A. New 5G Technologies and Capabilities

Table 3.1 shows that 5G will provide data rates potentially up to 20 gigabytes per second (gbps) (v. 4G: 20 megabytes per second (mbps)) and latencies as low as ~1 millisecond (ms) (v. 4G: 20–30 ms), and allow for considerably denser networks, of more than one million connections per square kilometer, which can support massive Internet of Things (IoT) deployments. In addition, 5G networks facilitate private networks (a network built for a specific organization – for example, on a university or corporate campus) and network slicing – that is, reserving part of the network for tailor-made applications for one or more specific clients.

5G also provides greater integration of useful capabilities that were formerly typically located outside cellular systems. For example, MEC (Mobile Access Edge Computing) incorporates processing capabilities at the edges of the network, in part to achieve broader system goals such as low latency. Integration also serves new applications through new specialized subsystems such as C-V2X (Cellular Vehicle to Everything) for safety coordination among vehicles, infrastructure, and other connected devices as well as UAS (Unmanned Aerial Systems) for 5G control of aerial drones.

These new subsystems stand to create value, which mobile operators, infrastructure vendors, users, and suppliers of the attendant new technologies may appropriate. Table 3.2 contains an exemplary list of markets and suppliers that could benefit.

B. New 5G Distributed Architecture Options

The providers of radio access networks (RANs) for 2G, 3G, and 4G have long provided deeply integrated solutions comprising network equipment, software, and services. The situation resembles that of IBM in the mainframe era, in that a small number of incumbents provide end-to-end solutions to mobile operators. These incumbents, along with their consolidated predecessors such as Nortel and Motorola Networks, helped create the mobile industry. At present, four firms dominate the RAN market. Of these, Huawei and ZTE dominate the market in the People’s Republic of China (PRC) but are excluded from several markets, including the United States, where Nokia and Ericsson enjoy an effective duopoly, as seen in their market revenue shares depicted in Figure 3.1.Footnote ³

Figure 3.1. RAN market revenue by OEM

Source: Base Station Market Poised for Strong Year Thanks to 5G and China According to Omdia, Microwave J. (Dec. 7, 2020)

As we see in Figure 3.2, however, 5G enables new underlying architecture options, including vRAN and Open RAN, which provide network operators with additional supplier options to help build their networks. This architectural opening resembles the move from IBM mainframe to Client/Server, which launched a computing revolution. This change potentially provides opportunities for highly competitive vendors to participate in the 5G system, making it more efficient, less expensive, and more innovative. Leading industry analyst Dell’Oro forecasts Open RAN as 15% of the market by 2026, while vRAN will be 5–10%, and combined these will represent 20–25% of the market within four years.Footnote ⁴

Figure 3.2. Comparison of RAN architecture options

Open RAN and vRAN enable network operators to move workloads to Commercial-Off-The-Shelf (COTS) equipment from leading computing vendors. Providers can base their equipment on world-class silicon, hardware, and software from some of the most competitive firms in the world, as well as from innovative startups, providing further diversification in the mobile value chain. Operators could use cloud service providers who can further enhance efficiency and handle these workloads as a service. Table 3.3 lists firms that potentially can participate as decentralized providers.

In the special case of Open RAN, operators can obtain radio units from competing vendors beyond the incumbent leaders such as Ericsson and Nokia. Qualcomm’s 5G RAN Platform for Small Cells and FSM200xx processors promises a ready avenue for small-cell entrant providers to provide competitive radio unit products. Enhanced competition in this area appears particularly desirable given the national security imperative to exclude Huawei and ZTE from many key markets.

These new architecture and vendor options promise to enable new players to enter the market for cellular services. Examples include Rakuten in Japan and, following their example, DISH in the United States. New entrants may be very helpful in the United States given the merger of T-Mobile and Sprint, which leaves just three major wireless carriers to serve the US market. The US market suffers from higher wireless prices than many other countries.Footnote ⁵

The nature of the new system enables new and superior applications. For example, the combination of 5G’s wireless nature and the improved performance through eMBB make 5G FWA a practical alternative to fixed broadband services based on FTTx and DOCSIS. This should benefit wireless carriers, FWA vendors, and broadband consumers.

In short, 5G brings an array of new technologies, scope improvements, features, and applications. While each may create value, the value may accrue to different market segments and participants in each case based on the nature of each improvement. Most improvements may chiefly benefit carriers, while others may chiefly benefit enterprises, vendors of particular types of equipment, software or services, or others. Interestingly, the expanded scope of 5G means that many mobile device vendors may be agnostic to many of these improvements since they are targeted at other markets and other parts of the mobile value chain.

C. The 5G Value Chain

A notable feature of the 5G value chain is that it involves multiple, specialized firms that act in a decentralized fashion, coordinated by standards and market interactions. In that respect, the 5G value chain is like the value chain of previous technological generations, only more advanced and perhaps more complex.

As is well known, the theory of patent holdup and royalty stacking predicts that a market characterized by multiple agents contributing to a standardized technology will be able to exploit monopoly power. According to the theory, the existence of multiple monopolies strangles markets and most of the price paid by consumers will redound to the profit margins of the technology development firms.Footnote ⁶

A testable implication of the theory is that implementers “see down” the game tree, and therefore refrain from making investments. It is therefore curious that implementers in 5G do not seem to be concerned about this possibility. They are making sizable investments. We think that they are likely drawing on the history of 3G and 4G in making their investment decisions.

A. Value and Distribution in Wireless Mobile Markets

Price theory observes that in equilibrium the price paid by consumers equals the value created by the entire production chain at the margin, as determined by the good’s demand curve. This is true whether the good or service is a pound of bread, a bundle of pins, or a phone.

Price theory also shows that the total revenues generated by sales in a market are distributed among the factors of production – those inputs that were involved in the production of the final good – based on the value each adds to total revenues at the margin. It follows that revenues are exhausted by the payments made to the input suppliers along the value chain. These revenues remunerate the opportunity cost of the inputs used in the production of the final good and any rent that factors of production receive.

We follow our 2021 paperFootnote ⁸ and illustrate these insights with a simple supply and demand graph. Figure 3.3 depicts the observed equilibrium in the smartphone market in 2016. For simplicity, we parameterize the world market with a single linear demand curve and assume that all consumers paid the average selling price of a smartphone.Footnote ⁹

Figure 3.3. Value and distribution in the smartphone value chain

In 2016, phone manufacturers sold 1.42 billion units for $425.1 billion, at an average selling price of $298.Footnote ¹⁰ Because consumers are free to buy a phone, the demand curve shows how much consumers value a smartphone at the margin. On average, $298 is what the least willing consumer in 2016 was willing to pay for a smartphone. Figure 3.3 also shows that most consumers valued their phones at more than $298 and obtained a net surplus when they bought a phone, the difference between their willingness to pay, as shown by the demand curve, and the market price. It follows that the total consumer surplus was equal to the area between the demand curve and the market price for phones. According to the demand curve depicted in Figure 3.3, consumer surplus in 2016 was equal to $784 billion.Footnote ¹¹

Figure 3.3 also shows how the revenues generated by the sale of smartphones were distributed among phone manufacturers and input suppliers. Roughly 20% of the revenues accrued to semiconductor manufacturers ($85 billion; $60 per smartphone, on average); 5% accrued to manufacturers of baseband processors ($22 billion; $15 per smartphone, on average); and 60% of the revenues ($254.1 billion; $178 per smartphone, on average) accrued to the producers of other inputs, such as the firms that made the cameras, gorilla glass, and housings, as well as the firms, such as Foxconn, that assembled the phones. Around 12% ($50 billion; or $35 per smartphone) reached the firms that sold the phones in the form of profits, most of which accrued to Apple.

IP is an asset, and royalties are the revenues that this asset generates. As can be seen in Figure 3.3, 3.4% of the revenue generated by the smartphone market reached the owners of patents ($14.2 billion, or roughly $10 per smartphone). Most of this ($12.4 billion) was earned by SEP owners. The remainder was largely earned by non-SEP patents, held by firms such as Microsoft (which earned royalties mainly on the patents on its operating system and software technologies), the patent pools that license audio and video codecs, and patent assertion entities that own the patents necessary to manufacture semiconductors.

Figure 3.4. Monopoly licensor sets the royalty rate

The distribution of the $425.1 billion in revenues among input providers reflects the choices that firms in the production chain made to substitute away from more expensive inputs toward less expensive inputs. Thus, firms at the end of the production chain, which designed and marketed the phones (for example, Samsung, Apple), combined inputs from many suppliers to minimize the costs of producing the smartphones that consumers valued. Similarly, the firms that produced the intermediate inputs and IP for those smartphones (for example, Corning, Ericsson) also combined inputs from many suppliers to minimize costs. Those suppliers, in turn, purchased the necessary inputs from firms even further up the production chain, and so on. Each input in the production chain had its own demand curve. That is, the demand curve each producer faced was derived from the demand for smartphones, and the elasticities of each demand curve depended in part on the possibilities for substituting away from that input. Consequently, firms along the production chain equalized the value created by each input at the margin with the input’s market price.

The share of each input in the $425.1 billion in revenues in the smartphone market was therefore the equilibrium outcome of a complex process of cost minimization. Because the output of an upstream firm is the input of firms further downstream, and all value stems ultimately from consumers’ willingness to pay, no stage of the production chain is independent of, and separable from, the others – prices are determined simultaneously in all of them.

What does price theory tell us about how to value the IP necessary to produce a smartphone? The royalty is the rental price of IP and is a function of the value that consumers were willing to pay for the capabilities created by those patented technologies, at the margin, and the possibilities that producers had to substitute away from using those IP assets toward alternative technologies. The finding that the patent holders earned 3.4% of the value of the average smartphone in 2016 has three complementary interpretations. First, the purchaser with the lowest willingness to pay for the average smartphone valued those technologies at the equivalent of just 3.4% of the price she paid for her smartphone. Second, there must have been alternative technologies toward which producers could eventually substitute. If that had not been the case, then the owners of the IP property would have operated as monopolists and charged far more than 3.4% of the value of a phone, a point to which we return later. Third, and importantly, IP owners did not enjoy market power; they could not constrain output to raise prices.

B Economic Rent and the Distribution of Value across the Stages of the Production Chain

Should competition authorities care about the level of royalties? Some argue that patent holders can charge royalties because they enjoy monopoly power, even beyond that granted by the patent. According to this argument, the profits that patent owners obtain are rents wrought by market power, and there is a natural role for competition policy. A different explanation, however, is that the revenues that patent holders obtain are the fair remuneration of their investment, given the risks they took when researching and developing the technology. In this view, the rents enjoyed by patent holders are Ricardian – their origin is that the selected technology creates more value per unit of input than the alternatives, not the exercise of market power. According to this argument, rents are the result of competition among technologies, and there is little, if any, role for competition policy.

Thus, the nature and origin of the rents made by patent holders are central to assessing whether competition policy must do something, if anything, about royalties. To appreciate the difference between a monopoly rent and a Ricardian rent, a few diagrams help.

Consider first a royalty set by a patent holder who holds a monopoly because nobody can produce without infringing the monopolist’s patent. To keep the argument simple, we assume that the monopolist licenses the technology to competitive manufacturers that can produce phones at constant cost c. As Figure 3.4 shows, when fixing the royalty rate, the monopolist patent holder confronts the market demand for the final good, and reasons that if she sets the royalty equal to R, competitive manufacturers will pass on the royalty and set a price p = c + R for a phone.

Thus, by fixing R, the monopolist patent holder controls the final price p. Also, by setting R, the monopolist patent holder controls the per-unit rent. Standard economic theory says, and Figure 3.4 shows, that the monopolist patent holder will increase R to contract output until the marginal revenue from selling phones is equal to the manufacturing cost c. In equilibrium, the price of a phone will be equal to p_M = c + R_M and then

$\frac{p_M-c}{p_M}=\frac{R_M}{p_M}=\frac{1}{\eta },$(3.1)

where η is the elasticity of the demand for phones. It follows that, just as for a standard monopolist, pricing is determined by the classic Lerner formula.

A Ricardian rent, by contrast, emerges when a firm produces more or higher quality output per unit of input than its competitors. Because a Ricardian rent remunerates a competitive advantage, it can emerge in a competitive market. Observed royalties have nothing to do with market power.

To appreciate this point, Figure 3.5 draws the standard average and marginal cost curves of a competitive licensee. The figure shows an innovation that increases quality and consumers’ willingness to pay by a factor λ > 1 over the alternative, and in equilibrium, products that use the technology command a market premium equal to (λ − 1)p^c. Thus, the licensee obtains additional revenues equal to the shaded rectangle. This rectangle is the Ricardian rent wrought by the technology, which the patent holder can appropriate through licensing and charging royalties. But its origin is the market premium (λ − 1)p^c, and its total amount is capped by the incremental value created by the technology.

Figure 3.5. Patents and Ricardian rents

As can be seen in Figure 3.5, the royalty is part of the licensee’s average cost. The licensee covers all her costs, and the good’s market price equals its long-run marginal cost. But because a better technology commands a price premium, which is equal to consumers’ differential willingness to pay, well-functioning markets naturally create the rents and rewards that incentivize investments in R&D. This rent may be transferred to the patent holder via per-unit royalties, a lump-sum payment, or a combination of both. Whatever the means whereby the Ricardian rents are transferred, they remunerate those technologies that are more productive and deliver more output or value per unit of input.

It follows that competition policy may have a role if the source of the rents is the exercise of market power – that is, if technology developers are able to raise the running royalty to create scarcity at the margin. By contrast, competition policy does not have much of a role if technology developers earn Ricardian rents, which emerge in competitive markets. The point we make next is that the observed level of the royalties charged by technology developers is informative about the source of the rents.

C. Monopoly Power and Royalty Stacking in the Mobile Phone Industry

As Figure 3.6 shows, a patent holder acting as a monopolist would exploit market power by restraining output, raising the market price of the final good, and extracting the monopoly rent through the royalty. A direct test for the existence of a monopoly therefore uses the Lerner margin, as shown in (3.1) to predict the level of the royalties that patent holders would charge if they acted as a monopoly. We apply this reasoning to the smartphone market.

Figure 3.6. Value and distribution in the smartphone value chain with a single Monopoly Patent Holder (2016)

Figure 3.6 shows the same demand curve as Figure 3.3 but assumes that patent holders act as a single profit-maximizing monopolist. The main result is that patent holders would have earned 66% of all revenues of the value chain, instead of 3.3%. Higher royalties would have multiplied the average selling price of a smartphone by a factor of almost three – from $298 to $844. Consequently, only 722 million units would have been sold, instead of 1.42 billion. Despite the decline in unit sales, however, total revenues would have risen from $425.1 billion to $609.4 billion. More than two-thirds of those revenues (about $400 billion) would have accrued to the patent holders. Those revenues would have been pure rent, as they exceeded the long-run cost of the inputs used to produce the patented technologies.Footnote ¹²

An influential literature, known as patent holdup and royalty stacking, implies that the situation should have been even worse than outlined in the prior paragraph. The theory claims that because multiple firms own the IP that is essential to make a phone interoperable and compatible, each can exploit its monopoly independently.Footnote ¹³ That is, the magnitude of the cumulative royalty in the smartphone value chain predicted by the theory of royalty stacking grows with the number of patent holders.Footnote ¹⁴ Monopoly will stack upon monopoly, strangling the industry. Formally, if the number of patent holders is n, then the Lerner margin is:

$\frac{p_s-c}{p_s}=\frac{R_s}{p_s}=\frac{n}{\eta }>\frac{1}{\eta }$(3.2)

In our 2017 paper,Footnote ¹⁵ we parameterize a standard royalty stacking model. We observe that in 2016, there were 21 identified patent licensors who received royalty revenue, and that the cumulative royalty yield predicted by the theory of patent holdup and royalty stacking is 79%. That is, if patent holders were each exploiting their monopoly power independently, they would receive four out of every five dollars paid for a smartphone.

A predicted royalty of 67% (a single monopolist) or 79% (21 monopolists) compares with the observed average cumulative royalty yield from the 21 identified patent licensors of 3.4% in 2016.Footnote ¹⁶ That is to say, the actual yield is more than 20 times lower than either the yield predicted by the theory of patent holdup and royalty stacking or by the predicted royalty that would be charged by a single profit-maximizing monopolist.

The implication is that patent holders in the smartphone value chain do not exercise any meaningful monopoly power. On the contrary, the remuneration that patent holders receive is a Ricardian rent. Thus, the evidence from 3G and 4G is that there is no evidence that standardization creates market power. Because of this, there is little ground to claim that competition authorities have any meaningful role to play in cellular SEP licensing as traditionally practiced within the smartphone industry.

D. 3G and 4G: A Functioning Licensing Market

As we have seen, in 2016, the average royalty yield for a smartphone was 3.4% of the average selling price of a smartphone. In our 2018 paper,Footnote ¹⁷ we estimated royalty yields from 2007 through 2016 and found that they showed remarkable stability. Figure 3.7 shows the average royalty yield since 2007 for 16 licensors, which accounted for 78.2% of all royalty revenues in 2016. Since 2009, we have data for 21 licensors, which accounted for 92.5% of all royalty revenues in 2016. As Figure 3.7 shows, both series are remarkably stable. The average cumulative royalty yield of firms with data since 2007 hovers between 2.1% and 3%; the average cumulative royalty yield of firms with data since 2009 hovers between 3% and 3.5%, falling only marginally during the last three years. Note that, as can be seen in Figure 3.8, the composition of sales between feature and smartphones changed significantly during the period, and the value of sales roughly doubled, and yet the average cumulative royalty yield remained stable.

Figure 3.7. Patent royalties as percentage of the value of mobile (smart and feature) phones shipped (2007–2016)

Figure 3.8. WW mobile subscriptions by cellular generation

Source: Ericsson Mobility Visualizer, Sept. 16, 2023

E. New Estimates of the Cumulative Royalty Yield

Since our research described above, we have continued to monitor the market for cellular SEP licenses, although we have modified our methodology so that we now focus on cellular SEP licensing specifically.Footnote ¹⁸ The leading cellular SEP licensors are Qualcomm, Nokia, Huawei, Ericsson, and Interdigital. These constituted 84% of royalties in 2016, the last year covered in our prior studies.Footnote ¹⁹

As can be seen in Table 3.4, these five licensors obtained $82 billion in royalty revenues during 2015–2022, on average about $10.2 billion per year. On average, royalties were $9.05 per smartphone. If we make room for other licensors, average royalties per mobile device are about $9.79, in line with the orders of magnitude in our previous research. Basically stated, the data are consistent with our earlier estimates. In conclusion, little seems to have changed since the period covered in our previous research – a profound observation given the passage of so many years and the corresponding changes in markets, technologies, products, and companies.

A. 5G Alternatives and Substitutes

B. Future Developments

During the lifespan of 5G, we can expect much improvement in 5G itself, as well as in some of these substitute and alternative technologies. Cisco’s Annual Internet Report observes the following trajectories for some key 5G alternatives and substitutes:Footnote ²³ Fixed broadband speeds will more than double by 2023, to 110 mbps (v. 46 mbps in 2018); Wi-Fi speeds from mobile devices will triple by 2023, to 92 mbps (v. 30 mbps in 2018); Wi-Fi hotspots will grow four times, to 628 million public hotspots (v. 169 million in 2018). Cisco forecasts these improvements in the context of rapidly growing quantities of users, connected devices, and changing use cases, including for UHD video and IoT devices.

For nascent technologies, the trajectory is more uncertain and speculative, but perhaps has greater potential upside. We look at three interesting cases to consider.

A. The Phenomenon of Differential Ambiguity in Matters of Contract Formation and Good Faith Negotiation

Building on the English common law tradition, but insulated from Europe by an ocean, American contract law appears to have evolved in its own distinctly didactic manner, which I attribute to the long shadow cast by Oliver Wendell HolmesFootnote ¹⁰ and his intellectual heir in American jurisprudence, Richard Posner.Footnote ¹¹ With respect to the principles of contract formation, that American demand for crisp answers implicitly conduces to what economists call activity rules and closing rules, which simplify the task of definitively confirming whether a meeting of the minds has or has not occurred (much in the spirit, as I shall explain in the following pages, of economists realizing that national governments needed to define unambiguous activity rules and closing rules if they were to succeed in creating a workable market mechanism to auction licenses for 3G spectrum). By default, American law provides a clear closing rule for determining when contract formation has failed. The SEP holder makes an offer that is legitimately FRAND. Either the offer is accepted, or it is rejected explicitly or by counteroffer or by the passage of a commercially reasonable period of time. The licensee is not permitted to initiate rounds of offer and counteroffer. Following the licensee’s failure to accept a legitimately FRAND offer, negotiations of course may continue between the parties, but no longer under the FRAND framework. Instead, those negotiations revert to the framework of public patent law.

In contrast to this American veneration of transactional efficiency (and a concomitant abhorrence of ambiguity or euphemism), the bodies of contract law of other jurisdictions (in Europe and the rest of the world) evidently do not offer, and do not aspire to offer, such black-and-white rules on whether and when a contract has been formed.Footnote ¹² For example, when and for whom does the duty of good faith negotiation commence, and how long does it remain in effect once an offer has been made? Some scholars of French law impute to the duty to negotiate in good faith an explicitly noneconomic origin.Footnote ¹³ As one treatise on French law has observed, incompatible interpretations among scholars of the purpose and effect of the doctrine of good faith indicate that “the notion of good faith and its use by the courts is likely to remain contested.”Footnote ¹⁴ If SEP holders, implementers, and their attorneys fail to recognize that the monochrome character of American contract law differs from the Technicolor character of contract law in many other nations, they will expose themselves to an unmarked hazard in SEP licensing negotiations and SEP litigation whenever American contract principles do not control.

That hazard exists in very practical business terms because ETSI plays such a dominant role in the setting of wireless standards, and its FRAND commitment is, of course, controlled by French contract law.Footnote ¹⁵ In contrast, New York law controls the RAND contract of another prominent SSO, the Institute of Electrical and Electronics Engineers (IEEE).Footnote ¹⁶ It is quite possible that the differences between French contract law and New York contract law – to take only one example – will produce substantively different conclusions about the legal duties owed under the FRAND or RAND contract in question. Indeed, the scope of a SEP holder’s obligations and the scope of the rights granted to third-party beneficiaries by virtue of a FRAND or RAND contract depend on both that contract’s actual language and the controlling law governing the interpretation (if needed) of that contract. The key point is that, whenever American contract law does not apply, to know when the negotiation has failed to achieve contract formation, we need a closing rule.

This perspective on contract formation causes me to disfavor and avoid using the terminology of “holdup” and “holdout” to describe the presence or absence of good faith during the negotiation to license SEPs pursuant to the FRAND contract. I find it simpler and more germane to ask whether, and when, the offeror and the offeree have discharged whatever duties they bear under the FRAND contract between the SEP holder and the SSO. What is called “holdout” is a manifestation of the failure of the controlling law to declare in a timely manner that the contract negotiations have become futile, such that the SEP holder has discharged its contractual duty to the SSO (and to the implementer as the third-party beneficiary of the SEP holder’s FRAND contract with the SSO). Calling the problem “holdout” supplies an epithet, but it does nothing to help answer the legal or economic question; to the contrary, that nomenclature is arguably counterproductive in the sense that it falsely suggests that the SEP holder must make some further evidentiary showing that “holdout” has occurred before it may pursue its legal remedies under the national law of the country that issued the patents in suit.

To begin the task of reducing legal and economic ambiguity concerning the determination of whether a SEP holder and an implementer have conducted a FRAND licensing negotiation in good faith, I propose here the formulation of a specific activity rule and a specific closing rule when American contract jurisprudence does not control interpretation of the FRAND contract in question. My proposed activity rule is that, in each round of offer and counteroffer – and to the extent that the SEP holder has not already discharged its contractual obligation to ETSI (such as by its already having made a legitimately FRAND offer at the very outset of the negotiation) – a party must revise its bid or ask price by the minimum agreed-upon increment for that party to be deemed still to be negotiating in good faith. My proposed closing rule is that a party will be deemed to have made its final offer or counteroffer if it does not, within a commercially reasonable amount of time after receiving an offer or counteroffer, sweeten its price relative to its price in the previous round of offer and counteroffer. These rules of market design are proposals, which will surely benefit from scrutiny and refinement by others, but these proposals should suffice to invite a needed discussion.

Considering how controversial and how consequential these issues have been in the licensing of SEPs for smartphones, I see no reason why they will prove to be simpler to resolve in the licensing of SEPs for connected cars, smart homes, and the multitude of other 5G devices that will constitute the Internet of Things.

Parties in litigation over SEPs in the United States and England routinely solicit the expert opinions of scholars on French contract law to assist the court’s interpretation of a SEP holder’s FRAND contract with ETSI. Yet, as I have remarked elsewhere, the public trial testimony, expert reports, and judicial decisions describing those expert opinions on French law cast doubt on the determinacy of French contract law.Footnote ¹⁷ An American observer might conclude that French contract law on its own is incapable of defining good faith, or at least that it is ill equipped to supply a definition. Principles defining good faith in the evidentiary records of these cases are simply not percolating through into the public domain to shed light on how parties should behave or how judges should judge. Economic insights from the field of mechanism design can cure that indeterminacy.

B. Activity Rules, Closing Rules, and “Best Practices” in Negotiations over FRAND-Committed SEPs

I have previously argued that time is of the essence in the implementation of a standard – in particular because to waste time in the introduction of an entirely new generation of products featuring standard-dependent technological innovations is to harm the public interest by sacrificing consumer surplus irreparably.Footnote ³⁵ To the extent that one can properly impute a duty (or covenant) to negotiate in good faith to an intended third-party beneficiary of ETSI’s FRAND contract with a particular SEP holder, that duty reflects the understanding that a public interest inheres in the expeditious negotiation of SEP licenses. Whether the implementer’s behavior after receiving a legitimately FRAND offer adheres to the standard of good faith will depend ultimately on how quickly the implementer seeks to close the bid–ask spread and converge on an agreement – which is to say, contract formation.Footnote ³⁶ Following that interpretation to its logical conclusion, the point at which the implementer ceases to sweeten its counteroffer from one round of the negotiation to the next defines the point of impasse.

Implicit in this rule is the understanding that the parties also must define how long a given round lasts during their negotiation. How long may a party take to sweeten its bid or ask? If the parties provide no answer of their own to this question, the default answer becomes “a commercially reasonable amount of time.” But, rather than have a court rule what amount of time is commercially reasonable, the parties can create considerable value by agreeing on a framework that is both more precise and more expeditious than what is merely commercially reasonable.

The obvious analogy here is to the activity charge required of a bidder to maintain its right to keep bidding in the simultaneous multi-round ascending auction for 3G spectrum licenses in the United Kingdom. Economist Paul Klemperer of Oxford, who advised the UK government, explains:

As I observed at the opening of this part, governmental agencies around the world have promoted “best practices” in negotiations over FRAND‑committed SEPs. The most useful thing left undone in such statements of best practices is to endorse the concept of an activity charge for good faith FRAND negotiations, and then to identify an unambiguous economic methodology for determining the minimum bid increment by which an implementer must sweeten its counteroffer to the SEP holder’s legitimately FRAND offer for the implementer to be deemed still to be negotiating in good faith.

The next most useful thing left undone in statements of best practices for good faith FRAND negotiation is to identify an unambiguous economic methodology for determining when the negotiation has ended in failure. Again, it bears emphasis that this question is legally relevant only for FRAND obligations not controlled by American-style contract principles of offer and acceptance – which, I have explained earlier, inherently have the admirable (but evidently underappreciated) quality of unambiguously defining a closing rule for a bilateral negotiation.Footnote ³⁸ Evan Kwerel, a highly respected economist who spent a career at the Federal Communications Commission and made important contributions to the design and execution of American spectrum auctions there, explained that “[t]he closing rule was one of the major [market] design issues for a simultaneous auction” for spectrum in the United States.Footnote ³⁹ Stanford professors Robert Milgrom and Robert Wilson, who shared the Nobel Prize in economics in 2020 for their work on auction theory, “proposed a simultaneous closing rule whereby the auction closes on all licenses only after a round has passed with no bidding on any license.”Footnote ⁴⁰ In contrast, conspicuously absent from the current conception of the FRAND negotiation is any guidance on when it reaches its end in terms of rounds of offer and counteroffer. Instead, the negotiation is like a baseball game with an infinite number of innings, or a poker game in which a player may remain in the hand without calling his opponent’s bet.

Because of the failure of SSOs or courts or other institutions in positions of authority to impose both an activity rule and a closing rule, a contentious negotiation for a FRAND license, if not controlled by the law of a jurisdiction having American-style principles concerning contract formation, will regrettably resemble Zeno’s Dichotomy paradox. The journey’s end becomes ever closer in incrementally smaller half steps, but it is never reached.

A. Defining the Fair Division of Surplus

John Rawls famously argued that “fair” means “just.”Footnote ⁴⁷ “Justice as fairness,” he asserted, “is an example of … a contract theory.”Footnote ⁴⁸ Rawls argued that “[t]he word ‘contract’ suggests,” among other things, “the condition that the appropriate division of advantages must be in accordance with principles acceptable to all parties.”Footnote ⁴⁹ One would expect the same of the FRAND contract. How can we go about imputing to the fairness component of the FRAND contract – a meaning that is intellectually rigorous in both legal and economic respects?

This question of the meaning of a fair price turns out to have very real legal ramifications in the present day. Rarely do I disagree with Judge Posner, but I do with respect to his view that “fair” is surplusage in the FRAND contract. Judge Posner, sitting by designation as the trial judge in Apple, Inc. v. Motorola, Inc. in 2012 in the Northern District of Illinois said that, in the context of FRAND, “the word ‘fair’ adds nothing to ‘reasonable’ and ‘nondiscriminatory.’”Footnote ⁵⁰ My previous writings have followed this convention of making no legal or economic distinction between FRAND and RAND royalties, though I have never excluded the possibility that someone might eventually make a compelling argument for why “fair” is not a throwaway word for parties to insert into a contract.Footnote ⁵¹ And so, for example, I have previously analyzed at length the differences between actual FRAND contracts and actual RAND contracts with respect to how fairness creeps into the constraint to license SEPs on nondiscriminatory terms.Footnote ⁵² This part of this chapter will show why courts should take the distinction between FRAND contracts and RAND contracts more seriously.

More than 30 years ago, Robert Frank of Cornell University proposed a precise economic definition that is directly relevant to the question of what makes a FRAND royalty fair:

Frank then explained the problem that unfairness presents: “People will sometimes reject transactions in which the other party gets the lion’s share of the surplus, even though the price at which the product sells may compare favorably with their own reservation price.”Footnote ⁵⁴

This reasoning is very close to the conclusion I had reached before benefiting, late in the process of revising an earlier incarnation of this text over the course of several years, from reading Frank’s 1988 book. Frank and I each find ourselves using Judge Posner as our foil, though for different reasons. Frank criticized Judge Posner’s writings through the mid-1980s, as denying what Frank argued was the considerable explanatory power of fairness considerations in law and economics.Footnote ⁵⁵ In contrast, I gently chide Judge Posner for overlooking roughly 25 years later that, by the private ordering of contract law, some SSOs had chosen to impose an obligation of fairness so that (according to my economic interpretation) those SSOs could nudge parties into exercising the degree of moderation in their negotiation demands that is necessary to achieve contract formation reliably and expeditiously.

The irony is that my interpretation of why the word “fair” must have an independent meaning within the FRAND contract is quintessentially Posnerian: A division of surplus that is perceived by both parties to be fair maximizes the probability of contract formation over some defined time horizon, which in turn immediately benefits the parties to the contract. Thus, fairness clearly promotes static allocative efficiency. Moreover, across time the fairness constraint on the division of surplus also benefits countless consumers, whom the grand edifice of the FRAND contract is surely intended to benefit (though not necessarily by the formal machinery of conferring on those consumers legally enforceable rights of a third-party beneficiary, as the FRAND contract does confer on implementers). As Joseph Schumpeter taught us, it is the consumption of innovative products in the future that delivers radical – not marginal – gains in consumer surplus.Footnote ⁵⁶ Thus, the fairness constraint promotes dynamic efficiency as well.

In this respect, Posner’s emphasis on efficiency and Frank’s emphasis on fairness are reconcilable. A lopsided division of surplus is a cost imposed on efficient transactions to the extent that it prevents some otherwise promising negotiations from achieving successful contract formation; if that cost can be eliminated or mitigated, a larger number of efficient transactions will occur. Therefore, regardless of whether one prefers to call it a quest for fairness or a quest for efficiency, an SSO’s constraint on the SEP holder that a royalty for its SEPs be fair is a privately ordered feature of contract – a self-imposed cattle prod – that contributes to a result that proponents of fairness and proponents of efficiency can both applaud.

B. Fairness and Contract Formation

One can formalize a simple theory of fairness and contract formation. Imagine a decision tree depicting the expected surplus of a contract negotiation as the sum of the expected values of two mutually alternative outcomes: EV = pS + (1 – p)0, where p is the probability of contract formation and S is the surplus created by a successful transaction. The size of the surplus S is separately identified by the bargaining range, which is bounded by the reservation prices of the parties to the negotiation. But the expected value of the surplus is necessarily smaller than S because the division of the surplus might cause one of the parties to walk away. A simple and intuitive formulation of the relationship comes from defining as R the ratio of the seller’s share of the surplus (X) to the buyer’s share of the surplus (Y): R = X / Y = X / (1 – X). R is bounded below by zero and above by infinity. As R approaches zero, p approaches zero. As R approaches infinity, p again approaches zero. In either case, it becomes more likely that contract formation will fail, and consequently the parties will forfeit the surplus from the transaction.

At this point, it is instructive to consider the Ultimatum Game, a bargaining game in which a player makes a single take-it-or-leave-it offer, rather than multiple offers and counteroffers.Footnote ⁵⁷ The game ends in either an agreement to the unaltered terms of the first offer or no agreement at all. If the second party rejects the offer, neither party benefits – the first party does not keep any portion of the asset but rather forfeits it all. Thus, both parties have an incentive to agree, and the division of surplus (which in this particular game is assumed to be a windfall, not a return on either party’s investment) will depend on a fair offer having been made. As I previously explained in 2013, the Ultimatum Game is interesting in analyzing the FRAND contract not because a FRAND negotiation represents an Ultimatum Game.Footnote ⁵⁸ After all, in FRAND licensing, there are typically repeated rounds of offer and counteroffer, the identities of the parties are known (perhaps because the parties have previously negotiated a licensing contract), and the reputation of the players matters because they will face the prospect of repeated play in subsequent licensing over future standards. Instead, the Ultimatum Game is interesting for FRAND licensing because the results of economic experiments based on the Ultimatum Game shed light on which divisions of surplus the parties to a stylized negotiation would consider fair. Surveying the experimental economics literature as it existed in 2000, Ernst Fehr and Simon Gächter reported that “[a] robust result in [the Ultimatum Game] experiment, across hundreds of trials, is that Proposers who offer the Responder less than thirty percent of the available sum are rejected with a very high probability.”Footnote ⁵⁹

If there are any positive spillovers for society as a whole from successful contract formation, as there of course would be if the contract is one for the licensing of patents essential to practice an industry standard, then those externalities are forfeited as well when the negotiation collapses. In contrast to the scenarios of negotiation impasse that I previously described, as R goes to one, p approaches one, and thus contract formation becomes increasingly certain. An impartial spectator nudging the parties to maximize the expected value of the surplus of their contemplated transaction would prescribe “maximize p with respect to R,” since S is already exogenously determined.

Although any possible division of the surplus created by voluntary exchange is mutually beneficial, that fact does not imply that every price along the bargaining range (which defines the locus of “reasonable” royalties) is equally likely to yield an agreement. How does a given split of the surplus between the SEP holder and the implementer influence the probability of their successful contract formation within a specified period of time? One interpretation of a fair royalty is that it leads more expeditiously to contract formation than some other division of the gains from trade. That is, the fairness component of the FRAND contract between the SEP holder and the SSO takes on independent meaning by giving teeth to the proposition that time is of the essence in achieving contract formation between the SEP holder and the implementer. Fairness promotes economic efficiency in the sense of hastening voluntary exchange, which is the prerequisite to the expeditious exploitation of the standard.

C. Dividing Surplus Fairly

The probability of a successful voluntary exchange increases as each party signals its willingness to accept a lesser share of the surplus that the transaction will create. Thus emerges a simple understanding of fairness, which can be expressed in comparative terms: The price corresponding to a given bilateral division of the surplus from a voluntary exchange is fairer than the price corresponding to some alternative bilateral division of that surplus if the first division is more likely than the second to lead the parties to agree to a successful transaction within some specified period of time.

My proposed definition of a fair price echoes, but is not identical to, certain themes found in the economic literature examining justice and fairness. Most notably, my definition resembles the proposition that fairness requires the approximately equal division of surplus, which Robert Frank proposed in 1988 in Passions within Reason.Footnote ⁶⁰ However, my rationale for that definition differs from what I understand to be Frank’s reasoning.

D. Licensing SEPs on Terms Consistent with the Fairness Constraint of the FRAND Contract between the SEP Holder and the SSO

By definition, any price within the bargaining range is mutually beneficial. But that fact does not imply that every such price is equally likely to yield an agreement. How does a given split of the surplus between the SEP holder and the implementer influence the probability of their successful contract formation within a specified period of time? Is the distinguishing characteristic of a fair royalty that it leads more expeditiously to contract formation than some other division of the same gains from trade?

To analyze this question, let us normalize the bargaining range so that it runs from 0 to 100. Normalizing the bargaining range simplifies the application of this analysis to different prospective implementers of the SEPs belonging to a given SEP holder. A license agreement struck at a normalized price of 0 gives the implementer 100% of the surplus. That is, an agreement at a normalized price of 0 is equivalent to a license bearing a royalty rate equal to the SEP holder’s minimum willingness to accept and not a penny more. In contrast, an agreement at a normalized price of 100 gives the SEP holder 100% of the surplus. That is, an agreement struck at a normalized price of 100 is equivalent to a license bearing a royalty rate equal to the implementer’s maximum willingness to pay and not a penny less.

For any license agreement struck at a normalized price between 0 and 100, each party will gain some of the surplus generated. For any possible agreement at a single given normalized price between 0 and 100, some probability exists that, within a specified period of time, the implementer will accept that price and enter into an agreement, and some different probability exists that the SEP holder will accept that same price and enter into an agreement. If both parties accept the same price, then an agreement is reached, and contract formation occurs.

The probability that the implementer will agree to terms decreases as the negotiated price moves farther from the SEP holder’s minimum willingness to accept and closer to the implementer’s maximum willingness to pay. Conversely, the probability that the SEP holder will agree to terms increases as the negotiated price moves farther from the SEP holder’s minimum willingness to accept and closer to the implementer’s maximum willingness to pay. The implementer has a “bid function” that determines the implementer’s probability of agreeing to terms (within a specified period of time) at any given price over the bargaining range. Similarly, the SEP holder has an “ask function” that determines the SEP holder’s probability of agreeing to terms (within the same specified period of time) at any given price over the bargaining range.

If the bid function and the ask function are symmetric, then the most common agreement will occur where the parties divide the gains from trade evenly. This 50–50 outcome is merely the arithmetic implication of the bid function’s being the mirror image of the ask function. It is important to emphasize that this result does not rely on the Nash bargaining solution, which predicts a 50–50 split of the surplus in a bilateral negotiation using cooperative game theory.Footnote ⁷⁵ Nor does this result rely on the familiar cake-cutting principle described by Rawls and others – “You cut, I choose”Footnote ⁷⁶ – which Peyton Young notes “is fair because the outcome is envy-free.”Footnote ⁷⁷

However, there is no economic reason to expect that the bid function and the ask function will be symmetric. In practice, risk aversion, discount rates, or other economic factors will influence the specific shapes for the bid function and the ask function.

The point royalty within the range of reasonable royalties upon which the SEP holder and the implementer will agree – that is, how they will divide the surplus from voluntary exchange – will be determined by the parties’ relative bargaining power. The party that suffers least from delaying the agreement – that is, the party that is most patient – will typically have more bargaining power. Parties can have different levels of “patience” during a FRAND licensing negotiation while still negotiating in good faith, and it is common for SEP negotiations to take multiple years. For example, a SEP holder that lacks liquidity might need an immediate resolution of the negotiations. Or the implementer might be on the verge of releasing a standard-compliant product and therefore quickly needs a license to the SEPs before releasing a noninfringing product. Conversely, the SEP holder might not need an immediate license to the SEPs, which would increase its bargaining power. The near-absence of injunctive relief in SEP infringement litigation and the limited likelihood of enhanced damages might lead one to conclude, all other factors remaining the same, that implementers are more “patient.” It is well established in the economic literature that the cost that each party bears from a delay is measured by its respective discount rate.Footnote ⁷⁸

A more precise model could use different assumptions concerning the probability of contract formation. For example, are the probabilities for the two parties independent of one another, or is each probability conditional on the expected reaction of the counterparty (and, if so, for how many future rounds of the negotiation)? These questions are appropriate to ask if an economist wants to model the probability of contract formation in precise mathematical terms – for example, along the lines of the Rubinstein bargaining model, which is based on noncooperative game theory.Footnote ⁷⁹ But my goal here is more modest and more heuristic. So those particulars about the precise nature of the probabilities are unnecessary to resolve to make the larger point (which I believe a judge or jury could readily understand intuitively) – namely, that it is reasonable to expect that the speed of contract formation will depend on the relative parity or disparity of the shares by which each party to a negotiation proposes to divide the surplus from a successful licensing transaction.

One proposed division of surplus might be substantially more likely than another to yield successful contract formation within a specified period of time spent negotiating. For example, it seems intuitively clear that a 60–40 split of the surplus would more readily be accepted by both parties than would a 99–1 split. If so, we would say that the 60–40 split is fairer than the 99–1 split.

What would be the threshold for a judge or jury to make the qualitative determination that a particular division of surplus would be unfair? Perhaps the experimental results of the Ultimatum Game, which I discussed earlier, will suggest a useful line of analysis.