are “hidden” contracts bad for buyers
TRANSCRIPT
COMPETITION AND THE QUALITY OF STANDARD FORM CONTRACTS: AN EMPIRICAL ANALYSIS OF SOFTWARE LICENSE
AGREEMENTS∗
FLORENCIA MAROTTA-WURGLER NYU Law School
Center for Law and Business
August 22, 2005
ABSTRACT
Most commercial transactions are governed by standard form contracts. But despite the ubiquity of standard forms, there has been little empirical study of their content and determinants. This paper provides a comprehensive analysis of 647 software license agreements, paying special attention to the role of competitive forces in shaping standard terms. Friedrich Kessler argued that companies in concentrated industries, or companies with dominant market share, take advantage of their position by imposing one-sided standard terms that limit their obligations to consumers; other scholars have argued that market structure is irrelevant to the standard terms offered to buyers. The results support the latter view. While software products’ prices are highly sensitive to competitive conditions such as concentration and market share, license agreement terms are not. By demonstrating that sellers with market power do not offer unusually harsh terms, the results suggest that an important aspect of the standard analysis of procedural unconscionability is misguided. More broadly, the analysis provides a wealth of descriptive detail about an important class of modern standard form contracts.
∗ Wagner Fellow in Law and Business, NYU Law School and Stern School of Business. Email:
[email protected]. Webpage: http://homepages.nyu.edu/~fm275/. I am grateful to Barry Adler, Bill Allen, Jennifer Arlen, Yannis Bakos, Clayton Gillette, Lewis Kornhauser, Roberta Romano, Peter Siegelman, and Jeff Wurgler for helpful suggestions, and Leonard Lee, Christine Murphy, and Yuriy Prilutsky for excellent research assistance.
I. INTRODUCTION
It is estimated that 99% of all commerical contracts are standard form contracts.1
Examples include the warranties that accompany electronic appliances, the myriad terms of
software license agreements, the details of cell phone service contracts, and the fine print on
concert tickets. While consumers are able to determine the quantity and perhaps the delivery
method of the goods they purchase, they are often left with little choice but to accept many
important secondary terms presented in non-negotiable boilerplate. Courts generally enforce
these standard form contracts as long as they are not procedurally or substantively
“unconscionable.”2
Many scholars believe that standard form contracts are unfair. Friedrich Kessler noted
that standard form contract terms, like a product’s price, could be an instrument of market power
as dominant corporations may impose terms that greatly limit their obligations to consumers.3
Subsequent commentators have suggested that competition is the main force that ensures the
efficiency of standard terms, since sellers in competitive markets must offer the combination of
price and terms that buyers prefer in order to survive. Indeed, courts invalidate provisions in
standard form contracts under the doctrine of unconscionability, of which an important factor is
whether the buyer is deprived of meaningful choice because the seller is a monopolist4 or has
significant market share.5
1 W. David Slawson, Standard Form Contracts and Democratic Control of Law Making Power [hereinafter
Slawson, Standard Form Contracts and Democratic Control], 84 Harv. L. Rev. 529 (1971). 2 See UCC §2-218. Procedural unconscionability refers to oppression used in the process of making a
contract, such as unequal bargaining power. Substantive unconscionability refers to oppressive or harsh content in a contract. Both factors alone are necessary but not sufficient for a court to invalidate a contract under the unconscionability doctrine, although this varies from jurisdiction to jurisdiction. See James J. White & Robert Summers, Handbook of the Law Under the Uniform Commercial Code 128 (1972).
3 Friedrich Kessler, Contracts of Adhesion - Some Thoughts About Freedom of Contract, 43 Colum. L. Rev. 629 (1943).
4 See, e.g., Pack v. Damon Corp., 320 F. Supp. 2d 545 (E.D. Mich.2 004) at 556 (finding no procedural unconscionability because the “[p]laintiff has not shown that GRVC was his only source for buying a new motor home, or that other potential sources required submitting disputes to arbitration”); Lozada v. Dale Baker Oldsmobile, Inc., 91 F. Supp. 2d 1087 (D. Mich. 2000) at 1100 (noting that “[i]n order to determine whether a contract is procedurally unconscionable, the court typically considers the relative bargaining power of the parties, their relative economic strength, the alternative sources of supply,”); Rozeboom v. Northwestern Bell Telephone Co, 358 N.W. 2d 241, 242 (S.D. 1984) at 242–45 (finding a term unconscionable because the seller was a monopoly and the buyer could not shop for alternatives).
5 See, e.g., Flores v. Transamerica HomeFirst, Inc., 93 Cal. App. 4th 846, 853 (Cal. Ct. App. 2001); Arnold v. United Companies Lending Corp., 204 W. Va. 229, 511 S.E.2d 854, 861 (1998) (“The relative positions of the parties, a national corporate lender on one side and elderly, unsophisticated consumers on the other, were ‘grossly unequal.’”); A & M Produce Co. v. FMC Corp., 135 Cal. App. 3d 473, 186 Cal. Rptr. 114, 125 (1982) (finding
2
The view that market power affects the quality of standard terms is far from universally
accepted, however. Several theorists have pointed out that if consumers prefer warranties instead
of disclaimers, for example, and if they would be willing to pay a premium for this protection, a
monopolist will offer warranties.6 Hence, only the monopolist’s pricing behavior and not the
standard terms it offered would be cause for concern or regulation. Others have suggested that
competition is irrelevant to the content of standard terms for a quite different reason, namely that
buyers rarely read the fine print anyway and hence do not make purchase decisions based on its
content.7
Despite the large literature exploring the nature of standard form contracts and the
practical importance attaching to their enforcement, there has been little systematic empirical
work analyzing standard form contracts. This paper makes two contributions. The first is to offer
a detailed empirical description of an important modern class of standard form contracts,
software end-user license agreements (EULAs). The second is to investigate whether EULA
terms are, or are not, related to the competitive conditions that prevail in different segments of
the software industry.
My analysis is based on a unique, hand-collected sample of 647 EULAs of
“prepackaged” (i.e., non-custom) software. The contracts in my sample include examples from
598 different software companies, including virtually all well-known software publishers and
hundreds of smaller firms. The software programs that the sample EULAs span 114 distinct
software markets, from antivirus to voice recognition, and involve products targeted at the
general public as well as medium- to large-business users.
procedural unconscionability because FMC had substantially more sales than A&M and thus much more bargaining power).
6I summarize this literature later in the paper, but noteworthy contributions include A. Michael Spence, Monopoly, Quality, and Regulation, The Bell Journal of Economics, Vol. 6, No. 2 (1975); Alan Schwartz, A Reexamination of Non-Substantive Unconscionability, 63 U. Va. L. Rev. 1074 (1977); George L. Priest, A Theory of the Consumer Product Warranty [hereinafter Priest, A Theory of the Consumer Product Warranty], 90 Yale L. J. 1297 (1981); Alan Schwartz & Louis Wilde, Intervening in Markets on the Basis of Imperfect Information: A Legal and Economic Analysis, 127 U. Pa. L. Rev. 630 (1979); Alan Schwartz & Louis Wilde, Imperfect Information in Markets: The Examples of Warranties and Security Interests, 69 Va. L. Rev. 1387 (1983); Richard Craswell, Passing on the Costs of Legal Rules: Efficiency and Distribution in Buyer-Seller Relationships, 43 Stan. L. Rev. 361, 363 (1991). See also Richard Posner, Natural Monopoly And Its Regulation, 21 Stan. L. Rev. 548 (1969).
7 Later in the paper, I summarize the perspectives of Victor P. Goldberg. Institutional Change and the Quasi-Invisible Hand, 17 J. L. & Econ. 461, 485 (1974); Avery Katz, Your Terms or Mine? The Duty to Read Fine Print in Form Contract, 21 RAND J. of Econ. 518, 533 (1990); Russell B. Korobkin, Bounded Rationality, Standard Form Contracts, and Unconscionability 70 Chi. L. Rev. 1203 (2003).
3
To characterize the content of EULAs, I constructed a simple index to measure overall
“buyer friendliness.” This index is based on 25 common terms that allocate rights and risks
between buyers and sellers of software. The terms address acceptance of the license, scope of the
license, restrictions on transfer, warranties and disclaimers of warranties, limitations on liability,
maintenance and support services, and conflict resolution. For each term that is more pro-seller
relative to the default rules of Article 2 of the Uniform Commercial Code (UCC), I assign a
negative one point score, and for each term that is more pro-buyer, I assign a positive one point
score. I assign a score of zero if the contract is silent in regards to the specified term, or if the
specified term matches the default rule. I measure the overall bias of a given EULA as the sum
of the 25 scores—the more positive is the sum, the more pro-buyer is the contract. While crude,
this approach captures the overall tone of the EULA in a way that allows for empirical analysis. I
also relax the assumption that all terms matter equally to buyers by studying seven subindexes
that isolate categories of related terms.
Not surprisingly, the analysis shows that EULAs are generally more tilted toward the
sellers relative to the default rules of Article 2. There is a great deal of variation in the sample,
however. Larger companies and younger companies present significantly more one-sided (i.e.,
pro-seller) contracts. Also, I find that EULAs associated with products targeted toward the
general public are not significantly more pro-seller than are the EULAs associated with business-
oriented products. Furthermore, since the sample includes 98 EULAs from 49 companies that
sell both consumer and business versions of the same product, I can cleanly test, and reject, the
notion that sellers actively discriminate between user types by offering different terms. These
findings cast doubt on the notion that the general public is especially at risk when it comes to
boilerplate. Such generalizations, as well as the detailed analysis of each of the 25 terms that I
follow, provide a number of important new facts about standard form contracts.
To measure the effect of market power, I assemble data on market share of individual
companies in different software markets represented in my sample as well as the overall
concentration of those markets. Market share data is not widely available, so I estimated market
shares and concentration measures using a recently developed methodology that uses
Amazon.com product “sales rankings” to estimate sales. I also obtained some additional market
share data from various published estimates.
4
I find that the overall quality of standard terms is essentially uncorrelated with
competitive conditions. While competition does significantly reduce product prices, it does not,
from the buyer’s perspective, improve EULA terms. More competitive markets within the
software industry offer similar standard form terms to highly concentrated markets; within a
given software market, firms with larger market share offer similar terms as minor players. There
are a few exceptions involving particular sets of terms, but the main conclusion is that most
EULA terms do not depend on competitive conditions in an important way. In summary, while I
am unable to defend absolute statements about whether EULA terms are or are not efficient,
what I can prove is a relative statement that EULA terms are not more biased when sellers have
market power.
One practical implication of these results is that courts should de-emphasize market
structure and market share as factors when determining whether a particular term is procedurally
unconscionable. Also, by illustrating the actual distribution of EULA terms in practice, the
descriptive analysis may allow judges, legislators and practitioners to better determine what is
and is not unusual and thus how to better identify suspicious provisions. Finally, a careful study
of software licenses may help focus recent debates over the desirability of creating a uniform
body of law for on-line contracting.
The paper proceeds as follows. Section II summarizes views about whether competitive
conditions affect the quality of standard form contracts and reviews the small body of relevant
empirical work. Section III provides a descriptive analysis of software license agreements.
Section IV examines the relationship between license terms and competitive conditions. Section
V concludes.
II. COMPETITION AND STANDARD FORM TERMS:
THEORY, PRIOR EVIDENCE, AND A NEW TEST
A. Standard Terms Depend on Market Structure
In a seminal article, Friedrich Kessler posited that sellers with market power exploited
consumers by imposing harsh standard terms in a take-it-or-leave-it fashion.8 He argued that
“standard form contracts are typically used by enterprises with strong bargaining power. The
weaker party, in need of goods and services, is frequently not in a position to shop around for
8 See supra note 3.
5
better terms, either because the author of the standard form contract has a monopoly (natural or
artificial) or because all competitors used the same clauses.”9 Thus Kessler feared that in the
absence of competition, consumers are likely to be stuck with both high prices and poor standard
terms.
Courts have also frequently articulated, and acted on, a belief that market power shapes
standard terms. In Henningsen v. Bloomsfield Motors10, for example, the New Jersey Supreme
Court, in striking down Chrysler’s restrictive warranty, emphasized the fact that only three
manufacturers, including Chrysler, controlled over 90 percent of the passenger car market.
Echoing Kessler, the court wrote that “the gross inequality of bargaining position occupied by
the consumer is apparent. There is no competition among car makers in the area of the express
warranty. Where can the car buyer go to negotiate for better protection? Such control and
limitation of his remedies are inimical to the public welfare and, at the very least, call for great
care by the courts to avoid injustice through application of strict common-law principles of
freedom of contract.”11 In a study of automobile warranties, William Whitford corroborated the
court’s view in concluding that the warranties offered were designed to minimize seller’s costs,
not to internalize buyers’ interests.12
Lewis Kornhauser developed the notion in Henningsen regarding tacit collusion among
car manufacturers. He argued that “where market concentration exists one will probably observe
[...] shoddy or less durable goods, or oppressive contract terms assigning risks to buyers that
might be borne by sellers when there is less market concentration.”13 The suggestion is that
sellers in concentrated markets may tacitly agree not to compete on certain dimensions, such as
warranty coverage, to facilitate price coordination and thus monopoly profits.
B. Standard Terms Are Independent of Market Structure
Other scholars disagree with the view that standard terms depend on competitive
conditions. In particular, some suggest that standard terms may be efficient even under
monopoly. These authors point out that a product’s standard terms are just one of many attributes
9 Id at 632. 10 32 N. J. 358 (1960). 11 Id at 391. 12 William C. Whitford, Law and the Consumer Transaction: A Case Study of the Automobile Warranty,
Wis. L. Rev. 1006 (2004) at 1062. 13 Lewis Kornhauser, Unconscionability in Standard Forms, 64 Cal. L. Rev. 1179 (1976) at 1169.
6
of a product and can be viewed as a dimension of product quality.14 And A. Michael Spence
shows that a profit-maximizing monopolist will offer whatever product quality is preferred by
the marginal consumer (the consumer who is just willing to pay the going price), since that level
maximizes her willingness to pay for the product.15 Thus as long as the preferences of the
marginal consumer are the same as those of the average consumer, even a monopolist will offer
optimal terms, albeit at a supra-competitive price.
In fact, this theoretical conclusion may hold even in the more realistic case that a majority
of consumers do not read standard terms. Applying the “informed minority” logic, Schwartz and
Wilde show that non-reading buyers will benefit from a minority of informed buyers whose
willingness to pay for the product is sensitive to the quality of the standard terms.16 If all buyers
have the same taste for quality, a monopolist that cannot discriminate between reading and non-
reading buyers may offer the terms preferred by all buyers.17
Other scholars agree that standard terms do not depend critically on market structure, but
argue that the terms offered are likely to be biased toward the seller. For instance, Victor
Goldberg points out that if a market is competitive with respect to price it does not follow that it
will be competitive with respect to terms.18 This is because shopping for terms—which involves
reading lengthy, hard-to-understand contracts—might be costlier than shopping for price.
Goldberg notes that “unless the firm intentionally makes the particular term an important selling
point [...] few, if any, customers will perceive the existence of variations in terms.”19 When
buyers only shop for price and not terms, firms will lower their contract quality to reduce price
and compete for buyers. In equilibrium, price will be low and the terms harsh. Goldberg and
others suggest that a few aggressive term shoppers might not suffice to correct for this market
failure. In partifcular, Cruz and Hinck argue that under realistic conditions, it would take not a
14 See ProCD v. Zeidenberg, 86 F.3d 1447 (7th Cir. 1997). In refusing to strike down the standard terms that
were shrinkwrapped in a software package, Judge Easterbrook reasoned that “[t]erms of use are no less a part of ‘the product’ than are the size of the database and the speed with which the software compiles listings,” at 1453.
15 See Spence supra note 6. 16 See Schwartz and Wilde, supra note 6. 17 On the other hand, if some characteristic of the good—e.g., warranty terms—does facilitate price
discrimination, quality is likely to be affected. See Spence, supra note 6, for a proof that a price-regulated monopolist will provide lower quality goods. See Richard Schmalensee, Market Structure, Durability, and Product Quality: A Selective Survey, Economic Inquiry 17(2) (1979) for a summary of scenarios under which a monopolist would provide lower product quality than that in a competitive industry with similar cost conditions.
18 Goldberg supra note 7. Schwartz and Wilde, supra note 6, make a similar argument. 19 Goldberg supra note 7 at 485.
7
few but rather a significant fraction of buyers in a market to cure imperfections in form
contracts.20
Avery Katz develops a related argument.21 He argues that, under certain conditions, not
reading standard terms is a dominant strategy for the small minority of buyers who place a high
value on the quality of standard terms. The intuition is that sellers will offer quality no higher
than necessary for these buyers to accept their offer. Anticipating this, such buyers will no longer
find reading worthwhile, thus dropping out of the group of readers. But this leads sellers to offer
the lowest quality required to attract the readers with the next-most quality sensitive preferences,
who decline to read, and so on. In equilibrium, nobody reads and terms are low quality.
Behavioral rationales have also been offered to explain consumers’ failure to read.
Russell Korobkin, for example, suggests that boundedly rational consumers will only focus on
the salient features of a product, such as price, quantity, and warranty terms.22 He concludes that
competition will lead salient terms to be optimal but cannot keep non-salient terms from being
biased toward the seller.
C. Prior Empirical Studies
Richard Schmalensee wrote in 1979 that “there is an obvious need for empirical work to
confront the implications of the theoretical literature with the data.”23 But even twenty-five years
later there still have been only a handful of studies that attempt to measure the effects of
competition on product quality, and only a fraction of these have examined the quality dimension
of most interest to commercial and contract law scholars, namely, standard form terms.
Studies finding a link between market concentration and general product quality include
one by Michael Mazzeo, who finds that flight delays are significantly greater, both in frequency
and in magnitude, on routes where only one airline provides direct service.24 He also finds that
increases in competition are associated with improved on-time performance. Stephen Foreman
and Dennis Shea also examine the airline industry and find similar results.25 Caroline Hoxby
20 R. Ted Cruz and Jeffrey J. Hinck, Not My Brother’s Keeper: The Inability of an Informed Minority to
Correct for Imperfect Information, 47 Hastings L. J. 635 (1995) at 675. 21 Katz supra note 7. 22 Korobkin, supra note 7. The view that buyers may focus on a subset of terms was discussed in Todd
Rakoff, Contracts of Adhesion: An Essay in Reconstruction 96 Harv. L. Rev. 1173 (1982). 23 Schmalensee supra note 17 at 194. 24 Michael J. Mazzeo, Competition and Service Quality in the U.S. Airline Industry, 22 Rev. of Ind. Org.
275 (2003). 25 Stephen Foreman & Dennis Shea, Publication of Information and Market Response: The Case of Airline
on Time Performance Reports, 14 Rev. Ind. Org. 147 (1999).
8
studies school competition and finds evidence that metropolitan areas with more school districts
have higher quality schools, as measured by indices of student achievement.26
As to studies that directly address the quality of standard terms, perhaps the first striking
feature of the literature is the lack of simple descriptive evidence on standard form terms, which
is logically the first step to examining the effects of competition. Moreover, the most relevant
studies are now quite dated. They include those of George Bogert and E. E. Fink, who examine
warranties of small samples of various goods, and of George Priest, who studies home appliance
warranties.27 Bogert and Fink provide anecdotal descriptions of warranty terms and suggest that
trade associations standardized warranty practices to achieve collusion (of the sort later
envisioned by Kornhauser), but present no systematic analysis.
In the study most closely related study to mine, Priest examined the warranties of 62
products of different types of household appliances, such as refrigerators and washing machines,
and collected market share and industry concentration data. He found that firms with large
market share within a particular industry did not offer more restrictive terms than smaller firms
within that same industry. He also found no relationship between the level of industry
concentration and the degree of warranty coverage.
While an important and highly-cited contribution, Priest’s study, and the prior literature
more broadly, leaves many open questions. First, Priest’s finding of little association between
warranty terms and competition contrasts with the studies of airlines and schools that show a
strong relationship, indicating the need for further study. Second, Priest’s study did not attempt
to verify that concentration affects price for the products that he studied. He thus cannot rule out
the critique that he finds little effect because the markets he examines are contestable, or that his
proxies for competitive conditions are too noisy.
Third, and perhaps most importantly, there is still little basic descriptive evidence of the
content of standard form agreements. The most detailed prior empirical studies of standard terms
consider only small samples of home appliance or automobile warranties. Richer examples of
26 Caroline M. Hoxby, Does Competition Among Public Schools Benefit Students and Taxpayers?, 90 Am.
Ec. Rev. 147 (2000). 27 George G. Bogert & E. E. Fink, Business Practice Regarding Warranties in the Sale of Goods, 25 Ill. L.
Rev. 400 (1930); Priest, A Theory of the Consumer Product Warranty, supra note 6. See also W. Whitford, Law and the Consumer Transaction, supra note 12; Jennifer L. Gerner & W. Keith Bryant, Appliance Warranties as a Market Signal? 15 J. Consumer Aff. 75 (1981). See also Whitford, supra note 12; Marcel Kahan & Michael Klausner, Standardization and Innovation in Corporate Contracting (or “The Economics of Boilerplate”), 83 Va. L. Rev. 713 (1997).
9
standard form contracts, such as those including forum selection clauses, limitations of liability,
and so forth, have yet to be described at all, let alone connected to competitive conditions.
D. A Test Using Software End-User License Agreements
To shed new light on standard form contracting practices in general and the role of
competition in particular, I use data from the software industry. I examine the end-user license
agreements, or EULAs, found with typical “pre-packaged” (non-customized) software products.
These contracts present a rich set of standard terms that are important to understand in their own
right, given that sales of all types of software exceeds $100 billion per year.28 In addition, a
careful study of software license agreements should help focus recent debates over the
desirability of the Uniform Computer Information Transactions Act (UCITA), a contract law
statute developed by the National Conference of Commissioners on Uniform State Laws in an
effort to create a uniform and cohesive body of law for contracts of computer information such
as software. UCITA generated heated debate.29 It was enacted in Maryland and Virginia, and
while efforts to encourage other states to adopt it were suspended in 2003, those who oppose it
continue to be concerned that courts are using this model act as a reference for deciding cases.30
Given that the literature includes few large-sample studies of standard form contracts, as
mentioned above, my analysis begins with a descriptive analysis of EULAs. I gather a large
sample of these agreements, including one contract (or at most two contracts) from almost all
28 Software and Information Industry Association, Software Industry Profile, June 2004: Packaged
Software, at http://www.siia.net/software/pubs/profile_0604.pdf. 29 For views in support of UCITA, see e.g., Richard Epstein, In Defence of UCITA, Financial Times, April
2 2003, at http://news.ft.com/cms/s/2f8a79ca-d1b5-11d8-83e4-0003ba5a9905.html; Clayton Gillette, Letter in Support of UCITA to NCCUSL Standby Committee on UCITA, Jan. 21, 2003, at http://www.nccusl.org/nccusl/ucita/ucita/GilletteEndorse.pdf; George Priest, Letter in Support of UCITA to Carlyle C. Ring, Esq. Chair, NCCUSL Standby Committee on UCITA, Jan. 15, 2003, at http://www.nccusl.org/nccusl/ucita/ucita/ProfPriestletter.pdf. For views against UCITA, see. e.g, Lawrence Lessig, Sign it and Weep, The Standard, Nov. 20, 1998; Bad Software: A consumer Protection Guide, (listing individuals and organizations that oppose UCITA, including attorney general, consumer advocacy organizations, and academics, among others), at http://www.badsoftware.com/oppose.htm; Jean Braucher, The Failed Promise of the UCITA Mass-Market Concept and its Lessons for Policing of Standard Form Contracts, 77 J. Small & Emerging Bus. L. 393 (2003).
30 See Va. Code Ann. 59.1-501.1 to 509.2 (Michie. 2001); Md. Code Ann. Comm. Law 22-101 to 22-816 (Supp. 2002). Although the main goal of UCITA is to delineate a body of law under which to enforce standard form contracts of electronic and information products, it also tangentially interacts with intellectual property issues. However, the intellectual property issues involved in software licensing are beyond the scope of this paper. See Daniel A. DeMarco & Christopher B. Wick, Now UCITA, Now You Don’t: A Bankruptcy Practitioner’s Observations on the Proposed Uniform Computer Information Transactions Act, 34 Am. Bankr. Inst. J. 23 (2004) (noting that “while it may appear that UCITA disappeared on Feb. 10, 2003 […] we can all expect to encounter UCITA issues, and courts guided by UCITA’s principles, more frequently as computer information transactions become more prevalent in today’s economy.”
10
well-known software publishers as well as examples from hundreds of smaller companies. For
each EULA, I then tabulate the presence of various standard terms, noting in particular the extent
to which the terms are biased, relative to the appropriate default rules, in favor of the seller or
(more rarely) the buyer. In this way I describe the “quality” of an important, modern class of
standard form contracts.
I then turn to the relationship between competitive conditions and EULA terms. I classify
each company’s product into one of 146 distinct software “markets” that Amazon.com uses to
organize its software, from animation to voice recognition. A key advantage of my setting is that
the product, “pre-packaged” software, is homogeneous enough to make comparison of EULAs
across these many markets meaningful. While software varies greatly in its purpose and function,
program failure is typically caused by generic factors such as bugs, buyer misuse, or
compatibility problems with the buyer’s hardware or other software. Consequently, sellers of
different types of software tend to protect themselves in similar ways. But at the same time, the
software markets themselves are characterized by a wide range of market structures. Some
markets contain one or a small number of clearly dominant players, while others are quite
unconcentrated. The essence of the test is to see whether the EULA terms offered to consumers
depend on the key competitive conditions, such as market concentration and the seller’s own
market share, that characterize the market for that product.
III. SOFTWARE EULAS: A DESCRIPTIVE ANALYSIS
A. Sample
I gather EULAs for two sets of “prepackaged” (non-custom) software products. The first
set includes one representative product, or at most two products, from software publishing
companies that sell their products online through their corporate website.31 The starting point for
finding such companies was the Software Industry Directory 2005 CD-ROM, a comprehensive
database of over 7,000 software development and publishing companies. Several thousand
companies in this Directory sold custom software, e.g., they were systems integrators, or were
merely resellers, not selling their own software products. For the remaining firms, I manually
31This selection criterion is tailored to the analysis in Florencia Marotta-Wurgler, Are ‘Pay Now, Terms
Later’ Contracts Worse for Buyers? Evidence From Software License Agreements, NYU Law and Economics Working Paper No. 05-10 (2005). The data gathering process for this part of the sample is described in more detail there.
11
checked each company’s website to determine whether its products were sold through its
website. If this was the case, I attempted to obtain a EULA by searching the company’s website
or by requesting it directly. When a flagship product was apparent, I gathered its EULA, else I
chose a product at random. This process led to 515 EULAs from 468 companies, since from 47
companies in this sample I gathered a EULA for a product oriented to the general public or home
office user as well as a EULA for a similar product aimed at a business-oriented user.
The second set of products consists of one EULA from each software company that sells
its products through Amazon.com and is not already represented in the first set of products—in
other words, the company sells software through Amazon.com but not necessarily through its
own corporate website. (As explained later, data from Amazon.com is necessary to estimate
market share and market concentration.) Here, to identify companies’ flagship products, I
selected those with relatively high Amazon “sales ranks.” I then obtained the EULAs for these
products. This led to 132 additional EULAs from 130 additional companies (for two companies
it was convenient to gather both consumer and business-product licenses).
The total sample includes 647 EULAs from 598 companies. Based on the above
description of the sample collection process, the reader might be concerned that because EULAs
were gathered for firms that sell on Amazon.com or through their own webpage, the sample is
subject to some type of selection bias. This should not be a major concern. Virtually all
prepackaged software companies of any significance sell their products through their website or
Amazon.com, and hence are represented in my sample. The sample is as comprehensive and
representative as is practical to assemble, and in any case is an order of magnitude larger than the
prior samples of standard form contracts that have been studied in the law and economics
literature.
For a EULA to be included in the final sample, I also required the availability of certain
company and product data. I collected basic company characteristics such as revenue and the
age, defined as 2005 minus the year of incorporation or founding, with a maxmimum age of 25
(since any pre-1980 operations were unlikely to have emphasized software publishing). I also
gathered data on whether the company is private or public. I obtained most of these data from
Hoover’s Company Directory. Table 1 summarizes the characteristics of the companies whose
EULAs are included in my sample. Mean sales are $499 million, but range from $50,000 to
12
$87.5 billion. Median sales are $2.2 million. The average age is 15 years, but the sample includes
both startups and long-established firms. Publicly traded firms make up 16% of the sample.
Several product characteristics are important to track. An obvious one is the product’s
price. Others concern the nature of the license. Most licenses in the sample are of unlimited
duration and for a single user. Because a multi-user license would likely increase price and may
come with different terms, I record whether the license allows for multiple users. Another
important characteristic is whether the license is “developer” or “standard.” Developer licenses
allow the buyer to use the software to develop derivative products, and hence are commonly
found with products that aid programmers in creating new software. Since some of the rights
granted under the developer EULA may differ from those under the standard EULAs, I record
this aspect of the license.
I also record basic aspects of the function of the software. I subjectively determine
whether the product is targeted to a consumer (or home business) user or a large business user.
This allows for a test of whether companies impose poorer standard terms on less sophisticated
buyers.32 I also classified each product into one of 146 “markets” used by Amazon.com to
categorize software. For products obtained from Amazon.com, this classification is provided,
and for products obtained from the Software Industry Directory 2005 sample, I matched the
product to the market manually.
Table 2 reports some key product summary statistics and also illustrates the breadth of
types of software represented in the sample. Of the 146 Amazon.com markets (excluding
children’s software and games, categories for which I do not gather EULAs), 114 are represented
by at least one EULA in my sample. For example, the category that Amazon.com calls the
“Business & Office > Business Accounting > Accounting” market is represented by 13 EULAs.
The average price of the products in this category is $1,302. Four of the EULAs in this category
are for consumer-oriented products, whose average price is $144, while the remaining nine
EULAs are for business-oriented products and the average price for these is $1,816. Overall, the
average price of the products in the sample is $763, a result influenced by a few very expensive
products. The median price is $200. Just under half of the products in the sample, 45%, are
oriented towards consumers or home businesses. The mean price of these consumer-oriented
32 For example, a product entitled Cyber Sentinel 3.0 Home Edition, designed to prevent children from
accessing adult sites from a home computer, is categorized as “consumer” software, while software products targeted to large firms, such as Client Management Services v1.30, is categorized as “business” software.
13
products is $142 (the median is $61) and the mean price for medium to large business-oriented
products is $1,268 (the median is $499).
B. Measuring Bias
The next step is to analyze the content of EULAs. For each EULA in the sample, I collect
data on 25 common standard terms which allocate rights and risks between buyers and sellers.
These terms can be grouped into seven main categories. The first category, Acceptance of
License, includes terms designed to alert the buyer of her options should she find the license or
product disagreeable. The second category, Scope of License, contains terms restricting the
buyer’s use of the software. The third, Transfer of License, includes terms that limit the buyer’s
ability to sell or transfer the software.
The fourth category is Warranties and Warranty Disclaimers and is comprised of terms
that delineate the degree and type of warranty protection offered to buyers. The fifth category,
Limitations of Liabilities, contains terms specifying the extent of the seller’s liabilities for
different types of buyer loss arising out of use of the software. It also covers remedies, if any, for
such losses. The sixth category, Maintenance and Support, takes into account whether the base
price of the software includes these services. The last category, Conflict Resolution, includes
terms that restrict a buyer’s choices regarding her decision of where to sue (forum selection
clauses), whether to have a jury trial (arbitration clauses), and how legal fees are to be allocated.
How to measure the “bias” of a given term? I shall briefly summarize my approach and
then proceed to details. Unlike sellers of many other mass-market products, software companies
generally license rather than sell their products.33 Other than the additional post-transfer
restrictions sellers can impose to buyers through licensing instead of selling, the legal
implications regarding enforcement of the EULA and its terms are the same as those regarding
regular standard form contracts. Numerous courts have held that the sale (or licensing) of
33 Originally, copyright and patent protection did not extend to software products. To prevent unauthorized
copying, software publishers relied on contract law and claimed their products as trade secrets. To be able to maintain the trade secret status, sellers then licensed the software with contractual provisions prohibiting buyers from copying or disclosing the source code of the software and requiring them to keep the information confidential. Today, sellers benefit from copyright and patent protection of their software. However, sellers still license software because this allows them to retain control over the software after they have delivered it to buyers. By restricting buyers’ rights to transfer or resell the software, sellers can avoid copyright’s “first sale” doctrine and price discriminate more effectively, among other advantages. See Mark A. Lemley, Intellectual Property and Shrinkwrap Licenses, 68 S. Cal. L. Rev. 1239 (1995).
14
software should be interpreted as the sale of a good within the meaning of the UCC.34
Consequently, when faced with a dispute over the validity of a software EULA or a particular
term contained therein, courts have relied on Article 2 of the UCC to determine its
enforceability. Thus, the “bias” of a given term can be measured against the benchmark provided
by the default rules of Article 2.
For each of the 25 terms that I track, I assign a negative one point score if the stated term
is more pro-seller than the default rules of Article 2, a positive one point score if it is more pro-
buyer relative to those rules, and a zero score if the contract is silent in regards to the specified
term or if the specified term matches the default rule. For example, a provision occasionally
contained in EULAs entitles buyers to receive software updates and upgrades for a specified
period after purchase. Since there is no default rule in Article 2 mandating such an entitlement
and because, other things being equal, a buyer would prefer a EULA that entitles her to receive
updates to one without such an entitlement, I give the presence of this term a positive one point
score in the overall index.
Once all 25 terms are scored in this manner, I simply add the scores to form an “overall
bias index.” This provides a rough overall measure of the EULA’s net buyer friendliness relative
to the Article 2 defaults. Of course, since each of the 25 terms is given the same weight by this
process, an assumption here of the EULA Bias Index is that all terms matter equally to buyers.
This might not be realistic. Buyers of certain products might not care about whether they are
allowed to modify the software but be more concerned about warranty protection; on the other
hand, it is possible to imagine certain buyers having the opposite preference. While for simplicity
I focus on the overall index, I also construct and analyze subindexes for each of the seven groups
of terms that I follow.
Table 3 lists all terms and describes how each is scored in detail. In deciding on which
terms to follow, I rely primarily on the sample EULA and associated discussion in the
practitioner-oriented manual by Overly and Kalyvas and the textbook on electronic commerce by
Mann and Winn.35 Although as complete as is practical, the terms that I follow do not encompass
34 See e.g. Advent Sys. Ltd. v. Unisys Corp., 925 F.2d 670, 676 (3d Cir. 1991); Step-Saver Data Systems v.
Wyse, 939 F.2d 91 (3d Cir. 1991); Downriver Internists v. Harris Corp., 929 F.2d 1147, 1150 (6th Cir. 1991). 35 Michael Overly and James P. Kalyvas, Software Agreements Line By Line, Aspatore, Inc., (2004);
Ronald J. Mann and Jane K. Winn, Electronic Commerce, Aspen Publishers (2002). There are also a number of specialized manuals and references created by IT and software lawyers describing typical software license agreements and their most common provisions.
15
every term found in license agreements.36 The typical EULA in the sample is about 1,500 words
long and some are several times that length, hence it is impossible to summarize the entire
content of all EULAs into a few numbers. The bias index should be viewed as a rough but
imperfect measure of all the terms in a software license agreement. I now introduce the seven
broad categories of terms.
B.1. Acceptance of License
Almost all sellers include a notice in the EULA alerting the buyer that she should not
install the software if she does not agree to its terms. However, a smaller fraction also inform
buyers that they can return the product for a full refund within a specified period of time. While
many sellers post their return policy somewhere on their website, sellers that provide this
additional notice in the license itself receive plus one point.
B.2. Scope of License
The next terms I consider involve the breadth of the definition of the licensed software.
Sellers often include future updates with the sale of their product for a limited period. Sellers
prefer to define their product as narrowly as possible in order to charge a fee for subsequent
enhancements or additions, while buyers prefer that these be included in the license.37 If the
EULA is silent as to the definition of the product being licensed, the term gets a neutral score,
but if it explicitly includes updates, enhancements, or updates, it receives plus one point.
The other terms in this category pertain to restrictions on use. Sellers may allow use of
the software solely “for internal business purposes” or “for non-commercial purposes.”
Restrictions on commercial use, such as the one at issue in ProCD, allow sellers to price
36 The terms that I follow are those that generally govern buyers’ “normal use” of the software. For
example, the index does not include EULA terms prohibiting buyers from using the software with the purpose of creating competing products (although much of this restriction overlaps with limitations on alterations or modifications). The index excludes terms dictating sellers’ use of buyers’ personal information, or provisions dealing with buyers’ privacy. Also excluded are terms describing copyright or patent rights, as they belong to the realm of federal intellectual property law. Some terms included in the index arguably conflict with some aspects of federal intellectual property law. For example, a term fully restricting product transfer may conflict with the “first sale” doctrine of the Copyright Act (17 U.S.C. § 109). Numerous courts have held that since software is licensed and not sold, the “first sale” doctrine does not apply to software transactions. See, e.g., Adobe Sys. v. Stargate Software, Inc., 216 F. Supp. 2d 1051 (2002) (holding that because the clickwrap license made software purchaser a licensee and not a copy owner, the first sale doctrine did not apply). This view, however, is not unanimous. Other courts have held that in mass-market transactions, the software is deemed to be sold and not licensed. Softman Prods. Co. v. Adobe Sys. Inc., 171 F. Supp. 2d 1075 (2001) (holding that software purchaser was copy owner, despite purported license). See also Mark A. Lemley, Beyond Preemption: The Law and Policy of Intellectual Property Licensing, 87 Cal. L. Rev. 111 (1999). Finally, I exclude terms that would be deemed unenforceable by any court, such as disclaimers of good faith.
37 See Overly & Kalyvas, supra note 35, at 30.
16
discriminate. Other things equal, buyers prefer a broader scope of permissible uses.38 I score a
negative point for the presence of any of such restriction. Second, I note whether sellers prohibit
buyers from altering or modifying the software. Sellers include these limitations to prevent
buyers from discovering the source code of the software, thus protecting their intellectual
property. Restrictions on modifications also halt buyers’ ability to make improvements to the
software, potentially stifling competition.39 Besides these motivations, buyers may simply wish
to have the ability to modify the software in order to customize it to better suit their particular
needs.40 I score a negative point for the presence of this term. I also note whether sellers prohibit
the creation of derivative works of the software. The sellers’ rationale for this restriction is
similar to the one banning modifications, and so is the rationale for the buyers’ preference that
this restriction is not present. Again, I score a negative point for the presence of this restriction.
B.3. Transfer of License
Some sellers impose restrictions on transfer of the product by prohibiting assignment,
leasing, renting, distributing, selling, or transferring. Sellers may include these restrictions to
price discriminate or to prevent selling the software to a particular class of purchasers (such as a
competitor).41 Assuming no difference in price, buyers prefer software to be freely transferable. I
give a negative point if the seller adds restrictions on transfer. I also note if sellers allow buyers
to make a one-time transfer of the software, as long as the subsequent holder agrees to be bound
by the terms of the license.42 I give a negative point if the seller does not allow a one-time
transfer exception.
B.4. Warranties and Disclaimers of Warranties
The first term in this sub-category notes whether sellers make any express, unrestricted
warranties. For example, a handful of vendors warrant that the software will work according to
the specifications included in the product’s manual, and do not limit the remedies available to
38 Id at 31. 39 The purpose for this kind of provision is usually to simplify the vendor-furnished maintenance services.
By prohibiting the buyer from altering the code of the software, the vendor is solely maintaining its own code. For a discussion of how restrictions on use affect competition, see Lemley, supra note 33, and Mark A. Lemley, Brief Amicus Curiae of American Committee for Interoperable Systems in ProCD v. Zeidenberg (7th Cir. 1996).
40 See Overly & Kalyvas , supra note 35, at 33. 41 See Mark A. Lemley, Peter S. Menell, Robert P. Merges, and Pamela Samuelson, Software and Internet
Law, Aspen Publishers (2003). 42 Under Section 117 of the Copyright Act, there is a prohibition of renting or lending the software under
the “first sale” doctrine. This section also allows buyers to make archival copies, however. For a detailed account on the first sale doctrine in software licensing agreements, see id at 306.
17
buyers in case of breach of warranty. In the literature, warranties have often been interpreted as a
signal of quality (sellers of low quality products would find it too costly to grant broad
warranties) or as an insurance and repair contract.43 Other things equal, buyers prefer express
warranties of the type described above because they provide insurance against losses caused by
product failure. The UCC awards the buyers repair costs as well as consequential and incidental
damages resulting from the breach of warranty.44 I treat the presence of express warranties as
more pro-buyer than the default rule.
I next consider whether the seller provides two common limited warranties. First, the
seller may warrant that the software will perform according to document specifications for a
limited period exceeding 30 days, as opposed to the unrestricted period involved in the
unrestricted warranty. As noted, buyers benefit from this form of insurance, and the longer the
period it covers the better. The warranty is “limited” because the remedy for breach is generally
restricted to either repair or replacement of the software. I note whether this kind of limited
warranty is offered. I also note whether sellers warrant that the media on which the software is
delivered (e.g., CD-ROM) and the accompanying documents are free from defects in materials
and workmanship for a limited period exceeding 30 days. Again, sellers are insuring against any
defects in materials supplied and offering to repair or replace in case of warranty breach. I treat
the presence of each of these limited warranties as pro-buyer since, other things being equal,
buyers would prefer limited warranties to no warranties.45
Sometimes sellers restrict these limited warranties by limiting their benefits only to the
original purchaser of the software. By requiring privity, sellers can protect themselves against the
potential moral hazard that arises when buyers anticipate early transfer of the software.
Consequently, this restriction affects buyers’ ability to transfer the software, and I record its
presence as a pro-seller term.
43 See Priest, A Theory of the Consumer Product Warranty, supra note 6. 44 See UCC §§2-713(1),-714(3),-715(1). 45 Note that these limited warranties might overlap to some extent with the coverage granted under implied
warranty of merchantability. A breach of this implied warranty entitles the buyer to recover consequential and incidental damages, which are much broader than the remedies provided by limited warranties. Id. Still, I consider a contract with limited warranties more pro-buyer than a contract without them as they arm the buyer with additional causes of action. More importantly, a buyer might find it easier to establish breach of a limited warranty than to argue that the product did not pass without objection in the trade. Also, as a practical matter, Table 4 shows that 90% of firms disclaim implied warranties. In this context, limited warranties are better than no warranties at all.
18
In general, sellers disclaim the implied warranties of merchantability and fitness for
particular purpose. The implied warranty of merchantability ensures that the good is “fit for the
ordinary purposes for which such goods are used.”46 The warranty for particular purpose is more
specific, as it only arises if the seller had reason to know the buyer’s specific intended use of the
software. As a cost-saving strategy, sellers may prefer to disclaim these implied warranties and
offer limited warranties instead. Buyers, on the other hand, prefer broader warranties and broader
remedies, all else equal. The presence of disclaimers are thus recorded as pro-seller.
For disclaimers of implied warranties to be valid, they must be “conspicuous” as defined
by the UCC.47 The disclaimer must be in capital letters, in bold, or under a title heading. To
disclaim the implied warranty of merchantability, the disclaimer must include the term
“merchantability.”48 Alternatively, sellers may disclaim all implied warranties by including “as
is” language.49 I note whether sellers comply with this aspect of the UCC requirement and treat
as pro-seller any disclaimer that is not conspicuous as defined by Section 2-316.50
The last disclaimer that I note is of particular importance to software products (or the
licensing of any intellectual property). Section 2-312 of the UCC has an implied warranty of
title. In intellectual property transactions, the implied warranty of title forces the seller to
guarantee that the software will not infringe on third parties’ intellectual property rights. Not
surprisingly, many software sellers disclaim this warranty, and I interpret such disclaimers as
pro-seller.
An important consumer protection law is The Magnuson-Moss Warranty Act, a
disclosure law enacted in 1975 to regulate the form and content of consumer product
warranties.51 The act does not mandate sellers to include warranties but rather requires that
46 UCC §2-314. 47 UCC §2-316. 48 UCC §2-316(2). 49 UCC §2-316(3)(a). 50 I track whether disclaimers in EULAs are conspicuous as a general proxy for “text-friendliness.” Many
courts have enforced disclaimers that were neither in capital nor in bold letters, or under a title heading. In Valley Paving, Inc. v. Dexter & Chaney, 42 U.C.C. Rep. Serv. 2d (Callaghan) 433 (2000) at 6, the court stated: “Here, the disclaimer of warranties is not all in capital letters or in contrasting typeface or color. But as the district court noted, the disclaimer "was not hidden in an obscure part of a long, complicated document […] Therefore, we conclude that the district court did not err in concluding that the disclaimer of warranties is conspicuous and properly disclaims the implied warranties of merchantability and fitness for a particular purpose.” For another case citing location of a disclaimer as sufficient to determine conspicuousness, see Agristor Leasing v. Guggisberg, 617 F. Supp. 902, 909 (1985).
51 15 U.S.C. §§2301-2312 (1976).
19
sellers, if they choose to provide warranties, draft them in clear language.52 In addition, the act
requires that warranties be available for inspection prior to purchase53 and that if an express
(whether full or limited) warranty is given, that implied warranties not be disclaimed for the
period of such warranty. I chose not to tabulate whether the warranties in my EULA sample are
Magnuson-Moss compliant because it is still unsettled whether Magnuson-Moss applies to
software products. The act applies to “any tangible personal property which is distributed in
commerce and which is normally used for personal, family, or household purposes.”54 To this
date, and despite lengthy discussion, there is still no consensus on whether software should be
interpreted as “tangible personal” property.55
B.5. Limitations on Liability
Sellers often dedicate a considerable portion of the EULA to disclaiming and limiting
remedies and liabilities. The first term here that I record is whether the EULA assigns the risk of
loss exclusively to the buyer, the seller, or allocates the risk between both. If the risk of loss is
assigned to the buyer under all circumstances, even for losses caused by factors under the seller’s
control, the clause is clearly pro-seller. I also note whether the contract allocates risks arising out
of the performance of the software. An example is system slowdown created by the use of the
product. In the software environment, performance risk is defined as the degree to which a
system or component accomplishes its designated functions within given constraints, such as
speed, accuracy, or memory usage.56 Hence, language such as “licensee assumes full
responsibility for the choice of product and its functions” is counted as pro-seller.
Damages associated with breach of a software license agreement might include lost
revenues, data recovery fees, or loss of reputation, among others. Sellers often limit damages to
some multiple of the product price. I consider a limit on damages that is less than or equal to the
product price to be a relatively pro-seller term, and a greater limitation (or no limitation) to be
52 15 U.S.C. §2302(a). 53 15 U.S.C. §2302(b)(1)(a). 54 15 U.S.C. §2301. 55 See e.g. Symposium, Warranty Protection for High-Tech Products and Services, Federal Trade
Commission (2000), at http://www.ftc.gov/bcp/workshops/warranty/. 56 See Connie U. Smith & Murray Woodside, Performance Validation in Early Stages of Software
Development, in Gelenbe, Erol, System Performance Evaluation: Methodologies and Applications, CRC Press (2000).
20
neutral. Except for restricting the recovery of unforeseeable or speculative losses, the UCC
contains no other limitations on damages.57
Sellers also limit damages by disclaiming consequential, special, and incidental damages.
As just mentioned, the UCC allows a buyer to recover these damages under breach of contract
only if they are foreseeable and not speculative. I interpret such disclaimers as pro-seller, since
they limit buyers’ ability to recover damages. I also note whether sellers disclaim liability for
foreseeable losses of licensee, which further limits buyers’ ability to recover.
On occasion, sellers also restrict the theories of liability on which buyers can base causes
of action against them. If the product defect causes physical injury or property damage, for
example, a buyer can bring an action in tort against the seller. If a seller waives all types of
damages under all theories of liability, whether contract, tort, or strict liability, the buyer will
have greater difficulty challenging this disclaimer. I thus interpret such waivers as pro-seller.
The last of the limitations on liability that I study are indemnification clauses.
Indemnification for intellectual property infringement is the most common licensor-provided
indemnity.58 On the other hand, the licensee may be required to indemnify the licensor against
claims arising from the licensee’s use of the software in a manner outside the uses permitted by
the EULA. I interpret the indemnification clause as neutral if there is no mention of indemnities
or if there is an explicit two-way form of indemnification. I record one-way indemnification
clauses as pro-seller or pro-buyer, as appropriate.
B.6. Maintenance and Support
Maintenance and support (M&S) provisions vary widely across firms and products. I
simply record whether M&S for 31 days or more is included in the product’s base price. Many
companies offer high levels of support for an additional fee, but my approach allows for a simple
comparison of standard terms across companies. A more practical reason to limit the detail I
record on M&S provisions is that these terms are not always discussed on the EULA itself but
rather on a separate agreement or on the website; our purpose here is to describe EULA terms.
B.7. Conflict Resolution
The last category of EULA terms involve dispute resolution. Choice of forum and law
clauses may allow sellers to save costs by increasing predictability of results. Alternatively,
57 See UCC Section 2-715(1) and (2)(a). 58 See Overly & Kalyvas, supra note 35, at 58.
21
sellers use them strategically to direct disputes toward business-friendly jurisdictions. Similarly,
by mandating arbitration, sellers may prefer the cost-saving aspect and expedience, but also wish
to eliminate the possibility of class actions. I interpret the specification of a particular forum or a
mandatory arbitration provision as generically less buyer friendly than the default, under which
buyers usually have more options about where to bring claims.59
I do not interpret choice of law provisions per se as being less buyer-friendly. Sellers
commonly employ them to economize on legal fees and to increase certainty in outcomes.60
However, I do note when the choice of law differs from choice of forum since this might indicate
a strategic gerrymandering of the dispute resolution process.61 I thus interpret such discrepancies
as pro-seller.
Finally, I record whether the EULA specifies who pays the sellers’ attorney fees.
Contracts in which the buyer is asked to pay for the seller’s attorney’s fees regardless of the
outcome of the dispute are obviously pro-seller.
C. EULA Bias Summary Statistics
Figure 1 shows the distribution of the overall bias index. As constructed, the maximum
attainable overall bias index score is 7, a number associated with an extremely buyer-friendly
EULA relative to the default rules of the UCC. The minimum attainable score is -20, which
would reflect a contract disclaiming all remedies, greatly restricting buyers’ use of the software,
and so on.62 The last row of Table 4 shows that in our sample, the average of the overall bias
index is -5.64, and (unreported) the minimum score in the sample is -15 and the maximum is 2.
The distribution has a nice bell shape. On average, EULAs are substantially more pro-seller than
59 See e.g. Lee Goldman, My Way and the Highway: The Law and Economics of Choice of Forum Clauses
in Consumer Form Contracts, 86 Nw. U. L. Rev. 700 (1992) (arguing that it would be efficient to assign the risk of litigating in distant forums to sellers, given that buyers rarely read boilerplate and sellers are better able to handle this risk).
60 See letter from Larry E. Ribstein and Bruce H. Kobayashi to the Secretary of the Federal Trade Commission at http://www.ftc.gov/bcp/worshops/warranty/comments/ribstein.htm (noting that, “with respect to the need to protect consumers from unfair contract terms that choose the applicable state law is not subject to the same potential damages as enforcing other contractual terms. The effect of enforcing a choice-of-law clause is to apply a system of regulation imposed by a state legislature’s and courts – not simply to enforce the rules preferred by one of the contracting parties.”) See also Larry E. Ribstein & Bruce H. Kobayashi, Uniformity, Choice of Law and Software Sales, 8 Geo. Mason L. Rev. 261 (1999).
61 See John A. Burke, Contract As Commodity: A Nonfiction Approach, 24 Seton Hall Legisl. J. 285 (2000).
62 One reason the range of possible scores is not centered at zero is that the default rules of the UCC tend to benefit buyers.
22
the default rules of the UCC, but there is a very wide range, with a handful actually being
slightly pro-buyer.
Table 4 shows the means of each term and the correlations between the scores of
different terms. The reader can scan for the means and correlations she finds most interesting. As
an overall view, of the (25*24)/2=300 total pairwise correlations among the 25 EULA terms, 210
are positive, and 91 of those are significant at the 5% level. The remaining 88 correlations are
negative, of which 32 are significant. Since most correlations are positive, this means that on
average, contracts that are pro-seller on one term tend to be pro-seller on other terms. Likewise,
contracts that are pro-buyer on one term tend to be more pro-buyer on other terms. These results
are intuitive, and provide some confidence that we are defining “bias” correctly.
Clustering is much stronger, however, among the seven bias subindexes. Out of the
(7*6)/2=21 pairwise correlations for the subindexes listed in Table 5, 19 are positive, and 10 of
these are significant at 5%. Two coefficients are zero. As expected, each of the subindexes is
significantly positively correlated with the overall bias index, in part because of their mechanical
relationship.
D. Price, EULA Bias, and Company and Product Characteristics
Table 6 starts the analysis by examining the extent to which product and company
characteristics alone can explain EULA bias and product price. The regressions in the left side of
the table use the natural log of price as the dependent variable. The samples in the first three
specifications include all EULAs, i.e. those of both consumer and business products, and
examine different sets of controls. As expected based on the summary statistics, the prices of
consumer products are significantly lower than business products. Also, multi-user and
developer licenses cost more than single-user ones, other natural results. Based on the first
specification, developer licenses are priced 41% higher than single-user products and multi-user
licenses are priced 66% higher.
Firm size, as proxied by the natural log of firm revenue, is significantly positively related
to product price, although the magnitude of the effect is modest. Intuitively, larger firms tend to
produce more sophisticated products, which cost more. On the other hand, controlling for firm
size, publicly-traded firms sell somewhat lower-priced products. Finally, firm age is not an
important determinant of price. Most of these conclusions are similar whether or not we include
“market” dummy variables (using the 146 Amazon.com categories, e.g. “Business & Office >
23
Business Accounting > Accounting”) or just “sector” dummy variables (using the 12 broad
Amazon.com categories, e.g. “Business & Office”). The inclusion of these dummy variables is
important because they allow me to control for many product functions that I am not able
measure directly, and that may influence price.
One striking finding of the first few specifications is the lack of evidence that EULA
terms, viewed as a whole, are priced in an important way. No coefficient on Overall Bias Index
is significantly positive, which is what we would have expected if consumers of certain types of
products were willing to pay more to get relatively pro-buyer terms. In other words, this result
casts great doubt on the view that consumers make careful tradeoffs between price and secondary
terms.63
Consumer and business buyers might differ in their ability to bear certain risks, or prefer
different levels of warranties, and these differences might be reflected in the prices and terms of
the products they buy.64 To account for this possibility, I split the sample by product/user type.
There are a few differences that emerge. The firm size coefficient is positively significant and
public firm status is negatively significant only for business products. Multi-user licenses are
more expensive for both business and consumer products, while somewhat surprisingly the effect
of developer licenses is strongest among consumer products (although there are very few such
products in the sample, which is why the standard error is high). Finally, the last specification in
the left side of the table uses a sample of 49 companies that offer both business and consumer-
oriented versions of a product. By controlling perfectly for company characteristics, we can
better see the effect of product attributes. But the results are similar to those from the full sample.
The right side of Table 6 explores the relationship between product and firm
characteristics and overall EULA Bias. What explains the wide variation in EULA terms? The
first three specifications of this second group of regressions again include all the products in the
sample. Consistent with the results using price as a dependent variable, there is no relationship
between a product’s price and the overall bias of its standard terms. There is some weak
evidence that developer licenses come with generally more pro-buyer terms; in the second
63 In results omitted to save space, I regress price on the seven subindexes, to examine whether any groups of terms are assiciated with price effects. See James J. White, Contracting Under Amended 2-207, Wis. L. Rev. 723 (2004) (suggesting that “for a nickel or a dime, almost all of us would give up our right to resell software and would agree to arbitrate” at 742). No obvious pattern emerges from these regressions. One class of terms, warranties, is associated with significantly higher prices, while another class of terms, limitations on transfer, is associated with a significantly lower prices.
64 Much of the concern about standard form contracts has focused on consumer products.
24
specification, they are associated with 1.22-point more pro-buyer EULAs. This makes sense,
since developer licenses generally allow buyers to create and distribute derivative works, and
thus are less restricted.
Company characteristics are more significant determinants of EULA bias. Larger firms
and younger firms both have more pro-seller EULAs. Possible interpretations for the size effect
include the fact that larger firms are more likely to be advised by counsel who are better able to
shield the firms from potential liabilities. Also, larger companies simply have more to lose, and
hence are likely to prefer more protection. The intuition for the age effect may involve seller
reputation. Older companies may have created a more homogenous customer base over time and
thus are better able to provide insurance for certain types of losses, as opposed to newer
companies with diverse customer bases. Alternatively, younger software companies may be
more sophisticated about the licensing aspects of software than the industry pioneers, who may
have designed their EULAs long ago and haven’t given much thought to them since. Or, younger
companies whose survival depends on the success of a small, unestablished product line may
prefer stricter terms out of risk-aversion. However, these are speculations.
These impressions from the full sample generally hold true for sub-samples of business
and consumer products as well. The last specification in the table, which again includes 49
companies which sell both business and consumer versions of a product, allows me to address
the frequent concern by commentators and consumer advocates that sellers tend to exploit
unsophisticated customers, offering them more restrictive contracts than they do to their savvy
business buyers.65 However, the coefficient on Consumer product in these specifications, while
negative, is very small and insignificant. The results show that sellers who could in principle
easily discriminate among buyer types do not, in fact, offer significantly more one-sided terms to
their less sophisticated consumers.
IV. DO COMPETITIVE CONDITIONS AFFECT LICENSE TERMS?
A. Measuring Competitive Conditions
We now turn from a descriptive analysis of EULA terms to an investigation of the role of
competition. Standard measures of competitive conditions include market share (at the company
65 More recently, concerns have arisen over the practice in which companies selling over the Internet ask
buyers to select either “home user” or “business” categories before being allowed to proceed with their purchases. The potential to discriminate is obvious.
25
level) and concentration ratios and the Herfindahl-Hirschman index (HHI) (at the market level).
Concentration ratios add up the total market share of the largest N firms in a given market. The
HHI is another measure of market concentration that is commonly used by the Justice
Department and the Federal Trade Commission to evaluate horizontal mergers and is preferred
by economists. It is obtained by adding the squares of the individual market shares of the firms
that compete in a given market. The reason why economists prefer HHI is that it accounts for
disparities in firm size in a given market. For example, a market in which one firm has 80%
market share would have the same four-firm concentration ratio as a market in which the largest
four firms each have 20% market share, but the former market would have a much higher HHI.
All of the standard measures are based on company-level market share data. Since market
share data is not publicly reported by companies in any consistent format, one needs to be
somewhat creative to estimate competition measures. I use two approaches in this paper: a
recently-developed methodology that is based on the “sales ranks” of a company’s products on
Amazon.com and an approach that simply gathers recent market share estimates published by
industry research companies.
A.1. Amazon.com Data
Market share is defined as the percentage of sales of a particular firm in a given software
market. One thus needs estimates of relative sales. Amazon.com is the largest online marketplace
in general and the largest seller of online software, listing nearly 20,000 products from 500
software companies. While Amazon.com does not make available the precise sales of each
product, it does provide a sales “rank,” whereby products with higher rates of sales achieve sales
ranks closer to rank #1 and lower rates of sales are closer to rank #20,000.
A number of papers have shown how Amazon.com product sales ranks can be translated
into estimates of product sales.66 This is possible because the relationship has been found to obey
a “power law,” meaning that the relationship between observable rank and the unobservable
quantity sold for product p is closely approximated by:
Salesp = a*Rankpb
The exponent b is negative, so a lower sales rank (closer to #1) is associated with higher sales.
The constant term a account for the fact that Sales is a rate. Thus, a is proportional to the length
66 See Judith Chevalier & Austan Goolsbee, Measuring Prices and Price Competition Online: Amazon.com
and BarnesandNoble.com, 1 Quant. Mktg. and Econ. 203 (2003); Erik Brynjolfsson, Yu Hu, & Michael D. Smith, Consumer Surplus in the Digital Economy, 49 Mgmt. Sci. 11 (2003).
26
of time over which Sales is measured. Given this relationship, and given the Amazon.com sales
rank data, a company’s market share can be estimated with knowledge of the exponent b, since a
drops out of the calculation.
Here is a simple example to illustrate the idea. Suppose that a given company c has two
products in a given software market m, e.g. antivirus, and those products’ sales ranks are 12 and
427. Suppose that other antivirus products sold by other competitors are ranked 4, 50, and
10,000. Then the company’s market share in the antivirus market can be estimated as:
MScm = (a12b + a427b) / (a4b + a12b + a50b + a427b + a10,000b) which simplifies to
MScm = (12b + 427b) / (4b + 12b + 50b + 427b + 10,000b).
The same approach can be used regardless of the number of competitors or the number of
products per competitor.
To implement this, I need b. Various approaches have been taken to estimate b. Judith
Chevalier and Austan Goolsbee use a simple experiment in which they obtain from a publisher
information about a particular book’s weekly sales. They then purchase several copies of that
book on Amazon.com over a 10-minute period and track the change in the sales ranking of the
book. This led to an estimate of b for books of -0.855. Erik Brynjolfsson, Yu Hu, and Michael D.
Smith estimate b by mapping weekly sales data for 321 books obtained from a publisher to the
Amazon.com sales ranking of each book. They estimate b as –0.871 for books. They also
reproduce the Chevalier and Goolsbee experiment and obtain an estimate of -0.916. Thus, for
books, estimates of b are quite stable. However, since there is no theoretical reason to expect the
b for books to be the same as the b for software, which is ranked separately, I use the estimate of
b from Anindya Ghose and Arun Sundarajan.67 After carefully tracking Amazon.com sales rank
fluctuations for hundreds of software products over a two-week period, they estimate b for
Amazon.com software as -0.828. I use this estimate.
I then gather Amazon.com sales rank data by hand for all the significant sellers in all of
the 114 Amazon.com software markets represented in my sample.68 I produce market share
estimates for all the top sellers in each market and then construct concentration ratios and HHI
67 Anindya Ghose & Arun Sundarajan, Pricing Security Software: Theory and Evidence (NYU Stern School of Business mimeo, 2005).
68 Because some markets encompass hundreds of software products, and because lower ranked products have a rapidly diminishing impact on market share estimates, I only record products’ sales ranks which are within a factor of 100 of the product with the highest sales ranking in that market.
27
for each market. In addition, 189 of the EULAs in my sample belong to companies who sell that
product, or products in the same market, through Amazon.com. Hence for these EULAs I can
estimate the market shares for the relevant company and the relevant market, providing a
different notion of competitive conditions than HHI.
A.2. Other Competition Data
As an additional source of competitive conditions data, I was able to obtain market share
estimates for the major players in several Amazon.com markets from recent issues of the Market
Share Reporter. This is an annual compilation of market share statistics that are reported in
business journals, newspapers, and brokerage house reports. I collected data on a few more
Amazon.com markets from the U.S. Business Reporter, an online subscription service that also
summarizes and republishes various market share estimates. From these data, I construct HHI
Other, i.e. a market-level HHI statistic based on these assorted market share data rather than the
Amazon data. The overlap between these market share estimates and the company-market
combinations in my EULA sample was not large enough to construct an alternative measure of
market share, however.
B. Competitive Conditions Summary Statistics
Table 7 shows the summary statistics of the competition data. The top panel shows that
the average market share (of those company-markets whose EULAs are represented in my
sample) is 12%, but the median is 4% and the range is from less than 1% to 100%. That is, a
handful of markets represented in my sample are completely dominated by a company whose
EULA I have collected for a product in that market. Other companies selling products in that
market have Amazon sales ranks that is larger (sales are smaller) by a factor of at least 100.
The second panel shows the market-level competition data. As mentioned above, HHI is
the sum of squared estimated market shares for all companies who sell in that Amazon.com
market. This variable ranges from 0.07 for some highly unconcentrated markets to 1.00 for a few
markets with completely dominant sellers. Again there is a wide spread, with the mean HHI of
0.34, a median of 0.28, and a standard deviation of 0.22.
To put these figures into perspective, FTC merger guidelines suggest that an
“unconcentrated” industry is one in which HHI is less than 0.1.69 Markets with an HHI between
69 See United States Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines
(1997), at http://www.usdoj.gov/atr/public/guidelines/hmg.htm.
28
0.1 and 0.18 are defined as “moderately concentrated,” while markets with an HHI above 0.18
are deemed “concentrated.” Using these definitions, I construct dummy variables for each
category based on the Amazon.com HHI estimates. The table shows that about 5% of the
observations in the sample fall into the Unconcentrated category, 21% are Moderately
Concentrated, and 74% are Concentrated. (Fortunately, the sample is sufficiently large that even
5% implies a meaningful number of observations in the Unconcentrated category.) Concentration
ratios give similar impressions and are omitted.
The HHI statistics computed from other (non-Amazon.com) are similar to those
computed from Amazon.com data. Other HHI is available for the markets of 70 EULAs are
represented in our sample. Although this sample is too small to draw any strong conclusions
from the regressions that will employ it, it may provide useful corroboration of the general
direction of the results obtained using the Amazon.com data.
C. Price, EULA Bias, and Competitive Conditions
We are now ready to investigate whether competition improves the quality of standard
form contracts. While theorists disagree on this point, there is no disagreement that a monopolist
will exercise market power by charging supra-competitive price. We therefore begin by testing
this prediction. This is an important step because it offers a way to independently confirm that
my competition measures are indeed capturing market power. It is possible that the proxies are
too noisy to be meaningful, or that the markets I examine are “contestable” and therefore market
shares and concentration are not useful.70 Thus it is useful to confirm that my competition
proxies are indeed related to price.
Table 8 shows specifications using product price as the dependent variable and
competition measures as explanatory variables. All specifications include the company and
product controls examined earlier, but the coefficients are not reported to save space. The first
group of models use HHI, the second group uses the FTC merger guidelines dummies, and the
third group uses HHI Other. Since these competition variables vary at the market level, it is not
possible to also include market dummies. The finest controls I can add are sector dummies. Thus,
70 The theory of contestable markets states that even firms in highly concentrated industries will behave
competitively as long as there are no (or very low) entry and exit costs. See William J. Baumol, John C. Panzar & Robert D. Willig, Contestable Markets and the Theory of Industry Structure, San Diego: Harcourt Brace Jovanovich (1982); William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Ec. Rev. 1 (1982).
29
these regressions are basically asking whether, within the “Business & Office” sector, for
example, software in markets that are relatively less competitive also have higher prices.71
The results largely confirm this prediction. The first specification uses all the EULAs in
the sample and shows that less competition is indeed associated with higher prices. Specifically,
it suggests that a market that goes from perfectly uncompetitive (HHI of 0) to perfectly
monopolistic (HHI of 1) will see software prices rise by an average of 34%, all else equal. In my
subjective opinion, this magnitude seems sensible. The regressions also suggest that the effect of
HHI is strongest in consumer-oriented products.72
The second group of models, which include dummies for Moderately Concentrated and
Concentrated (according to FTC guidelines) allow us to look for a nonlinear effect of HHI. The
regressions imply that the main effect on price comes as the market goes from Unconcentrated
(the omitted category) to Moderately Concentrated. This is associated with a 44% increase in
price across all products. The effect of Moderatly Concentrated is close to the same as the effect
of Concentrated, therefore there is not much additional increase in price as a market goes from
the former (HHI at least .10 but below .18) to the latter (HHI above .18).
The third group of models use HHI Other as the market concentration measure. Although
the samples are much smaller here, the effect of HHI Other is strongly positive and even larger
than the effect of HHI. While I do not place much emphasis on these results because of the low
number of observations, it is reassuring that they are at least consistent with the other results.
The fourth group of models use market share as the competition measure. These
regressions, which include market dummies, are asking whether, within a given market, the firms
that have high market share charge higher prices than those with lower market share. Recall that
Kessler argued that firms with larger market share would offer more exploitative terms; here we
look at price. This proposition also receives some support, mainly from the consumer-oriented
products sample. Those estimates imply a very large effect of market share on price. The effect
from the business-oriented products subsample is positive but insignificant. One possible
71 When estimating effects of aggregate variables, such as market level data, on individual units, the error
terms for products within each market may not be independent. See Brent R. Moulton, An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on Micro Units, 72 Rev. Econ. & Stat. 334. Therefore for all regressions that I include a variable that varies only at the market level, I correct the standard errors using Stata’s “cluster” command.
72 In results omitted to save space, I also regressed ln price on various concentration ratios formed from Amazon.com data and found a similar pattern of results.
30
interpretation is that “business” buyers are more prone to shop around within a market in order to
get a lower price.
As an aside, the fact that my concentration and market share data can be directly
connected to higher prices marks an important contribution of my analysis relative to that of
Priest.73 While Priest finds that various competition measures do not help to explain consumer
product warranty terms in various appliance markets, he does not verify that his measures
reliably explain price, either. Thus his analysis, while careful and pathbreaking, remains subject
to the critique that (i) his competition measures are too noisy to be useful or (ii) the appliance
markets he examines are essentially contestable, and hence data on concentration and market
share will never be useful. In contrast, the results of Table 8 allow us to move forward with
reassurance that the competition measures are valid.
Table 9 turns to the relationship between competitive conditions and overall EULA bias.
The specifications parallel those in Table 8, the only difference is that now the dependent
variable is the overall bias index. Contrary to those who view the presence of less competitive
conditions in a market as likely to lead to more restrictive boilerplate, the results demonstrate
that there is essentially no important relationship between competitiveness and the overall quality
of standard terms.
The first three specifications show that higher concentration, in the form of higher HHI,
does not lead to more pro-seller terms. In fact, the only significant relationship between contract
bias and concentration goes in the “wrong” direction: for consumer products, higher
concentration is associated with slightly more pro-buyer terms. Even this weak relationship
disappears, however, when I use the concentration categories suggested by the FTC in place of
the HHI variable.
The top of Figure 2 illustrates this conclusion graphically. The bars in the figure illustrate
the average bias of EULAs in markets with different concentration levels (unlike the regressions,
the effects of control variables are not included). As suggested by the regressions, EULAs from
Moderately Concentrated and Concentrated markets are only slightly more pro-seller overall
than EULAs drawn from Unconcentrated markets.
When I consider the effect of Other HHI on EULA terms, I find further evidence that
concentration does not lead to worse terms. In fact, when Other HHI is high, terms are
73 Priest, A Theory of the Consumer Product Warranty, supra note 6.
31
significantly more pro-buyer. While this result is based on relatively few data points and thus
may be spurious, it holds in both consumer and business software subsamples. But at the very
least, one can confidently conclude from these regressions and those preceding that that the
effect of market concentration centers on price, not standard terms.
Finally, the last specifications of the table show that there is no important negative
relationship between market share and contract terms, either. In particular, within a given
market, it is not the case that firms with high market shares offer significantly worse terms. In
consumer software, there is a negative but insignificant effect of market share; in business
software, there is a positive but insignificant effect. The bottom part of Figure 2 illustrates the
relationship between estimated market shares and overall bias. The figure suggests that firms
with 20-50% market share offer slightly worse terms than other firms, but there is no pattern as
we increase market share from <1% all the way to 10-20%. Further, the effect is small even
before including the effects of any market, company, or product control variables.
At this point, we can conclude that market power, however defined, does not lead to
meaningfully worse standard terms overall, although it does increase product price. As noted
earlier, a built-in assumption of the overall bias index is that all contract provisions matter
equally to buyers. However, buyers may care only about a particular set of terms.74 For instance,
a home user might not care about whether a software’s license agreement forbids that he create
derivative works, but he might care about being able to contact technical support if a problem
arose. Russell Korobkin posits that bounded rationality and other behavioral biases lead buyers
to focus only on “salient” terms such as, e.g., price and warranties.75
Table 10 explores this by repeating specifications (1) and (10) from Table 9 but using the
seven bias subindexes as the dependent variables. As can be seen from a quick scan of the table,
HHI is not significantly related to any of the seven subindexes. Market share shows results of
mixed sign. Within markets, firms with higher market shares are more likely to give extra notice
regarding the acceptance of the license and return instructions, however they tend to give slightly
worse terms on the dimensions of warranties and conflict resolution. However, these latter
results are not significant when we perform the same regressions on business and consumer
74 See e.g. Rakoff, supra note 22. 75 See Korobkin, supra note 7.
32
products in isolation, suggesting that they too are weak at best. In summary, even when we look
at individual categories of terms, the effect of competition is weak or nonexistent.
V. CONCLUSIONS AND IMPLICATIONS
Standard form contracts are pervasive, but while a large literature discusses their content
and enforceability at a theoretical level, there has been little systematic empirical investigation of
their content and economic determinants. This paper is a first attempt to systematically measure
differences across standard form terms in an important industry, software, and use this
methodology to explore hypotheses about the determinants of standard form agreements, in
particular regarding the role of competitive conditions.
The first main contribution of this paper is to provide a detailed description of a large
sample of standard form contracts, software end-user license agreements. I analyze a sample of
647 EULAs from various software markets and measured the variation of 25 common terms,
such as warranties and limitations on liability, to create an overall index of contract bias. An
immediate conclusion is that the vast majority of the contracts in our sample are more pro-seller
relative to the default rules of Article 2 of the UCC. While EULA terms vary greatly across
software markets, I find that larger and younger firms tend to have more pro-seller terms than
smaller and older companies. I find no evidence that firms offer worse terms with software
targeted to the general public versus software targeted to the large business user.
The second primary contribution is a careful analysis of competitive conditions on
standard form contract terms. The debate over whether competitive conditions are correlated
with contact quality dates back to Kessler, who argued that standard terms were an instrument
that companies with market power used to impose poor or restrictive terms on buyers. Variations
of this argument recur in the literature on standard terms, with many scholars suggesting that
competition is what ensures that standard terms reflect buyers’ preferences. Others have disputed
this conclusion. However, almost all of the debate thus far has been theoretical. I find that
competitive conditions are essentially uncorrelated with the quality of standard terms. More
competitive segments of the software industry offer similar standard form terms to highly
concentrated segments of the software industry. Similarly, within an industry, firms with larger
market share offer similar terms to minor players. These results cast doubt on the notion that
competitive conditions play an important role in the setting of standard terms.
33
It is worth repeating that an assessment of whether EULAs or standard terms in general
are efficient (or good or bad in general) require making a judgment about the quality of these
contracts in an absolute way. This paper does not attempt to do this. What this paper does
measure is the descriptive content of EULAs relative to the default rules of Article 2 of the UCC;
and the differences between EULAs in competitive and uncompetitive markets.
The results also offer several practical implications for judges, regulators, and
practitioners who need to determine whether a term may be procedurally unconscionable. By
illustrating the range of EULA terms in practice, the results can help determine whether a given
contract or a particular term is unusual relative to industry practice. Also, by demonstrating that
sellers with market power do not offer unusually biased terms, the results show that an important
aspect of the standard analysis involved in the determination of procedural unconscionability is
misguided. Market power does not suffice to ensure that standard terms are particularly pro-
seller. Finally, the results add some hard facts to the recent debates over the desirability of a
uniform body of law for on-line contracting.
34
BIBLIOGRAPHY
Baumol, William J. “Contestable Markets: An Uprising in the Theory of Industry Structure.”
American Economic Review 72 (1982): 1-15.
Baumol, William J., Panzar, John C., and Willig, Robert D. Contestable Markets and the Theory
of Industry Structure. San Diego: Harcourt Brace Jovanovich, 1982.
Bogert, George Gleason and Fink, E.E. “Business Practice Regarding Warranties in the Sale of
Goods.” Illinois Law Review 25 (1930): 400-17.
Braucher, Jean. “The Failed Promise of the UCITA Mass-Market Concept and its Lessons for
Policing of Standard Form Contracts.” Journal of Small and Emerging Business Law 77
(2003): 393- 423.
Brynjolfsson, Erik, Hu, Yu, and Michael D. Smith. “Consumer Surplus in the Digital Economy:
Estimating the Value of Increased Product Variety.” Management Science 49 (2003):
1580-96.
Burke, John A. “Contract As Commodity: A Nonfiction Approach.” Seton Hall Legislative
Journal 24 (2000): 285-325.
Chevalier, Judith and Goolsbee, Austan. “Measuring Prices and Price Competition Online:
Amazon.com and BarnesandNoble.com.” Quantitative Marketing and Economics 1
(2003): 203-22.
Cooter, Robert and Ulen, Thomas. Law and Economics. 3rd ed. Glenview, IL: Scott, Foresman &
Co, 2000.
Craswell, Richard. “Passing on the Costs of Legal Rules: Efficiency and Distribution in Buyer-
Seller Relationships.” Stanford Law Review 43 (1991): 361-98.
Cruz, R. Ted and Hinck, Jeffrey J. “Not My Brother’s Keeper: The Inability of an Informed
Minority to Correct for Imperfect Information.” Hastings Law Journal 47 (1995): 635-
76.
DeMarco, Daniel A., and Wick, Christopher B. Now UCITA, Now You Don’t: A Bankruptcy
Practitioner’s Observations on the Proposed Uniform Computer Information Transactions
Act, Am. Bankr. Inst. J. 34 (2004): 23-4.
Epstein, Richard. “In Defence of UCITA.” Financial Times. April 2, 2003.
35
Foreman, Stephen E. and Shea, Dennis G. “Publication of Information and Market Response:
The Case of Airline on Time Performance Reports.” Review of Industrial Organization
14 (1999): 147-162.
Gerner, Jennifer L. and Bryant, W. Keith. “Appliance Warranties as a Market Signal?” Journal
of Consumer Affairs 15 (1981): 75-86.
Gillette, Clayton. Letter in Support of UCITA to NCCUSL Standby Committee on UCITA.
http://www.nccusl.org/nccusl/ucita/ucita/GilletteEndorse.pdf. January 21, 2003.
Ghose, Anindya and Sundarajan, Arun. “Pricing Security Software: Theory and Evidence.”
Unpublished manuscript. New York: NYU Stern School of Business, 2005.
Goldberg, Victor P. “Institutional Change and the Quasi-Invisible Hand.” Journal of Law and
Economics 17. (1974): 461-85.
Goldman, Lee. “My Way and the Highway: The Law and Economics of Choice of Forum
Clauses in Consumer Form Contracts.” Northwestern University Law Review 86 (1992):
700-741.
Hoxby, Caroline M. “Does Competition Among Public Schools Benefit Students and
Taxpayers?” American Economic Review 90 (2000): 1209-38.
Kahan, Marcel, and Klausner, Michael. “Standardization and Innovation in Corporate
Contracting (or ‘The Economics of Boilerplate’).” Virginia Law Review 83 (1997): 713-
68.
Kaner, Cem and Pels, David. “Bad Software: A Consumer Protection Guide.” At:
http://www.badsoftware.com/oppose.htm.
Katz, Avery. “Your Terms of Mine? The Duty to Read the Fine Print in Contracts.” RAND
Journal of Economics 21 (1990): 518-33.
Kessler, Friedrich. “Contracts of Adhesion – Some Thoughts About Freedom of Contract.”
Columbia Law Review 43 (1943): 629-42.
Klick, Jonathan. “The Micro Foundations of Standard Form Contracts: Price Discrimination vs.
Behavioral Bias.” Florida State University Law Review 32 (2005): forthcoming.
Korobkin, Russell B. “Bounded Rationality, Standard Form Contracts, and Unconscionability.”
University of Chicago Law Review 70 (2003): 1203-95.
Kornhauser, Lewis A. “Unconscionability in Standard Forms.” California Law Review 64
(1976): 1151-83.
36
Lemley, Mark A. “Intellectual Property and Shrinkwrap Licenses.” Southern California Law
Review 68 (1995): 1239-94.
Lemley, Mark A. “Brief Amicus Curiae of American Committee for Interoperable Systems.”
ProCD v. Zeidenberg (7th Cir. 1996).
Lemley, Mark A. “Beyond Preemption: The Law and Policy of Intellectual Property Licensing,”
California Law Review 87 (1999): 111-72.
Lemley, Mark A. Peter S. Menell, Robert P. Merges, and Pamela Samuelson. Software and
Internet Law. Gaithersburg, MD: Aspen Law & Business, 2003.
Lessig, Lawrence. Sign it and Weep. The Standard, November 20, 1998.
Llewellyn, Karl N. The Common Law Tradition: Deciding Appeals. Boston: Little, Brown, and
Company, 1960.
Llewellyn, Karl N. “Book Review of Prausnitz: The Standardization of Commercial Contracts in
English and Continental Law.” Harvard Law Review 52 (1939): 700-11.
Mann, Ronald J. and Winn, Jane K. Electronic Commerce. Gaithersburg, MD: Aspen Law &
Business, 2002.
Marotta-Wurgler, Florencia. Are ‘Pay Now, Terms Later’ Worse For Buyers, Evidence From
Software License Agreements, NYU Law and Economics Working Paper No. 05-10
(2005).
Mazzeo, Michael J. “Competition and Service Quality in the U.S. Airline Industry.” Review of
Industrial Organization 22 (2003): 275-296.
Meyerson, Michael I. “The Reunification of Contract Law: The Objective Theory of Consumer
Form Contracts.” University of Miami Law Review 47 (1993): 1263-1333.
Moulton, Brent R. “An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables
on Micro Units.” Review of Economics & Statistics 72 (1990): 334-38.
Overly, Michael and Kalyvas, James P. Software Agreements Line by Line. Boston: Aspatore
(2004).
Posner, Richard. “Natural Monopoly and its Regulation.” Stanford Law Review 21 (1969): 548-
643.
Posner, Richard. Economic Analysis of Law. 6th ed. Gaithersburg, MD: Aspen Law & Business,
2003.
37
Priest, George. “A Theory of the Consumer Product Warranty.” Yale Law Journal 90 (1981):
1297-1352.
Priest, George. Letter in Support of UCITA to NCCUSL Standby Committee on UCITA.
http://www.nccusl.org/nccusl/ucita/ucita/ProfPriestletter.pdf. January 15, 2003.
Rakoff, Todd. “Contracts of Adhesion: An Essay in Reconstruction.” Harvard Law Review 96
(1982): 1173-1284.
Ribstein, Larry E. and Bruce H. Kobayashi. “Uniformity, Choice of Law and Software Sales.”
George Mason Law Review 8 (1999): 261-306.
Schmalensee, Richard. “Market Structure, Durability, and Product Quality: A Selective Survey.”
Economic Inquiry 17 (1979): 177-96.
Schwartz, Alan. “A Reexamination of Non-Substantive Unconscionability.” University of
Virginia Law Review 63 (1977): 1053-83.
Schwartz, Alan & Wilde, Louis. “Imperfect Information in Markets: The Examples of
Warranties and Security Interests.” University of Virginia Law Review 69 (1983): 1387-
467.
Schwartz, Alan & Wilde, Louis. “Intervening in Markets on the Basis of Imperfect Information:
A Legal and Economic Analysis.” University of Pennsylvania Law Review 127 (1979):
630-82.
Slawson, W. David. “Standard Form Contracts and Democratic Control of Law Making Power.”
Harvard Law Review 84 (1971): 529-66.
Slawson, W. David. “The New Meaning of Contract: The Transformation of Contracts Law by
Standard Forms.” University of Pittsburgh Law Review 46 (1984): 23-74.
Smith, Connie U. and Woodside, Murray. “Performance Validation in Early Stages of Software
Development.” In System Performance Evaluation: Methodologies and Applications,
edited by Erol Gelenbe. Boca Raton, FL: CRC Press, 2000.
Software and Information Industry Association. Software Industry Profile: Packaged Software.
http://www.siia.net/software/pubs/profile_0604.pdf. (June, 2004).
Spence, A. Michael. “Monopoly, Quality, and Regulation.” The Bell Journal of Economics 6
(1975): 417-29.
Federal Trade Commission, “Warranty Protection for High-Tech Products and Services.”
Symposium transcript: http://www.ftc.gov/bcp/workshops/warranty/. 2000.
38
U.S. Department of Justice and Federal Trade Commission. “Horizontal Merger Guidelines.”
http://www.usdoj.gov/atr/public/guidelines/hmg.htm. (1992, revised 1997).
White, James J. “Contracting Under Amended 2-207 (Freedom from Contract Symposium).”
Wisconsin Law Review 2004 (2004) 723-51.
White, James J. and Robert S. Summers. Handbook of the Law Under the Uniform Commercial
Code. 3d ed. St. Paul, MN: West, 1988.
Whitford, William C. “Law and the Consumer Transaction: A Case Study of the Automobile
Warranty.” Wisconsin Law Review (1968) 1006-.
39
TABLE 1
SOFTWARE COMPANY SUMMARY STATISTICS NOTE.—Means and standard deviations are based on 647 contracts from 598 software companies. Firm data, including total revenue, public versus private status, and years since incorporation, were obtained primarily from Hoover’s company directory and the 2005 Software Industry Directory database. Age refers to the number of years since incorporation, as measured as of the year 2005. The maximum age is set to 25, since firm operations before 1980 were unlikely to emphasize software publishing.
Mean (s.d.)
Minimum
Median
Maximum
Revenue ($)
499,000,000
(4,700,000,000)
1,000
2,200,000
87,500,000,000
Age (Years)
15
(6.57)
1
14
25
Public Company
.16
(.37)
0
0
1
TABLE 2
SOFTWARE PRODUCT SUMMARY STATISTICS NOTE.—Number of EULAs in the sample and average software prices by market and user type. Markets are based on Amazon.com software product categories. Price is the price of the product for which the EULA is obtained. Consumer products are those oriented towards the general public or small business users, while business products are those oriented toward medium- or large-size business users.
All Consumer Business Market
N EULAs
Mean Price
($)
N EULAs
Mean Price
($)
N EULAs
Mean Price
($) Business & Office > Business Accounting > Accounting 13 1302 4 144 9 1816 Business & Office > Business Accounting > Check Printing 2 257 1 20 1 495 Business & Office > Business Accounting > Payroll 3 304 0 . 3 304 Business & Office > Business & Office Mgmt. Software 15 1248 4 91 11 1669 Business & Office > Communication > E-mail > E-mail Clients 5 237 2 57 3 356 Business & Office > Communication > E-mail > E-mail Servers 1 350 0 . 1 350 Business & Office > Communication > E-mail > Security & Filtering 5 548 4 47 1 2550 Business & Office > Communication > Fax 3 253 1 50 2 355 Business & Office > Communication > Remote Access 4 194 2 174 2 214 Business & Office > Communication > Terminal Emulations > Linux & Unix 0 . 0 . 0 . Business & Office > Communication > Terminal Emulations > Macintosh 0 . 0 . 0 . Business & Office > Communication > Terminal Emulations > Windows 6 702 2 120 4 993 Business & Office > Database 34 879 8 344 26 1044 Business & Office > Document Mgmt. > Document Tracking & Inventory 8 497 3 117 5 725 Business & Office > Document Mgmt. > Scanning & OCR 4 162 2 75 2 249 Business & Office > Office Suites 4 714 3 119 1 2500 Business & Office > Presentation > General 5 271 3 171 2 422 Business & Office > Presentation > Flowcharts 8 1321 2 148 6 1712 Business & Office > Presentation > Marketing 0 . 0 . 0 . Business & Office > Project Mgmt. 18 1016 2 297 16 1105 Business & Office > Reports & Forms 15 499 6 300 9 631 Business & Office > Schedule & Contact Mgmt. > Calendars 5 567 1 225 4 652 Business & Office > Schedule & Contact Mgmt. > Organizers & Address Books 12 211 7 144 5 306 Business & Office > Schedule & Contact Mgmt. > Phone Books on Disc 1 40 1 40 0 . Business & Office > Spreadsheet 16 571 5 248 11 718 Business & Office > Training & Tutorials 4 195 3 144 1 350 Business & Office > Word Processing 7 276 4 62 3 561 Education & Reference > Arts & Culture 4 38 4 38 0 . Education & Reference > Encyclopedias & Dictionaries 2 50 2 50 0 . Education & Reference > Foreign Languages 5 193 5 193 0 . Education & Reference > Geography 1 499 1 499 0 . Education & Reference > History 0 . 0 . 0 . Education & Reference > Mapping 12 2543 6 86 6 5000 Education & Reference > Religious Software 12 151 8 73 4 307 Education & Reference > Science 16 3117 7 141 9 5431 Education & Reference > Script & Screenwriting 2 219 1 239 1 200 Education & Reference > Secondary Education > English & Grammar 6 192 3 63 3 321 Education & Reference > Secondary Education > Math 3 100 2 70 1 159 Education & Reference > Secondary Education > Science 0 . 0 . 0 . Education & Reference > Test Preparation 8 111 6 69 2 239 Education & Reference > Typing 0 . 0 . 0 . Education & Reference > Writing & Literature 3 118 2 77 1 200 Graphics > 3-D 15 1975 2 925 13 2137
Graphics > Animation 3 248 2 247 1 250 Graphics > CAD 11 1802 3 242 8 2386 Graphics > Home Publishing > Calendars 1 65 1 65 0 . Graphics > Home Publishing > Clip Art 2 25 2 25 0 . Graphics > Home Publishing > Fonts 2 3207 0 . 2 3207 Graphics > Home Publishing > Greeting Cards 0 . 0 . 0 . Graphics > Home Publishing > Mailing Labels 3 365 2 50 1 995 Graphics > Illustration 4 89 3 86 1 99 Graphics > Image Capture 7 254 5 63 2 732 Graphics > Photo Editing 10 659 6 52 4 1569 Home & Hobbies > Cooking & Health 6 36 6 36 0 . Home & Hobbies > Fashion 2 54 1 99 1 10 Home & Hobbies > Gardening & Landscape 2 99 2 99 0 . Home & Hobbies > Genealogy 4 47 4 47 0 . Home & Hobbies > Hobbies 6 132 4 132 2 132 Home & Hobbies > Home Design 3 120 1 250 2 54 Home & Hobbies > Legal 2 285 1 149 1 420 Home & Hobbies > Movies & Television 2 119 2 119 0 . Linux > Business & Office > Database 0 . 0 . 0 . Linux > Business & Office > Word Processing 1 40 1 40 0 . Linux > Communication > Terminal Emulations 0 . 0 . 0 . Linux > Graphics > 3-D & Animation 0 . 0 . 0 . Linux > Graphics > Home Publishing 0 . 0 . 0 . Linux > Networking > Firewalls 0 . 0 . 0 . Linux > Networking > Local Area Network 0 . 0 . 0 . Linux > Networking > Virtual Private Network 0 . 0 . 0 . Linux > Operating Systems & Utilities > Backup 3 153 1 60 2 199 Linux > Operating Systems & Utilities > Virus Protection 0 . 0 . 0 . Linux > Programming > Database 0 . 0 . 0 . Linux > Programming > Development Utilities 2 272 1 50 1 495 Linux > Programming > Programming Languages 0 . 0 . 0 . Linux > Web Development > Web Effects 1 499 0 . 1 499 Linux > Web Development > Web Page Editors 0 . 0 . 0 . Linux > Web Development > Web Site Hosting 0 . 0 . 0 . Networking > Directory Servers 11 666 4 223 7 919 Networking > File & Print Servers 3 198 1 396 2 98 Networking > Firewalls 2 332 1 395 1 270 Networking > Local Area Networks 3 126 1 20 2 179 Networking > Netware 6 367 0 . 6 367 Networking > Network & Enterprise Mgmt. 13 3919 1 29 12 4243 Networking > Security 9 580 3 70 6 835 Networking > TCP-IP 3 3215 0 . 3 3215 Networking > Telephony 4 85 3 93 1 60 Networking > Virtual Private Networks 5 1170 0 . 5 1170 Operating Systems > BeOS 0 . 0 . 0 . Operating Systems > DOS 0 . 0 . 0 . Operating Systems > Linux & Unix 5 493 3 73 2 1124 Operating Systems > Macintosh 0 . 0 . 0 . Operating Systems > Microsoft Windows 0 . 0 . 0 . Personal Finance > Investment Tools 0 . 0 . 0 . Personal Finance > Money Mgmt. 4 337 4 337 0 . Personal Finance > Tax Preparation 3 32 2 22 1 50 Programming > Database > DB2 0 . 0 . 0 . Programming > Database > Filemaker 0 . 0 . 0 . Programming > Database > Oracle 1 15000 0 . 1 15000 Programming > Database > SQL 2 820 0 . 2 820 Programming > Development Utilities > Code Testing 12 659 1 139 11 706
42
Programming > Programming Languages > C & C++ 14 677 4 213 10 862 Programming > Programming Languages > COBOL 0 . 0 . 0 . Programming > Programming Languages > FORTRAN 4 797 1 795 3 798 Programming > Programming Languages > Java 6 1148 1 99 5 1358 Programming > Programming Languages > Visual Basic 2 349 0 . 2 349 Utilities > Backup 13 157 8 62 5 309 Utilities > Cross Platform 5 141 2 84 3 178 Utilities > File Compression & Decompression 1 95 0 . 1 95 Utilities > File Conversion 4 101 4 101 0 . Utilities > Memory Mgmt. 1 28 1 28 0 . Utilities > Partitions 5 620 2 67 3 989 Utilities > PC Maintenance 14 591 5 68 9 882 Utilities > Screen Savers 4 287 2 25 2 549 Utilities > Virus Protection 18 161 11 40 7 351 Utilities > Voice Recognition 3 382 1 200 2 473 Video & Music > CD Burning & Labeling > CD Burning 3 70 3 70 0 . Video & Music > CD Burning & Labeling > CD Labeling 2 29 2 29 0 . Video & Music > Digital Audio > MIDI 1 70 1 70 0 . Video & Music > Digital Audio > Music Creation & Sequencing 5 265 3 99 2 514 Video & Music > Digital Audio > Sound Editing 0 . 0 . 0 . Video & Music > Digital Audio > Sound Libraries 1 695 1 695 0 . Video & Music > Digital Video > Compositing & Effects 0 . 0 . 0 . Video & Music > Digital Video > Video Editing 5 377 4 371 1 399 Video & Music > DVD Viewing & Authoring 6 59 5 61 1 50 Video & Music > Encoding 4 105 4 105 0 . Video & Music > Instrument Instruction 7 60 7 60 0 . Video & Music > MP3 Software 1 20 1 20 0 . Video & Music > Music Appreciation 3 60 3 60 0 . Video & Music > Music Notation 4 163 3 152 1 199 Web Development > e-Commerce 12 685 5 202 7 1030 Web Development > Internet Utilities > Electronic Publishing 1 599 0 . 1 599 Web Development > Internet Utilities > Instant Messaging 2 240 0 . 2 240 Web Development > Internet Utilities > Internet Phone 4 137 2 60 2 215 Web Development > Internet Utilities > Security & Filtering 7 237 4 46 3 491 Web Development > Internet Utilities > Shared Internet 0 . 0 . 0 . Web Development > Internet Utilities > Streaming Video 5 204 1 100 4 230 Web Development > Professional Development 19 929 4 652 15 1003 Web Development > Web Browsers 2 1012 1 30 1 1995 Web Development > Web Effects > Animation 0 . 0 . 0 . Web Development > Web Effects > Objects 1 399 0 . 1 399 Web Development > Web Page Editors > Intranet 1 1430 0 . 1 1430 Web Development > Web Page Editors > Tutorials 1 249 1 249 0 . Web Development > Web Page Editors > WYSIWYG 3 714 2 574 1 995 Web Development > Web Site Hosting > Application Servers 2 650 0 . 2 650 Web Development > Web Site Hosting > Web Servers 0 . 0 . 0 . Web Development > Web Site Hosting > Web-Site Analysis 1 699 0 . 1 699 All Markets
647
763
290
142
357
1268
43
TABLE 3
EULA TERMS AND BIAS: METHODOLOGY NOTE.—The table describes the terms tabulated for the EULAs in the sample and how each term is scored for purposes of measuring the overall buyer (licensee) vs. seller (licensor) bias of the contract. Negative scores capture pro-seller terms and positive scores capture pro-buyer terms. Zero scores capture neutral terms or (in case the term is not discussed in the particular contract) terms that would correspond to the default rule.
Acceptance of License
Does license alert consumer that product can be returned if she declines terms?
1=yes 0=no
Scope of License
Does definition of "licensed software" include regular updates such as enhancements, versions, releases, etc.?
1=yes 0=no mention -1=no
Are there restrictions on use?
0=no or no mention -1=yes (e.g., for business-oriented products, "for business purposes" or "internal purposes only" language; for consumer-oriented products, restrictions on commercial use)
Can licensee alter/modify the program?
0=yes or no mention -1=no
Can licensee create derivative works?
0=largely unrestricted or no mention -1=strict prohibition, derivative works owned by licensor, or need permission of licensor
Transfer of License
Are there limitations on transfer?
0=no or no mention -1=some or full restrictions (licensee cannot assign, transfer, lease, sublicense, distribute, etc.; or, needs written consent of licensor)
Can licensee transfer the software to an end user who accepts the license terms without licensor’s prior permission?
0=yes or no mention -1=no
Warranties and Disclaimers of Warranties
Are there express warranties? 1=yes 0=no
Is there a limited warranty stating that software is free from defects in materials and workmanship or that the software will work according manual specifications in force for 31 days or more?
1=yes (recorded as the number of days of coverage) 0=no
Is there a limited warranty stating that the media of software distribution and documentation are free from defects in force for 31 days or more?
1=yes (recorded as the number of days of coverage) 0=no
Do the limited warranties apply only to the original purchaser? 0=no or no mention -1=yes
45
Is the disclaimer in caps, bold, or otherwise conspicuously presented?
0=yes or no disclaimers appear -1=no
Disclaims IWM and IWFPP or contains "AS IS" language?
0=no -1=yes
Disclaims warranty that software will not infringe on third parties’ intellectual property rights?
0=no -1=yes
Limitations on Liability
Who bears the risk of loss?
0=licensor, for losses caused by factors under licensor’s control, or no mention -1=licensee
Who bears the performance risk? 0=licensor (for causes under licensor's control), or no mention, or licensee (for uses expressly forbidden by licensor) -1=licensee (language “licensee assumes responsibility of choice of product and functions,” etc)
Disclaims consequential, incidental, and special damages? 0=no or no mention -1=yes
Are damages waived under all theories of liability (contract, tort, strict liability)?
0=no or no mention -1=yes
Disclaims liability for foreseeable losses of licensee? 0=no or no mention -1=yes
What is the limitation on damages?
0=no mention or cap on damages greater than purchase price -1=cap on damages less than or equal to purchase price
Is there an indemnification clause?
1=indemnification by licensor 0=no, no mention, or two-way indemnification -1=indemnification by licensee
Maintenance and Support
Does base price include M&S for 31 days or more?
1=yes 0=no or no mention
Conflict Resolution
Forum specified? 0=court, choice of licensee, or no mention -1=specific court or mandatory arbitration
Law specified?
0=same as forum or no mention -1=yes and different from forum
Who pays licensor’s attorney fees?
0=paid by losing party or no mention -1=paid by licensee
TABLE 4
EULA TERMS AND BIAS: SUMMARY STATISTICS NOTE.—Mean, standard deviation, and correlation statistics for each of the terms comprising the overall bias index described in Table 3. Negative scores capture pro-seller terms and positive scores capture pro-buyer terms. Correlations in bold are statistically significant at the 5% level.
Correlation
Mean (s.d.)
Label
x1
x2
x3
x4
x5
x6
x7
x8
x9
x10
x11
x12
x13
x14
x15
x16
x17
x18
x19
x20
x21
x22
x23
x24
x25
Acceptance of License
Does license alert consumer that product can be returned if she declines terms?
.49 (.50)
x1 1
Scope of License
Does definition of "licensed software" include regular updates such as enhancements, versions, releases, etc.?
.07 (.46)
x2 .05 1
03 02 1
04 03 1
1
Are there restrictions on use? -.20 (.40)
x3 . -.
Can licensee alter/modify the program?
-.63 (.48)
x4 -.17 . .
Can licensee create derivative works?
-.37 (.48)
x5 -.13 .00 .06 .48
Transfer of License
Are there limitations on transfer? -.93 (.25)
x6 -.20 .05 1
1
.01 .27 .17
Can licensee transfer the software to an end user who accepts the license terms without licensor’s prior permission?
-.47 (.50)
x7 .10 .09 .11 .28 .18 .25
Warranties and Disclaimers of
Are there express warranties? .04 (.21)
x8 -.02
1
1
.07 -.12 .03 .00 -.06 .02 1
Is there a limited warranty stating that software is free from defects in materials and workmanship or that the software will work according manual specifications in force for 31 days or more?
.30 (.46)
x9 .21 .06 -.08 .04 -.02 -.12 .11 .17
Is there a limited warranty stating that the media of software distribution and documentation are free from defects in force for 31 days or more?
.32 (.47)
x10 .27 -.06 -.01 -.09 -.08 -.13 .00 -.05 .20
Do the limited warranties apply only to the original purchaser?
-.04 (.19)
x11 -.11 .03 1 .04 .02 .05 .02 -.03 .04 -.08 -.14
Is the disclaimer in caps? -.21 (.41)
x12 .04
1
1
.02 -.11 -.13 -.18 .02 -.06 -.03 .00 -.02 -.03 1
Disclaims IWM and IWFPP or contains "AS IS" language?
-.90 (.29)
x13 -.20 .02 .02 .24 .18 .22 .16 -.05 -.13 -.12 .07 .12
Disclaims warranty that software will not infringe on third parties’ intellectual property rights?
-.38 (.49)
x14 -.09 -.08 .06 .18 .25 .09 .09 .09 -.02 .00 .06 -.24 .26
Limitations on Liability
Who bears the risk of loss? -.16 (.37)
x15 .07 1
1
1
1
.12 1
-.04 .06 .04 .12 .07 .10 .01 .12 .05 -.04 -.08 .08 .08
Who bears the performance risk? -.27 (.44)
x16 -.08 .01 -.07 .20 .08 .04 .04 .05 .10 -.05 -.01 .09 .11 -.01 .02 1
Disclaims consequential, incidental and special damages?
-.89 (.31)
x17 -.19 .00 .02 .21 .20 .26 .15 .07 -.15 -.11 .04 -.05 .58 .27 .07 .08
Are damages waived under all theories of liability (contract, tort, strict liability)?
-.34 (.47)
x18 -.07 .07 .17 .27 .30 .12 .22 .04 -.05 -.07 .04 -.16 .19 .30 .06 -.02 .25
Disclaims liability for foreseeable losses of licensee?
-.72 (.45)
x19 -.20 .04 .03 .21 .21 .20 .09 .05 -.14 -.05 .04 -.15 .35 .29 .07 .08 .54 .26
What is the limitation on damages?
-.59 (.49)
x20 -.17 -.02 .28 .22 .19 .20 -.01 -.05 -.06 .07 -.12 .27 .24 .00 .05 .32 .35 .26
Is there an indemnification clause? -.12 (.42)
x21 .00 .03 .03 .01 .12 .06 .07 .08 .12 .13 -.02 -.04 .09 .18 .12 .03 .04 .08 .04 .01 1
Maintenance and Support
Does base price include M&S for 31 days or more?
.68 (.46)
x22 .00
.07 .03 .01 .02 -.01 .09 -.03 .05 -.03 .02 -.06 -.02 .12 -.09 -.04 .01 .07 .04 .04 .01 1
Conflict Resolution
Forum specified? -.32 (.46)
x23 .02
-.05 .13 .20 .14 .12 .18 -.08 -.03 .03 .05 -.22 .18 .30 .10 .01 .18 .23 .15 .23 .15 .09 1
Law specified? -.00 (.07)
x24 .07 .01 .02 -.04 -.01 .02 .03 .01 .04 .05 -.01 -.04 .02 -.01 .09 .01 .02 -.05 .04 .06 .03 -.05 .10 1
Who pays licensor’s attorney fees?
-.01 (.09)
x25 .09 .01 .05 .03 -.03 .02 .06 .02 .06 .06 -.02 .00 .03 -.07 -.04 .03 .03 .09 .02 .07 .06 -.06 .13 -.01 1
Overall Bias Index (sum of 25 above terms)
-5.64 (3.07)
.08 .20 .22 .51 .48 .31 .50 .11 .21 .13 .07 -.08 .45 .48 .26 .24 .47 .52 .45 .49 .32 .22 .46 .08 .12
47
TABLE 5
EULA TERMS AND BIAS: SUBINDEX SUMMARY STATISTICS NOTE.—Means, standard deviations, and correlations for each of the seven subindexes that comprise the overall bias index described in Table 3. Negative scores capture pro-seller terms and positive scores capture pro-buyer terms. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.
Correlation
# of Terms Included
Mean (s.d.)
Acceptance of License
Scope of License
Transfer of
License
Warranties
and Discl. of
Limitations on Liability
Maintenance and Support
Conflict
Resolution Acceptance of License
1
.49 (.50)
1
Scope of License
4
-1.13 (1.05)
-.10***
1
Transfer of License
2
-1.40 (.61)
.00
.33***
1
Warranties and Disclaimers of
7
-.88 (1.01)
.11***
.05
.09**
1
Limitations on Liability
7
-3.09 (1.49)
-.18***
.38***
.31***
.26***
1
Maintenance and Support
1
.68 (.46)
.00
.06
.07*
.03
.02
1
Conflict Resolution
3
-.33 (.50)
.04
.18***
.20***
.11***
.30***
.07*
1
Overall Bias Index
25
-5.64 (3.07)
.08**
.63***
.54***
.53***
.79***
.22***
.46***
TABLE 6
REGRESSIONS: EULA BIAS AND PRODUCT AND COMPANY CHARACTERISTICS NOTE.—Regression results. In the left of the table, the dependent variable is the natural log of the price of the software product for which the EULA is collected. In the right, the dependent variable is the overall bias index of the EULA. Higher values indicate pro-buyer bias; lower values indicate pro-seller bias. Samples include all EULAs (All), EULAs of consumer-oriented products only (Consumer), EULAs for business-oriented products only (Business), or EULAs of the subset of companies for which I have collected licenses for both a consumer and business-oriented product (Cos. w/both). Other product characteristics include a dummy indicating a consumer-oriented product and dummies for multi-user and developer licenses (the default is single-user license). Company characteristics include the size of the company as proxied by the natural log of revenue, the natural log of the age of the company since incorporating as of 2005, and a dummy for publicly traded companies. Fixed effects may include Amazon.com software sectors (e.g. “Business & Office”), Amazon.com software markets (e.g. “Business & Office > Business Accounting > Accounting”), or company (for the companies for which I have collected two EULAs). Standard errors are in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Dependent Variable: Ln Price Dependent Variable: Overall Bias Index
All All All Consumer Business Cos.
w/both All All All Consumer BusinessCos.
w/both (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Overall Bias Index .01
(.02) .01
(.02) .00
(.02) .02
(.02) -.02 (.02)
-.04 (.10)
Ln Price .08(.10)
.07 (.10)
.01 (.12)
.24 (.24)
-.12 (.16)
-.08 (.21)
Consumer -1.78*** (.10)
-1.67*** (.10)
-1.47*** (.10)
-1.44*** (.16)
.33 (.30)
.10 (.31)
.13 (.32)
. . -.39(.38)
Multi-User .66*** (.17)
.66*** (.17)
.89*** (.18)
1.09** (.50)
.88*** (.21)
1.51*** (.35)
-.50 (.43)
-.39 (.43)
-.45 (.49)
-1.62 (1.70)
-.28 (.53)
-.14 (.60)
Developer .41** (.18)
.37* (.20)
.37* (.22)
1.47** (.61)
.34 (.26)
-.09 (.71)
.69 (.46)
1.22** (.52)
.94 (.58)
1.92 (2.07)
.46 (.64)
.14 (1.02)
Ln Revenue .07***
(.02) .08*** (.02)
.08*** (.02)
.05 (.03)
.13*** (.04)
-.22*** (.06)
-.21*** (.06)
-.18*** (.06)
-.20* (.11)
-.12 (.09)
Ln Age .15* (.09)
.14 (.09)
.05 (.09)
.05 (.12)
.14 (.14)
1.63*** (.22)
1.47*** (.22)
1.51*** (.24)
1.06*** (.41)
1.58*** (.35)
Public -.29* (.15)
-.27* (.15)
-.19 (.16)
.14 (.20)
-.48* (.25)
-.40(.39)
-.44 (.40)
-.47 (.43)
.25 (.69)
-1.31** (.63)
Fixed Effects . Sector Market Market Market Company . Sector Market Market Market Company
N 647 647 647 290 357 98 647 647 647 290 357 98
R2 .43 .47 .60 .50 .45 .88 .01 .14 .30 .42 .39 .94
TABLE 7
COMPETITIVE CONDITIONS SUMMARY STATISTICS NOTE.—Estimates of competitive conditions across different software markets. For 189 EULAs in the sample, the product is sold on Amazon.com. Market share is an estimate of the company’s share of all Amazon.com sales in that Amazon.com market. The estimate is based on Amazon.com salesranks as described in the text. HHI is Herfindahl-Hirschman Index. For each Amazon.com market, it is estimated as the sum of the squared market shares of all companies who sell through Amazon.com. Unconcentrated is a dummy indicating that HHI is less than 10. Moderately concentrated is a dummy indicating that HHI is between .10 and .18. Concentrated is a dummy indicating that HHI is greater than .18. HHI Other is an alternative estimate of the Herfindahl-Hirschman Index which is based on other published market share estimates, collected from various sources as described in the text.
N obs
Mean (s.d.)
Minimum
Median
Maximum
Panel A. Company Market Share
MS
189
.12
(.19)
.00
.04
1
Panel B. Market Competition
HHI
647
.34 (.22)
.07
.28
1
Unconcentrated
647 .05 (.22)
0 0 1
Moderately Concentrated
647 .21 (.41)
0 0 1
Concentrated
647 .74 (.44)
0 1 1
HHI Other 70 .40 (.30)
.05 .36 .87
TABLE 8
REGRESSIONS: PRICE AND COMPETITIVE CONDITIONS NOTE.—Regression results. The dependent variable is the natural log of the price of the software product for which the EULA is collected. Samples include all EULAs (All), EULAs of consumer-oriented products only (Consumer), or EULAs for business-oriented products only (Business). Competitive conditions at the market level are measured by Herfindahl-Hirschman Indexes based on Amazon.com market share estimates (HHI), dummies for Moderately Concentrated and Concentrated markets (the excluded group is Unconcentrated), and HHI based on other published sources. Competitive conditions at the company level are measured using Amazon.com market shares (MS). Product controls in all models include the overall bias index, a dummy indicating a consumer-oriented product, and dummies for multi-user and developer licenses (the default is single-user license). Company controls in all models include the natural log of revenue, the natural log of the age of the company, and a dummy for publicly traded companies. Fixed effects include Amazon.com software sectors (e.g. “Business & Office”) or Amazon.com software markets (e.g. “Business & Office > Business Accounting > Accounting”). Standard errors are in parentheses. In specifications that include competitive conditions measured at the market level, the standard errors are adjusted for clustering at the market level. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Dependent variable: Ln Price All Consumer Business All Consumer Business All Consumer Business All Consumer Business (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
HHI
MS 1.02
.34* (.21)
.63** (.30)
.24 (.27)
Mod. Conc.
.44* (.23)
.30* (.18)
.63 (.39)
Conc.
.45** (.18)
.42*** (.16)
.63** (.31)
HHI Other .57** (.25)
.79*** (.16)
1.17*** (.22)
(.65) 1.96* (.98)
.31 (1.67)
Co. and Product Controls
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fixed Effects Sector Sector Sector Sector Sector Sector Sector Sector Sector Market Market Market
N 647 290 357 647 290 357 70 36 34 189 117 72
R2 .47 .12 .18 .47 .11 .18 .40 .31 .31 .83 .76 .83
TABLE 9
REGRESSIONS: EULA BIAS AND COMPETITIVE CONDITIONS
NOTE.—Regression results. The dependent variable is the overall bias index of the EULA. Higher values indicate pro-buyer bias; lower values indicate pro-seller bias. Samples include all EULAs (All), EULAs of consumer-oriented products only (Consumer), or EULAs for business-oriented products only (Business). Competitive conditions at the market level are measured by Herfindahl-Hirschman Indexes based on Amazon.com market share estimates (HHI), dummies for Moderately Concentrated and Concentrated markets (the excluded group is Unconcentrated), and HHI based on other published sources. Competitive conditions at the company level are measured using Amazon.com market shares (MS). Product controls in all models include the overall bias index, a dummy indicating a consumer-oriented product, and dummies for multi-user and developer licenses (the default is single-user license). Company controls in all models include the natural log of revenue, the natural log of the age of the company, and a dummy for publicly traded companies. Fixed effects include Amazon.com software sectors (e.g. “Business & Office”) or Amazon.com software markets (e.g. “Business & Office > Business Accounting > Accounting”). Standard errors are in parentheses. In specifications that include competitive conditions measured at the market level, the standard errors are adjusted for clustering at the market level. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Dependent variable: Overall Bias Index All Consumer Business All Consumer Business All Consumer Business All Consumer Business (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) HHI
MS -.91
-.21 1.23* (.57) (.74)
-.86 (.76)
Mod. Conc.
-.76(.61)
-.87 (.82)
-.49 (.97)
Conc.
-.59(.58)
-.55 (.76)
-.41 (.94)
HHI Other 3.17*** (.36)
4.10*** (.77)
2.99*** (.61)
(2.20) -2.00 (4.09)
2.37 (4.04)
Co. and Product Controls
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fixed Effects Sector Sector Sector Sector Sector Sector Sector Sector Sector Market Market Market
N 647 290 357 647 290 357 70 36 34 189 117 72
R2 .14 .15 .19 .15 .15 .19 .30 .32 .55 .65 .68 .86
52
TABLE 10
REGRESSIONS: EULA SUBINDEX BIAS AND COMPETITIVE CONDITIONS NOTE.—The dependent variables are the seven subindex bias scores: Acceptance of License, Scope of License, Transfer of License, Warranties and Disclaimers of Warranties, Limitations on Liability, Maintenance and Support, and Conflict Resolution. Higher values indicate pro-buyer bias; lower values indicate pro-seller bias. The samples include all EULAs with required data. In odd-numbered models, competitive conditions are measured by Herfindahl-Hirschman Indexes based on Amazon.com market share estimates (HHI). In even-numbered models, competitive conditions at the company level are measured using Amazon.com market shares (MS). Product controls in all models include the overall bias index, a dummy indicating a consumer-oriented product, and dummies for multi-user and developer licenses (the default is single-user license). Company controls in all models include the natural log of revenue, the natural log of the age of the company, and a dummy for publicly traded companies. Fixed effects include Amazon.com software sectors (e.g. “Business & Office”) or Amazon.com software markets (e.g. “Business & Office > Business Accounting > Accounting”). Standard errors are in parentheses. In specifications that include competitive conditions measured at the market level, the standard errors are adjusted for clustering at the market level. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Dep. Var.: Acceptance of
License
Dep. Var.: Scope of License
Dep. Var.: Transfer of
License
Dep. Var.: Warranties and Disclaimers of
Dep. Var.: Limitations on
Liability
Dep. Var.: Maintenance and
Support
Dep. Var.: Conflict
Resolution (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) HHI
.02 (.10)
.04(.18)
-.21(.08)
.19(.18)
-.32(.23)
.08(.10)
-.01(.09)
MS .81* (.43)
.76(.87)
.28(.52)
-1.67* (.86)
.04(1.04)
-.07(.34)
-1.07*** (.40)
Co. and Product Controls
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fixed Effects Sector Market Sector Market Sector Market Sector Market Sector Market Sector Market Sector Market
N 647 189 647 189 647 189 647 189 647 189 647 189 647 189
R2 .05 .55 .07 .54 .08 .58 .10 .50 .12 .70 .07 .62 .12 .61
FIGURE 1. EULA Bias - Distribution
0
10
20
30
40
50
60
70
80
90
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Overall Bias Index
Num
ber
of E
UL
As
Pro-Seller Default Pro-Buyer
FIGURE 2. EULA Bias and Competitive Conditions
-10 -8 -6 -4 -2 0 2
Unconcentrated
Moderate
Concentrated
Not Listed
<1%
1-5%
5-10%
10-20%
20-50%
>50%
Pro-Seller Default Pro-Buyer
Am
azon
HH
I Gro
upA
maz
on M
arke
t Sha
re
Overall Bias Index