still does not reflect the cross-ownership or other alliances and collaborations between different companies in the market. However, Noam (2016) also proposed an additional index to capture the cross-ownership concentration, which he calls the media power index (MPI). He explains that

the [MPI] aggregates the squared shares of the same firm in different national media markets, adjusting for market size. It is the sum of a company’s market shares in the markets it operates, summed up across the various markets in which the company operates, weighted by market size. (p. 24)

But, again the MPI still does not capture the collaborations and alliances between the firms. What this demonstrates is that, despite numerous attempts to measure the concept of ‘concentration’ in media markets, we are still left with a blind spot when it comes to addressing collaboration between firms, most notably a measure of the extent to which these firms collaborate through different types of alliances and joint agreements.

Thus, this article will focus on how different media companies collaborate. From our point of view, the possibility of tracking the alliances and collaboration of the media industries is an important blind spot for the traditional or commonly used concentration indices. It should be noted, however, that scholars working within the critical political economy of communication tradition have long been concerned with such issues (Bagdikian, 2004; Garnham, 2000; Meehan, 2005; Murdock and Golding, 1973; Wasko, 1982). In fact, Meehan (2005) points out that

all Big Five were connected through client/supplier relations, whether via licensing of network programming, development of deals, cable channel carriage, or cable channel programming that ranged from television series to films to music videos. While each company had its own mix of such relationships, none of the Big Five lacked such entangling contracts: these relations, allowed under deregulation, fostered the intertwining of corporate interests and codependence of operations. (p. 80)

Similarly, Garnham (2000) argues that in the case of communication industries and companies it is crucial to understand how these companies share ‘resources rather than . . . exchanging discrete, substitutable, and therefore competitive, products or services’ (p. 59). Thus, the media companies work more as networks, because they optimize their value when everyone is connected. In the case of communications ‘these networks of distribution are particular importance, since communication is primarily about the exchange of meanings within a flow of symbols’ (Garnham, 2000: 59). At the same time, mapping networks between media firms offer a more complete picture of the overall media industries, while also allowing for an illustration of shared interests between firms internationally rather than evaluating media markets separately within national contexts. Indeed, Eli Noam has also attempted to map the international alliances between media firms, as exemplified by his ambitious Who Owns the World’s Media, in which he applied the HHI and CR4 index at worldwide level for the first time (Noam, 2016).

In this study, we track the networks that are generated by the alliances and collaborations of the Big Five ² media conglomerates companies (Bagdikian, 2004; Birkinbine et al., 2017; Meehan, 2005) in the context of digital convergence and ongoing mergers around the globe. However, this context is set, to some degree, by public policies that either promote or inhibit such a scenario from existing. Indeed, this has been pointed out by political economy of communication scholars when they argue that media systems are a direct output of explicit public policies. In words of McChesney (2004), ‘most dominant media firms exist because of government-granted and government enforced monopoly broadcasting licenses, telecommunication franchises, and rights to content. Competition markets in the classic sense are rare; they were established or strongly shaped by the government’ (p. 19). In other words, media market structures are shaped by the public policies of the government for action or omission. Following Des Freedman (2014), ‘when we think about questions of ownership . . . policy should be understood both as empirical fact and as an ideological tool’ (p. 86). For this reason, studies of the interlocks between large firms can be valuable to media policy makers, so they have better tools of analysis to evaluate the particularities of concentration in the media industries.

Furthermore, for those policymakers who claim to be interested in maintaining competitive markets and free competition among firms, Garnham (2000) counters by arguing that concentration

is a general problem for those in favour of competitive markets because it leads to oligopoly or monopoly where a small number of firms accumulate sufficient market power to manipulate the market in their favour at the expense of consumers. (p. 54)

At the same time, Baker (2007) explains why media ownership patterns ought to be a priority for public policy and identifies two main reasons for opposition to media ownership concentration: (a) it does not achieve a more democratic distribution of communication power and (b) it does not provide a democratic safeguard against abuses of economic, political, and cultural power. In other words, Baker argues that increased concentration inevitably leads to a public loss of democratic power, and this loss of democratic power comes from a lack of diversified media sources. However, at the same time, we recognize and agree with Stuart Hall (1986) when he claims that the analysis of media ownership and concentration is not a sufficient ‘explanation of the way the ideological universe is structured, but it is a necessary starting point. It gives the whole machinery of representation its fundamental orientation in the value-system of property and profit’ (p. 11).

Finally, before specifying the method used in our analysis, it is worth pointing out that mapping interlocks between media corporations is not necessarily entirely new, as some of the previous discussion reveals. Arsenault and Castells (2008), for example, also mapped the different partnerships, cross-investments, board members, and managers between the Big Five media firms plus Google, Microsoft, and Apple as well as Bertelsmann. Figure 1 comes directly from their study, and it demonstrates different types of interlocks between the firms. Their article in general demonstrates the networks of collaboration between these firms, while focusing on investments and partnerships. Maps like these are certainly valuable tools in visualizing the interlocks between firms. However, the availability of network analysis software, which is primarily used to conduct social network analyses, opens up a new possibility for mapping relationships between firms. In the following section, we explain how we conducted this study, as well as some of the potential methodological shortcomings of the approach.

OPEN IN VIEWER

Figure 1.

Key interlockings between multinational media and diversified Internet corporations.

Method

For the purposes of this study, we used network analysis techniques to map the joint ventures between the largest American media firms. This study focused solely on joint ventures because these properties represent shared interests among the top firms. As such, these relationships provide the clearest example of how the top firms cooperate within media markets rather than competing against one another in terms of ownership. The top firms selected for analysis were Comcast, the Walt Disney Company, Time Warner, National Amusements (owner of CBS and Viacom), and 21st Century Fox/News Corporation. However, in the interest of measuring the consequences of approved and announced mergers in the media industries, we mapped the joint ventures among the firms before the mergers as well as how these relationships would be affected after the mergers. Most notably, we mapped joint ventures prior to AT&T’s acquisition of Time Warner and again after the merger. We also completed pre- and post-merger analyses for the proposed merger involving 21st Century Fox.

To construct the dataset necessary for the network analysis, we relied on publicly available documents. This practice has been used consistently by scholars working within the political economy of communication tradition (Meehan, 2005; Murdock and Golding, 1973; Wasko, 1982, 2001). The United States Securities and Exchange Commission (SEC) requires all publicly traded companies to file 10K annual reports that provide information about a company’s areas of business, board of directors, pending lawsuits, properties owned, and select financial information, among other disclosures. In this study, we primarily relied on the Form 10K filings for each of the largest media companies (Comcast, the Walt Disney Company, Time Warner, CBS, Viacom, 21st Century Fox, and News Corporation). The one exception here is National Amusements, which is a privately held company that owns both CBS and Viacom. However, to gather information about the joint ventures of CBS and Viacom, we were able to rely on the Form 10K annual reports. We then recorded National Amusements’ control of both companies in the database for the network analysis. Furthermore, we also included the 10K forms from AT&T once the merger with Time Warner was approved.

The database that provided the foundation for our network analysis was coded in the following ways. From the Form 10K documents, we coded all the disclosures related to ownership or shared interests with other firms. These measures were separated into two primary spreadsheets: one for node data and the other form edge data. Within the node table, any other company that was mentioned in the 10K forms was assigned a unique number and a tag for the industry within which it operates. For each of the large firms, we also gathered data on annual revenues, net assets, total income, and geographic location of the company’s headquarters, and we also included a separate column for any notes that provide greater context or clarification for the quantitative data.

We then mapped the relationships among the nodes in a separate spreadsheet for edge data. The edge data were constructed using the following measures: source, target, type, weight, and notes. Source data included any other company mentioned in the Form 10K from the large firms. The large firm was then recorded in the target column. For example, the Walt Disney Company was listed as the target, while the source would be ESPN. The type of relationship captured by this measure was labeled as ‘subsidiary’ since ESPN is a jointly owned subsidiary of the Walt Disney Company and the Hearst Corporation. The weight measure captured the percent ownership assigned to each firm. In the ESPN example, the Walt Disney Company was assigned a weight value of .80 because of its 80% ownership stake and Hearst was assigned .20 for the remaining 20%. Finally, the notes column was where we listed the source of the information. In nearly all cases, the Form 10K was the source of the information. However, in certain cases, we needed to consult additional publicly available documents to verify or clarify otherwise limited information contained in the Form 10K.

The node data and edge data provided the foundation for the network analysis. However, this database captured more than simply shared interests among the large firms; in effect, this database provides the beginnings of mapping the corporate structures of the large firms, which includes subsidiary properties that are 100% owned by the large firms. For the current study, however, we were simply interested in the shared interests among firms, which were mapped by parsing the dataset to include only joint ventures. Moreover, after parsing the dataset to only include joint ventures, we could then create network maps to demonstrate how those relationships would be altered once additional mergers or acquisitions were approved. Therefore, we simply needed to replicate the network analysis after altering the source and target data for those properties included in the proposed merger.

To complete the network analysis, we used Gephi, which is an open-source network analysis and data visualization software. The software not only allows for a visualization of network relationships, but it also allows for some customization in the way that network relationships and nodes are presented. For this study, the size of the node is determined by the company’s total annual revenues. While annual revenues are only one indicator of a company’s size, they can provide some perspective on how the annual revenues of AT&T, which is a new entrant into the market for ‘traditional’ entertainment media, compare to other media firms in the industry Comcast or the Walt Disney Company. Furthermore, this type of visualization can also reveal the existing ties between so-called ‘media giants’ (Birkinbine et al., 2017) and other large telecommunications firms, like Verizon. Indeed, one of the other beneficial options within Gephi is the ability to highlight specific nodes of the network to reveal the specific interconnections that exist between a single node and its affiliates. Consequently, we are able to isolate a single company, like AT&T, to reveal the degree to which it will now share interests with other firms in the industry through joint ventures and other shared equity interests.

In sum, network analysis provides some clear advantages for mapping industrial structures. First, researchers can get a macro-level perspective on the overall structure of the industry. Primarily for the purposes of this study, we can visualize the joint ventures and shared equity interests of firms within the industry. Doing so can illustrate the degree to which supposedly competing firms are actually cooperating with one another. Second, researchers can also focus on a specific node in the network to reveal the ways in which it is connected to other nodes in the network. This micro-level perspective can be used to determine the degree to which a specific company shares interests with other firms in the industry and controls particular markets. Moreover, this type of analysis can also be used to illustrate the effects of a new entrant into the industry. Particularly because regulators focus on market share and total revenues of top firms, these types of shared interests are not often represented in any regulatory consideration of overall market structure. At the same time, this analysis could be extended to map the interactions between technological and information industries, for example, Google, Apple, Facebook, and Amazon.

Results

Figure 2 provides an illustration of the joint ventures between the largest media firms in the United States before AT&T’s merger with Time Warner. From this initial illustration, we can already observe a couple of interesting points. First, keep in mind that the size of each node in the network is determined by the total revenues reported by each company for the fiscal year 2017. With even a cursory glance at the illustration, readers can quickly see that the telecommunications firms, AT&T and Verizon Communications, are the largest nodes in the network. This should underscore the degree to which these firms already outpace media firms when considering their annual revenues. The largest media firms visible in this map are Comcast, which has holdings in both telecommunications and media, the Walt Disney Company, and Sony Corporation. National Amusements appears as a smaller node in the network because its revenues are divided among its two primary subsidiaries, CBS Corporation and Viacom. Similarly, after splitting into two companies in 2013, both News Corporation and 21st Century Fox would appear comparatively small on this map. Indeed, since News Corporation is now largely cut off from joint ventures with other firms, we have not included it on the illustration.

OPEN IN VIEWER

Figure 2.

Map of joint ventures between media firms in the United States, 2017, before merger.

The second point to make about this initial figure is the fact that even before its proposed merger with Time Warner, AT&T was already integrated into the media industries. For example, AT&T was already connected to the Sony Corporation through a joint venture in the Game Show Network. Similarly, AT&T was also connected to Televisa, which is based in Mexico City and is the largest entertainment multimedia conglomerate in the Hispanic American and Spanish-speaking world. According to the financials reported by Market Watch (2019), Televisa reported annual revenues of approximately equivalent to 5.3-billion USD in 2018.

Finally, as Figure 2 demonstrates, the large media firms are networked and have overlapping interests despite the fact that they are supposed to be competitors. These shared interests take on two predominant forms: on one hand, the large firms may be directly connected through joint ventures, but the more common occurrence is for these more direct relationships to be masked by a series of holding companies or other subsidiaries. However, when one traces the ownership structure of these individual companies, they ultimately lead back to one of the major media firms. This general tendency is notable here, but we will examine some examples of this practice in greater detail below. Before moving onto those detailed examinations, however, it is useful to analyze the consequences of the AT&T–Time Warner merger in greater detail.

Figure 3 illustrates the joint ventures between the large media firms after the AT&T merger with Time Warner. This particular map was also recalibrated to specifically isolate the large firms more effectively for analysis. The most consequential connection illustrated on this map is AT&T’s direct tie to Time Warner. After the merger, AT&T renamed Time Warner as WarnerMedia, which is also reflected on this map. The extent to which AT&T becomes connected to the other media firms after its acquisition of WarnerMedia can be seen by looked at a more detailed map of these relationships.

OPEN IN VIEWER

Figure 3.

Map of connections between media firms in the United States, 2017, after merger.

Figure 4 provides a more detailed look at AT&T’s interlocks with the other media firms after its acquisition of WarnerMedia. Figure 4 shows those shared interests within only one degree of separation. These immediate connections include WarnerMedia and its interlocks to Warner Brothers film studio, its partial ownership stake of Hulu LLC and the CW television channel, and its ownership of Central European Media Enterprises, Ltd., which is a media company that operates in Bulgaria, the Czech Republic, Romania, and the Slovak Republic.

OPEN IN VIEWER

Figure 4.

Detailed map of AT&T’s first-degree connections via WarnerMedia.

Figure 5 expands the picture of AT&T’s ties to the other major firms by tracing the ownership patterns of its immediate connections. By doing so, we can illustrate how AT&T has ties to all the other Big Five media companies through joint ventures and other shared interests. A few connections are particularly noteworthy in this figure. Comcast is connected to AT&T through its joint ownership of Hulu and Fandango Media. Hulu is also the property that connects AT&T to the Walt Disney Company and Twenty-First Century Fox, although this is the only shared property between those firms. However, given the Walt Disney Company’s pending merger with Twenty-First Century Fox, Disney will consolidate its ownership interests in Hulu. This is especially noteworthy within the context of the latest moves by AT&T to take over WarnerMedia, Disney’s merger with Twenty-First Century Fox, and Comcast’s purchase of Sky. All of these moves seem to suggest there is a growing importance of strengthening holdings in audiovisual content production and distribution, which could be driven, in part, by the emergence of other significant players like Netflix.

OPEN IN VIEWER

Figure 5.

Detailed map of AT&T connections to other major media firms.

Also noteworthy from Figure 5 is the connection between AT&T and other large media corporations based outside the United States. In this figure, AT&T is connected to the Sony Corporation through Sony Pictures Television, which owns 58% of the Game Show Network, while AT&T owns the remaining 42%. Similarly, AT&T is connected to Televisa through its AT&T International and Sky Central America and Caribbean subsidiaries. These examples demonstrate how a series of holding companies or separate operating segments can mask relationships between firms.

Finally, we can also isolate the most important nodes that provide crucial points for facilitating interconnection between the firms. Figure 6 identifies these nodes, while also transforming them into a color-coded pie chart to illustrate the various ownership stakes that each of the large media firms holds in the joint venture. The two nodes that provide the greatest degree of interconnection are Hulu LLC and Rede Telecine. Hulu currently has four different companies with ownership stakes, as discussed earlier, although this may change after Disney’s pending acquisition. For the time being, the Walt Disney Company, AT&T, Comcast, and 21st Century Fox all have ownership stakes in Hulu. Similarly, Rede Telecine provides a crucial point of interconnection by bringing together 21st Century Fox, Grupo Globo from Brazil, Comcast, National Amusements, and MGM Studios in one joint venture. Also pictured here is the CW, which provides a link between National Amusements and AT&T.

OPEN IN VIEWER

Figure 6.

Most important nodes connecting the network.

Discussion

The purpose of this article was to map the joint ventures and other shared interests among the major media firms by using network analysis techniques. We also visualized how those relationships were affected after AT&T’s acquisition of Time Warner, which was subsequently renamed WarnerMedia. The purpose of doing so was to illuminate one of the major blind spots when assessing the extent of market power in the media industries. Commonly accepted indices of market power, specifically the HHI and CR4 indices, only account for market shares and total revenues held by the top firms. However, these indices also assume that firms are competing against one another in the market. What our analysis illustrates is the way that these firms often share economic interests, thereby undercutting the assumption of a fully competitive market.

Furthermore, we wanted to illustrate how network analysis techniques could be used to map relationships among firms. Even though we only focused on joint ventures between the largest media firms in the United States, we think that this type of method could be expanded to map many different types of relationships that exist between such firms. These include co-production agreements, licensing deals, distribution agreements, and other agreements, but it could also easily be adapted to map interlocks between members on Boards of Directors. Similarly, the method could also be expanded to produce maps of all properties owned by each company, which would give a comprehensive picture of how many media ownership are housed within the corporate structures of these large firms.

However, there are a few limitations to the full implementation of such studies, and these limitations can explain some of the shortcomings of the present study. Whereas social network analysis has proven to be a useful tool in analyzing large sets of relationship data, these studies often make use of existing datasets that can be obtained or constructed with relative ease as long as one has the technical capability to obtain the data. Unfortunately, no similar database or repository exists for data about corporate interlocks. As discussed in the methodology section of this article, we needed to rely primarily on government documents, specifically the 10K annual report filings. These documents contain a wealth of information about individual companies, but they do not provide a comprehensive report on all of a company’s joint ventures and subsidiaries, let alone other types of agreements. In addition, the data provided are not structured in a way that makes it conducive to easy retrieval using automated scripting techniques. As such, our initial dataset needed to be constructed manually, and the initial analysis yielded nearly 400 separate node data points and more than 400 edge data points.

Further exacerbating the difficulty of obtaining reliable data is the fact that international regulations differ from country to country. This is particularly difficult in countries that do not have regulations about public disclosures but even when such regulations exist, companies may not report full or accurate data. Therefore, the reliance on self-reported data or data from third party reports undercuts the validity of such figures.

Despite these methodological shortcomings, we think that this initial application of the network analysis technique has given us a valuable look at the alliances and collaboration between media companies. Furthermore, the methodological difficulties do not preclude further attempts to conduct similar analyses at a much broader scale, nor should these difficulties preclude any effort to automate data collection. Doing so could ensure that similar types of analysis could be much more efficient in the future. If we were able to develop a reliable methodology, then we could replicate this study by including additional companies as well as historical data, which would allow us to determine how alliances, ownership structures, and other forms of collaboration have changed over time.

What has been provided here suggests that network analysis can be effectively applied to the study of media industries, specifically to analyze corporate interlocks through joint ventures. Not only does our network analysis demonstrate that these interlocks exist, but it also shows the extent to which they pervade the seemingly ‘competitive’ media market. Indeed, the data visualization provided by our network analyses seems to reinforce Meehan’s (2005) observation that ‘The Big Five collaborate to maintain the transindustrial stability sanctioned by deregulation, by integrating numerous media industries under their aegis and by building alliances with each other across their operation’ (p. 81). Even if the specific companies involved may have changed since Meehan conducted her analysis, the general principle remains the same. Understanding this general principle, as well as visualizing the specific contours of the alliances between media firms, can be a useful tool in assessing the overall health of the media market. Where traditional indices of industrial concentration fall short, however, is their inability to map such alliances. This is precisely where network analysis can supplement these traditional measures to illuminate these blind spots and more effectively measure levels of concentration in the media industries.

Perhaps most importantly, however, the measures of media concentration and the illustration of media networks ought to move beyond the level of simple description. In drawing from the tradition of critical political economy, we want to emphasize that from our perspective the network analysis and network theory ought to be situated within the historical context of power relations and asymmetrical structure of global capitalism. In tracing the growing size and power of corporations historically, we can see that these trends are enabled by regulatory regimes that have either failed or refused to intervene in waves of ongoing mergers and acquisitions. Indeed, one of the hallmarks of the critical political economy tradition has been that of praxis (Mosco, 2009). As such, these measures could be used to intervene in policy debates by providing empirical evidence of noncompetitive markets during review of potential mergers between media firms. When combined with insights from other scholars who emphasize the unique characteristics of media industries, such as the importance of a plurality of voices (Noam, 2009) and the public good qualities of media systems (Pickard, 2014), these measures can be used to emphasize the importance of maintaining a healthy public sphere and ensuring a plurality of voices, both of which are prerequisites for a functioning democracy.