Patent pools, vertical integration, and downstream competition

• Date: 01 March 2019
• Author: Markus Reisinger,Emanuele Tarantino
• Published date: 01 March 2019
• DOI: http://doi.org/10.1111/1756-2171.12266

RAND Journal of Economics

Vol.50, No. 1, Spring 2019

pp. 168–200

Patent pools, vertical integration,

and downstream competition

Markus Reisinger∗

and

Emanuele Tarantino∗∗

Patent pools are commonly used to license technologies to manufacturers. Whereas previous

studies focused on manufacturers active in independent markets, we analyze pools licensing to

competing manufacturers, allowing for multiple licensors and nonlinear tariffs. We find that the

impact of pools on welfare depends on the industry structure: whereas they are procompetitive

when no manufacturer is integratedwith a licensor, the presenceof vertically integratedmanufac-

turers triggers a novel trade-off between horizontal and vertical price coordination. Specifically,

pools are anticompetitive if the share of integrated firms is large, procompetitive otherwise. We

then formulate information-free policies to screen anticompetitive pools.

1. Introduction

A patent pool is an agreement among patent owners to license a bundle of patents to each

other or to third parties (Quint, 2008). From the 1890s to the 1940s, many dynamic manufacturing

industries in the United States had a patent pooling arrangement (Lerner and Tirole, 2007).

Following a number of unfavorable antitrust rulings, pools essentially vanished between the

∗Frankfurt School of Finance & Management; m.reisinger@fs.de.

∗∗University of Mannheim and CEPR; tarantino@uni-mannheim.de.

We thank the Editor (Kathryn Spier) and two anonymous referees for very helpful comments and suggestions. This

article also benefited from comments by Guillermo Caruana, Jay Pil Choi, Cl`

emence Christin, Vincenzo Denicol`

o,

Ramon Faul´

ı Oller, Johannes Koenen, Anne Layne-Farrar, Volker Nocke, Martin Peitz, Daniel Quint, Regis Renault,

Patrick Rey, Thomas Rønde, Klaus Schmidt, Florian Schuett, TimSimcoe, Yossi Spiegel, Konrad Stahl, Javier Suarez,

Chengsi Wang, Martin Watzinger, and Ali Yurukoglu. Aja Pe´

ak provided excellent research assistance. We also thank

participants at seminars in CEMFI (Madrid), CREST (Paris), DICE (D¨

usseldorf), European Commission DG COMP

(Chief Economist Team), Humboldt University Berlin, University of Bologna, University of Frankfurt, University of

Munich, and at the Annual Meeting of the German Committee for Industrial Economics, Barcelona GSE Summer

Forum, Workshop on Competition in Vertical Market Structures (D¨

usseldorf), MaCCI IO Day (Mannheim), MaCCI

Summer Institute on Competition Policy (Bamberg), MaCCI Annual Conference 2017, and PetraliaApplied Economics

Workshop conferences. Tarantino gratefully acknowledges financial support from the Collaborative Research Center

(CRC) Transregio 224 Bonn-Mannheim. A previous version of this article was titled “Patent Pools in Input Markets.”

The article also supersedes a previous version published as CEPR Discussion Paper no. DP11512 (2016). All errors are

ours.

168 C2019, The RAND Corporation.

REISINGER AND TARANTINO / 169

mid-1950s and the mid-1990s. This changed in 1995, after the release of new guidelines on

the licensing of intellectual property by the Department of Justice (DOJ) and Federal Trade

Commission (FTC). In the years that followed, American authorities approved patent pools tied

to major technologies in electronics, information technology, and medicine.1

A lenient approach to patent pools is in line with the observation that products in these

industries build on a large number of complementary patents. This forces licensees to navigate

a patent thicket: a web of overlapping claims that may preclude the commercialization of a new

product because buyers need to get patents from multiple sources. These patent thickets trigger

a problem of horizontal double marginalization: when multiple licensors sell complementary

patents, each of them does not consider that lowering its price has a positive effect on other

licensors’ profits, due to an increase in the demand for the bundle. Thus, prices of complementary

patents are optimally set by a patent pool (Shapiro, 2001).2

The theoretical literature that followed has studied patent pools mainly in environments

in which licensors and licensees are separated firms and licensees are active in independent

markets (e.g., among others, Lerner and Tirole, 2004; Quint, 2014; Choi and Gerlach, 2015;

Boutin, 2016; Rey and Tirole, forthcoming). The general message of these articles is that pools

comprising complementary patents tend to be procompetitive. However, their modelling approach

overlooksthat patent pools are ubiquitous in markets where patent owners deal with manufacturers

competing with each other on the product market. Moreover,as documented by Layne-Farrar and

Lerner (2011), many patent pools’ members are vertically integrated.3This suggests that the

available models cannot capture an important feature of existing pools, as they cannot relate the

market structure to the competitive consequences of these agreements.

We propose a new theory of patent pools in which monopoly licensors offer to license their

perfectly complementary patents to manufacturers who are rivals in the downstream market.

In the model, the market power of a licensor is constrained by the presence of an inefficient

status-quo technology. We then study the welfare consequences of the formation of patent pools

in industries where licensees compete against each other and licensors and licensees may be

vertically integrated. We also give guidance on the policies that are best suited at screening

anticompetitive pools. Specifically, in line with recent literature (Lerner and Tirole, 2004, 2015;

Boutin, 2016; Rey and Tirole, forthcoming), we consider information-free policies that break the

anticompetitive outcome without relying on the specific number of firms or market structure in

the industry.

To develop our model, we build on two main theoretical pillars. First, firms can engage in

interlocking vertical relationships. This assumption allows for the formation of an interconnected

web of trades between licensors and manufacturers, thus capturing a salient feature of patent

thickets.4Second, patent owners use two-part tariffs. This reduces contractual inefficiencies and

allows for larger production as licensees take advantage of declining average costs. There is also

evidence documenting that most licensing programs charge fixed fees in addition to variable

royalty rates (Gilbert, 2011).5

We characterize the equilibrium of the resulting common agency game involving multiple

licensors and manufacturers, product-market competition, and nonlinear tariffs. Weconsider two

1A nonexhaustive list includes the MPEG, DVD, Bluetooth, Firewire, 3G-Mobile, and laser eye surgery

technologies.

2This result builds on Cournot’s (1838, 1897) insight that a monopoly raises welfare when products are perfect

complements. Another thorny consequence of patent thickets is that a new product unintentionally infringes on existing

patents (see Choi and Gerlach, 2015).

3For example, as of August 2018, in the MPEG-4 Visualpatent pool, which licenses critical technologies used in

personal computers and mobile devices, 22 of the 32 licensors are vertically integrated. Also, in the DVD6C licensing

group, which sells patents tied to the DVD technology, four of the eight licensors are integrated.

4For instance, within the pool on the 3G-Mobile technology, Samsung, LG, and Sony, among others, cooperated

with vertically specialized licensors, such as Bosch, on the pricing of the bundle of patents. At the same time, they

competed on the product market.

5Wediscuss our assumptions in more detail in Section 3.

C

The RAND Corporation 2019.

170 / THE RAND JOURNAL OF ECONOMICS

main scenarios: the one with vertical separation and the one with vertical integration of one or

more licensors. We show that the pool’s competitive consequences crucially depend on industry

structure.

First, pools are procompetitive when no licensor is vertically integrated. Although this

conclusion is consistent with the one on pools licensing to noncompeting manufacturers, a new

mechanism drives our result: without a pool, each licensor considers the effect of a reduction in

its royalty rate only on the part of a manufacturer’s profit that it can extract through the fixed fee;

however, it does not take into account that other licensors will also demand a positive fixed fee.

As royalty rates of all licensors will be positive then, each of them does not fullybenefit from the

marginal increase in a manufacturer’s profit when lowering its royalty rate (i.e., our offer game

features contracting externalities). By contrast, the pool is the only entity demanding a payment

from manufacturers and therefore reaps the full benefit. It follows that its incentive to reduce

patent prices is larger. Interestingly, and in contrast to conclusions from models focusing on

noncompeting manufacturers, this mechanism also implies that a larger number of patent owners

not only leads to a higher aggregate patent price, but also to a higher individual patent price.

We then consider pools formed by verticallyintegrated licensors. Differently from common

wisdom, we show that a pool of perfectly complementary patents can raise patent prices. Specif-

ically, the pool is anticompetitive if the share of integrated firms in the industry is large, and

procompetitive otherwise.6This result is the consequence of a noveltrade-off between horizontal

and vertical price coordination: on the one hand, as integrated manufacturers are members of

the pool through their upstream licensors, all licensors, even the nonintegrated ones, reduce their

patent prices to these manufacturers. The pool therefore allows licensors to internalize the con-

tracting externalities (horizontal price coordination). On the other hand, all licensors benefit from

increased profits of vertically integrated manufacturers, and therefore have an incentive to soften

price competition. They achieve this by raising the patent prices to nonintegrated manufacturers

(vertical price coordination).

If the number of vertically integrated licensors is large, the effect of vertical coordination

dominates. Without the pool, as integrated licensors do not demand a fixed fee from manufac-

turers, a nonintegrated licensor can reap a large part of the manufacturer’s profit increase when

lowering its royalty rate. Thus, patent prices are low. With the pool, vertical price coordination

implies that nonintegrated licensors demand high royalties from nonintegrated manufacturers,

thus making the pool anticompetitive. By contrast, with only few integrated licensors, the royalty

rates of nonintegrated licensors are relatively high without the pool because of the presence of

contracting externalities. Therefore, the pool is procompetitive due to the effect of horizontal price

coordination. All these findings are robust to the nature of competition (prices or quantities), the

number of firms in the industry, the contractual environment, and the degree of differentiation

between final products.7

Whereas the existing literature on patent thickets and royalty stacking suggests that a pool

of complementary patents reduces patent prices (e.g., among others, Shapiro, 2001; Lemley and

Shapiro, 2013), we showthat, depending on industry str ucture, such a pool can be anticompetitive.

In fact, there is evidence demonstrating that pools, absent regulation, may refuse to license

their technologies (Lampe and Moser, 2014). More importantly, there is case evidence on pool

members’ use of provisions restricting downstream competition (Gilbert, 2004), and on the

practice of granting affiliated manufacturers with a privileged access to the pool’s patent bundle

(Harris, 2003; Flamm, 2013).8This case evidence exactly matches the theoretical mechanism

yielding our anticompetitive result.

6In the baseline model, we focus on the case with two manufacturers to bring out this result within a parsimonious

setting. Wethen prove that all results carry over to a setting with a general number of manufacturers.

7Our focus on perfect complements puts us in a situation in which pools are generally procompetitivewhen buyers

are active in separate markets (i.e., do not compete). It follows that our anticompetitive result is likely to arise even in

settings where patents are imperfect complements.

8Wewill come back to this evidence in Section 7.

C

The RAND Corporation 2019.