Communications Insights: Trends and Cool Stuff

The conceptual foundations and the economics Network Neutrality

Remarks as Prepared for Delivery

Good Morning and thank you for that warm introduction.  It is truly a pleasure for me to speak to you this morning.  Before I begin, I would like to thank Stefano Quintarelli for inviting me and the Fondazione Ugo Bordoni for hosting me here today.  I would also like to thank Sebastiano Trigila and his staff for their terrific help in arranging this event.

We would like to think of Network Neutrality as being in the vanguard – one the leading edge issues in communications.  Well, while the implementations may be new, the ideas are old.  For example, in our research project at the WIK, we found a reference to an 1848 law in New York State which mandated that telegraph companies provide non-discriminatory service to individuals and to competing telegraph companies.  Substitute the word datagram for telegram, and you have a Network Neutrality regulation.

Yet, the debate goes back even further.  What is truly at the core of the debate is common carriage versus private carriage. It is in fact particularly fitting that we are here today in the Eternal City to discuss this ideas. Common Carriage – the duty to service all customers without discrimination – traces its origins to the laws of the Roman Empire.

As such, common and private carriage express fundamental and competing notions of fairness versus economic efficiency. We think it is unfair to give preferential treatment to certain customers (those who are willing to pay more).  At the same time, we also think it is economically inefficient to mandate a single (or limited set of) Internet access options for everyone, including those who are willing to pay more for premium services. So, how does one determine which means of achieving differentiated services is socially permissible and economically desirable?

The answer to that question depends on a variety of things.  The question itself is like a Rorschach test.  Look at Network Neutrality, tell me what you see, and I will tell you whom you work for, your educational background, your political ideology, your stock portfolio, and maybe even something about your personality.  This is perhaps why the debate in the United States created strange and almost surreal coalitions including the religious right, the American Civil Liberties Union, left-of-center netroots organizations and the Gun Owners of America.  Opponents on both sides of the issues agree on one thing, if their agenda is not adopted as regulatory policy, innovation and investment will be stifled and consumer welfare and overall economic activity will suffer.  Much of the public policy debate has been the external face of a private bargaining problem involving upstream content providers and downstream network providers.

The term Network Neutrality has come to describe all manner of sin.  There is no generally agreed upon definition for Network Neutrality and deviations or violations against the principle.   It is a catch-all phrase that emerged in the United States over the past decade and has come to reflect a number of potential behaviors that some have considered to be anticompetitive. Network Neutrality implies that all IP packets should be treated more-or-less the equally, and the debate reflects concerns that they might not be in the future – that a network operator might somehow apply different treatment to IP packets associated with different services, applications, destinations or devices.

The term Network Neutrality is, at best, a loaded one. It implies that any deviation from its principles is not neutral, and in some way unfair. In many instances, it is perfectly fair to provide preferential treatment to some network uses, say premium customers or emergency services. However, the opponents of Network Neutrality had devised a counter term which is almost equally misleading. In recent discussions involving the US FCC, the term network management practices has come to take the place of the term Network Neutrality.  While the semantics of this term might be more objective, it does not accurately describe the nature of the problem.  Network management refers to a much broader scope of activities including business practices, sales and marketing, security, fault tolerance, and capacity planning. Since the issues present in this debate represent only a narrow slice of network management.  This term was not picked at random, but comes from a US FCC decision which I will discuss a bit later.

Furthermore, Network Neutrality covers a broad range of behaviors, which at the extreme are clear acceptable or not clearly not acceptable.  Again, how do we decide which are tolerable which are not. The behaviors at the ends most people would agree are or are not acceptable.  The challenge comes in determining which of the behaviors in the middle are acceptable.  I fully expect that I will show this slide one day and be told by some Estonian botnet herder that customer requested SPAM filters are not acceptable under Network Neutrality.  This challenge is made more difficult by the fact that these behaviors can be insidious and difficult to detect.  Network operators are loathe to disclosing the precise contours of their network management for legitimate reasons of competitive advantages and network security.  Most Internet subscribers are not sophisticated enough to detect deviations from Network Neutrality on their own.

Well, further compounding the problem is the fact that Network Neutrality presents the classic problem of the elephant and the blind men. This may be due to the fact that a single network practice may give rise to several different economic, policy and technological issues, involving several different classes of persons. Given that there is little agreement on what exactly Network Neutrality is and there is no common language to describe the issues, I set out to create taxonomy designed to articulate and characterise different types of Network Neutrality issues.  It is my hope that this set of classifications will become widely used because it will give us a langua Franca to describe the things we disagree about.  I hope that this taxonomy will make the debate more objective and approachable, and help to reduce the ever-present hyperbole.

In order to describe the nature of relationships in an interconnected-multilateral-all-IP network world, we can characterize relationships among network participants and service providers as being one three dimensions: vertical, horizontal, or diagonal. Vertical conflicts are those between players in the value chain – content sources, network operators, and endusers. Issues between subscriber and provider are typically those of market power. Horizontal conflicts are those between persons in the same link of the network value chain. Horizontal conflicts arise between, says, endusers competing for finite bandwidth in the access network. The horizontal issues which arise are related to pool problems or issues of bargaining power.  Common pool resources are one type of economic good which is rivalrous and non-excludable.  That is a fancy way of saying that the inherent characteristics of the good make it such that one person’s enjoyment of it, diminishes another person’s, and there is no means of preventing individuals from consuming it. In common pool resources, people tend to consume the good without respect to their negative impact on others, leading to a level of consumption which might reduce overall benefit from the good.

Our so-called diagonal conflicts arise when the actions of one party affect the benefits received by a party on a completely autonomous network. Diagonal issues present the most fertile ground for the Network Neutrality debate and issues of first impression.  We observe many of the experiences of Network Neutrality arise when Internet-based networks insinuates itself into the communications and contractual relationships of users who do not directly subscribe to it. Consider the following situation depicted in the Figure, were entities 1 and 2 are access networks and A, B, and C are endusers. Further suppose that Networks 1 and 2 are interconnected, and endusers A and B are both customers of 1. C is a customer of Network 2. We use the vertical dimension describe the network participant’s place in the value chain, i.e., network provider or enduser. An example of a vertical issue would be Network 1’s blocking or degrading of non-affiliated content to its customers A and B. If endusers A and B are competing for network resources, this would be an example of a horizontal issue. Similarly, deviations from Network Neutrality in terms of interconnection between Networks 1 and 2 would also be along a horizontal dimension.  Somewhat more complex is an example of an diagonal issue which might be the case where Network 1 would block or degrade enduser C’s (who is not its customer) ability to exchange messages with its customer, enduser B. An example of this is the Comcast case which I will discuss later on.

Most people discuss the “speed” data networks.  This term is not precise. The more precise description is the end-to-end delay or data transfer rate.  For the Network Neutrality debate, jitter, and packet loss are also particularly relevant measures of network performance.

This transfer rate is much a function of latency and delay.  Delay in an IP network can be compared to waiting times on a ski lift. In the figure, skiers face a variable queuing delay when standing in line to get on the ski lift. The length of the variable delay is based on the number of skiers trying to get on the lift, and the number of skiers that can board per minute. Each skier faces a fixed delay (latency) once on the chair, based on the length of the lift cable and the constant speed at which it moves.  (I assume that the speed of the cable can not be increased). In this analogy, overall delay is the total time that it takes to get up the lift from the moment that one enters the queue at the bottom of the ski slope to the time that one disembarks at the top of the mountain.

The latency can be viewed as being roughly the time to get a packet through the network if there were no contention for resources whatsoever. For each communications link, it is a function of the speed of light through the transmission medium of which the link is comprised, and the length of the link (or the ski lift cable in this example). As long as the end-to-end communications path does not change, the latency of an end-to-end communication can be viewed as being a constant.

Queuing delays, however, are highly variable. For each transmission link in a packet’s path, the packet can be viewed as having been placed on a waiting line (a queue) while it seeks to gain access to the transmission link. The routers that forward packets have substantial buffers in which they maintain waiting lines of packets; however, these buffers are not unlimited. If the offered load is greater than the capacity of the link, the waiting line will get longer, and will eventually overrun the size of the buffer, no matter how large. Routers respond to this condition by simply discarding excess packets, which normally causes no harm for data applications such as email.

Nothing can make the network (or the ski lift) go faster. Rather, prioritised queuing determines which packets are delayed more (or dropped altogether) when the network is heavily loaded or overloaded. Consider again the analogy of the ski lift. When the waiting lines are long, we might wish to be favoured by a prioritised express line.  So, in discussions of Network Neutrality, it is not who gets to go faster, rather who gets to go first.

If we all were the same there would be little concern for Network Neutrality or for regulation for that matter.  But we are different and we differ in a variety of dimensions.  These two diagrams show how different applications require different things from the network.  Some applications such as two-way video require high bandwidth and low overall delay.  While email, might only require low bandwidth and can tolerate higher latency.  At present, we are putting heterogeneous demands on what was largely designed as a homogeneous network.  This is sure to cause conflicts as heterogeneous users contend for limited capacity.

Let’s talk for a moment about access networks.  In discussing Network Neutrality issues, we tend to focus exclusively on access network issues because this is where competition issues tend to be the most problematic.  Consider a plain-vanilla broadband internet connection.  All of the capacity in the circuit can be used for just about any IP-based application.  Now consider a more complicated cable- or fiber-based network.  While it may appear to be a single network, it is really three, or so, networks sharing the same wire or fiber.  Indeed, there might be three completely separate sets of network equipment attached at either end of the line – Internet modem, cable box and phone terminal.  When we talk about Network Neutrality, we tend to focus solely on the Internet component.  But the other networks, certainly the video programming component is not neutral.  Certain frequencies in the network have been reserved, and prioritized for certain applications.  So, why such a focus for the Internet component? Perhaps this is because it is the most important for the future of communications.

Many of the concerns that have been raised in regard to Network Neutrality relate to behaviors that, in the absence of market power, would tend to enhance consumer welfare. In a competitive market, these practices would be entirely appropriate.  Competition is one of four essential ingredients for mitigating the negative impacts of deviations from Network Neutrality.  The other three are: well informed consumers; low switching costs; and economic feedback loops that provide proper incentives.  The presence of these four factors means that regulators have less work to do in order to ensure that welfare-enhancing differentiation does not devolve into welfare-diminishing discrimination.  Network neutrality also brings into play economic issues including price discrimination, network externalities, transaction costs, switching costs, two-sided markets, and the economics of vertical foreclosure.  Only some of which I will talk about today.

Opponents of Network Neutrality intervention frequently point to the most common problems and claim that this is a two sided market, which if left to its own devices will sort out an efficient, market clearing price.  A two sided market is a fairly common arrangement where a market maker, must successfully bring together participants in two classes of persons in order to make a profit.  The classic example is the bar owner who may give away free drinks at a “lady’s night” ensure a sufficient number of women in order to attract men as paying customers.  In networks, this might be the cable television provider who must attract a certain number of viewers in order to attract program provides and to be able to sell advertising, and vice versa.  In some cases, the programmer is selling advertising and in some cases the cable provider is.  Two sided markets can be complex in that subsidizing one side with zero, or even negative prices, can be rational and even welfare enhancing.  Now imagine that we are talking about the Internet.  Here the threat to Network Neutrality is the market maker attempting to extract a rent from the programmer’s advertiser – someone with whom he does not have privity of contract.  This is, to use the language of our three dimensions, a diagonal issue.  The fear of this “toll gate” problem is probably what brought together strange coalitions in the US.  It probably would not fly in a competitive market, since it has little to do with bringing the two sides together.  Where there is little or no competition, it is most likely an unacceptable practice.

Let me change gears a bit and talk about Network Neutrality in the US.  Why did it take so long for Network Neutrality to become a real problem in the US? Deviations have been technically feasible for more than ten year.  Telecommunications firms in the US were not permitted to discriminate (prior to 2005), but cable operators were unrestricted. The problem has heated up in the US coincident with the decline of competitive broadband alternatives for consumers.  This was also coincident with an aggressively deregulatory time period.

In August of 2005, the FCC adopted a Broadband Policy Statement. This statement does not have the enforceable weight of a Commission rule, but the Commission committed to incorporating these principles into future policymaking. The policy principles in the Statement were intended to “ensure that broadband networks are widely deployed, open, affordable, and accessible to all consumers.” The Statement further set out four entitlements for consumers which it felt necessary to further this goal:

• consumers are entitled to access the lawful Internet content of their choice.
• consumers are entitled to run applications and use services of their choice, subject to the needs of law enforcement.
• consumers are entitled to connect their choice of legal devices that do not harm the network.
• consumers are entitled to competition among network providers, application and service providers, and content providers.”

There are inherent tensions and ambiguities in these principles.  Further, in order to make these policies enforceable it is necessary to craft meaningful rules to properly address them.  Further, the Policy Statement reflects current views of the Commissioners at the time it was issued. Currently, only two of the five FCC Commissioners who were present when the statement was adopted are currently serving. One Commissioner is already planning to leave to join the Rural Utilities Service.  Will their votes on future proceedings be the same as those of the commissioners who signed the Policy Statement? What happens when the next Commissioner is appointed?  Hearings for two or three Commissioner, including the Chairman will be held in about three weeks.

I feel compelled to talk about the Comcast Peer to Peer case.  There are, however, other cases which might be more interesting. but this is where the FCC has already attempted to apply these four principles. What Comcast was doing was using re-set packets to stop the transmission of files in various peer-to-peer networks, notably BitTorrent.  What is most interesting is that Comcast was only resetting the connections when non-Comcast subscribers were receiving files from Comcast subscribers.  Again this is a diagonal Network Neutrality issue.  The reason Comcast engaged in this practice was probably to conserve network capacity for the files its subscribers deliberately wanted to exchange, not just offered to passively exchange.  This would improve network performance for its paying customers without Comcast having to invest in adding additional capacity.  This behavior was in fact rather difficult to detect at first since most Comcast subscribers who could complain, would only observe better performance. The practice was only first notice by a network engineer who was a Comcast subscriber.  He was making his uncopyrighted Barbershop Quartet music making freely available on BitTorrent.  This is not the kind of music which would have a huge following; however, he was surprised to find that absolutely no one was downloading it. The fact that Comcast felt it could do this suggests that it felt it had some form of power. Further, Comcast had not disclosed this practice to its subscribers.  It is questionable whether if disclosed, Comcast subscriber would change providers.  However, if every ISP were to engage in this practice, then all peer-to-peer networking would collapse.

After an AP story, in November 2007, Free Press, a public-interest organization, filed a complaint against Comcast and petition for declaratory ruling with the FCC.  The FCC launched a proceeding against Comcast for failing to follow its Policy Principles, claiming jurisdiction to enforce federal policy.  The FCC never alleged a violation of law or one of its rules.  Instead, launched what was really an administrative proceeding combined with an enforcement action.  In the adjudicatory proceeding, the FCC sought to determine whether Comcast’s actions were violating consumers’ right to “run applications and use services of their choice,” and the degree to which Comcast’s action might constitute “reasonable network management practices.” In addressing the latter concern, the FCC inquired whether Comcast’s network management practices were “carefully tailored to its interest in easing network congestion.”

This might strike you as an odd approach.  Even so, the FCC does have the legal authority to do so, and there is precedent for this.  In the end, the FCC found that Comcast was violating its policy and ordered it to:

• precisely disclose to the Commission its network management practices;
• submit a compliance plan with interim benchmarks; and
• disclose to the Commission and to the public the details of its future network management practices.

In the end, the Commission called on Free Press, Vuze and others in the Internet community to vigilantly observe the practices of ISPs and to bring complaints about those behaviors to the Commission.  I think that the FCC’s approach produced an appropriate result in the instant case, despite procedural concerns. Perhaps this is because the Comcast violation was egregious.  I do not think this is a good way for European regulators to go in the future, or the US for that matter.

Network Neutrality concerns have been raised in Europe (for example, by the BBC’s iPlayer).  Network Neutrality manifests itself very differently, and much less problematically, in Europe than it does in the United States.  To begin with, Europe today enjoys a far more competitive broadband market than does the United States. On average, more than half of all retail DSL lines in Europe are provided by competitive entrants.  That said, trying to address Network Neutrality challenges through ex ante regulation is likely to prove extremely difficult.  It is no different from the problems the FCC has faced.  Thus, the first line of defense for European policymakers should instead be to avoid the problem altogether by maintaining the competitiveness of the underlying markets.  Occasional or sporadic problems related to Network Neutrality might be addressed ex post through the exercise of competition law.  Finally, the European regulatory framework affords European regulators more and better tools for addressing the problems.

In sum, we are still faced with the following dilemma:  on the one hand it is very hard to describe ex ante what is a serious and significant violation of Network Neutrality. On the other hand, it is very hard to make broad, sweeping concepts actionable or enforceable as rules.

The most effective way to promote welfare-enhancing differentiation as opposed to anti-competitive discrimination is the presence of effective, sustainable competition. This is, by no means, a guarantee. As several US experts correctly observe, the US has four major wireless carriers, but no truly open mobile networks. Nonetheless, in the absence of competition to constrain the behavior network providers, some form of regulation will be required. Once the regulator is required to dictate allowable and unallowable network practices, all of the options are unattractive. The market and technology moves too quickly for administrative or legislative management, making effective competition the best fix.  In the presence of effective competition, informed consumer choice and low switching costs, the market will punish welfare-diminishing discrimination.

This is easier said than done.  Managing competition is difficult.  This is especially true where there is an entrenched, powerful incumbent.  This leads to politicized rulemaking and even regulatory capture. Nonetheless, the profit motive is a sufficiently corrupting influence that it must be constrained. The best constraint is the behavior of numerous competing suppliers.  The absence of these constraints, the regulator must impose obligations. Or better yet, figure out how to introduce or reintroduce effective, sustainable competition.

The regulator should view itself as not dividing up a pie among rent-seeking incumbents.  Rather, it should seek policies which make the whole pie larger, by encouraging more participants, more network providers, and more content suppliers to going the network.

I thank you for your time and attention.