In my previous article in Intermedia (September 2015) I focused on the regulatory and competitive consequences of increasing inter-modal (or horizontal) competition among converging communication infrastructures (based on copper, fibre, coax cable and radio spectrum) on ex ante regulation of communication networks. This convergence of communication networks is one consequence of the development of IP (internet protocol) networks. Another is that networks and services are no longer tied to each other, ie. a network does not specialise in providing a particular service. Instead, content and service provision is network agnostic – similar services can be provided over all networks. This also means that the provision of services is no longer restricted to firms that operate communication infrastructure.
This article analyses the regulatory and competitive consequences of the interdependence between network operators and IP service providers that drives demand and supply in ECMs (electronic communication markets). This interdependence is driven by two key facts:
- Subscriber demand for access to broadband (or IP) networks is 'derived', ie. based on the demand for services provided over these networks rather than the networks themselves. In particular it is driven by supply of a wide variety, high-quality services beyond those directly provided by network operators, such as voice.
- It is important for IP service providers to be able to reach as many broadband (or IP) network subscribers as possible. This is because IP service providers earn revenues from user subscriptions and/or advertising targeted at these users.
It is the combination of networks and services that creates joint value for end-users. I start my analysis from a blank slate (ie. a market with no regulation) and the factors that would determine outcomes in this world. This is useful to understand if and when regulation may be required on a forward looking basis and the potential consequences of existing regulation.
Absent any regulation, network operators and IP service providers would bargain on how to split the joint value they create. The outcome of this commercially negotiated process would reflect the relative bargaining positions or market power of the negotiating firms. I use a Nash bargaining framework to explore the bargaining factors that determine outcomes in this complex dynamic of interdependence to produce joint value and competition for a share of the joint value.
THE NASH BARGAINING FRAMEWORK
The seminal work by John Forbes Nash in 1950 considers bilateral bargaining between a buyer and seller, each with some market power.1 This bargaining game is characterised by distinct components which include:
- Actors or players
- Possible actions of these actors or players
- The payoffs for each outcome/agreement that the game can take
- The information sets available to each actor or player.
Following Nash, a stable equilibrium to this negotiating setting can arise where both parties respond to each other with their 'best response' (ie. each party maximises its own gains given its expectations of the other parties' actions and thus has no incentive to deviate from the 'best response').
Absent regulation, operators and IP service providers would bargain on how to split value.
Nash shows that under relatively simple assumptions, this means that the players will maximise the difference between the payoffs of reaching an agreement and the payoffs of not reaching an agreement. As such, the value to the parties of agreement relative to non-agreement determines the bargaining power and hence the division of any surplus (or joint value) created by reaching an agreement. I consider the four bargaining components in the context of the current game – the interaction between network operators and IP service providers.
Actors or players
The relevant players in the ECM game are network operators and IP service providers. I define these players as follows:
- Network operators supply physical connectivity to end users – an IP/broadband pipe – and some services over these networks such as voice and, increasingly, media services.
- IP service providers deliver services using these broadband pipes and usually do not own and/or run broadband access networks. IP service providers can offer a range of services such as communication and information sharing services (Skype, FaceTime, WhatsApp etc.); audio and video entertainment services (Netflix, YouTube, Spotify, Apple Music etc.); and a plethora of other services offered by internet platforms (Google, Facebook, Amazon etc.).
- Both network operators and IP service providers enjoy a degree of market power. This market power is based on control of 'infrastructure inputs' (usually by network operators) and 'digital inputs' (usually by IP service providers):
- 'Infrastructure inputs' include various network inputs required to run an electronic communications network. These network inputs include passive and active elements. The passive elements can include ducts/poles, fibre, copper, coaxial cables and radio spectrum; and the active elements can include electronics to control and transmit IP data over fibre, copper or coaxial cable. In addition buildings (exchanges) and cabinets are required to store the electronic equipment.
- 'Digital inputs' include proprietary expertise (like search and pricing algorithms), resources (like computer and data scientists), user generated data (like search and browsing history, device location) and media content (user generated and professionally produced content).
Infrastructure and digital inputs have different characteristics in terms of the pace of development, availability, scalability, network effects and upfront costs (see box overleaf). These differences are reflected in the degree of market power provided by these inputs:
- On the one hand, the faster pace of development (for example of algorithms compared with physical access networks), availability (digital inputs are non-rivalrous in use), and low upfront investment for digital inputs compared with infrastructure inputs may mean that digital inputs are easier to acquire and that IP services are relatively more contestable compared with network services. This means that entry and exit by firms is easier.
Market power is based on control of infrastructure inputs and digital inputs.
- On the other hand, the ability to scale digital inputs globally (compared with regionally or nationally for network inputs) and exploit a first mover advantage and network effects at the
company level (compared with at the industry level for network inputs) may mean that particular digital inputs are not be easily replicable. These factors mean that some IP service providers may have substantial market power (at least temporarily).
The availability and control of digital and/or infrastructure inputs are important determinants of the possible actions and outside options of network operators and IP service providers (outside options are other options available to the players in case of non-agreement between any two players). And these possible actions and outside options will be important determinants of the bargaining power of these two players.
Possible actions (access to inputs and outside options)
- Access to digital and infrastructure inputs
As discussed above, in a Nash bargaining framework, the lower the value lost due to a non-agreement the stronger the party's bargaining power. In a bilateral bargaining framework between a network operator and IP service provider, the ability to start offering services provided by the other party (expand into the other's 'territory') will be an important determinant of bargaining power. This is because it reduces the loss due to non-agreement (assuming no outside options for the moment). The extent to which a network operator or IP service provider is able to expand its scope of operations and offer services provided by the other party will depend on its access to digital and infrastructure inputs.
For example, if a network operator and Skype disagree on transmission terms, the network operator can still provide voice and messaging services (services provided by Skype) as the network operator already provides these services and owns the inputs required to do so. However, a network operator may not be able supply internet search, email or social media services of comparable quality and functionality to those provided by IP service providers. It may not have access to all the digital inputs (algorithms, search data history etc.) required to provide these services. Similarly, most IP service providers will not find it commercially viable to start offering network services. This is because the high upfront fixed investments of setting up communication networks and the relatively long time period (often over 15 years) to recover these investments are likely to act as a barrier to entry.
The entry does not need to take place but should be credible to increase bargaining power.
Some form of limited entry into each other's 'territory' may be feasible and indeed important to demonstrate the ability to credibly reduce the value lost due to non-agreement. This is because in a bargaining context what matters is the network operator's (IP service provider's) belief about the incentives and ability of the IP service provider (network operator) to expand into the other's territory. The entry does not actually need to take place but should be credible to increase a party's bargaining power.
differenceS between infrastructure and digital inputs
Pace of development
- The pace of development/innovation of digital inputs is faster compared with infrastructure inputs, given the potential for continuous user feedback and 'learning by doing' in the case of digital inputs. For example a search algorithm can be continually improved based on user feedback on search results produced by the algorithm. Moreover such advances are usually made by individual firms whose business model is driven by the introduction of new services with the aim of disrupting existing business models.
- In contrast, innovation and development of network inputs takes place at a slower pace. For example the rollout of new broadband technologies like 4G and DOCSIS takes years to develop and rollout. Moreover these innovations are at the industry level, not company specific. This means there is less differentiation in broadband compared with IP services.
Availability – Rivalrous versus non-rivalrous use
- Infrastructure inputs are rivalrous in use and their supply is limited (ie. there are natural limits on the availability of these inputs) – for example two network operators cannot use the same radio spectrum simultaneously, and the supply of radio spectrum suitable to operate mobile networks is limited at any given time.
- Digital inputs are non-rivalrous – the same algorithm or data set can be used by two IP service providers simultaneously – and in principle the supply of digital inputs is not constrained. New algorithms can be invented/developed using scientific expertise and data can be collected from users.
- Infrastructure inputs scale by geographic area. Scale economies and subscriber network effects are local and/or national in scope.
- Digital inputs can scale globally which means that IP service providers can exploit global scale economies in production and subscriber acquisition.
- Network effects for a service exist when more users make a service more valuable for existing as well as potential new users. Network effects usually operate at the industry level for network operators as subscribers on different broadband networks can interact directly with each other. They can do this based on voice or messaging services provided by network operators or using IP services (for example messaging services like WhatsApp).
- Network effects for IP service providers usually operate at the company level as different communication or social media services (for example WhatsApp or Facebook) are not interconnected. I note that this is not always the case – for example users of different email services can contact each other.
- Network services require high upfront investment in infrastructure inputs. For example to set up a mobile network or rollout a fibre access network. Once undertaken these investments are generally irreversible, ie. exit without incurring a substantial write-down on the assets purchased may not be possible.
- In contrast the availability of cloud services like Amazon Web Services means that the upfront costs required to set up an IP service are low and can scaled up and down with the business. The absence of high and sunk fixed costs also means that exiting the market is easier (less costly). See amzn.to/2038bcJ for details of Amazon Web Services.
Indeed there are instances of such limited entry. For example Google is investing in fibre access networks in selected US cities,2 and considered bidding for the 700 MHz range mobile spectrum in the 2007 FCC auction to ensure that the spectrum was designated as 'open', ie. mobile phones using software from any manufacturer and software developer are able to run on the network3. Similarly network operators increasingly supply media services and on-demand content, services which are similar to those provided by IP service providers such as Netflix, YouTube and Amazon.
- Outside options
The interdependency between network access and service providers will usually span multiple providers and multiple platforms. In the presence of outside options, non-agreement with one party leaves open the possibility of reaching an agreement with another party which provides similar IP services or an alternative network. Hence the value of non-agreement will also depend on the availability of outside options in addition to the ability to expand into the other's 'territory'.
It is possible to apply the bargaining framework to a wider number of players that bargain simultaneously.4 This reflects the fact that there are other options available, and this will also be an important determinant of the value of agreement with a particular party compared with non-agreement with the same party.
- Outside options for IP service providers – substitutable network providers
The outside options for IP service providers will depend on the number, type and size of networks available.
The higher the number of network operators available, the more potential network partners with which an IP service provider can come to an agreement with and the less likely it is to lose from non-agreement with any one operator. This assumes that different networks are good substitutes, and in general this will be true for the same type of network (for example fixed access) as the supply of broadband connectivity by different network providers is reasonably equivalent.
However, there will be differences among different types of networks – for example between fixed and mobile networks in terms of the geographic locations at which access is provided (higher for mobile networks) and capacity provided (higher for fixed networks). Hence the substitutability of different outside options (networks) for a particular IP service provider will depend on whether the IP service is best delivered using mobile networks, fixed networks or both. For example the navigational features of Google maps or taxi hailing services like Uber work best on mobile networks. Conversely, media rich services like Netflix work better over fixed networks (especially when delivered to large screens).
Finally, the larger the subscriber base of a network operator the more valuable it is to an IP service provider (because IP service providers benefit from being able to reach the maximum number of broadband subscribers possible). Hence smaller and larger operators may not be good substitutes because the value lost due to a non-agreement with the large operator will be higher.
- Outside options for network operators – substitutable IP service providers
Cooperation with IP service providers can help network operators differentiate their service offerings to help attract new subscribers, reduce churn, and potentially increase revenues by driving data consumption. However, different IP services will not be the same in this respect. Some IP services will be more attractive to end users and more effective differentiators than others. This means that the value of agreement (relative to non-agreement) will not be the same for different IP service providers.
The value of non-agreement will also depend on the availability of outside options.
For example, IP service providers with a large number of users (locally or globally) will benefit from greater network effects valued highly by broadband subscribers, and be more valuable to network operators. Social networks like Facebook and LinkedIn are good examples of the advantages of such direct network effects. For other services like Netflix and Amazon it is not other users but the presence of content providers or sellers that attracts subscribers, and the number of subscribers in turn attracts content providers and sellers. The scale of such 'indirect' network effects may also be important in determining the value of an IP service provider to a network operator. IP service providers that add more value than others will be in a stronger bargaining position.
Pay-offs from agreement
The value of agreement to each party will also vary with the type of agreement reached between a network operator and IP service provider. The type of agreement may vary among preferential, non-discriminatory and partial access.
- Preferential access
A network operator could give an IP service provider preferential access to its subscribers through different types of (commercial) agreements. For example, a network operator can use traffic management policies (like QoS differentiation) to prioritise traffic for a particular IP service provider (like OTT video) to ensure a better user experience5 or apply 'zero rating' for the IP service provider's data traffic (ie. these data do not count towards the subscriber data cap). Similarly, an IP service provider could agree to supply its services exclusively to a network operator's subscribers (this is unlikely in practice given IP service providers benefit from reaching and targeting as many users as possible via competing broadband networks). Another option would be for an IP service provider and network operator to agree to bundle services together. For example, Vodafone in the UK bundles Spotify, Netflix, Now TV, or Sky Sports with certain mobile subscriptions.6
As BEREC, the European regulators body, notes, there are an increasing number of partnerships between IP service providers and network providers in Europe though "the partnerships do not yet seem to provide any exclusive feature or content compared to if purchased and used outside of the context of the partnership".7
- Non-discriminatory access
This means that the IP service provider does not have preferential access to a network operator's subscribers. When generalised to all IP service providers it is the underlying principle behind different forms of net neutrality rules such as the FCC's open internet order, the European Parliament and Council's measures concerning open internet access, and TRAI's (Telecom Regulatory Authority of India) prohibition of discriminatory tariffs for data service regulation.8
In effect, net neutrality rules enforce non-discriminatory access to a network operator's subscribers and prohibit other forms of vertical contracting.
- Partial access
Partial access means that the IP service provider's access to the network operator's subscribers is partially blocked or degraded. For example the IP service provider may lower the transmission quality and/or block some functionalities of the service. Recent proposals to introduce 'ad blocking' at the network level by some mobile networks (such as Digicel and Three)9 can be seen as an example of network operators trying to gain a larger share of the joint value by degrading the transmission of some IP services.
The value of agreement to each party will vary from high to low depending on the type of agreement reached. The highest value is likely to be associated with preferential access, then non-discriminatory access and finally partial access.
The ability of each party to estimate and generate value from an agreement will also depend on the information set that each party has, and the freedom to experiment with different business models. The information set refers to each party's knowledge about the value of agreement compared with disagreement. A complete information hypothesis is when both bargainers know the value of agreement and disagreement.
However, this may not hold in practice and a party with private information about its outside options and/or the value it can derive from the agreement will be in a stronger bargaining position. Different business models (subscription, free or some combination of the two) may be used to generate value in ECMs, and these require different information sets.
In a subscription driven business model (with no other revenue sources), the value generated from the service (and from each customer) is explicitly stated and hence observable and measurable. This business model does not require extensive data on users. A free service does require extensive data on users. User generated data (location information, search history, browsing history, social network interactions etc.) are captured by the service provider and used to serve targeted advertising.
It is this targeted advertising that generates revenues and there may be no direct monetary payment from users to the service provider. This means that it is difficult to observe or measure the value created by a specific user or sets of users (for example the subscribers of a particular network operator).
The framework highlights the complexity as all factors interact simultaneously.
Network operators mostly use subscription based business models to generate revenues whereas many IP service providers use the free business model (or combinations of free and subscription). The absence of direct monetary payments and the provision of multiple IP services (using different business models) means that it may be difficult for a network operator to estimate the value added by any one IP service provider. Network operators may also not be able to accurately estimate the value of user generated data.
As discussed above, network operators and IP service providers are mutually dependent on each other – broadband connectivity expands the reach/ availability of IP services and in turn IP services increase the functionality of broadband connectivity.
The Nash bargaining framework highlights that the competitive dynamics in ECMs is determined by the level of competition among network operators (an important determinant of the outside options for IP service providers) and among IP service providers (an important determinant of the outside options for network operators); and by bargaining on the type and value of agreement. The framework also highlights the complexity of this interaction as all these factors interact simultaneously and do not work in isolation.
In this framework, existing or proposed regulations in ECMs will affect the bargaining dynamic between network operators and IP service providers. This is because regulation (for example wholesale network access regulation, net neutrality or data privacy rules) will strengthen the bargaining position of one party over the other (see also box above). Three key takeaways based on the Nash bargaining framework are as follows.
1 Competition law which allows for case specific analysis may be a better policy instrument.
The complexity of the interaction between network operators and IP service providers and the many factors that affect bargaining outcome means that each case will be different. For example the market power held by network operators is likely to vary across countries and will depend on the level of infrastructure and service competition. For IP service providers, the level of market power and the source of this market power is likely to vary across firms. Given the diversity in the sources of market power, which may or may not result in competition problems, it may be difficult to design proportionate ex ante regulatory remedies. A case specific application of competition law may be a better policy instrument.
INTERACTION BETWEEN REGULATION AND THE BARGAINING FRAMEWORK
POSSIBLE ACTIONS OF THE PLAYERS
- Wholesale access regulation: The presence or absence of wholesale network access regulation is an important determinant of the availability of infrastructure inputs, and hence the outside options (alternative network operators) available for IP service providers.
- Interconnection: Interconnection determines whether network effects operate at the industry or company level. The presence of company level network effects combined with the potential to scale globally is an important determinant of the replicability of digital inputs used by a particular firm and hence the outside options (alternative IP service providers) available for network operators.
PAYOFFS FOR EACH OUTCOME/AGREEMENT
- Net neutrality rules: These rules may constrain the freedom to enter into different types of vertical contracts (preferential, non-discriminatory and partial access) and hence the payoff from each outcome/agreement.
INFORMATION SETS/BUSINESS MODELS
- Data privacy rules – for example the EU's e-privacy directive and data protection regulation (GDPR) and recent broadband consumer privacy proposal by the FCC – constrain the ability to collect and use private data by both network operators and IP service providers. See Europe and FCC.
- Asymmetries in the application of these rules (in terms of restrictions placed on the collection and use of private data) between network operators and IP service providers will affect the information sets available to these players and the ability to experiment with different business models – for example free versus subscription.
For example it not clear that an 'a priori' ban on different types of vertical agreements imposed by net neutrality rules is justified. This is because in some instances there may be benefits from vertical contracting arrangements which place some restraints on the two parties. A way forward may be to provide both network operators and IP service providers with some flexibility when implementing these rules – for example, allowing preferential access agreements when these do not adversely affect competition in the relevant market.
2 A priori, it may be difficult to associate particular bargaining outcomes with consumer harm.
The bargaining framework and interaction between network operators and IP service providers only relates to the share of the joint value captured by each party. Only if the division of the surplus results in a reduction of downstream competition for consumers (by network operators or IP service providers) and/or there are negative dynamic effects such as a reduction in investment or innovation is there a case for consumer harm and hence intervention.
- 3 If there is potential consumer harm, and regulation is proposed, then the effect of regulation on the bargaining power of both network operators and IP service providers should be considered.
If regulation is warranted then it should be based on an analysis of the availability of all (wholesale) digital and infrastructure inputs, the effect of the proposed regulation on the bargaining power of both network operators and IP service providers, and how this translates into better outcomes for consumers.
Finally, I note that the analysis above is not exhaustive. It does not consider all the factors that may affect the bargaining position of network operators and IP service providers. For example, differences in time preferences (discounting) may also be relevant.10 Other economic models from cooperative game theory such as the Shapley value (interpreted as the expected marginal contribution of the IP service provider or network operator to joint value created) may provide valuable insights.
References 1 Nash Jr. F (1950). The bargaining problem. Econometrica 18 (2): 155-62. bit.ly/1ovGuv9 2 See fiber.google.com/about for rollout plans. 3 Levy S (2011). In the Plex: How Google thinks, works, and shapes our lives. pp 222-3. 4 See for example: Krishna V and Serrano R (1996). Multilateral bargaining. Review of Economic Studies 63: 61–80. 5 See for example: Nokia Siemens Networks (2013). Netflix over a QoS enabled LTE network. White paper. bit.ly/1qqcbb4
6 See Vodafone's entertainment page: vdfn.co/1dHLhVb 7 BEREC (2015). Report on OTT services. p32. bit.ly/1RXNQ8u 8 Details on these regulations: US – fcc.us/1V4LBke Europe – bit.ly/1UDThtC India – bit.ly/1Q2YC68 9 Cookson R (2015). Digicel first mobile group to block ads in battle against Google. Financial Times, 30 September 2015. bit.ly/1qkK9hr Cookson R (2016). Three to become first European mobile operator to block ads. Financial Times, 19 February 2016. bit.ly/200JAoF
10 For example, Ariel Rubinstein further advances the Nash model to show the importance of the value of time to the players. This modifies the basic model to include the time preferences (discounting) of the players as well as a step-wise structure – ie. it is possible for multiple iterations of the offer/reject cycle. Under this modification, it is possible for the more patient player to exploit the other's impatience and extract a higher share of the surplus. See: Rubinstein A (1982). Perfect equilibrium in a bargaining model. Econometrica 50 (1): 97–100. bit.ly/1N5de4S
- Wednesday, 01 June 2016