abstract - conference.druid.dk · participants in the us economy - the context of our study and...

Paper to be presented at the DRUID Academy Conference 2019 at Aalborg University, Denmark

January 16-18, 2019

The effect of opportunistic litigation on firms’ collaborative behavior

Miryam Martín SánchezUniversity of the Balearic Islands

Department of Business [email protected]

AbstractThe effect of opportunistic litigation on firms’ collaborative behaviorMiryam Martin-SanchezUniversity of the Balearic [email protected]

The aim of this paper is to examine the effect of opportunistic litigation held by patent assertionentities (PAE) on firms’ collaborative behavior. Using a comprehensive panel data of the U.S. publiclytraded firms operating in high-tech industries from 2003-2008, we generate the following results: First,we find that companies form more alliances after being sued by a patent assertion entity. Second, weidentify the underlying mechanism in this relationship. Additional costs impose by opportunisticlitigation lead companies to divert resources from their main activity. In order to buffer the negativeimpact of this reallocation, firms use alliances to obtain similar resources to those that they havereallocated towards litigation.

Keywords: Opportunistic Litigation; Patent Assertion Entities; Patent Troll; Alliance Formation;Collaborative Behavior; Treatment effects; Intellectual Property Rights; Patents

THE EFFECT OF OPPORTUNISTIC LITIGATION ON FIRMS’

COLLABORATIVE BEHAVIOR

Miryam Martin-Sanchez

University of Balearic Islands (Spain)

[email protected]

Abstract

The aim of this paper is to examine the effect of opportunistic litigation held by patent

assertion entities (PAE) on firms’ collaborative behavior. Using a comprehensive panel data

of the U.S. publicly traded firms operating in high-tech industries from 2003-2008, we

generate the following results: First, we find that companies form more alliances after being

sued by a patent assertion entity. Second, we identify the underlying mechanism in this

relationship. Additional costs impose by opportunistic litigation lead companies to divert

resources from their main activity. In order to buffer the negative impact of this reallocation,

firms use alliances to obtain similar resources to those that they have reallocated towards

litigation.

Keywords: Opportunistic Litigation; Patent Assertion Entities; Patent Troll; Alliance

Formation; Collaborative Behavior; Treatment effects; Intellectual Property

Rights; Patents

2

1. Introduction

Patent Assertion Entity (PAE) activity is a growing concern amongst policymakers - which is

reflected in an extensive set of legal changes1- and managers - who actively participate in

lobbying campaigns2. The main issue is the drain of resources that these agents impose on all

participants in the US economy - the context of our study and primary PAEs' targeted country

(Reitzig, Henkel, & Heath, 2007). As an example, it is estimated that the aggregate losses to

the defendants involved in opportunistic litigation from 2000 to 2010 reached $87.6 billion

(Bessen, Meurer, & Ford, 2011).

PAEs are opportunistic litigators whose business model revolves around obtaining excessive

damage awards through enforcing patents in one-shot trials (Reitzig, Henkel, & Schneider,

2010). These agents are firms who acquired the ownership of a patent from third parties -

frequently from a bankrupt company (Dekkers & Tietze, 2014) - without the intention of using,

commercializing or granting a license (McDonough III, 2006). Instead, these opportunistic

litigators engage resources in scanning alleged infringers and hide their intellectual property in

order to be infringed (Reitzig et al., 2007). They keep waiting until a firm has engaged in a

costly sunk investment and after that, enforce their property rights against the manufacturer

who has inadvertently infringed the PAE’s intellectual property rights. What they try to

provoke is a hold-up situation, avoiding that the operating company can “invent around.” By

applying what we call a “keep-and-seek” strategy, PAEs maintain their intellectual property

rights hidden and focus on identifying corporate “negligence” or monitoring deficiency (Reitzig

et al., 2007).

1 Reflecting the policy markers’ concern, from 2013-2017 US Congress has considered over a dozen of bills

proposing to regulate the assertion of patents: The Innovation Act (H.R. 3309), the Patent Transparency and

Improvements Act (S. 1720), the Patent Quality Improvement Act (S. 866), the Patent Abuse Reduction Act (S.

1013), the Patent Litigation Integrity Act (S. 1612), the Innovation Protection Act (H.R. 3309), the Patent

Litigation and Innovation Act (H.R. 2639), the SHIELD Act (H.R. 845), the Stopping the Offensive Use of Patents

Act (STOP Act) (H.R. 2766), and the End Anonymous Patents Act (H.R. 2024).

In June 2013, the Executive Office of the President issued a report entitled Patent Assertion and U.S. Innovation.

In March 2016, the President’s Council of Economic Advisers released again another issue brief that functions as

an update to the 2013 Patent Assertion report.

In October 2016, Federal Trade Commission issued a report entitled Patent Assertion Entity Activity.

The United States Patent and Trademark Office (USPTO) has developed training programs for examiners and

judges in order to avoid useless patents.

In 2011, the President Obama signed the Leahy-Smith America Invents Act (AIA), a piece of legislation designed

mostly to deter trolls’ activity. 2 Some companies, particularly IT firms, have formed several lobbies, such as Coalition for Patent Fairness. An

example of the activities of this kind of organizations is the letter from Adidas, Ag, et al. to Member States of the

European Union, et al (Sept. 2013) or the letter from the Alliance of Automobile Manufacturers, et al. To Harry

Reid, Majority Leader, U.S. Senate et al. (July 17, 2013).

3

Consistently with the adverse practices described above, researchers has identified the

following two categories of potential harms 3. On the one hand, patent assertion activity can

impose litigation costs that are “unrealistic” (Reitzig et al., 2007, p. 134) in relation to the

patented technology at issue (Henkel & Reitzig, 2008; Reitzig et al., 2010).On the other hand,

an increase in firms’ additional costs due to opportunistic litigation can divert technical talent,

managerial attention and other corporate resources away from the main activity of the operating

company, a result that decreases the social welfare (Bessen, Ford, & Meurer, 2011; Cohen,

Gurun, & Kominers, 2014; Tucker, 2011). These findings suggest that companies are in a

vulnerable situation when they are involved in PAE litigation.

Viewing strategic alliances as cooperation links that improve the strategic position of the firms

by providing resources that they lack (Gulati, Nohria, & Zaheer, 2000), we examine whether

companies use alliance formation as a strategic alternative to maintain their competitive

position while they are facing to opportunistic litigation. From the resource-based perspective,

we understand companies as bundles of tangible (i.e., financial assets) or intangible

(i.e.managerial skills) resources that can be shared through alliances. That is, through

partnerships agreements, accused companies can obtain resources that buffer external

pressures, being particularly relevant during periods of exogenous shocks (Miner, Amburgey,

& Stearns, 1990). In this regard, opportunistic litigation can be understood as an adverse event

that eventually may become companies more vulnerable due to a reallocation of resources to

paying litigation costs. Therefore, the goal of this research is to examine the impact of

opportunistic litigation on alliance formation.

Our study yields interesting insights. We find that targeted companies engage in more alliance

agreements after a lawsuit bring by a PAE. Moreover, in order to go further, we drill this

relationship by explicitly considering average and individual effects. We also test our

assumption of the transfer of resources to deal with litigation costs. In doing so, we

operationalize some features of the legal process as well as some indicators of the firms'

financial health as proxy variables.

We believe advancing this understanding is important for several reasons. First, while previous

studies suggest that patent assertion activity hurts innovation in general terms, and particularly,

3 Note that particularly at the early beginning of PAE debate, several authors - see Geradin, Layne-Farrar, &

Padilla (2012); McDonough III (2006) and Shrestha (2010) as an examples - justified the activity of these agents.

Nevertheless, Pénin (2012) pointed out that most of the divergences between proponents and opponents are the

result of the improper use of the terms non-practicing entity, patent trolls and patent brokers as synonyms.

4

decreases firms’ R&D expenditure (see Bessen, Meurer, & Ford, 2011; Cohen, Gurun, &

Kominers, 2014; Smeets, 2014, among others), we know little about their effects on firms’

strategic decisions. Yet, to our knowledge, no study has assess patent assertion activity on

firms’ collaborative behavior despite its importance to maintain firms’ competitive advantage.

Second, an overwhelmed legal system and the pressures exerted by some managerial lobbies

have lead policymakers to propose some bills to regulate the assertion of patents but that

implicitly lead to protect (inadvertent) infringers (Tucker, 2011). This fact has led to a

controversial debate among stakeholders from different industries. Alternatively, it could be

interesting to inform policymakers of strategic decisions that operating companies implement

so that policymakers can institutionally ease their implementation (i.e., promoting inter-firms

collaborative agreements with tax incentives). Finally, we perform this study in an empirically

way, which should be highlighted, taking into consideration the difficulty in measuring PAE

activity that has repeatedly been emphasized in the literature (Bessen, Meurer, et al., 2011;

Fischer & Henkel, 2012; Pénin, 2012).

Drawing on a unique dataset of the U.S publicly traded companies operating in high-tech

industries from 2003 to 2008, we first estimate a Poisson regression model that includes an

endogenous-treatment variable. In a second stage, we estimate a matching procedure to examine

the average and individual level impact of PAE litigation on alliance formation. We also focus

on the distribution of the impact. Finally, we use this individual level to test our assumption

about the role played by litigation costs and firms’ resource endowment.

This paper is structured as follows. Section 2 outlines the theoretical insights and the

background discussion that drive our proposition about the positive effect of opportunistic

litigation on firms’ collaborative behavior. Section 3 describes the data, variables and methods.

Section 4 overviews our results and Section 5 discusses our main contributions, implications of

our study and future research.

2. PAE litigation, legal costs, and collaborative behavior

In this section, we proceed to explore the (potential) PAE effect on firms’ collaborative

behavior. We argue that PAE activity contributes positively to alliance formation. In order to

go further with this statement, we examine potential mechanisms to explain this relationship:

characteristics of lawsuits and firms’ resources endowment.

Alliances are defined as voluntary cooperative arrangements among independent firms,

designed to exchange or share resources to achieve goals that are mutually beneficial (Gulati,

5

1998; Kogut, 1988). They can take different forms, ranging from collaborative R&D

agreements to co-marketing arrangements, and can occur as a result of a wide range of motives

(Lavie, 2007; Powell, Koput, & Smith-Doerr, 1996). Precisely, a large portion of the literature

has focused on why firms form alliances. Such cooperative agreements can provide companies

with several advantages such as cost sharing, risk reduction and rapid response to unexpected

situations (Lavie, 2007)

From a theoretical perspective, a well-known approach to alliance formation is resource-based

framework (Eisenhardt & Schoonhoven, 1996; Ireland, Hitt, & Vaidyanath, 2002; Miotti &

Sachwald, 2003). Through this approach, companies are seen as bundles of heterogeneous

resources. Differences in firms' resource endowment are related to differences in firms'

performance. Thus, firms’ resource profile plays an important role. In this sense, Eisenhardt &

Schoonhoven (1996) propose that alliances are formed when firms are in a vulnerable position

or when firms are in such strong social position that have enough resources to attract partners.

Focusing on the vulnerable position, this occurs when companies are in a challenging market

situation, or they are undertaking expensive or risky strategies.

In this sense, the literature about opportunistic litigation suggests that the detrimental effects of

PAEs increase transaction costs and uncertainty (Pénin, 2012), dissipate social value reducing

manufacturers’ incentives to innovate (Bessen, Meurer, et al., 2011; Reitzig et al., 2007) and

decrease the rate of invention (Turner, 2015). In addition, patent assertion activity can be

overcompensated in relation to the real impact of the patent infringement (Reitzig et al., 2007).

As consequence, an increase in litigation costs can divert technical talent, managerial attention

and other corporate resources away from the main activity of the operating company (Bessen,

Ford, et al., 2011; Cohen et al., 2014; Federal Trade Commission, 2016; Tucker, 2011). These

findings imply that companies are in a weaker position after being sued than they would be if

they were not litigated. Precisely, the vulnerability situation is clearly reflected in the statement

of Congressional testimony of John Boswell, the General Counsel of SAS, the world’s largest

privately held software company, regarding one of the legal struggle that the company faces

against a patent troll (Executive Office of the President, 2013).

“(…) If SAS ultimately wins this case it will be a Pyrrhic victory at best. We

spent $8 million and huge amounts of developer time and executive time etc.,

for what? This victory does not resolve the other patent troll cases that we

face, or will face in the future. This $8 million and the millions more we are

6

spending on other cases is money SAS no longer has to invest in people,

facilities, research, or product development; and we are a relatively small

player in this world... It does not cost much to be a troll and to make broad,

vague demands. On the other hand, the risk to the company receiving a troll

threat is enormous.”

This statement summarizes the costs of litigation that companies face. Although, in general

terms, it is perceived that the more litigious environments increase firms’ risk of being exposed

to significant direct and indirect costs, particularly, in those lawsuits filed by PAEs, the cost of

litigation falls more heavily on the accused infringer (Tucker, 2011). To achieve a better

understanding of this previous premise, it is necessary to understand the operating logic of the

majority of technological industries in the absence of opportunistic litigation. Traditionally,

patents had been used as a way of protecting the firm’s innovation against others, so if a

company A sued the company B, B could answer with a counterclaim against A. When both of

them are manufacturers, both face to equal (potential) losses, at least initially. Basically,

companies had established implicit “non-aggression pacts” (Reitzig et al., 2007).

However, with the advent of opportunistic litigators, this “non-aggression status quo” is broken.

As PAEs do not actually manufacture anything, and often, they are lawyers (Columbia &

Blasberg, 2006), they bear fewer costs regarding discovery and preparing for trial (Chien,

2009). These facts cause that lawsuits filed by PAEs are clearly asymmetric in terms of costs.

In addition to legal cost (i.e., attorneys’ fees), companies also face other costs of indirect nature.

Thus, in the above fragment, Boswell identifies some of them such as the wasting time for

developer and executives. Distractions to management is a common cost mentioned in the

literature (i.e. Bessen & Meurer, 2007, 2008). Thus, Bessen and Meurer (2008) pose that

business activities could be interrupted because managers and researchers must commit a

significant amount of time in preparing for litigation and appearing in court. Moreover, other

indirect costs have been already identified. On the one hand, financial cost arising from greater

risk, including risk of bankruptcy (Bessen & Meurer, 2007; Feldman & Price II, 2014; Kiebzak,

Rafert, & Tucker, 2016). On the other hand, delays in implementing marketing strategies

(Bereskin, Hsu, Latham, & Wang, 2017) or leakages on confidential information (Bessen &

Meurer, 2007).

Therefore, the inability of targeted firms to defend against opportunistic litigation entails

potential losses that hamper innovation and imply (potential) low performance (Bessen, Ford,

7

et al., 2011; Lemley, 2008; Merges, 2009; Pohlmann & Opitz, 2013). Clearly, the presence of

PAEs leads companies to face additional costs and risks that they did not have in the pre-

litigation situation. From this weaker position,, alliances can provide companies to critical

resources to buffer PAE effects. For example, following the classification of (Das & Teng,

2000, p. 24), alliances can provide, at least, four categories of important resources. First,

financial resources. The availability of capital could smooth profit pressures and ensure

predictable resources flows (Miner et al., 1990). Second, technological resources. Greater

access to a technological resource such as superior R&D capability which could help targeted

firms to cope with the detrimental effects of PAEs on innovation. In this regard, Cohen et al.

(2014) analyzing a sample of publicly listed firms, find a negative impact of trolls’ attacks on

R&D expenditure. Similarity, Bessen & Meurer (2014) conclude that troll litigation reduces

innovation incentives for any type of company. Third, alliances also can provide physical

resources. As an example (Das & Teng, 2000) mention new distribution channels. Fostering

distribution networks could mitigate some indirect costs related to marketing delays. Finally,

managerial resources obtained through alliance agreements could help targeted firms’ managers

with time-management and decision-making.

All these insights together lead us to argue that patent litigation driven by PAEs have a positive

impact on alliance formation. As we argue, alliance formation could be understood as a tool to

obtain resources when companies are in a vulnerable situation because of opportunistic

litigation. Therefore, the mechanisms that we propose are related to the characteristics of the

lawsuits and the firms’ resource endowment.

3. Data and methodology

In what follows, we describe our empirical strategy to test the ideas developed above. Firstly,

we give information on the data sources and about sample formation. In a next step, we explain

the variables. Finally, we specify our model and expose the applied methodology to investigate

the relationship between patent assertion activity and alliance formation in a subsequent period.

3.1. Data

We obtained data from five different sources. We integrated financial data from Compustat

with data on corporate subsidiaries from CorpWatch, with litigation data from RPX website

and with information on alliances agreements from SDC Platinum as well as information about

patents retrieved from PATSTAT database. After this process, we obtain a firm-level data from

8

the US. listed companies classified by standard industrial classification (SIC) (identified by

two-digit SIC categories: 28, 35, 36, 38, 48 and 73).

The period of our sample covers from January 1, 2003, to December 31, 2008. The reason to

choose this period is twofold. First, the identification of the troll’s activity is roughly dated at

the beginning of the 2000s. Second, the availability of the different databases.

3.1.1 Financial data

We identified from Compustat all firms which are publicly traded, from January 1, 2003, to

December 31, 2008, operating in the following two-digit SIC categories: 28, 35, 36, 38, 48

and 73). These sectors are chosen because a general consensus exists in delimiting high-tech

industry as the main target of PAEs (Allison, Lemley, & Walker, 2009; Henkel & Reitzig,

2008a; Pénin, 2012; Reitzig et al., 2007).

This process yielded a total of 3,433 firms split into six different industries with 1,188 firms

(34.61%) in chemicals and allied products (28), 128 firms (3.73%) in industrial and commercial

machinery and computer equipment (35), 521 firms (15.18%) in electronic and other electrical

equipment and components, except computer equipment (36), 479 firms (13.95%) in

measuring, analyzing, and controlling instruments; photographic, medical and optical goods;

watches and clocks (38), 258 firms (7.52%) in communications (48) and, 859 (25.02%) in

business services (73).

3.1.2 Data on corporate subsidiaries

A potential concern about the previously firm-level data is that we only capture publicly

traded companies. The reason is that the availability of financial information for this type of

companies is higher because they have a legal obligation to publish their annual accounts.

Although this fact does not seem to be a significant shortcoming taking into account that

these firms are responsible for the major (US) R&D (Bessen & Meurer, 2005), it could lead

to underestimating the number of litigious that these companies suffer. For this reason, we

retrieved all the subsidiaries of the firms identified in Compustat. We collect data on

subsidiaries provided by Corpwatch. CorpWatch API uses parsers to retrieve the subsidiary

relationship information from Exhibit 21 of the 10-K forms required to file by the U.S.

Securities and Exchange Commission (SEC) as part of the company’ annual report.

Moreover, the use of this data enables us to detected potential changes of name of the firms

http://sec.gov/

9

which will be relevant to apply the name-matched matching procedure that will be described

in Section 3.1.6. The available information of this source starts at 2003. For this reason, our

sample uses the year 2003 as the start date.

One caveat is that companies do not submit information in a standardization format which

makes more difficult to parse the information. Nevertheless, it is considering that, at least,

90% of the subsidiary companies are correctly parsed. A manual check was conducted to

minimize errors in the identification of companies and subsidiaries.

3.1.3 Litigation data

We collect data on patent litigation from RPX’s website (https://insight.rpxcorp.com/). This

dataset provides free information about each patent litigation lawsuits filed in each of the 94

US. Federal District Courts captured from Public Access to Court Electronic Records

(PACER). We collected information about 58,282 lawsuits from 1978 to March 2017.

Declaratory Judgments cases were deleted from the analysis following the same argument of

Smeets (2014). That is, in these cases, patent holders are mostly the defendant which means

that a manufacturing firm, the alleged infringer, holds the role of plaintiff.

It is worth mentioning that we were able to classify those lawsuits according to the primary

SIC of each of the defendants after performing a name-matched matching procedure (see

section 3.1.6).

A potential limitation is that we only use data on observed patent litigation. Unfortunately,

no comprehensive dataset of demand letters and other informal patent assertions by

opportunistic litigators exits. Nevertheless, as discussed by Cohen et al., (2014), informal

patent assertions are in decline and is expected to continue falling principally due to two main

drivers. First, the decreasing credibility of informal patent assertions, since a patent troll had

a viable case, it would sue instead of sending letters which is a less credible signal. Second,

the rise of legislation to assume responsibility for unsubstantiated demand letters.

We desegregated each case by defendant-level, which represents a significant improvement

over previous studies that are not able to identify defendants properly. For example, Kiebzak

et al. (2016) concede that they are not able to perform a manual review and Smeets (2014)

just identifies the primary defendant.

3.1.4 Data on partnership agreements

https://insight.rpxcorp.com/

10

We retrieved information from SDC Platinum database, provided by Thomson Reuters, which

compiles more than 140000 publicly announced alliances from 1984. SDC Platinum database

is showed as the more consistent database considering its alternatives (MERIT_CATI or RCAP,

amongst other) (Schilling, 2009). The companies of our sample signed a total of 4,106 alliance

agreements from 2003 to 2008. We did not extend the study after 2008 since, after this year,

the majority of the alliances were not confirmed by SDC, being classified in other categories

such as "pending."

3.1.5 Data on patents

We obtain firm-level patent information from the Worldwide Patent Statistical Database

(PATSTAT). This database contains patent information issued by the United States Patent and

Trademark Office (USPTO).

3.1.6 Sample

This list of publicly traded companies and their subsidiaries retrieving from CropWatch enabled

us to match all datasets. To do so, we employed a name-matched matching procedure. This way

of combining databases is essential to merge databases where there are no codes to identify

companies, particularly, for data on R&D, data on intellectual property rights and data on the

financial status of firms (Thoma et al., 2010).

Although we believe that we are the first scholars to create complete litigation, financial, and

patent dataset applying the name harmonization to defendants, our use of name-matched

matching procedure in management research is far from unique (see e.g. Agrawal, Bhattacharya,

& Hasija, 2016; Bessen, Ford, et al., 2011; Galasso, Schankerman, & Serrano, 2013; Thursby,

Fuller, & Thursby, 2009; Zobel, Balsmeier, & Chesbrough, 2016).

Basically, it is a three-step process: a parsing stage, a matching stage and a filtering stage (Raffo

& Lhuillery, 2009). 4 In a first phase, we standardize the names of companies that comprising

the different datasets. The standardization is performed adapting the work developed by Hall

(2008) 5 and improved by Bessen (2010) 6 . Subsequently, the matching is carried out by

constructing a matching scored (the Levenshtein distance or edit distance) between strings. To

4 They can also be referred as “data pre-processing”, “name cleaning” and “harmonization results” (Magerman,

Van Looy, & Song, 2006) or in the context of a software prototype, first software module, referring to the cleaning

phase; second software module regards the matching phase and third software module as a refinement process

(Thoma et al., 2010). 5 Retrieving from: http://eml.berkeley.edu/~bhhall/pat/namematch.html 6 Retrieving from: https://sites.google.com/site/patentdataproject/Home

11

do that, we apply a vectorial decomposition of texts. We implement a mechanism based on the

division of name strings into sequences of characters (“tokens”) (used, e.g., by Galasso et al.,

2013). Basically, this approach generates a measure of similarity between different company

names. The algorithm classifies as perfectly matching, those names which degree of quality is

equal to 1.

We combined our automatic matching algorithm with an extensive manual review. We sought

a secondary source to verify the affiliation for each remaining name. We check all the resulting

matching pairs which degree of equality is above 0.8. Table 1 shows the number and percentage

of names that we are able to match through the nine different name-matched matching procedure

that we carried out. We report the number of perfect pairings (classified as 1 by the algorithm)

and the number of company’s names that were classified as 1 after being checked manually.

<< INSERT TABLE 1 ABOUT HERE >>

We execute this procedure for retrieving and joining all information about litigation, patents

and alliance agreements with our list of publicly traded companies and their subsidiaries

retrieving from Compustat and CropWatch.

3.2 Variables

In this section, we explain all the variable included in the model. First, we define the dependent

variables. Second, the independent variable is explained. Third, we develop an explanation

about control variables. All the definitions of variables are contained in Table 2.


Dependent variable

The explanatory variable in our model, Strategic Alliances, is a variable that captures the

number of alliances a company i has established in a year t. As we discuss in Section 2 alliances

can provide access to different type of resources that companies lack. For this reason, this

research uses all kind of alliances, R&D alliances, and non-R&D alliances.

Main explanatory variables

The independent variable directly associated with the discussion above is Opportunistic

Litigation (Equation 2). Opportunistic Litigation is a binary variable that takes the value 1 if a

12

company has been attacked by an opportunistic litigator. Opportunistic Litigation also is a

dependent variable in the first stage of our endogenous binary-variable model.

To identify opportunistic litigators (patent trolls), we use as a start point the same approach of

Kiebzak, Rafert, & Tucker (2016, p. 222). They consider a patent troll as a “frequent litigator.”,

that is entities that have filed a considerable number of lawsuits. In our empirical approach, we

operationalize this approach from a conservative perspective, using as a threshold, the 90th

percentile. Thus, opportunistic litigators are companies which number of attacks (lawsuits)

filed is above this score. To do that, we counted the number of lawsuits that a company filed

per year. This empirical approach to identifying potential opportunistic litigators is also in line

with Galasso et al., (2013) who identify as “serial litigants” (p.303) those companies that are

on the top decile of intellectual property right purchases.

The main shortcoming of this measure, as Kiebzak et al., (2016) admitted, is that could cover

operating firms that for whatever reason are fighting for their intellectual property rights in a

court, and however, it could exclude those opportunistic litigators that are more selective in

choosing their targeted firms. For this reason, we performed an identification process of

opportunistic litigators. First, we discarded as opportunistic litigators, all the subsidiaries of the

companies that we have identified as operating firms using Compustat. Second, we built our

own list of opportunistic litigators by collecting all the scattered examples that we have found

in the literature. In addition, we manually check all the plaintiff’s names that were on the top

percentile of our sample. Finally, although no comprehensive directory of opportunistic

litigators exits, we contrasted our initial sample based on the measurement of frequent litigator

with a 20% of the random sample of Non-Practicing Entity (NPE) Litigation Dataset. This

dataset contains a tentative classification of Non-Practicing Entity (NPE), by categorizing

plaintiffs in each case into one of 13 types of patent asserter. As the purpose of this measure is

to identify those companies that use their intellectual property rights opportunistically, only

companies classified in 1 (acquired patents), 4 (corporate heritage) and 5 (individual/inventor

started company) categories are considered (see Miller (2015) for a more detailed explanation

about this classification).

By implementing all these steps, we believe that we obtain a more sophisticated measure of

opportunistic litigator than the original one only based on frequent litigators. In doing so, we

identified some companies that initially were categorized as opportunistic litigators, and they

are not, as well as, we recognized new opportunistic litigators whose levels of litigation are

13

lower than percentile 90 of our sample. Finally, we were able to come up with a tentative list

of companies that (potentially) show opportunistic behavior.

Although the variable described above, Opportunistic Litigation, is the variable that we use in

all the models, we tried to clean more this way to capture potential opportunistic litigators by

adding some extra filters. Thus, we matched the plaintiff names to an apparently reliable list,

available on Internet, which7 compiles potential opportunistic litigators from public sources of

reported frequent litigators. However, this source warns that the list can suffer from potential

inaccuracies. We also applied an “association rule” by considering that in those case with

various plaintiffs where at least one has been previously identified as an opportunistic litigator,

the rest of plaintiffs that are litigating together with this company, are also opportunistic

litigators. The underlying logic is that operating companies that are suffering the detrimental

effects of trolls are not likely to join forces with this agents, at least theoretically. The obtained

variables of applying these extra filters were also used in all the models that we propose and,

we obtained analogous results to those that we present in this paper8.

The main explanatory variable in the first equation of the endogenous dummy-variable model

refers to lawsuits filed in the Eastern District of Texas. We create this variable by summing the

total number of lawsuits in which an operating company is involved per year that has been filed

in Eastern District of Texas regardless the nature of the plaintiff (opportunistic litigator or not).

Control variables

Following, we describe the control variables that we include in the model.

R&D Intensity. This variable is related to the level of technological sophistication of a firm and

captures the firm-level dedication to knowledge creation (Hall, Jaffe, & Trajtenberg, 2005). We

measure this variable as the proportion of operating expenditure devoted to R&D activities of

the firm i for the period t.

7 https://trollala.com/possibletrolls.php 8 It is worth mentioning that OLitigation is the independent variable in the first step of the endogenous dumy

variable model, the probit model.

14

Stock of Patent Applications (ApplicationStock). This variable measures the stock of patent

applications by applying the perpetual inventory method. It is calculated as:

ApplicationStockit = Applicationsit + (1 – δ) ApplicationStockit-1

Where Applicationsit represents the productivity of the firm at generating potential inventions

given its accumulative knowledge capital, and it is defined as the application stock of firm i at

the year t. Here, we follow convention and resort to the traditional 15% of depreciation rate

per year (Hall et al., 2005; McGahan & Silverman, 2006). We include this variable using the

logarithm value.

Firm size. As it is common for firm-level studies (see Arora, Athreye, & Huang, 2016;

Rosenkopf & Almeida, 2003 among others), we control for possible scale effects. This is a

variable to measure the size of the firm i for the year t as a logarithm of the number of

employees.

Fixed Assets Stock. The proportion of the net fixed assets with regard to the total firms’ assets.

Potential partners could be interested in the combination of specialized assets through the

formation of alliances (Colombo, Grilli, & Piva, 2006).

Country of Incorporation. A dummy is included to control by the location of the public

company.

Attacks. Number of plaintiffs (opportunistic litigators or not) that have sued the company i for

the year t.

Not Patent. A binary indicator equal to 1 if the firm i does not have any patent at t, and 0

otherwise.

Sectorial activity and time dummies. Industry dummies and time dummies and are also

incorporated to control for time-fixed effects and industrial fixed-effects. These are sources of

unobserved heterogeneity that might affect alliance formation.

3.3 Empirical approach

3.3.1 Impact of opportunistic litigation on alliance formation

15

The objective of the empirical analysis is to determine whether the opportunistic litigation

affects the number of collaborative agreements that companies reach after being sued by a

PAE. In order to do that, we propose the following model:

Strategic Alliancest+1= f (Opportunistic Litigation, R&D intensity, Stock of Patent Applications, Firm size,

Fixed Assets Stock, Country of Incorporation)

where Strategic Alliancest+1 is the outcome of interest, the propensity to form alliances,

which is the number of alliances that a company establishes in the subsequent period (the

following year). The variable in interest is Opportunistic Litigation, a binary variable

indicating the existence of opportunistic litigation. The estimated coefficient of this variable

is expected to be positive according to our explanation in Section 2. The rest are variables

commonly used in the alliance formation literature (see Section 3.2 for a more detailed

explanation). Subscript i refers to firm and t to period.

As Opportunistic Litigation is an endogenous variable, we apply a Poisson regression model

that deal with endogenous binary-treatment variables (whether the company has received an

attack). This permits to analyze a model on which there is a dichotomy variable that is

endogenous for the outcome. That is, these models are able to deal with correlations between

the unobservables that affect the treatment and the potential outcomes (Terza, 1998).

Operationally, this is a two-step model. The first stage consists of the model predicting the

probability to be litigated by an opportunistic litigator. In particular, the following model is

estimated:

Equation (1):

Opportunistic Litigation it = β0 + β1 Texas Casesit-1 + β2 Attacksit-1 + β3 R&D intensityit-1 + β4 R&D_intensityit-

12 + β5 Firm Sizeit-1 + β6 Firm Sizeit-1

2+ β7 Not patentit-1 + β8 Stock of Patent Applicationsit-1 + β9 Fixed Assets

Stock it-1 + β10 Country of Incorporationit-1 + uit

Where Opportunistic Litigation it is the outcome of interest, a dummy equal to 1 if the firm

was litigated by an opportunistic litigator in a particular year. In our attempt to identify the

causal effect, our identification strategy focuses on Texas Cases variable. There is no a reason

to think that receive a claim from the Eastern District of Texas should affect the alliance

formation process directly but, if our troll forum-shopping conjecture is true, should predict

16

very good the fact of being attacked by a PAE. In addition,we also propose that the number

of plaintiffs that annually sues a company (Attacks) as an explanatory variable.

Equation (2):

E(yit|xit, Opportunistic Litigation) = exp(xit β1 + δ l.Opportunistic Litigation)

where Yit is the outcome, Strategic Alliances. The variable in interest is Opportunistic

Litigation, a binary variable indicating the existence of opportunistic litigation (equal to 1 if

the firm has been litigated). Xit is a set of control variables commonly used in the alliance

formation literature , and β is a vector of parameters to be estimated. In order to control for

sources of time and industry-varying heterogeneity, we include a set of control dummies in

both regressions.

As a third method, we estimate the average impact of opportunistic litigation on alliance

formation. We follow previous studies by using a matching approach to estimate this

relationship (Smeets, 2014). This methodology generates the counterfactual outcome by

identifying non-treated twin firms (non-attacked), which are similar in the rest of exogenous

features. As Chapman, Lucena, & Afcha (2018) explain that matching generates the average

effect by comparing the alliance formation outcome when a firm is sued by an opportunistic

litigator to the counterfactual alliance formation outcome if the firm would not have been

attacked. We follow the literature standard by applying propensity score nearest neighbor

matching (PSM) (Caliendo 2005).

3.3.2 Differential impact of litigation

In the second stage of our analysis, we turn our attention to the differential impact of

opportunistic litigation on alliance formation. The concept of differential gains suggests that

distinct companies experience different impacts on alliance formation from litigation, which

could be in terms of the direction (i.e., positive or negative) and/or magnitude of the impact.

To identify the individual impact effect, we employ the approach deploy by (xxxx pendiente)

and recently applied in the assessment of R&D policies (Chapman, Lucena, & Afcha, 2018).

To implement this methodology, we first identify the individual level impact of litigation on

alliance formation for each treated firm as follow:

𝛼𝑖𝑇𝑇 = 𝑌𝑖 − �̂�𝑖

𝐶

17

Where 𝛼𝑖𝑇𝑇 stands for the individual effect. That is, the difference between the alliance

formation of a treated firm i (Yi) and the counterfactual level of alliance formation the threated

firm would have had in the absence of litigation (�̂�𝑖𝐶). Examining the distribution of 𝛼𝑖

𝑇𝑇

permits insight into the extent of differential effects generated by opportunistic litigation on

alliance formation.

Robutness check. We complement our study by applying the general method of moments

(GMM) as a robustness check (see Apendix A.Table 1)

4. Results

Table 3 presents descriptive statistics for all variables used in the two equations of the analysis.


Table 4 shows the results of the Poisson regression analysis following the specification

described in Section 3.2. Examining the results (Equation 2), we note that they provide support

to our assumption that being sued by an opportunistic litigator impact positively on subsequent

alliance formation. The next step is to estimate the average treatment effect on the litigated

companies. Thus, in estimating marginal effects, we find that litigated companies will establish

1.40 more alliances that they would if they were not sued (in a one year period).


Previously to the Equation 2, we first run a probit model to obtain the predicted probability of

being sued by an opportunistic litigator (Equation 1).

Next, we match firms based on the propensity score. Table 5 shows that significant pre-

matching differences exist between the treated and the control groups in all characteristics.

However, as can be seen in Table 6, these differences have been removed after the matching

procedure. Explained differently, comparing (pseudo) R-squared of the two models, before

(0.349) and after (0.013) of applying the matching procedure, show that any remaining

differences in alliance formation can now be attributed to opportunistic litigation. Using this

methodology, the number of alliances after receiving an attack (period t+1) is, on average, a

35,07% higher.

18



After identifying the individual level impact of opportunistic litigation on alliance formation,

we now go further in the comprehension of 𝛼𝑖𝑇𝑇 by examining its distribution. As can be seen

in the Figure 1, the effect has a heterogeneous nature.

<< INSERT FIGURE 1 ABOUT HERE >>

Focusing on this positive effect, we now examine the theorized effect of resources and litigious

costs (Table 7). To do that, we use the individual level of opportunistic litigation on alliance

formation as our explained variable in an ordinary least squared regression model. As

independent variables, we propose a set of variables that have been identified by the previous

literature as a dimension of the complexity of a legal process, and consequently, associated with

more intensive litigation cost. Thus, in the Model 1 (Table 7), we follow Kesan and Ball (2006)

who use the length of the patent lawsuits, as the number of days from the opportunistic litigator

sued the alleged infringer to the judge’s ruling, as a proxy for patent litigation cost. In an

analogous context, Galasso (2010) also use the length of the patent lawsuit as a proxy for the

speed of technology diffusion through licensing. The underlying idea is that those lawsuits that

last longer generate more litigation costs to the defendant because of the persistence of the

effect (Smeets, 2014).


In Model 2, we use the number of legal documents filed in a patent lawsuit as another way to

capture the complexity of the legal process. These documents can involve a quite range of topics

such as magistrate orders or complaints by one of the litigants. In this sense, this measure could

capture time spent by defendants' lawyers working on an opportunistic lawsuit (Kesan and Ball,

2006). From the same approach, we also propose the number of asserted patents and accused

products respectively (Models 3 and 4) as a proxy for patent litigation costs.

Because of the high correlation between these four variables, we analyze each of these

dimensions of the litigation cost in independent models.

19

Moreover, we also include some indicators of the firms’ financial health. On the one hand, we

include the leverage ratio (debt-to-total assets ratio) that indicates the level of assets that are

being financing by debt.On the other hand, Working capital is defined as the net amount of

short-term assets and is the result of the difference between current assets minus current

liabilities. This variable is commonly used as an indicator of the firm’s short-term operating

liquidity. As is common in the financial constraints literature, we scale this variable by

firms’assets (Czarnitzki, Hottenrott, & Thorwarth, 2011).

We observe a positive and statistically significant effect for all the variables related to the

dimensions of litigations except for the length of the lawsuits. Regarding financial indicators,

the working capital has a positive and significant effect.

5. Discussion and conclusions

This study investigates the impact of opportunistic litigation on firms’ alliance formation. First,

the analysis reveals a positive and significant relationship (Model 1). That is, when companies

are litigated by PAEs, they engage in a greater number of alliance agreements in the subsequent

period. Second, we find that litigated companies establish 1.40 alliances more than they would

form if they were not sued. Third, our results further reveal the mechanism that drives this

positive relationship. We argue that opportunistic litigation generates additional costs for

accused companies. In order to buffer these negative effects, operating firms use alliances to

obtain similar resources to those that they had reallocated to litigate. In line with (Smeets,

2014) Kesan and Ball, 2006), we find that the costs of litigation are positive and statistically

significant (Model 3, 4 and 5). Interestingly, the Model 2 shows that the length of the lawsuit

has not a significant effect. A potential explanation is that the number of days could be

capturing inherent delays in the US legal system instead of the complexity of the opportunistic

lawsuits. Regarding the financial health of the company, working capital has a negative and

significant impact on all the models (Model 2, 3, 4 and 5). This result shows that potential

liquidity problems act as a driver of alliance formation in an opportunistic litigation context.

However, the amount of debt (leverage ratio) is not significant.

Our results complement and extend previous studies focused on opportunistic litigation, by

performing the first research that demonstrates that a positive effect of PAE litigation on firms

collaborative behavior. Some anecdotal suggestions suggest a potential negative effect of PAE

litigation on firms’ collaborative behavior. As an example, Bessen et al., (2011) mention that

20

litigation may endanger cooperative development because of a potential “contagion effect.”

PAEs could eventually sue suppliers and customers or any other party that had made, used o

sold the patented technology without permission. Conversely, (Tucker, 2011), by studying a

drop in sales after PAE litigation, determines that such slowdown is not due to a decrease in

demand. Instead, the decline in sales is due to a decision of the own litigated company. This

result seems to reject the potential “contagion effect.” Nevertheless, a negative effect exists as

our results show (Figure 1) and should be studied more deeply.

We also advance understanding about the mechanisms that drive the alliance formation,

particularly, from the resource-based perspective.

In addition, we perform this study in an empirically way, which should be highlighted, taking

into consideration the difficulty in measuring trolls’ activity that has repeatedly been

emphasized in the literature (Bessen, Meurer, et al., 2011; Fischer & Henkel, 2012; Pénin,

2012). Moreover, in order to solve the endogeneity problem of the proposed model, we

uncovered that the number of cases filed in the Eastern District of Texas (variable Texas Cases)

works as an instrument. This variable is related to the probability to be litigated by a PAE, but

we do not have evidence that this variable affects the propensity to form alliances. Our sample

shows that from 2003-2008, the Eastern District of Texas was the court with the most decisions

involving PAEs.

This study also pretends to be useful to policymakers. On the one hand, this work helps

lawmakers to have a better understanding of the impact of trolls on innovation as well as it

contributes to extending the knowledge about the potential strategic decisions that companies

adopt when they are litigated. Knowing in advance that litigated companies use alliances as a

way to maintain their competitive position when they have to divert resources to litigate, should

lead policymakers to design measures for promoting and easing alliances.

Future research could improve and build on this paper in several directions. First, as we

mentioned above, researchers should analyze the negative impact of opportunistic litigator on

alliance formation. Second, although we think that our method has systematically captured the

majority of PAEs operating in the high-tech industry, it is not possible to know if they are all.

In this sense, we agree to (Kiebzak et al., 2016) that more work is needed to establish how to

identified PAEs.

21

Notwithstanding these limitations, we believe that this research provides significant

contributions to the relation between opportunistic litigation and alliance formation.

References

Agrawal, A., Bhattacharya, S., & Hasija, S. (2016). Cost-Reducing Innovation and the Role of

Patent Intermediaries in Increasing Market Efficiency. Production and Operations

Management, 25(2), 173–191. http://doi.org/10.1111/poms.12391

Arora, A., Athreye, S., & Huang, C. (2016). The paradox of openness revisited: Collaborative

innovation and patenting by UK innovators. Research Policy, 45(7), 1352–1361.

http://doi.org/10.1016/j.respol.2016.03.019

Bereskin, F. L., Hsu, P.-H., Latham, W. R., & Wang, H. (2017). So Sue Me! The Value

Implication of Patent Litigation. Ssrn, (302). http://doi.org/10.2139/ssrn.3090551

Bessen, J., Ford, J., & Meurer, M. J. (2011). The private and Social Cost of Patent Trolls.

Boston University School of Law Working Paper (Vol. 45).

Bessen, J., & Meurer, M. J. (2005). The Patent Litigation Explosion. B.U.S.L. Law and

Economics Working Papers Series.

Bessen, J., & Meurer, M. J. (2007). The Private Costs of Patent Litigation. SSRN.

Bessen, J., & Meurer, M. J. (2008). Patent Failure: How Judges, Bureaucrats, and Lawyers

Put Innovators at Risk - By James Bessen and Michael J. Meurer. Princeton University

Press Published. http://doi.org/10.1111/j.1541-1338.2009.00438_2.x

Bessen, J., & Meurer, M. J. (2014). The direct cost from NPE disputes. Cornell Raw Review,

99(2), 387–424.

Bessen, J., Meurer, M. J., & Ford, J. (2011). The Private and Social Costs of Patent Trolls.

Regulation, 34, 26–35. http://doi.org/10.2139/ssrn.1930272

Chapman, G., Lucena, A., & Afcha, S. (2018). R&D subsidies & external collaborative

breadth: Differential gains and the role of collaboration experience. Research Policy,

47(3), 623–636. http://doi.org/10.1016/j.respol.2018.01.009

Chien, C. V. (2009). Of Trolls , Davids , Goliaths , and Kings : Narratives and Evidence in the

Litigation of High-Tech Patents. North Carolina Law Review, 85, 1571–1616.

http://doi.org/10.1525/sp.2007.54.1.23.

Cohen, L., Gurun, U. G., & Kominers, S. D. (2014). Patent Trolls: Evidence from Targeted

Firms. Harvard Business School Working Paper. Retrieved from

https://dash.harvard.edu/handle/1/13350439

Cohen, L., Gurun, U. G., & Kominers, S. D. (2017). Patent Trolls: Evidence From Targeted

Firms. NBER Working Papers. Retrieved from http://www.nber.org/papers/w20322

Colombo, M. G., Grilli, L., & Piva, E. (2006). In search of complementary assets: The

determinants of alliance formation of high-tech start-ups. Research Policy, 35(8 SPEC.

ISS.), 1166–1199. http://doi.org/10.1016/j.respol.2006.09.002

Columbia, S. C., & Blasberg, S. L. (2006). Beware patent trolls. Risk Management, 22–27.

Retrieved from http://vmo-blog.com/files/106598-99438/Beware_Patent_Trolls.pdf

22

Czarnitzki, D., Hottenrott, H., & Thorwarth, S. (2011). Industrial research versus

development investment: The implications of financial constraints. Cambridge Journal

of Economics, 35(3), 527–544. http://doi.org/10.1093/cje/beq038

Das, T. K., & Teng, B.-S. (2000). Instabilities of Strategic Alliances: An Internal Tensions

Perspective. Organization Science, 11(1), 77–101.

http://doi.org/10.1287/orsc.11.1.77.12570

Dekkers, R., & Tietze, F. (2014). Excavating the role of NPEs in the innovation process: Did

we start a mission possible? In ICMIT 2014 - 2014 IEEE International Conference on

Management of Innovation and Technology (pp. 111–118).

http://doi.org/10.1109/ICMIT.2014.6942410

Eisenhardt, K. M., & Schoonhoven, C. B. (1996). Resource-Based View of Strategic Alliance

Formation : Strategic and Social Effects in Entrepreneurial Firms. Organization Science,

7(2), 136–150.

Executive Office of the President. (2013). Patent Assertion and U.S. Innovation.

Federal Trade Commission. (2016). Patent Assertion Entity Activity. An FTC Study.

Feldman, R., & Price II, W. N. (2014). Patent Trolling - Why Bio & Pharmaceuticals Are at

Risk. Stanford Technological Law Review, 773, 773–808. http://doi.org/Feldman, Robin

and Price, William Nicholson, Patent Trolling — Why Bio & Pharmaceuticals Are at

Risk (February 14, 2014). 17 Stan. Tech. L. Rev. 773, 2014; UC Hastings Research

Paper No. 93. Available at SSRN: https://ssrn.com/abstract=2395987 or

http://dx.doi.org/10.2139/ssrn.2395987

Fischer, T., & Henkel, J. (2012). Patent trolls on markets for technology - An empirical

analysis of NPEs’ patent acquisitions. Research Policy, 41(9), 1519–1533.


Galasso, A., Schankerman, M., & Serrano, C. J. (2013). Trading and enforcing patent rights.

RAND Journal of Economics, 44(2), 275–312.

Gulati, R. (1998). Alliances and networks. Strategic Management Journal, 19(4), 293–317.

http://doi.org/10.1002/(SICI)1097-0266(199804)19:4<293::AID-SMJ982>3.0.CO;2-M

Gulati, R., Nohria, N., & Zaheer, A. (2000). Startegic Networks. Strategic Management

Jorunal, 21(Special Issue), 199–201.

Hall, B. H., Jaffe, A., & Trajtenberg, M. (2005). Market value and patent citations. RAND

Journal of Economics, 36(1), 16–38.

Henkel, J., & Reitzig, M. (2008). Patent sharks and the sustainability of value destruction

strategies. Academy of Management Proceedings, 1–6.

http://doi.org/10.5465/AMBPP.2008.33653927

Ireland, R. D., Hitt, M. A., & Vaidyanath, D. (2002). Talent Management As a Source of

Competitive Advantage. Journal of Management, 28(3), 413–446.

http://doi.org/http://dx.doi.org/10.1016/S0149-2063(02)00134-4

Kiebzak, S., Rafert, G., & Tucker, C. E. (2016). The effect of patent litigation and patent

assertion entities on entrepreneurial activity. Research Policy, 45(1), 218–231.


Kogut, B. (1988). Joint ventures: Theoretical and empirical perspectives. Strategic

23

Management Journal, 9(4), 319–332.

Lavie, D. (2007). Alliance Portfolios and Firm Performance: A Study of Value Creation and

Appropriation in the U.S. Software Industry. Strategic Management Journal, 28(12),

1187–1212. http://doi.org/10.1002/smj

Lemley, M. A. (2008). Are Universities Patent Trolls? Fordham Intellectual Property, Media

and Entertainment Law Journal, 18(3), 611–631. Retrieved from

https://vpn.uibk.ac.at/+CSCO+0h756767633A2F2F796E6A2E73626571756E7A2E7271

68++/fordham-intellectual-property-media-and-entertainment-law-journal/iplj.htm

Magerman, T., Van Looy, B., & Song, X. (2006). Data Production Methods for Harmonised

Patent Statistics : Patentee Name Harmonisation. EUROSTAT Working Paper and

Studies (Vol. 1). Luxembourg. http://doi.org/10.2139/ssrn.944470

McDonough III, J. F. (2006). The Myth of the Patent Troll: An Alternative View of the

Function of Patent Dealers in an Idea Economy. Emory Law Journal, 56(07), 189.

Retrieved from http://ssrn.com/abstract=959945

McGahan, A. M., & Silverman, B. S. (2006). Profiting from technological innovation by

others: The effect of competitor patenting on firm value. Research Policy, 35(8 SPEC.

ISS.), 1222–1242. http://doi.org/10.1016/j.respol.2006.09.006

Merges, R. P. (2009). The Trouble with Trolls: Innovation, Rent-seeking, and Patent Law

Reform. Berkeley Technology Law Journal, 24(4), 1583–1614.

Miller, S. P. (2017). Introduction to the Stanford NPE Litigation Dataset.

Miner, A., Amburgey, T., & Stearns, T. (1990). Interorganizational linkages and population

dynamics: buffering and transformational shields". Administrative Science Quarterly, 35,

689–713.

Miotti, L., & Sachwald, F. (2003). Co-operative R&D: Why and with whom? An integrated

framework of analysis. Research Policy, 32(8), 1481–1499.

http://doi.org/10.1016/S0048-7333(02)00159-2

Pénin, J. (2012). Strategic uses of patents in markets for technology: A story of fabless firms,

brokers and trolls. Journal of Economic Behavior and Organization, 84(2), 633–641.

http://doi.org/10.1016/j.jebo.2012.09.007

Pohlmann, T., & Opitz, M. (2013). Typology of the patent troll business. R and D

Management, 43(2), 103–120. http://doi.org/10.1111/radm.12003

Powell, W. W., Koput, K. W., & Smith-Doerr, L. (1996). Interorganizational Collaboration

and the Locus of Innovation: Networks of Learning in Biotechnology. Administrative

Science Quarterly, 41(1), 116. http://doi.org/10.2307/2393988

Raffo, J., & Lhuillery, S. (2009). How to play the ‘Names Game’: Patent retrieval comparing

different heuristics. Research Policy, 38(10), 1617–1627.


Reitzig, M., Henkel, J., & Heath, C. (2007). On sharks, trolls, and their patent prey-

Unrealistic damage awards and firms’ strategies of ‘being infringed’. Research Policy,

36(1), 134–154. http://doi.org/10.1016/j.respol.2006.10.003

Reitzig, M., Henkel, J., & Schneider, F. (2010). Collateral damage for R and D

manufacturers: How patent sharks operate in markets for technology. Industrial and

24

Corporate Change, 19(3), 947–967. http://doi.org/10.1093/icc/dtq037

Rosenkopf, L., & Almeida, P. (2003). Overcoming Local Search Through Alliances and

Mobility. Management Science, 49(6), 751–766.

http://doi.org/10.1287/mnsc.49.6.751.16026

Schilling, M. A. (2009). Understanding the alliance data. Stategic Management Journal

Management Journal, 30, 233–260. http://doi.org/10.1002/smj.731

Smeets, R. (2014). Does patent litigation reduce corporate R&D ? An analysis of US public

firms. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2443048

Thoma, G., Torrisi, S., Gambardella, A., Guellec, D., Hall, B. H., & Harhoff, D. (2010).

Harmonizing and combining large datasets - An application to firm-level patent and

accounting data. NBER Working Papers Series, (15851), 1–29.

Thursby, J., Fuller, A. W., & Thursby, M. (2009). US faculty patenting: Inside and outside the

university. Research Policy, 38(1), 14–25. http://doi.org/10.1016/j.respol.2008.09.004

Tucker, C. E. (2011). Patent Trolls and Technology Diffusion. SSRN (online). Retrieved from

http://ssrn.com/abstract=2136955 or http://dx.doi.org/10.2139/ssrn.2136955

Turner, J. L. (2015). Patent Thickets, Trolls and Unproductive Entrepreneurship. Retrieved

from http://ssrn.com/abstract=1916798

Zobel, A., Balsmeier, B., & Chesbrough, H. (2016). Does patenting help or hinder open

innovation ? Evidence from new entrants in the solar industry. Industrial Corporate

Change, 25(2), 307–331. http://doi.org/10.1093/icc/dtw005

25

Figure 1. Differential Effect of Opportunistic Litigation on Alliance Formation.

0.1

.2.3

.4.5

Pro

po

rtio

n o

f firm

s

-5 0 5 10Magnitude of the impact of opportunistic litigation on defendants' subsequent alliance formation

26

Table 1. Name-matched Matching Procedure Results.

Number of pairings Share of total pairings (%)

Matched with RPX

Total matched 2.566 100

Perfect matched 2.324 90.57

Manually matched 242 9.43

Matched with SDC

Total matched 2.462 100

Perfect matched 2.234 90.74

Manually matched 228 9.26

Matched with PATSTAT

Total matched 6,643 100

Perfect matched 5,585 84.07

Manually matched 1.058 15.93

27

Table 2. Definitions of Variables.

Variables Stage of

Analysis Type Description

Source

(created

from)

Strategic Alliances 2 Real Number of alliances that a company

established SDC Platinum

Opportunistic Litigation 1-2 Binary 1 if a company has been attacked by

an opportunistic litigator RPX

R&D Intensity 1-2 Real R&D expenditures divided by the total

operating expenditures of the firm Compustat

Stock of Patent

Applications 1-2 Log

Stock of patent applications (in log

values) PATSTAT

Firm size 1-2 Log Logarithm of the number of

employees Compustat

Fixed Assets Stock 1-2 Ratio Ratio of net fixed assets to total firms'

assets Compustat

Country of Incorporation 1-2 Binary

Dummy variables for controlling

whether the country of the public firm

in the U.S.

Compustat

Texas Cases 1 Real Number of lawsuits filed in the

Eastern District of Texas RPX

Attacks 1 Real Number of plaintiffs RPX

Not Patent 1 Binary 1 if the firm does not have any patent Patstat

SIC dummies 1-2 Binary 6 binary variables based in 2-digit SIC

Code Compustat

Year dummies 1-2 Binary Dummy variables for the time-fixed

effect models Compustat

28

Table 3. Summary Statistics.

Variable Mean SD Min. Max.

Strategic Alliances 0.338 1.014 0 34

Opportunistic Litigation 0.032 0.177 0 1

R&D Intensity 0.259 0.296 0 9.352

Stock of Patent Applications 48.038 430.841 0 1,7267.36

Firm size 5.035 20.472 0 344.36

Fixed Assets Stock 0.142 0.164 0 1

Country of Incorporation 0.816 0.388 0 1

Texas Cases 0.076 0.580 0 18

Attacks 0.432 2.765 0 146

Not Patent 0.590 0.492 0 1

SIC dummies

SIC 35 0.037 0.189 0 1

SIC 36 0.152 0.359 0 1

SIC 38 0.140 0.347 0 1

SIC 48 0.075 0.264 0 1

SIC 73 0.250 0.433 0 1

29

Table 4. Poisson Regression Model that Includes the Endogenous Binary-Treatment Variable.

Equation 1. Probit Model Equation 2. Poisson Model

Model 1 Model 1

Variables (Equation 1) (Equation 2)

Texas Casesit-1 0.227***

(0.055) Attacksit-1 0.032***

(0.008) R&D Intensityit-1 1.310*

(0.732) R&D Intensity Squareit-1 (log) -1.911

(1.200) Firm Size (log) 0.277**

(0.113) Firm Size Square it-1 (log) -0.005

(0.007) Not Patent it-1 0.061

(0.119) Stock of Patent Applications it-1 (log) 0.132***

(0.029) Fixed Assets Stock it-1 -1.283***

(0.361) Country of Incorporation it-1 0.247**

(0.114) R&D Intensity 0.148 (0.119)

Stock of Patent Applications (log) 0.160***

(0.019)

Firm Size (log) 0.275***

(0.020)

Fixed Assets Stock -0.812***

(0.265)

Country of Incorporation 0.657***

(0.091)

1.Opportunistic Litigation 1.540***

(0.137)

Sectorial dummies (2 digits) Included Included Year dummies Included Included Constant -4.610*** -3.826***

(0.481) (0.173)

/athrho -1.441***

/lnsigma (0.355)

rho -0.147***

sigma (0.045)

Observations 6,375 6,375

Robust standard errors in parentheses *** p<0.01, ** p<0.05, *p<0.1

Dependent variable Eq (1): Litigation probability

Dependent variable Eq (2): Stategic Alliancet+1

30

Table 5. Descriptive Statistics for Treated and Control Firms Separately.

Variables Treated Firms Control Firms Difference P-value

(N=394) (N=11,763)

Control variables

Texas Cases 0.898 0.034 0.863 ***

Attacks 3.541 0.332 3.540 ***

R&D Intensity 0.167 0.258 -0.091 ***

R&D Intensity Square 0.04243 0.14137 -0.099 ***

Firm Size (log) 8.670 5.556 3.114 ***

Firm Size Square (log) 79.414 35.825 43.589 ***

Not Patent 0.172 0.514 -0.342 ***

Stock of Patent Applications (log) 4.630 1.875 2.755 ***

Fixed Assets Stock 0.141 0.124 0.017 **

Country of Incorporation 0.811 0.843 -0.032 *

SIC 35 0.168 0.045 0.123 ***

SIC 36 0.348 0.204 0.144 ***

SIC 38 0.045 0.157 -0.112 ***

SIC 48 0.041 0.026 0.015 *

SIC 73 0.348 0.234 0.115 ***

Outcome variable

Alliance Formation 1.594 0.341 1.25 ***


31

Table 6. Matching Results.

Variables Treated Firms Control Firms Difference P-value

(N=185) (N=6,270)

Control variables

Texas Cases 0.330 0.277 0.053

Attacks 1.545 1.310 0.235

R&D Intensity 0.168 0.166 0.003

R&D Intensity Square 0.042 0.046 -0.004

Firm Size (log) 8.178 8.111 0.067

Firm Size Square (log) 70.882 69.496 1.386

Not Patent 0.222 0.207 0.014

Stock of Patent Applications (log) 3.813 3.806 0.007

Fixed Assets Stock 0.134 0.121 0.013

Country of Incorporation 0.784 0.821 -0.037

SIC 35 0.124 0.149 -0.024

SIC 36 0.357 0.329 0.028

SIC 38 0.054 0.047 0.007

SIC 48 0.038 0.047 -0.010

SIC 73 0.389 0.391 -0.002

Outcome variable Alliance Formation 0.989 0.642 0.347 ***


32

Table 7. The Effect of Litigation Costs and Firms’ Resource Endowment.

Variables

Individual Alliance Formation Impact of Opportunistic litigation

Model 2 Model 3 Model 4 Model 5

Days in Litigation (log) 0.217

(0.131) Patents in Suit (log) 0.409**

(0.178) Number of Dockets (log) 0.269**

(0.111) Accused Products (log) 0.310**

(0.123)

Leverage Ratio -0.948 -1.066 -0.622 -0.984

(0.963) (0.969) (0.932) (0.963)

Working Capital -2.007*** -2.093*** -1.802*** -2.138***

(0.627) (0.608) (0.589) (0.618)

Country of incorporation 0.648** 0.658** 0.699** 0.650**

(0.297) (0.288) (0.301) (0.276)

SIC 35 -1.187 -1.342 -1.284 -1.414

(1.306) (1.287) (1.275) (1.266)

SIC 36 -1.262 -1.433 -1.410 -1.486

(1.295) (1.283) (1.278) (1.258)

SIC 38 -0.884 -0.943 -0.931 -1.051

(1.296) (1.244) (1.252) (1.239)

SIC 48 -1.892 -2.004 -2.110 -1.993

(1.408) (1.381) (1.422) (1.315)

SIC 73 -0.950 -1.017 -1.060 -1.078

(1.332) (1.296) (1.300) (1.269)

2005.year 0.704* 0.653 0.748* 0.876**

(0.422) (0.395) (0.409) (0.422)

2006.year 0.500 0.380 0.513 0.560

(0.455) (0.438) (0.457) (0.453)

2007.year 0.835** 0.830** 0.859** 0.891**

(0.368) (0.385) (0.371) (0.384)

2008.year -0.027 -0.122 -0.047 -0.183

(0.257) (0.259) (0.257) (0.271)

Constant 0.261 1.341 0.150 1.452

(1.498) (1.193) (1.174) (1.159)

Observations 183 183 183 183

R-squared 0.159 0.169 0.180 0.182

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

33

Appendix A. Table 1. Robusness check

(1) (2)

VARIABLES xb bo

LStrategicAlliance 0.042***

(0.013) LdPAE_6 0.167**

(0.084) LRD_intensity_op 0.238**

(0.117) Llog_stock_patent_app 0.167***

(0.032) Llog_employees 0.275***

(0.031) Lshare_fixed -0.758***

(0.290) Lloc46 0.638***

(0.123) 35.SIC_2 -0.480***

(0.143) 36.SIC_2 -0.906***

(0.109) 38.SIC_2 -0.702***

(0.100) 48.SIC_2 -0.909***

(0.294) 73.SIC_2 0.104

(0.098) 2005.year 0.113

(0.070) 2006.year 0.353***

(0.070) 2007.year 0.194**

(0.077) 2008.year -0.011

(0.076) Constant -3.572***

(0.197)

Observations 7,147 7,147

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

abstract - conference.druid.dk · participants in the us economy - the context of our study and...

Documents