Patenting and Licensing of Financial Innovations 1 by Praveen Kumar and Stuart M. Turnbull C.T. Bauer College of Business University of Houston July 17, 2006 1 We thank Josh Lerner for helpful comments. Tian Zhao and Guowei Zhang provided excellent research assistance. All remaining shortcomings are our responsibility.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Patenting and Licensing of Financial Innovations1
by
Praveen Kumar and Stuart M. Turnbull
C.T. Bauer College of Business
University of Houston
July 17, 2006
1We thank Josh Lerner for helpful comments. Tian Zhao and Guowei Zhang provided excellent researchassistance. All remaining shortcomings are our responsibility.
Abstract
Recent court decisions, starting with the State Street decision in 1998, allow business methods to
be patentable and now give �nancial institutions the option to seek patent protection for �nancial
innovations. We develop a dynamic model that incorporates salient aspects of the adoption and
dissemination of �nancial innovations. We �nd, somewhat surprisingly, that patent protection
may be detrimental to long-term pro�ts because of embedded real options for further innovations
and market expansion that are especially important for �nancial �rms. Moreover, a strategy of
patenting and then licensing may not be e¤ective because of expertise-related constraints of the
licensees. Our framework helps understand the success of a wide class of innovations including
swaps, credit derivatives, and pricing algorithms; it also helps �nancial institutions decide whether
it is optimal to exercise the patentability and licensing option.
Keywords: Business Methods; Financial Innovations; Patents; Licenses; Real Options
JEL classi�cation codes: G20, L10, O31
1 Introduction
Until recently, new products and services in the �nancial arena, ranging from back o¢ ce processing
systems to new methods to hedge liabilities, were generally considered not to be eligible for patent
protection, because of the business method exception to patentability (see below). However, a
number of recent court decisions, starting with the celebrated State Street decision,1 allow business
methods to be patentable. These legal decisions have ushered a new patentability paradigm that
poses an immediate decision challenge for �nancial institutions, because they now have the option
to seek patent protection for �nancial innovations. The possibility of patenting also brings with it
the option of licensing.
Financial innovations have occurred at a torrid pace (e.g., Tufano (1989) and Lerner (2002)),
and its economic importance is widely recognized (Merton (1992)). Yet, there are few available
decision frameworks that address the optimal exercise of the patenting and licensing options for
�nancial innovations. Such frameworks are especially needed because of certain unique features of
�nancial innovations.
There are systematic reasons why innovation in �nancial instruments and services may be much
more subject to public exposure than innovation elsewhere. First, there is often non-con�dential
regulatory scrutiny. Second, depending on the form of innovation, a �nancial institution may invite
other institutions to participate in order to reduce its risk exposure. This allows potential imita-
tors to learn about the innovation. Third, to generate increased demand, the end users must be
educated about the product. Again, information is disseminated about the product. Fourth, to
generate market liquidity the innovating institution needs the participation of other institutions.
The importance of developing liquid secondary markets for �nancial innovations distinguishes the �-
nancial services industry from other industries, such as computer software and telecommunications,
where developing a large market base is also critical.
However, there are barriers to entry and imitation in the �nancial industry that provide incen-
tives for innovation even with the endemic threat of exposure and imitation. More so than in most
other industries, specialized human capital and allied organizational assets are central to e¤ective
absorption of certain �nancial innovations � such as the development of a new kind of security or
implementation of especially complex formulas. Moreover, there is substantial heterogeneity among
1State Street Bank & Trust Co. v. Signature Financial Group Inc. 149 F.3d 1368, 1375 (Fed. Cir. 1998). Meurer(2002) provides a discussion of business method patents and related legal issues.
the end users of �nancial innovations. Consequently, if the innovating institution has a competitive
advantage with respect to in-house expertise, then it can continue to earn rents from the innova-
tion through �cream-skimming�or market-segmentation, even though the various versions of basic
innovation are o¤ered by imitators.
Indeed, while �nancial innovations share some structural features with innovations in other
industries where network externalities are important (see, e.g., Economides (1996))� again, for
example the computer software industry � �nancial innovations have certain unique characteristics.
In particular, in the �nancial services industry, the innovating institution need not be the exclusive
or even the most dominant vendor in order for the innovation to be pro�table. Rather, the innovator
can allow other �nancial institutions to o¤er various versions of the innovation to share risk, increase
market depth, liquidity and price transparency, while using its human capital and expertise-related
advantages to pro�tably trade with high-valuation end users.
In this paper, we identify and articulate important special features of �nancial innovations and
present a model� with applications� that helps address the basic question: to patent or not to
patent a �nancial innovation; and, if the innovation is patented, whether to license the innovation.
Our main contribution is to show � both theoretically and empirically � that for an important
class of �nancial innovations imitation is not detrimental to the innovating �nancial institution; in
fact, the innovating institution may optimally or strategically forego patent protection. By foregoing
patent protection and targeting the more sophisticated end of the market, innovating institutions
can reconcile the somewhat con�icting imperatives of developing liquid markets and earning rents
from the innovation. Interestingly, this conclusion is robust to the availability of a patent-and-
license option, especially if the licensees degrade the value of the second-generation innovations
for buyers because of expertise- or human-capital-related constraints � a factor that is especially
relevant for �nancial innovations, but is typically not emphasized in the innovations literature.
To �x ideas, we examine some common types of �nancial innovations and consider whether
patent protection may be warranted. If the innovation is strictly for internal use, such as a back
o¢ ce system, then there appears to be little at issue about whether to protect it either as a trade
secret2 or as a patent, apart from legal and administrative costs.
Many forms of innovation provide a direct service to clients that are visible to competitors.
2A trade secret is any information that can be used in the operation of a business or other enterprise and that issu¢ ciently valuable and secret to a¤ord an actual or potential economic advantage to others. See Sections $$39 to$$45 of the Unfair Competition Act.
the exception to strike the patent. In this case the patent was found invalid for lack of novelty and
the �invention�not because it was improper subject matter for a patent.
This judgment changed the law in two ways. First, if a software invention produced a useful,
concrete and tangible result, then the mathematical algorithm exception does not apply. Second,
it recognizes that business methods are patentable. Note that Judge Rich stressed that business
methods must meet the other legal requirements for patentability: novelty (the method must be
new) and non obviousness (the method must not be obvious to a person knowledgeable in the area).
The State Street decision was a¢ rmed in 1999 in A.T. & T. v. Excel Communications Inc.9. There
the Federal Circuit clari�ed the scope of the mathematical algorithm exception, establishing that
it is what an algorithm does �not how it does � that determines whether the subject matter is
patentable.
Following the State Street decision, the criterion for patentability rests on whether an invention
has a novel and practical application, as measured by the production of useful, concrete or tangible
results. Under this approach is the recognition that abstract ideas are not patentable10.
An abstract idea by itself never satis�es the requirements of eligible subject matter under 35
U.S.C. Section 101. An abstract idea, when practically applied to produce a useful, concrete
and tangible result, satis�es Section 101.
A practical application must be within the useful arts, employing technology if needed to realize
the practical application.
The test for practical application involves a determination of whether there is: (1) speci�c, sub-
stantial and credible utility in the speci�cation; (2) whether the invention produces a �con-
crete�result; and (3) whether the result is tangible, that is more than a merely mathematical
construct or a disembodied data structure, for example.
The extent of protection and whether it will be easy to design around the claims of the patent
depend on the particular invention. The Doctrine of Equivalence provides protection against models
that represent a �minor�extension11. The doctrine of equivalence is involved when a patent claim is
9See 172 F3d 1352, 50 USPQ 2d 1447 (Fed. Cir. 1999), cert dnd 120 S.Ct. 368 and remanded 1999 WL 1050064,52 USPQ2d 1865 (Oct. 1999).10This section draws on the comments of the American Intellectual Property Law Association (2002).11 In Festo Corp. v. Shoketsu Kinzoku Kogyo Kabushiki Co. (SMC), No. 00-1543, 2002 WL 1050479 (U.S. May
28, 2002) the Supreme Court attempted to clarify the applicability of the Doctrine of Equivalence with respect toamendments in any prosecution history estoppel.
7
not literally infringed, but substantially infringed. In most cases the infringing product has similar
but not identical elements as these described in the patent.
2.2 Recent History of Patent Awards
We follow the approach described in Lerner (2001, 2002) and examine patents in �ve subclasses of
classi�cation 705 for the period January 1971 to September 200512. In Table 1 we identify patents
that were listed as having �nance as the main classi�ed. In Part A, we record the history of patents
prior to the State Street decision and in Part B post the decision. It is clear that there has been a
regime shift, with more than double the number of patents awarded post the decision, than during
the 28 years prior to the decision.
In Table 2, we identify the companies that have been the most active in successfully seeking
�nancial patents. We split the companies into non-�nancials (Part A) and �nancials (Part B). The
split is some what arbitrary, as we have put General Electric into non �nancials. Comparing the
two parts, �nancial corporations have secured the most �nancial patents, as is to be expected. This
raises the question of the nature of the �nancial innovations that have been patented.
2.3 The Nature of Financial Patents
In Table 3, we list a number of patents awarded to di¤erent �nancial institutions. Part A, gives
examples of the 705/35 classi�cation. The patents typically describe a method and system to
perform some type of task, such as providing account values in an annuity with life contingencies,
management of an investment fund over a speci�ed horizon, dynamic portfolio bench marking and
a method for hedging one or more liabilities associated with a deferred compensation plan. These
types of patents are all for systems and methodologies that allow the �nancial institution to o¤er
services to clients using the systems/methodologies covered by the patent. As such, the patent
provides either a potential barrier to entry or a licensing opportunity. Other patents cover back
o¢ ce type of applications or methodologies to perform a particular task.
12The subclasses are:705/35: �nance (e.g. Banking, investment and credit).705/36: portfolio selection, planning or analysis.705/37: trading, matching or bidding.705/38: credit risk processing, loan processing.705/4: Insurance �calculation of annuity rates, investment of insurance company assets, the management of
risk through �nancial instruments and related topics.
8
Part B, gives examples of the 705/36 classi�cation. The patents typically describe a method
to perform some type of task, such as a computer implemented method for interacting with the
user of a trading computer, a method to improve the analysis of the performance of mortgage and
closed end loan portfolios, a data processing system and method to track di¤erent types of market
bench marks, and a method for pricing options using importance sampling and strati�cation Monte
Carlo simulation. The �nancial institution will bene�t from the methodology by either employing
it itself, selling the services generated by the product or by licensing its use.
In Part C, the 705/37 classi�cation, the patents cover a diverse range, examples being: a
compute method for optimizing portfolios of multiple participants that facilitate trades as swaps
among multiple parties that keep trades out of the market, a computerized order method and system
for tracking orders implemented on a trading �oor exchange, a system and method for managing a
plurality of stock option accounts each for a plurality of participants, and an investment company
that issues conventional shares and exchange traded shares in the same fund.
In Part D, the 705/38 classi�cation, the patents typically describe a method to perform some
type of task, such as a system that emulates the thought process of a loan o¢ cer in the risk
assessment and completion of a loan package, a system that assists customers in identifying their
credit needs and picking available products, a neural network that imitates a credit manager�s
evaluation and decision process to control loss and guide business expansion, and a method and
system for real time credit approval. Again, the patent is describing some type of back o¢ ce
function that gives the institution a comparative advantage.
In Part E, the 705/4 classi�cation, the patents typically describe a method to perform some
type of task, such as a system and method for providing retail clients with insurance policies
for foreign exchange risk, a system that administers a mortgage and life insurance combination
program, a method for evaluating the insurability of a potential insurable risk, and a data mining
based underwriting pro�tability analysis.
To summarize, we have randomly sampled a number of �nancial patents awarded to �nancial
institutions or �rms that provide �nancial services. The patents fall into the category of a system
and/or method that undertakes a particular task that provides either a direct service to the �nancial
institution or can be transformed to provide a revenue generating service to clients. The patent
acts as a barrier to entry.
9
2.4 Licensing13
An institution can always patent an innovation and then license the use of the innovation to other
institutions. Alternatively, for certain types of innovations that lack the uniqueness to qualify for
patent protection, and require the participation of other institutions to help develop a market, the
innovator can register the name of the product as a service mark14 and license its use. In the
case of �nancial innovations, licensing brings its own set of complications arising from pricing and
antitrust laws.
For example, in April 2003, Morgan Stanley and J. P. Morgan (the owners) launched their
jointly owned Trac-X credit default indexes15. By November 2003, it had been licensed to eleven
dealers. Trade information (volume, size) could not be shared for antitrust reasons. This prevented
a per trade or a per notional variable license fee. Determination of a realistic value for a �xed fee is
not easy, given the di¢ culty of quantifying the license value to di¤erent dealers. There is an upper
bound, as the size of any �xed fee would have to be relatively low, given that the barriers to entry
for indexed products is low. A small �xed fee could be charged to help o¤-set some of the legal and
regulatory compliance costs.
The owners did however derived indirect bene�ts. First, the prestige value associated with being
a market innovator. Second, if a licensee wanted to introduce a new structure based on Trac-X,
they had to receive permission from the owners. This meant possibly revealing information about
the new structure and possibly revealing proprietary trading information to the owners. Third,
as both Morgan Stanley and J. P. Morgan had agreed to make markets in the index, they were
able to observe market �ow information. End users in shopping around for quotes, would normally
approach these dealers. Other investment dealers would also approach these dealers for quotes,
when trying to hedge their positions. Fourth, they controlled the selection of the names in the
indexes.
13We gratefully acknowledge discussion with Lisa Watkinson, Lehman Brothers (New York), aboutsome of the issues associated with licensing �nancial products.
14A service mark is similar to a trademark, except that a trademark promotes products while service marks promoteservices. See Lanham Trademark Act 15 U.S.C.A.$$1051-1127.15The case history is reported in Du¢ e (2004) and Chacko, Dessain, Sjoman, Maruani and Hao (2005).
10
3 Innovation Characteristics
Senior management must decide on a policy that describes what forms of innovations will be
patented. In Table 3, the patents usually describe some type of method or system that provides
the �nancial institution with a comparative advantage. These can be grouped into three broad
categories: (a) undertaking some form of back o¢ ce function; (b) facilitates a service that can be
o¤ered to clients or an improvement in the technology of an existing service; and (c) a methodology
of performing a particular task.
Innovations that fall into categories (a) and (c) may be di¢ cult for outsiders to directly observe.
However, there is di¤usion of knowledge. In the �nance industry there is mobility of labor and when
workers move, they take their knowledge with them and can employ it in their new employment.
There are also industry conferences and trade magazines, where people describe their work. Con-
sequently, there are a number of di¤erent avenues for the di¤usion of knowledge. Without patent
protection, competitors may quickly reverse engineer the product and o¤er a competing product.16
For innovations that fall into category (b), the innovation facilitates a service that is o¤ered to
clients. Consequently, the institution advertises the service. Rival �rms can attempt to reverse
engineer the innovation that facilitated the service. A patent provides protection to the innovator,
allowing it to earn rent.
In this section, we describe the characteristics of innovations for which patent protection is
unnecessary or undesirable.
3.1 The Form of Innovation
To be concrete, we consider the innovation to be some form of �nancial instrument, that will appeal
to a wide section of end users. We will give examples shortly. In the �rst round of transactions,
the innovator learns how to appropriate price the instrument and how to hedge. In the process,
the institution earns rent that rewards it for its innovation. It also earns itself a reputation as
a market leader. News of the innovation spreads among competitors and imitators start to o¤er
similar products. Consequently, rents dissipate and the innovator ends up earning a fair rate of
return. This is the type of �rst mover advantage argument described by Herrera and Schroth (2003)
(HS). However, the situation in practice is often far more complex, than that described by HS.
16 If an innovation is protected by a trade secret and a third party reverse engineers its innovation, the innovationwill not have protection if the third party was not bounded by the trade secret.
products to disrupt the installation of user base for rivals (Farrell and Saloner (1986)).
However, pro�t-generation in successful innovations in network markets and �nancial innova-
tions arguably di¤er in at least one important aspect. Successful innovations in network markets
generate pro�ts by becoming dominant standards and increasing the size of the user base. In such
�winner-take-all�markets, axiomatically, there is little scope for the simultaneous existence of di¤er-
ent versions of the basic technology provided by di¤erent vendors. By contrast, successful �nancial
innovations are often characterized by various derivative innovations (of the basic innovations) be-
ing simultaneously o¤ered by di¤erent �nancial institutions. The innovating institution typically
earns pro�ts not by grabbing the entire market, but by expropriating the most pro�table trading
segments through a �rst-mover advantage based on expertise.19
�Cream-skimming� by the innovating institution in the derivative innovations market, after
initially developing the liquidity and transparency of the market by inviting the participation of
other institutions, is thus a relatively unique aspect of an important class of �nancial innovations.
Importantly, this aspect of �nancial innovations poses a dilemma for the innovating institution:
whether to patent (and possibly license) the innovation or to forego patent protection. In the next
section, we present a model that analyzes this decision problem in an industry equilibrium.
4 A Real Options Model of Financial Innovations
The model has three dates. At the start (t = 0), the �nancial institution (denoted I) expends
an amount C0 to develop a new form of a derivative. The derivative can be purchased from the
institution and is e¤ective for one-time period. That is, if investors wish to obtain recurring bene�ts
from the derivative, they must re-purchase in every time period. The institution has a client base or
initial market ofM0 investors, each of whom obtain a a basic value of � from using the derivative at
each date, and this parameter is common knowledge. We assume that I is the high-quality provider
and the leader in the market because of its proprietary intellectual capital and its pool of specialized
human capital. Therefore, in addition to the derivative, I can provide additional services regarding
the derivative that add value to the buyers: for example, from expertise in the structuring of the
19Herrera and Schroth (2003) argue that �nancial �rms that innovate possibly enjoy a sustainable �rst moveradvantage. They learn how to structure a particular type of deal. Repeated deals provide the innovator with moreexperience on the structuring of deals. While the term sheet for the deal may be public knowledge, the actualstructuring details are private. Consequently, the innovator can earn rents for a period before imitators learn how tostructure similar deals eroding spreads.
15
derivative; the resolution of legal and regulatory issues; providing ancillary technology to compute
the required cash �ows to di¤erent stake holders, and the development of the necessary pricing and
hedging methodologies. These additional services are valued at �. Thus, if the derivative premium
is �, then the net bene�t to the investor is, �+ � � �.
To develop the market beyond the initial client base, the �nancial institution needs to advertise
the derivative to potential end users. But, in this process, imitators learn about the derivative and
compete with the institution at (t = 1). These imitators also advertise to their end users. We will
assume that there are N imitators and each imitator has Z end users, who obtain the basic value �
per period from using the derivative. The imitators, however, are not in a position to provide any
additional services to the end users. Hence, the net bene�t to the customers of investors is, � � �.
The provision of specialized services is costly; for example, for �nancial instruments, the bulk
of the delivery costs are specialized labor wage costs. Because such highly quali�ed labor is in
short-supply, we assume that the unit costs are convex in the market size: thus, if the market size
is M; then the unit costs are, 12c¯M2; for some given parameter c
¯> 0: The imitators at (t = 1),
on the other hand, act competitively and face a common constant unit cost function, �cM; where
�c < �:
4.1 The Patenting Decision
The innovating institution, I, can forestall imitation by patenting the derivative valuation process.
If the institution does patent the new derivative, then the market size for the derivative at date
t = 1 is limited by the institution�s own advertising and client reach, namely M0. We denote the
patenting decision through the binary variable P 2 f0; 1g, where P = 0 indicates that the �rm
obtains a patent, while P = 1 denotes the alternative.
The institution I expects that the innovation will lead to further innovations. It anticipates
that in the process of communicating with end users and observing their value generation from the
derivative, it will learn about (a) how to improve the market for the derivative and (b) the potential
demand for new forms of innovations related to the derivative. E¤ectively, the institution has real
options for further innovation.
The end user may view the future innovation either as a complement or a substitute. For
example, if the initial innovation is a credit default swap and the future innovation is an option on
the credit default swap, the end user might invest in both. Alternatively, if viewed as a substitute,
16
the end user switches from investing in the credit default swap to investing in options on the swap.
Another example of a substitute would be a collateralized debt obligation and the future innovation
being a synthetic collateral debt obligation on a credit index, the later having more transparent
pricing. In this paper, we treat the future innovation as a substitute, though the analysis readily
extends to the case of a complement.
We model these real options by assuming that at date t = 1, I makes an investment C2 in a
future innovation that will materialize at date t = 2, with a probability q, 0 < q < 1 The new
innovation increases the buyer valuation to � > �+�: Because I is the originator of this innovation,
it has a monopoly over its delivery at date t = 2; taking as given the total market size for the initial
innovation (or derivative) at the end of the previous period (date t = 1). The unit cost function for
this innovation for I is 12c0M2. Note that in general the cost of further innovation,C2, will depend
on the initial patenting decision. If the market has su¢ ciently developed in size and knowledge,
then costs may be lower. The size of the market will tend to be larger in the absence of patenting.20
Under certain situations, I; as a monopolist, may wish to restrict the market it serves. In this
case, the residual market can still buy the original derivative in the market place. However, we
assume that by date t = 2; the market for the original derivative is competitive, and all producers
face common constant unit costs of production and delivery of, c¯: We therefore incorporate the
idea, well documented by the empirical literature, that the original innovation eventually becomes
a commodity over time as the expertise and specialized inputs required for its production and
delivery become publicly known and freely available, respectively� see Tufano (1989). Indeed, we
also assume that the opportunity to earn rents from this class of derivatives itself expires at the
end of date t = 2, although we can easily allow a competitive market in the product class to remain
over the horizon, without materially a¤ecting our results:21
Finally, all players are risk-neutral. The time-interval between adjacent dates is � and the
instantaneous risk-neutral discount rate is r:
4.1.1 Analysis
20Similarly, we would expect Bayesian updating to occur for the probability of a successful innovation, q, given thatat date (t = 1), I can observe the state of market development. For the present, we ignore the Bayesian updating.21More realistic touches, such as allowing the buyer value from using the original derivative to atrophy over time
(because of possible obsolescence) can be easily incorporated at the cost of additional notation, but without materiallya¤ecting our results.
17
We solve the model through backward induction, starting at date t = 2: Let, MT1 denote the total
number of investors purchasing the initial innovation at the end of date t = 1: Clearly,MT1 depends
on whether I patented the innovation at date t = 0 or not. That is,
MT1 =
8<: M0; if P = 0
M1 �M0 +NZ; if P = 1(1)
We �rst consider the case of no patenting at the initial date: P = 1. If I successfully develops
the innovation, then it faces a market where the buyers�reservation utility is determined by their
ability to purchase the initial innovation at the price c¯. Let �� � � + �: Buyers will only purchase
the new innovation at a price �2, if (���2) � (���c¯) The constrained pro�t maximization problem
facing I, conditional on having a successful innovation at date t = 2, is to choose a derivative
premium �2 and market size Q2 to:
Maxf�2;Q2g
��2Q2 �
1
2c0Q22
�; s:t:; (i) �2 � �� (�� � c¯); (ii) Q2 �M1 (2)
In (2), we recognize the upper limit on the price due to the reservation utility of the buyers. As
this pricing constraint will be binding in any optimal strategy for I; we straight forwardly compute
the optimal price of the new innovation and its market share as:
Q�2(P = 1) = Min
�M1;
�� (�� � c¯)
c0
���2 = �� (�� � c
¯) (3)
This policy yields the pro�ts,
��2(P = 1) = Q�2(P = 1)
��� (�� � c
¯)� c
0Q�2(P = 1)
2
�(4)
Note that these pro�ts are positive because, by assumption, � > ��: Thus, I will invest in developing
the new innovation if and only if
C2(P = 1) � exp(�r�)[q��2(P = 1)] (5)
We turn next to the case where I has taken out a patent at date t = 0: P = 0: In this case, I
maintains a monopoly over the market,M0. Of course, I may still wish to segment this market into
18
buyers who receive the second-generation innovation, at a premium of �2, and buyers who receive
the original innovation, at a premium of �2. Buyers of the latest innovation therefore will purchase
as long as (�� � �2) � � � �2: Hence, conditional on successfully developing a second-generation
innovation, the optimization problem of I is now to,
Maxf�2;Q2g
�[�2Q2 �
1
2c0Q22] + [(�2 � c¯)Max(0;M0 �Q2)]
�; s.t., �2 � �� (�� � �2) (6)
The objective function (6) shows how the market gets endogenously segmented between the
�rst and second generation innovations. Analysis of the maximization problem yields the optimal
pricing and market segmentation policies:
Q�2(P = 0) = Min
�M0;
�� (�� � c¯)
c0
���2 = �; ��2 = �� (7)
These policies yield the pro�ts:
��2(P = 0) = Q�2(P = 0)
��� c
0Q�2(P = 0)
2
�+ (�� � c
¯)Max(0;M0 �Q�2(P = 0)) (8)
Thus, I will invest in developing the new innovation if and only if
C2(P = 0) � exp(�r�)[q��2(P = 0)] (9)
We can delineate two sets of conditions for whether it is optimal to patent or not. It then follows
from (1), (4) and (8) that,
Proposition 1 Suppose that M1 >��(���c
¯)
c0 : Then, ��2(P = 1) > ��2(P = 0) if M0 is su¢ ciently
small relative to ��(���c¯)
c0 :
Proposition 1 con�rms the intuition that if imitators bring in a su¢ ciently large number of
buyers into the market at date t = 1; then the pro�ts from a successful new-generation innovation
are higher for I if it does not patent the initial innovation. This result also indicates that, for a
su¢ ciently large ZN , allowing imitation is more likely to be optimal for I if the second-generation
innovation signi�cantly improves buyer value compared to the unit cost, that is, (� � ��)=c0 is
high and/or if there is a signi�cant cost reduction between the two innovations, that is, c¯/c0 is
19
high. However, if (� � ��)=c0 is low and/or if there is a signi�cant cost increases between the two
innovations, that is, c¯/c0 is low, then we have the reverse case:
Corollary 1 If M0 >��(���c
¯)
c0 , then ��2(P = 0) > ��2(P = 1):
We turn next to analysis at date t = 1: We �rst consider the case of no patenting. Because I is
the market leader, it chooses a premium and market size, with the imitators serving the remaining
market at the break-even price of �c. An end user will buy from I only if �1 � �c � �; where �1
is the premium charged by I: Hence, I�s constrained pro�t maximization problem is to choose a
derivative premium �1 and market size Q1:
Maxf�1;Q1g
��1Q1 �
1
2c¯Q21
�; (10)
s.t., (i) �1 � �+ �c; (ii) Q1 �M1 (11)
The reservation utility constraint in (11) will be binding in the optimal policy. Hence, the solution
to (10)-(11) is,
Q�1(P = 1) = Min
�M1;
�+ �c
c¯
���1(P = 1) = �+ �c (12)
The pro�ts with the optimal policy, ��1(P = 1); are given by
��1(P = 1) = Q�1(P = 1)
��+ �c� c¯Q
�1(P = 1)
2
�(13)
If a patent has been taken out at date t = 0, that is, P = 0; then at date t = 1; I has a monopoly
over the provision of the initial innovation. Thus, I will charge the premium �1, subject to the
constraint that �1 � �� and serve its pro�t maximizing market:
Maxf�1;Q1g
��1Q1 �
1
2c¯Q21
�; s.t.�(i) �1 � ��; (ii) Q1 �M1 (14)
The optimal policies are therefore,
Q�1(P = 0) = Min
�M0;
��
c¯
���1(P = 0) = �� (15)
20
The pro�ts from the price-quantity strategy speci�ed in (15), denoted by ��1(P = 0) :
��1(P = 0) = Q�1(P = 0)
��� � c¯Q
�1(P = 0)
2
�(16)
Now, we can directly compare I�s pro�ts at date t = 1, based on the patent decision at the
previous date. Intuitively, this comparison trades-o¤ the higher pro�t margin and lower market size
with patent protection against the lower margin and higher market size without patent protection.
Patenting strictly dominates the alternative at date t = 1 if M0 is at least as large as the optimal
monopoly market size for I. This is stated formally in the following proposition:
Proposition 2 If M0 � ��c¯; then ��1(P = 0) > �
�1(P = 1):
Note that Proposition 2 holds is independent of ZN: That is, if I�s initial end user base is
not too small, then patenting strictly dominates the alternative from the viewpoint of date t = 1;
irrespective of the market extension provided by imitators. On the other hand, the logic of
Proposition 2 can be reversed if M0 is su¢ ciently small relative to ZN: That is,
Proposition 3 If M0 is su¢ ciently small relative to ZN; then ��1(P = 1) > ��1(P = 0):
Propositions 2 and 3 clarify the essential con�ict between patenting and allowing imitation:
patenting increases pro�ts on the initial innovation, but may restrict pro�ts� relative to an open
imitation environment� from the second-generation or subsequent innovation.
Indeed, Proposition 3 implies that if the innovating institution�s initial market size, M0; is
su¢ ciently small, then it is bene�cial not to patent in order to increase the value of the real option
of the subsequent innovation. That is, irrespective of the size of ZN; ifM0 is su¢ ciently small, then
it is likely that the optimal policy is to forego patenting. Usually, if there is a �break through,�
one expects innovations creating large buyer value per unit cost, i.e., a large (��=c), to be more
patentable. The following Corollary shows that this is not always the case.
Corollary 2 There exists some 0< M¯ 0< ��c¯such that ��1(P = 1) > �
�1(P = 0) whenever M0 <
M¯ 0:
This result is somewhat counter-intuitive because it implies that, for a given M0, patenting is
less likely to be optimal if the initial innovation creates large buyer value per unit cost of delivery.
Usually, one expects innovations creating greater buyer value (per unit cost) (��=c¯)to be more
21
patentable. However, this intuition overlooks the fact that high-value initial innovations increase the
innovator�s short-run pro�ts without a patent, while also (at least weakly) increasing the innovator�s
pro�ts from subsequent innovations. Another way of stating this point is that it may be sub-optimal
to patent signi�cant �break throughs,� especially if these break throughs can give rise to further
innovations. Some types of �nancial innovations fall into this category. If patented, so that there
are no other suppliers, the market is too small. To be viable, the market needs other suppliers and
this can be achieved by not patenting.
We now analyze the determinants of the optimal patenting decision in further detail by com-
paring the present value of pro�ts, from the view point of date t = 0. First, with patenting22
a pure quality ladder, each point in the sequence increases the product quality by a �xed amount,
while this is not the case in our model. More importantly, our model allows heterogeneous buyer
valuation of the innovation sequence, based on di¤erences in quality of supporting services, unlike
the quality ladder model where the quality superiority of the innovations is �xed for the industry
(e.g., O�Donoghue et al. (1998)).
5 Applications of the Model
The model developed in the last section can be expressed in the form
��0(P ) = PV0(P ) + PV0[option(P )] (21)
where for a given patenting policy denoted by the symbol P , the term PV0(P ) represents the
present value of the initial innovation and PV0[option(P )] the present value of options associated
with subsequent innovations that depend on the initial innovation. The model provides a framework
to assist executives in determining whether or not to patent an innovation. We now apply this model
to the examples considered in Section 3.2.
5.1 Case Study One: Swap Innovations
Consider the case of interest rate swaps. In the 1980�s, if the innovating institution had patented the
idea of an interest rate swap, preventing other institutions from competing23, then it must estimate
the present value of the cash �ows from marketing swaps to end users: (PV0(P = 0)). Central to
the analysis is an estimate of the size of the market (M0), the growth in the market24, the cost of
servicing each transaction (c¯) and the value added (��). The cost of the innovation (C0), depends
on the costs associated with designing the contract, addressing legal issues associated with the
exchange of cash �ows in the presence of counterparty risk, addressing regulatory issues, designing
the back o¢ ce, designing hedging strategies and having the sta¤ to run a swaps desk.
If the institution does not patent, then information about the swap contract and the potential
pro�ts will be disseminated, attracting other institutions (N) to enter the market, each able to
23The institution could have allowed other institutions to o¤er swaps under licensing agreements. This wouldhowever have hindered the development of the market. The pricing of the licensing agreement would also be an issue- see the discussion in Section 2.4.24 In the model developed in the last section, we did not address this issue in order to avoid complication.
27
reach a client base (Z). The operating cost per contract are (�c) and the value added (�). The
institution I is assumed to have an advantage in execution, at least over some initial period and
the value added to end users is (�� = � + �; � > 0). The size of the potential market has now
expanded to (M1 = M0 +NZ). The institution must estimate the present value of the cash �ows
in this more competitive environment. This is denoted by PV0(P = 1).
In the Herrera and Schroth (2002) analysis, it is argued that it is possible for PV0(P = 1) >
PV0(P = 0), implying that I may not require patent protection to recoup the costs of innovation.
For �nancial innovation involving �nancial instruments, the situation is usually more complicated.
It is possible for PV0(P = 1) < PV0(P = 0), yet it is still optimal for I not to seek patent
protection. The di¤erence arises from the present value of subsequent innovations motivated by
the initial innovation.
Here future innovations may take many forms. One example is instead of exchanging �xed
for �oating payments, exchange �oating for �oating payments referenced to two di¤erent interest
rates25. Another example would be to trade options on swaps. The success of future innovations
depends on the acceptance of the initial innovation. End users must be aware of the bene�ts of
using swaps. The size of the market a¤ects the liquidity of the market. If I had patented the
innovation, it reaches a market of size M0. This may a¤ect the costs of introducing new forms of
swaps, as it needs to educate end users about the merits of swaps, the liquidity of the swap market
may be quite limited, restricting its development. It also must address the legal and regulatory
issues that arise from the new forms of innovation.
If I had not patented the innovation, the size of the swap market will be enhanced and with
more end users there will be more knowledge about the product. Institutions in the swap market
will also learn from each other about the pricing and hedging of swaps26. Consequently, the costs
of introducing a new form of swap should be lower compared to the patent case: C2(P = 0) >
C2(P = 1). The institution I must estimate the present value of the option to undertake further
innovation: PV0[option(P )]. Therefore I is now in a position to calculate the net present value of
initial innovation plus the option for further innovation.
Note that in our model we assumed that I is the innovator for subsequent innovations. This is
not necessary. Usually there are a small number of leading institutions that act as market leaders,
25For example, exchange LIBOR payments for Federal Fund payments.26 In investment banking there is high mobility of labor, so knowledge is readily di¤used. There are also industry
publications and conferences resulting in the dissemination of knowledge.
28
each being able to capture some rent over some �nite period. The analysis can incorporate this
possibility. The analysis readily extends to cover the case of multiple innovations.
The analysis for credit default swaps is similar, so we omit the details.
5.2 Pricing Algorithms
This example is quite di¤erent in nature from the previous example. The algorithm could be for
pricing of options using simulation27 or it could be a risk model. For example, the RiskMetrics
algorithm �rst developed by J. P. Morgan for risk management or the risk model developed by
Lehman Brother for analyzing the risk characteristics of �xed income portfolios -see Naldi, Chu
and Wang (2002). To analyze this type of innovation, the �rst step is to identify the objectives of
the innovation project. If the pricing algorithm is developed to be part of a package of algorithms,
it is hoped that it increases the value added, (�), to end users. If the pricing algorithm is developed
to speed up pricing for, say, risk management, it helps to lower the cost per unit (c¯). In both
cases the option for further development may be non-existent. Patenting in either case provides a
barrier preventing competitors copying the innovation. The economic analysis is straight forward
in theory, if not in practice.
For the case of the algorithm being a risk model, we �rst consider the J. P. Morgan case.
Here the initial motivation for the innovation is dictated by the need to meet Basel I regulatory
requirements. It could either purchase the necessary software or develop in-house. The advantage
of a leading institution developing in-house is the �exibility it allows to incorporate new structures
into a risk management system28. Viewed in isolation, J. P. Morgan would have bene�tted from
obtaining a patent, as other institutions would be forced to bear the full costs of development:
PV0(P = 1) < PV0(P = 0). However, they did not apply for a patent (this was after the State
Street Decision), instead they followed a policy of full disclosure and became a market leader.
The option for further innovation in this case is to capitalize on the development of the software
by transferring it over to a separate risk management entity. This stand alone entity generates
revenue by providing risk management consulting to other institutions and corporations. By making
the development open and becoming an industry standard enhances the value of the option. In
this case PV0[option(P = 1)] > PV0[option(P = 0)].
27Patent 6,381,586, granted to International Business Machines, prices options using importance and strati�edsampling Monte Carlo simulation.28Software vendors can be quite tardy in responding to clients requests.
29
For the case of the algorithm being a risk model for �xed income securities, it is developed
as an aid to clients who manage their �xed income portfolios relative to one of the institution�s
bond indices. The bene�t to the institution is that it lowers the costs to their clients in managing
their portfolios and it is hoped that they will continue to use the institution for trading. To be
acceptable to clients, the model must be transparent, so details are public knowledge. A patent
may prevent competitors from developing similar models. Absent a patent, the institution still
has an advantage, because it has monopolistic access to the bond index data. The options for
further innovation might entail extending to risk model to the many di¤erent indices that are used
in practice.29
To undertake a formal analysis, the institution must quantify the bene�ts of the innovation, the
value of subsequent innovations and the e¤ects of patenting on these values. While this is extremely
di¢ cult, the analysis represented by expression (21) at least provides a framework.
6 Summary and Conclusions
Following the State Street Decision recognizing that business methods can be patented, there has
been a rapid growth in the �ling of patents for �nancial innovations. Moreover, the fact that
business methods can now be patented presents �nancial institutions with an option to obtain
patent protection for di¤erent types of �nancial innovations. However, the literature is only just
beginning to explore whether patent protection for �nancial innovations is warranted; this is an
important issue because �nancial innovations di¤er from innovations in other areas due to certain
market and regulation related aspects that are unique to the �nancial industry. In particular, public
exposure of innovations, ease of imitation, importance of educating end users, and leveraging on
the participation of other market makers to reduce the costs of adoption are some of the features
that are especially important in the industry.
We undertake a systematic review of the major characteristics of recent �nancial innovations.
We �nd that certain types of �nancial innovations have been successful because they were not
protected by patents and imitation was allowed. We identify the characteristics of such innovations
through a simple but illuminating model. Our analysis highlights the real options for subsequent
innovations and market expansion that are embedded in certain types of �nancial innovations as
the primary determinant of whether patent protection is warranted; furthermore, we also examine
29Many of these indices are designed to meet particular needs of a client.
30
the role of licensing in �nancial innovations. We then apply this model to a number of case studies
to show that it provides a framework to address the di¤erent characteristics of �nancial innovations
in order to reach a decision about the appropriate patenting decision.
To our knowledge, our paper is among the �rst analyses to systematically examine the determi-
nants of long-term pro�ts from �nancial innovation. This analysis is not only central to the optimal
patenting decision, but is also of independent interest as a model that characterizes �nancial inno-
vations in an industry equilibrium. Thus, our framework can be extended in a variety of directions
to examine the sources of competitive advantage in this important industry.
31
Appendix
Proof of Proposition 1 Put � � ��(���c)c0 : We want to �nd a set of additional conditions under
which L � ��2(P = 1) � ��2(P = 0) > 0: Let, � �h(�� � c)� c0M0
2
iM0: Then, under the conditions
annunciated in the Proposition, we can write
L =c0�2
2� c0�M0 � �
=c0
2[(� �M0)
2 �M20 � 2�=c0]
Let us �nd the critical value of � for which L = 0.
� �M0 = (+=�)(M20 + 2�=c
0)1=2
Now
M20 + 2(�=c
0) = M20 +
2
c0
�(�� � c)� c
0M0
2
�M0
=2
c0(�� � c)M0
Hence if,
� �M0 > [2
c0(�� � c)M0]
1=2; (22)
then L > 0: Thus, if M0 is su¢ ciently small then (22) will be satis�ed.
Proof of Corollary 1 In this parametric range, we have internal solutions to the pro�t maximizing
output condition in both cases: P = 0 or P = 1: This implies that
��2(P = 0) =1
2c0[�� (�� � c
¯)]2 + (�� � c
¯)M0
��2(P = 1) =1
2c0[�� (�� � c
¯)]2 (23)
Given that �� �c¯, then the result follows.
Proof of Proposition 2 Given the conditions of the Proposition, then as � > �c, if follows that
M1 � (�+ �c)=�c. This implies an interior solution, so that ��1(P = 1) = (�+ �c)2=�c. Hence
Pre the State Street DecisionYear Total 705/35 36 37 38 705/4
1971 2 0 0 2 0 0
1972 2 0 0 0 1 1
1973 1 0 1 0 0 0
1974 0 0 0 0 0 0
1975 0 0 0 0 0 0
1976 1 0 1 0 0 0
1977 0 0 0 0 0 0
1978 0 0 0 0 0 0
1979 0 0 0 0 0 0
1980 2 0 0 0 2 0
1981 0 0 0 0 0 0
1982 2 0 2 0 0 0
1983 2 0 1 1 0 0
1984 0 0 0 0 0 0
1985 0 0 0 0 0 0
1986 3 0 3 0 0 0
1987 5 0 1 2 1 1
1988 13 3 3 2 4 1
1989 5 1 1 0 1 2
1990 8 0 2 4 1 1
1991 7 0 2 3 2 0
1992 11 3 4 3 0 1
1993 24 3 9 6 5 1
1994 9 1 0 5 1 2
1995 3 0 0 0 0 3
1996 10 0 1 4 1 4
1997 25 4 3 5 7 6
1998 67 16 17 13 8 13
Total 202 31 51 50 34 36
Part B
Post the State Street Decision
Year Total 705/35 36 37 38 705/4
1999 104 25 23 17 19 20
2000 134 38 29 33 16 18
2001 78 16 17 25 11 9
2002 72 15 18 22 7 10
2003 73 16 11 24 10 12
2004 26 2 5 11 2 6
Total 487 112 103 132 65 75
Table 2
Patents Awards: by Major Companies
(Excluding Financial Institutions)
1971 to 2004Company Major Finance Minor Finance Total
IBM 10 22 32
Hitachi 3 16 19
Reuters 13 0 13
Fujitsu 6 9 15
NCR 3 9 12
GE 7 1 8
Optimark 7 0 7
Casio 3 2 5
SUN 1 4 5
Xerox 2 3 5Main Finance: major classi�cation 705:/35/36/37/38/4
Minor Finance: minor classi�cation 705:/35/36/37/38/4
Table 3
Summary of Patent Awards
Patents Awards: to Financial Institutions
1971 to 2004Company Major Finance Minor Finance Total
Citigroup 15 11 26
Merrill Lynch 20 1 21
Walker Asset 3 3 6
Cantor Fitzgerald 4 0 4
Goldman Sachs 4 0 4
Thomson 2 2 4
American Express 2 1 3
College Saving Bank 3 0 3
Financial Engines 5 0 5
Foreign Exchange Transaction Services 3 0 3
Golden 1 Credit Union 3 0 3
Morgan Stanley 3 0 3
Capital One 2 0 2
Huntington Bancshares 2 1 3
Bank of America 1 0 1
Wells Fargo 1 0 0Main Finance: major classi�cation 705:/35/36/37/38/4
Minor Finance: minor classi�cation 705:/35/36/37/38/4
Table 4Types of Patents Held By Financial Institutions
Part A 705/35PatentNum-ber
Assignee Description
6,611,815 Lincoln National Life Insur-ance Co.
A data processing method for admin-istering an annuity product having aguarantee of lifetime payments.
6,799,167 Decision Analytics A bench mark portfolio is providedto be customizable to an investmentportfolio where the customization isdynamic.
6,766,303 Goldman Sachs & Co. A method for hedging a deferred com-pensation liability.
6,336,102 Wells Fargo InstitutionalTrust Company
Method and system for managementof an investment fund over a speci�edlife for that fund.
Part B 705/36
6,625,583 Goldman, Sachs, & Co. A computer implemented method forinteracting with a user of a tradingcomputer.
6,249,775 Chase Manhattan Bank, NewYork, NY
A method designed to improve analy-sis of past and future performance ofloan portfolios.
6,381,586 International Business Ma-chines Corporation
Pricing of options using importancesampling and strati�cation/QuasiMonte Carlo
6,092,056 Morgan Stanley dean Witter,New York, NY,
A data processing system and methodfor implementing and control of a �-nancial instrument which is listed fora period of time. The instrumentis based on an underlying basket ofstocks optimally selected to track anestablished capital market.
Part C 705/37
6,393,409 Morgan Stanley Dean Witter& Co.
Computer method and apparatus foroptimizing portfolios of multiple par-ticipants.
6,505,175 Goldman, Sachs & Co. A computerized order centric methodand system for tracking orders imple-mented on a trading �oor exchange.
6,269,346 Merrill Lynch A system and method for managing aplurality of stock option accounts eachfor a plurality of participants.
6,879,964 Vanguard Group, Inc. Investment company that issues aclass of conventional shares and a classof exchange traded shares in the samefund.
Part D 705/386,112,190 Citibank, N. A., New York,
NYA method and system for assimilatingdata, applying mechanisms, and emu-lating the though process of a credito¢ ce for commercial credit analysis
5,765,144 Merrill Lynch & Co., Inc A system for interactively interview-ing and educating a customer regard-ing their credit needs and the availableproducts.
5,696,907 General Electric Co. A method and system for performingrisk and credit analysis of �nancial ser-vice applications with a neural net-work.
6,405,181 NextCard, Inc. A system and method for providingreal time approval of credit over a net-work
Part E 705/45,884,274 Walker Asset Management Lt.
Partnership.A system and method for providinga foreign exchange insurance policythat determines a premium based ona number of risk factors.
5,819,230 HomeVest Financial Group,Inc.
A method and system for tracking andfunding asset purchase and insurancepolicy.
4,975,840 Lincoln National Risk Man-agement,Inc.
A method and apparatus for evaluat-ing a potential insurable risk.
5,970,464 International Business Ma-chines Corporation.
A computer implemented method ofunderwriting pro�tability analysis anddelivers the analytic process to a widecross section of insurance decisionmakers.