licensing-in technology markets: how does the strength of ...€¦ · stronger patents should lead...
TRANSCRIPT
Paper to be presented at the
35th DRUID Celebration Conference 2013, Barcelona, Spain, June 17-19
Licensing-in technology markets: How does the strength of patents
influence firm strategies and competition? Empirical evidence from the
2004 Indian Patent ReformsAnand Nandkumar
Indian School of BusinessStrategy
AbstractThere are two views of how stronger patents influence product market competition. The first is that since strongerpatents increase monopoly power in the upstream technology market, they will also increase market concentration in theassociated downstream product markets. The second is that since stronger patents decrease transaction costsassociated with licensing, they increase licensing activity, which enables even firms with no proprietary technology toenter product markets. In this paper, using the recently enacted patent reforms in India, we test these contrary views.Our empirical results reveal the intuition that on average, patenting activity increases with stronger patents more formultinationals relative to domestic firms and in disembodied markets relative to embodied markets. Consequently, weshow that competition increases with stronger patents and by more in disembodied markets, while licensing activityincreases with stronger patents and more so in disembodied markets.
Jelcodes:O34,L10
Licensing-in technology markets: How does the strength of patents influence
firm strategies and competition? Empirical evidence from the 2004 Indian
Patent Reforms
ABSTRACT
There are two views of how stronger patents influence product market competition. The first is that since stronger
patents increase monopoly power in the upstream technology market, they will also increase market concentration in
downstream product markets. The second is that since stronger patents decrease transaction costs associated with
licensing, they increase licensing activity, which enables even firms with no proprietary technology to enter product
markets. In this paper, using the recently enacted patent reforms in India, we test these contrary views. Building on
the literature on Markets for Technology (MFT), we build a model that identifies conditions under which stronger
patents can actually stimulate competition as opposed to retarding it. We show that on one hand, stronger patents
increase entry by firms that do not have the ability to produce proprietary technology because stronger patents
facilitate entry by in-licensing. On the other hand, for firms that do not have complementary capability but can
generate proprietary technology, stronger patents make them more attractive to out-licensing rather than entering
product markets. Assuming that multinationals have high technical capability and relatively low complementary
capability and that domestic firms have low technical and high complementary capability, we hypothesize that
stronger patents should facilitate licensing between multinationals and domestic firms, especially those that are
incapable of producing proprietary technology. Our empirical results reveal the intuition that on average, patenting
activity increases with stronger patents more for multinationals relative to domestic firms and in disembodied
markets relative to embodied markets. Consequently, we show that competition increases with stronger patents more
in disembodied markets, while licensing activity increases with stronger patents and more so in disembodied
markets.
1. Introduction
With the advent of international trade agreements, such as TRIPS, there has been a push towards
harmonizing Intellectual Property Regimes (IPRs) around the world. This harmonization of IPRs the
world over has been very controversial in part because there is little empirical work to guide how the
strength of IPR influences firm strategy and competition. While on one hand, the literature argues that
stronger patents should lead to less competition because stronger patents imply more market power within
firms (Nordhaus 1969), on the other hand, stronger patents should facilitate arm’s length trade in
technology (Arora, Fosfuri and Gambardella 2001; Arora and Ceccagnoli 2006) or, Markets For
Technology (MFT) and entry by firms not endowed with technological capability, leading to more
competition. In this paper, using the recently enacted patent reforms in India, we explore how stronger
IPRs influence entry, licensing, and competition.
The literature suggests that stronger IPRs grants innovators temporary monopoly power to
innovators in order to incentivize innovators to produce more innovations, which consequently leads to a
concentrated market structure (Nordhaus 1969, 1972). However, for the monopoly in technology
ownership to translate into a concentrated market structure, firms would have to be vertically integrated:
they should produce their own proprietary technology as well as produce and sell products in the product
market, and their technology strategy should be unaffected by stronger patents. This is contrary to the
view that stronger patents should also reduce transaction costs and should facilitate arm’s length licensing
of technology (Arora et al. 2001; Arora and Ceccagnoli 2006). In addition, the literature also discusses the
conditions under which technology leadership can translate into product market leadership (Teece 1987).
These insights suggest that stronger patents in theory can also facilitate vertical disintegration, trade in
technology, and encourage entry by firms particularly by those without technological capability and
consequently increase competition. In sum, the key to whether stronger patents encourage competition
depends on how it affects the technology strategies of firms. In this paper, we empirically explore how
stronger patents influence technology strategies and consequently competition. In particular, we explore
the conditions under which stronger patents can influence technology strategies of firms and encourage
entry and competition rather than discouraging it, as has often been presumed by one stream of literature.
Critical to understanding how stronger patents influence competition is whether technological and
downstream capabilities are fragmented. We consider a context in which multinationals have relatively
superior technological capability, while due to “liability of foreignness” (Zaheer 1995), domestic firms
have superior market access and downstream capabilities. In such a scenario, stronger patents should have
two effects on firms’ strategies. First, multinationals or firms with superior technical capability that do
not have complimentary capability to cater to local markets can exploit the existence of markets for
technology and license technology to others with superior complementary capability or to domestic firms.
Therefore, stronger patent protection at the margin should encourage innovation and out-licensing
especially by multinationals. In addition, stronger patent protection should facilitate greater entry by
domestic firms. Stated otherwise, it should allow firms with inferior technical capability to in-license and
enter product markets on account of having downstream capability, which in the absence of strong patent
protection would have been impossible. All of this is likely to result in higher competition especially in
embodied markets or markets in which technology can easily be licensed.
Our contribution is to clarify how firm heterogeneity, more precisely how fragmentation or
concentration of firm capabilities, such as technological and downstream capabilities, alters the effects of
IPRs on competition. Our paper thus contributes to a small but important subset of empirical literature on
how IPRs influence technology strategies of firms (Gans, Stern and Hsu, 2002). In addition, given that a
change in IPR is an instance of an institutional change, our additional contribution is to provide another
empirical instance of how institutional changes alter firm strategies, particularly technology strategy and
consequently, competition (Peng 2003,Tan and Peng 2003; Kale and Anand 2006; Majumdar 2008;
Piramal 1996; Chari and David 2012; Khanna and Palepu 2006; Khanna, Kogan and Palepu 2006).
We take advantage of the recent patent reforms instituted in India to test our hypotheses. We
compare how the strengthening of patent reforms differentially affected the strategies of different types of
firms in two types of industries: embodied and disembodied industries. To this end, we make use of a
novel dataset that comprises of patents filed with the Indian Patent Office (Indian patents, henceforth) pre
and post reforms that were enacted in the beginning of 2005.
This paper is organized as follows: the following section reviews literature and develops our
hypotheses. We then briefly provide a background of patenting in India along with a description of the
data sources used in this paper, which is followed by our empirical analysis and findings. We conclude
with a discussion of our findings.
2. Literature
Our main contribution in this paper is to investigate the conditions under which strengthening
patent protection facilitates increases rather than decreases in competition. In addition, our secondary
contribution is to highlight how an institutional change can have profound impact on technology
strategies of both domestic and multinational firms using changes to IPR as an example.
A. Strength of patents and technology strategy
Prior research argues that stronger patents facilitate arm’s length trade in technology (Teece 1986; Arora
et al. 2001). Innovations face a disclosure problem (Arrow 1962; Mowery, 1983; Williamson, 1991), and
patent protection plays a key role in innovators’ decision to license technology inputs (Arora and
Ceccagnoli 2006). In order to verify the quality of the innovation before paying for it, a potential licensee
would need the inventor to disclose the innovation. However, once the inventor discloses the invention,
the licensee would have little incentive to pay for it. Patents provide protection to the potential inventor
from this disclosure issue. As Teece, (1986) notes, out-licensing is likely when the owner of proprietary
technology lacks complementary assets, such as manufacturing and marketing, under a strong patent
regime.
Empirical support for this view has, however, been mixed. Shane (2002), using a sample of
Massachusetts Institute of Technology inventions, found that patentability of inventions increases the
likelihood that the owner of proprietary technology licenses to an incumbent. Similarly, using data from
the chemical industry in which patents are believed to be effective, Anand and Khanna (2000a, 2000b)
found that there were relatively more technology deals in that sector. Similarly, Yang and Maskus (2001)
also found that that stronger IPRs are associated with greater licensing by U.S multinationals, while Smith
(2001) determined that U.S. firms are more likely to export or directly manufacture than to license
technology in countries with weak patent regimes. Recently, using a Carnegie Mellon Survey, Arora and
Ceccagnoli (2006) exploited variation in the strength of patents between industries and empirically tested
if patent effectiveness leads to more licensing. They asserted that increases in the effectiveness of patent
protection increases licensing propensity, but only when firms lack specialized downstream assets
required to commercialize technologies. Contrary to this evidence, Cassiman and Veugelers (2002) did
not find that more effective patents encourage Belgian firms to enter into collaborative R&D
arrangements. Similarly, Fink and Maskus (2005) revealed a very weak relationship between the strength
of patents and licensing with German data. Finally, Lee (2006) demonstrated that stronger patents do not
increase arm’s length licensing transactions; while stronger patents increase royalty payments received by
U.S. firms from its affiliated subsidiaries, it had no effect on the royalty payments received from
unaffiliated parties.
We extend these studies in two ways. First, most studies focus on how patents influence
technology strategies of firms. Very few studies, if any, concentrate on how changes in technology
strategies influence entry barriers and thus competition. Our primary focus in this paper is not simply on
how the strength of patents influences technology strategies. Our focus is also on how the changes in
technology strategies driven by the strength of patents influences entry and competition. Our second
contribution to the literature is more pertinent to the research that deals with how the strength of patents
influences international licensing. We not only explore how the strength of patents in a country influences
licensing by multinationals to firms unaffiliated to it, but we also examine how international licensing
influences entry barriers and competition in the licensee’s country (host country henceforth)—using India
and the changes to India’s patent regime as an empirical context. Previous research suggests that
multinationals in general are more proficient in generating technology (Keller 2001; Eaton and Kortum
1996), while their “liability of foreignness” (Zaheer 1995) implies that domestic firms are more capable
of manufacturing and distribution of the fruits of the technology—products that are based on those
technologies. Given this setup, we explore whether strengthening patents induces more licensing by
multinationals to unaffiliated firms and its influence in lowering entry barriers and stimulating
competition.
B. Strength of patents and competition
Much of early theoretical work assumes an unambiguous relationship between the strength of patents and
the rate of innovation (Gilbert and Shapiro, 1990; Waterson, 1990; Williams, 1994). Theoretically, there
is a well-known tradeoff that is implicit while using patents to stimulate innovation. Stronger patent
protection results in static losses and dynamic gains (Nordhaus, 1962, 1969). The static losses from
stronger patent protection arises from conferring upon firms higher (temporary) monopoly power, which
comes at the cost of the consumers; since a monopolist faces a downward sloping demand curve, it would
set a price that is higher than the competitive price, which is likely to lead to welfare losses. Dynamic
gains from stronger patent protection can arise from stronger patents providing long run incentives to
innovation. By increasing appropriability of innovations, patents provide long term incentives to
innovation. Hence, patents in the long run might actually increase consumer surplus by increasing the
basket of products and services that are available for consumption. While this aspect of dynamic gains
from patenting has attracted significant empirical work in the literature, there appears to be a lack of
consensus on whether stronger patent protection stimulates domestic R&D activity in the long run. While
many studies find no effect of the strength of patents on domestic innovation (see for instance Lerner
2009; Branstetter 2001), others show that the stronger patents increases the rate of domestic innovation
(see for instance Evenson and Kanwar 2001; Chen and Puttitanun 2005).
A preponderance of research in this stream presumes that dominance in technology also translates
to dominance in product markets, an assumption that has been shown not to be true in many industries
(Gambardella and Torrisi 1998). Put differently, implicit in this argument is that most firms are vertically
integrated: firms not only produce their own proprietary technology, but they also embody capabilities to
produce and distribute products that emanate from their proprietary technology. This, as we have earlier
noted, is quite contrary to the literature that suggests that the strength of patents can shrink the vertical
boundaries of firms. While we focus on how IPR influences entry, we also explicitly test the presumption
of whether monopoly power in technology translates to market concentration in downstream product
markets. Stated otherwise, we test if the strength of patents alters the vertical boundaries of firms and its
consequences on entry barriers and competition.
C. Institution and strategy
Finally, our paper also relates to the recent literature in strategy that examines how institutional
changes influence firm strategies. Institutional changes create new challenges or provide new
opportunities and consequently alter the basis of competitive advantage. While it is true that resources and
capabilities of a firm determine its strategies, recent work suggests that firm strategies are also responses
to institutional changes or variations in the context in which they function (Wright et al. 2000; Meyer and
Peng 2005). Institutions play an interactive role by constraining or enabling a set of organizational actions
(Ingram and Silverman 2002). We hypothesize that stronger IPRs should reduce transaction costs (North
1990) and should facilitate arm’s length licensing. We show how strengthening of Intellectual Property
Regimes (IPR) influence a domestic firms' decisions to in-license technology versus developing it by
themselves. In the case of multinationals, we examine how stronger patents influence whether to
participate in product markets or in MFT. Thus, an additional contribution of this work is to show how
firms respond to changing institutions. By reducing transaction costs and facilitating exchange of
technology, stronger patents influence a firm’s technology strategy. We show how firms with superior
technical capability that do not have complimentary capability can exploit the existence of stronger patent
protection to develop and license technology to others with superior complementary capability. Likewise,
it facilitates firms with inferior technical capability to enter product markets through in-licensing.
3. Model
We assume that each firm requires two types of inputs or capabilities to enter the product market. These
inputs are technology and manufacturing, which can broadly be thought of as including other types of
complimentary capabilities such as marketing and distribution. We assume away differences between
product markets in order to develop a tractable model. We assume that there are two types of firms,
domestic and multinational firms. Multinational firms are assumed to be endowed with technology
(Keller 2001; Eaton and Kortum 1996), and they have to decide whether to participate in the product
market or the MFT. Domestic firms, however, are heterogeneous in their ability to generate technology.
We model this heterogeneity by assuming that only a proportion of them can generate proprietary
technology by investing in R&D.
We assume that all firms differ in their ability to manufacture the end product to suit local
markets. We assume that firms are homogenous in other dimensions, in the interest of analytical
tractability. We develop a model in which both the product market and the market for technology are in
equilibrium in which all firms enter at the same time.
Notation and assumptions
As stated above, firms differ in their ability in how efficiently they can produce and sell products.
In general, although efficiency is multifaceted, we assume that more efficient firms produce and sell a
higher quantity denoted by q. Thus, q is a summary measure of the differences in the ability to
manufacture the end product. We assume that 0�q�Q is a random variable that is distributed F(q), which
reflects the ability of domestic and multinational firms to produce and sell a given product.
The cost function for firms in the product market is cq, where c is the marginal cost. Ignoring
product heterogeneity, we assume that the demand for the end product D(p) is decreasing in p, the price of
the product. Domestic firms are of two types: a proportion � are assumed to be capable of being capable
of generating their own proprietary technology by incurring k, while the remaining are incapable of
generating proprietary technology.
We adopt a reduced form approach to model the technology market. Each licensor that
participates in MFT earns license revenues of L. Licensees buy technology at a price of �. We do not
model how � is determined but instead assume that � is simply assumed to be a decreasing function of the
total number of licensors, M. Stated otherwise, �(M) with �’(M)<0. All licensors incur a cost of E, which
reflect the costs of writing and enforcing contracts. We assume that stronger IPRs make it cheaper to
enforce licensing contracts. Moreover, we also assume that it is cheaper to write and enforce contracts in
markets in which technology is more amenable to be sold in disembodied form (disembodied markets,
henceforth) costs. In order to reflect these assumptions, E(�,�) reflects the cost of writing and enforcing
contracts to a licensor, where � and � denote the strength of patents and the extent to which technology
can be sold in disembodied form in a market, respectively. We assume that ���� � � and
���� � �. Moreover,
we assume that � and � are complements—in disembodied markets, stronger patents further decrease
transaction costs, or ������� � �. This assumption is critical for our analysis of how the effect of stronger
patents differs between embodied and disembodied markets. There are � potential multinationals, of
which mF enter the licensing market and nF enter the product market. Likewise, there are a total of �
domestic firms, of which mD of them enter the licensing market, nD enter the product market with self-
generated proprietary technology, and �D enter the product market by in-licensing foreign technology. We
denote the total number of product market entrants as N= nD+ nF+ �D and the total number of technology
market entrants as M = mD+ mF.
Finally, in order to understand patenting behavior, we assume that technology holders that are
also licensors patent their innovations, while only a � proportion of technology holders that are also
producers patent their innovations. This is in line with past literature that producers typically have a
variety of mechanisms to protect their innovations and patenting may not be the dominant method of
preventing misappropriation of intellectual property (Cohen et al. 2000). Thus, if M represents the total
number of licensors in an economy and NT the number of producers that are also technology holders, the
total number of patents in an economy is =M+�NT. We denote the total number of producers as N.
Profits
With these assumptions, profits of a multinational firm that participates in the product market earns a
profit of�AB C DE F ���, while a multinational firm that participates in the technology market earns
A� C � F �D�� ��. Domestic firms, those that choose to enter the product market by in-licensing earn a profit of
�� C DE F ��� F �D��. Those that decide to enter the product market by inventing their own proprietary
technology earn �B C DE F ��� F �. Finally, domestic firms that choose to license, earn �� C � F�D�� �� F �.
Domestic firm’s entry decisions
Domestic firms can enter the product market in two ways by generating proprietary technology or by in-
licensing it from a licensor. Domestic firms that generate proprietary technology will enter product
markets when they have “high” manufacturing capability. This is when (p – c)q – k > L – E – k or when
� � ��� �!. Thus, the probability that a domestic firm will enter the product market by generating
proprietary technology is just " #$ F % &��� �!'(. Similarly, the probability that a domestic firm that has
the ability to generate proprietary technology will out-license rather than use it to enter the product market
is only "% &��� �!'.
Domestic firms that do not have the capability to generate proprietary technology will enter by in-
licensing if they can make positive profits by entering the product market by in-licensing technology from
a licensor. Formally, this condition is just (p-c)q – �(M) >0 or when � � )DA� �! . Thus, the probability of a
domestic firm without the capability to generate proprietary technology entering the product market is just
D$ F "� #$ F % &)DA� �!'(.
Multinational firms’ entry decisions
Since multinationals firms are endowed with technology, such firms will enter the product market if they
have a ‘high’ level of manufacturing capability and participate in the MFT otherwise. Stated differently,
they will enter product markets when � � ��� �! or participate in the technology market otherwise. This
probability is only $ F * &��� �!'.
Market Equilibrium
Market equilibrium involves two interrelated markets: the product market and the technology market.
Equilibrium in the product market implies that the quantity supplied by producers must equal the quantity
demanded. The quantity supplied in the product market is the total quantity supplied by the participants in
the product market. This condition is given by
+DE� C , -D$ . "�#$ F % &��� �!'( . D$ F "�#$ F % &)DA� �! '(/. (1)
In the technology market, market clearing is more subtle. Any given licensor can sell as many
licenses as required. The equilibrium condition is that the total license revenues (of all technology
suppliers) should equal the total licensing payments. The former is L multiplied by the number of
licensors, and the latter is the number of licensees multiplied by the license price �(M). This condition is
given by
�, 0D$ . "�% &��� �!'1 C �D��,D$ F "�#$ F % &)DA� �! '(. (2)
To get some intuition into these market clearing conditions, Figure 1 shows how M and p are
related for each market clearing condition. The PP curve represents equilibrium in the product market,
and the TT curve represents equilibrium in the technology market. Both the PP and TT curves are
downward sloping, while the slope of the PP curve is steeper than that of its TT counterpart (all proofs are
omitted from this draft). The intersection represents market equilibrium, with p*and M
* as the equilibrium
price and technology suppliers.
The model has a number of testable predictions (The formal statements and proofs are not
provided in this draft). Here, we state the results verbally. As shown in Figure 1, the model predicts that
stronger patents reduce the cost of entering into arm’s length transactions to license technology. This
should result in an increase in the number of firms that out-license technology, M.
Lemma 1: 2A32� � �4�Stronger patent protection increases the number of technology licensors.
The model implies that an increase in � increases the probability of entry into the product market
by firms endowed with technology, as well as those that are not. An increase in � has two effects: it
decreases the license price, �, and reduces p, the product price. For producers endowed with technology,
an increase in � makes licensing less attractive: the firm indifferent between product and technology
market entry now prefers the product market. For firms without technology, the net effect is to encourage
entry, which by definition is in the product market. In sum, an increase in � should stimulate greater entry
by such firms into the product market.
Result 1: 252� � �4�A decrease in � increases product market entry.
Recall that in our setup, we had assumed that all licensors patent due to the disclosure problem
associated with technology, while only a proportion of producers patent their technology or =M+�N.
Given that both N and M are increasing in �, the total number of patents should also be increased when �
is higher. However, also recall that more multinationals have technology; in our setup, all multinationals
hold proprietary technology, but only a proportion of domestic firms hold technology. When patents are
stronger, the result is disproportionately larger increases in patenting for multinational firms relative to
domestic firms when patent protection is strengthened. Stated otherwise, if D denotes the number of
patents held by domestic firms and F the number of patents filed by multinationals, then �67�� � �68
�� . Note
that this result is merely an artifact of there being more potential multinational entrants relative to
domestic entrants, but rather because of the fact that multinationals have a greater probability of entering
technology markets relative to domestic firms.
Result 2a: 9ρ9θ � �4�Stronger patents increase the number of patents.
Result 2b:�67�� � �68
�� . Stronger patents increase the number of multinational patents to be more than
domestic patents.
The model also suggests that the license payments in a market increase with �. This is because
stronger patents increase the number of potential licensors in a market, which in turn stimulates even
domestic firms with reasonable manufacturing capability to enter the product market despite not having
the ability to generate proprietary technology. Letting A be the license payments in an industry, we
formally state this as a hypothesis below.
Result 3:9μ9θ � �4�An increase in � increases industry license payments. In-licensing is more prevalent
when patents are stronger.
Embodied vs. disembodied markets
Embodied technology in general is sticky to the inventing firm and is hard to out-license. On the contrary,
disembodied technology can be de-coupled from the inventor and can be licensed out relatively easily.
Thus, disembodied technologies are likely to have relatively lower transaction costs (Williamson 1985)
and are more amenable to licensing. Recall that in out setup, it costs more for firms to license out
embodied technology given that the licensing costs are increasing the extent to which a technology is
disembodiable reflected by �. All else equal, given that the transaction costs associated with licensing
disembodied technology are likely to be lower, there should be more licensors and hence patents in
markets in which technology is relatively disembodiable. Given that in our setup, multinationals are
technologically superior on average, more multinationals are likely to participate in markets for
technology relative to domestic firms. As a result, the number of patents held by multinationals relative to
domestic firms should be higher in markets in which technology is easily disembodiable or, 9Φ:9α � 9Φ;
9α .
Since greater number of licensors stimulates entry of even domestic firms that do not have the ability to
generate proprietary technology, disembodied markets should also see greater licensing activity. We
formally state these results below.
Lemma 2: 252� � �4�An increase in � increases product market entry.
Lemma 3: 262� � ��and
�67�� � �68
�� 4�The number of patents is higher in markets in which technology can be
sold relatively easily in a disembodied form. The difference between the number of patents held by
multinational firms and domestic firms is higher in disembodied markets than in embodied markets.
Lemma 4: 9μ9α � �4 An increase in � increases industry license payments. In-licensing is more prevalent
when in disembodied markets.
Effect of stronger patents in embodied versus disembodied markets
We had assumed that the decrease in transaction costs of licensing is in disembodied markets relative to
embodied markets. This implies that the increase in the number of licensors M with an increase in � is
higher in disembodied markets relative to embodied markets. Stated otherwise� ��A���� � �. The relatively
larger increase in M also implies that the increase in patenting should be larger in disembodied markets.
The fact that transaction costs associated with licensing technology are even lower when patents are
stronger and when the nature of technology is easily disembodiable means that there is just greater entry
into the product market by firms that are not capable of producing proprietary technology by in-licensing
technology. Thus, in disembodied markets, the increases in both licensing and product market should be
larger. Also, since in our setup, multinationals are technologically superior on average, even more
multinationals are likely to participate in markets for technology relative to domestic firms in
disembodied markets relative to embodied markets. As a result, the difference in the number of patents
held by multinationals relative of domestic firms should be higher in markets in which technology is
easily disembodiable or, 9�ρ:9α9θ � 9�ρ;9α9θ4�We formally state these as hypotheses below:
Result 4a:2�52�2� � �4 The increase in product market entry with an increase in � is more in disembodied
markets relative to embodied markets.
Result 4b: 2�ρ2�2� � � and
9�ρ:9α9θ � 9�ρ;9α9θ4 The increase in patenting with an increase in the strength of
patents is higher in disembodied markets relative to embodied markets; an increase in the strength of
patents increases the difference between the number of patents held by multinational firms and domestic
firms by more in disembodied markets than in embodied markets
Result 4c: 2�<2�2� � �4�The increase in license payments with � is more in disembodied markets relative to
embodied markets.
4. Data and variables
We tested our hypotheses using India as an empirical context. Our dataset consisted of 924
industry-year observations, which are made up of 33 industries followed over 28 years, from 1980–81
through 2007–2008.4 We assembled our dataset from five sources: the Annual Survey of Industries (ASI),
the Indian Patent Office (IPO), USPTO, and the World Bank.
First, we uses the aggregate data from the Annual Survey of Industries (ASI) from 1980–81
through 2007–08. Since the ASI is the principal source of industrial statistics in India, it has been used by
several researchers to empirically explore various facets of industrial development in India (see for
instance Hsieh and Klenow 2009). Second, the IPO, we acquired all the patents that were filed between
1974 and 2007. These amounted to 99441 Indian patents. We then classified each assignee on a patent as
domestic corporation, foreign corporation, Indian individual, or a foreign individual. In our regressions,
we combined these categories into two: Indian if the assignee was an Indian resident or foreign if the
assignee on the patent was a foreign resident. We were unable to classify 7790 patents. Third, we used the
���������������������������������������� �������������������4 The 33 industries are those we were unambiguously able to match to US SIC codes.
USPTO to develop measures on the extent to which an industry is comprised of disembodiable
technologies. To this end, we used all US utility patents granted between 1974 and 2006 from the NBER
patent database. In addition, we utilized data on US patent assignments from Google patents. This data
source comprises all patent assignments from 1974–2012. Fourth, we also used the Prowess database
compiled by Center for Monitoring of Indian Economy (CMIE) to construct industry level licensing and
R&D measures. Since this database is compiled only from 1994, we made use of this data from 1994–
2007. Finally, we also acquired data on IP cases that were heard by the Supreme Court, High Court, or the
Intellectual Property Appellate Board in India8 from India Kanoon, a publicly accessible website run by
the Michigan Law School (www.indiakanoon.org) and other court websites operated by the Ministry of
Legal Affairs, India.
Before dwelling on our measures, we provide a brief summary of the recent changes to the Indian
patent law to highlight the unique features some of which we exploit in our empirical analysis. In 1994,
the Indian government signed the Agreement on Trade Related aspects of Intellectual Property Rights
(TRIPs treaty) and committed to making it the patent law consistent with global patent laws. A decade
later, India, through the enactment of the Patent Act of 2004, completed the transition to a new patent
regime consistent with the TRIPS mandate. The 2004 law allowed the granting of product and process
patents for a term of 20 years from the date of filing. Also, the act enabled the setting up of a specialized
judiciary like that CAFC in the US to hear IP cases, by setting up the Intellectual Property Appellate
Board (IPAB). In our empirical analysis in specifications that use time variation to understand the effect
of stronger patents in industries other than chemicals and pharmaceuticals, we used a reform dummy
variable which takes a value of 1 if the focal year was after 2005, to identify the effect of patent reform.
In the case of chemicals and pharmaceuticals, the reform dummy takes a value of 1 from years 1994–
2008.
We now describe our empirical measures in detail.
Dependent variables
Patenting activity: We measured the amount of patenting activity using the number of granted patents that
were filed in an industry in a year (year of application, and log patents, henceforth). By industry and year
of application, we aggregated the number of patents filed with the IPO.
Yearly variation in applications that are eventually granted is also likely to pick up variation in
the amount of time the IPO takes to grant a patent. Given that the average time taken by the IPO is about
4 years, we restricted our sample to all granted patents from 1974–2007
���������������������������������������� ���������������������The Intellectual Property Appellate Board is a special court that adjudicated IP cases in India.
Entry: We measured the total number of entrants in an industry using the total number of active units in
an industry in a fiscal year. Since our data is a balanced panel, a change in the number of unit across years
reflects the number of net entrants in an industry in any given year. In regressions, we used the natural log
of this measure as a dependent variable.
Industry license payments: We constructed this data from the Prowess database. For every year and
industry using the Prowess dataset, we first aggregated the total license payments. Similarly, we also
aggregated the revenues for every year industry combination. Then, we scaled the total industry licensing
revenues by dividing it by industry sales.
Independent variables and controls
Strength of patents: We used two variables that proxy for the strength of patents. The first is the reform
dummy variable, which takes a value of 1 if the focal year was after 2004 for industries other than
chemicals and pharmaceuticals. In the case of chemicals and pharmaceuticals, the reform dummy takes a
value of 1 from years 1994–2008. Note that since this variable varies exclusively with time, we will not
be able to separately identify the effect of stronger patents from unobserved time variations.
Hence, as an additional robustness check, we created an alternative measure that is based on the
total number of Intellectual Property (IP) cases in an industry in a year. We calculated this variable by
aggregating the total number of IP cases by industry and year. This proxy is based on the fact that the
number of IP litigations in an industry in a year also reflects the effectiveness and importance of IP in that
industry (Lanjouw and Schankerman 2001). We use this alternative proxy to test the robustness of our
principal results.
Disembodied nature of an industry: We measured the extent to which an industry is disembodied using
the total number of US patent assignments in an industry, lagged by a year. We calculated this variable as
follows. We first acquired all of the US patent assignments from 1974–2006 from Google Patents and
mapped each assigned patent to an industry using the IPC-NIC concordance described above. We then
calculated the total number of utility patent assignments in an industry by grant year. Finally, we also
acquired the total number of utility patents issued by the USPTO in a year from the NBER patent
database from 1974–2006. Our principal measure for the extent to which technologies in an industry is
disembodiable was the proportion of utility patent assignments over the total number of utility patents for
an industry-year combination.
We also used the proportion of US utility patents held by universities in an industry in a year as
an alternative proxy for the extent to which an industry is comprised of disembodiable technologies in a
year. We constructed this measure by separately calculating the total number of utility patents held by
universities and total number of utility patents in an industry, respectively. We then calculated our
alternative measure by dividing the total number of utility patents held by universities by the total number
of utility patents in an industry, for every industry year combination.
Total licensors: Our proxy for the total number of licensors is the lagged (by one year) total number of
Indian patents filed by foreign residents (foreign Indian patents henceforth) in an industry in a year. While
our theory suggests that the total number of licensors in an industry comprises both multinationals and a
proportion of domestic firms, multinationals are likely to be more technically proficient than domestic
firms in our context. Moreover, the literature also points out that many multinationals face "liability of
foreignness" in foreign markets (Zaheer 1995), which consequently hampers their product market
performance in those markets. As a result, variation in foreign patenting is likely to capture variation in
the number of potential licensors of technology.
We also explore the robustness of our results using an alternative measure of licensing that we
constructed using ASI data. This measure is constructed from the reported non-operating expenditure,
which includes royalty payments, R&D investments, printing and stationery, and staff welfare expenses.
We rely on this measure to construct our alternative proxy as follows. We first regressed non-operating
expenses in log against proxies for R&D investments, printing and stationery, staff welfare expenses, and
managerial remuneration. To this end, we regressed non-operating expenses in log against industry
patents held with the IPO in a year in a region (lagged by a year in logs), sales revenues, total number of
employees, total supervisor staff (all in logs) along with industry, region, and year dummies. We then
calculated our proxy for license payments that varies by year, industry, and region by using the
coefficients on year, industry, and region dummies.
Instruments
We utilized an instrument that exogenously varies the number of foreign Indian patents in an industry in a
year. Our instrument is the number of patents filed by foreign corporation in the US (foreign US patents,
henceforth), lagged by a year. The literature suggests that the export intensity of a multinational firm is
inversely related to changes in domestic market opportunity (Ito and Pucik 1993). To the extent that
patenting in India reflects the intent of multinationals to exploit the market opportunity in India either
directly or through licensing, variation in patenting in India should be inversely related to variation in the
extent of market opportunity outside India. Given that a preponderance of the Indian patents are filed by
US entities, foreign US patents proxies for the market opportunity in the US.
Controls
Year dummies: We also controlled for time effects using 27 time dummies, one each for every year in our
sample, 1981 through 2007. Since our licensing data starts only from 1994, we only used 13 dummies in
regressions that select industry license revenues as a dependent variable.
Fiscal year dummies: In regressions in which we employed the aggregate number of units in a year as a
dependent variable, we used 27 fiscal year dummies, one each for every fiscal year in our sample, instead
of the year dummies described above.
Industry fixed effects: We also controlled for industry effects using 32 industry dummies, one each for
every NIC code.
Proxy for the supply of scientific talent: We utilize the 5-year moving average number of engineering
college enrollments proxy for the supply of scientific talent in India. In regressions, we selected the
natural log of this measure as a control. However, since this variable only varies by time, we only used it
in regressions in which we did not include time or fiscal year dummies.
Proxy for the size of the market opportunity: Larger markets may attract more multinational entrants and
hence differences in patenting activity also reflect the size of the market opportunity. We controlled for
the size of the market opportunity using the gross domestic product (GDP) per person employed. This
variable is the GDP divided by the number of people employed in India. As with engineering college
enrollments, this variable also varies only by time. We hence only used it in regressions in which we did
not include time or fiscal year dummies.
Empirical analysis
We started with providing evidence for our hypotheses via simple comparison of means. In order to test if
the effects were different in disembodied industries versus embodied industries, we split our sample into
two based on whether the level of disembodiment was higher or lower than the average level of
disembodiment in the sample. For the purposes of this analysis, we classify an industry in “high”
disembodiment if the proportion of patent assignments in that industry was greater than the mean
proportion of patent assignments or embodied otherwise.
Patenting activity
We started by exploring how the number of patents changed after the patent reform. We had hypothesized
that patenting activity should increase with stronger patents and also that they should increase more in
disembodied markets relative to embodied markets (Results 2a and 4b). We had also hypothesized that an
increase in the strength of patents should increase multinational patenting by more relative to domestic
patenting (Result 2b).
In order to test these hypotheses, log Indian patents as a dependent variable, we implemented a
set of OLS regressions to explore the effect of reform on patenting activity in Table 4. We started with
using reform dummy to test the effect of reform in Specification 1. We also included the proportion of
U.S patents that were assigned in an industry, which is our proxy to measure the extent to which an
industry is disembodied. In addition, we also controlled for industry effects using 33 industry dummies.
In Specification 2, in order to test Result 4b, we compared the proportion of US patent assignments with
the reform dummy. In Specification 3, given that our main independent variable, the reform dummy, does
not vary by time, we additionally include time-varying macro-economic variables, such as the amount of
scientific talent available in India (in natural log) and 5-year average federal education spending in logs.
Results of Specification 1 suggest that in the period before the reform, the number of granted
patents filed in an industry in a year was about 67% lower on average. Further, an industry with one
standard deviation higher assignments is likely to see 10.1% higher patenting activity. Specification 2
suggests that the increase in patenting activity was greater in industries that had a greater proportion of
US patent assignments. Results suggest that in industries with one standard deviation higher proportion of
US patent assignments, the increase in patent activity was about 5% more. Specification 3 suggests that
the inclusion of macro-economic variables do not alter our principal results by much.
In Specification 4, we use a time-varying proxy for the strength of patents; we used the number of
Intellectual Property (IP) litigations by industry and year instead of the reform dummy. This variable, as
described earlier, varies both by year and industry, which enables us to include 28-year fixed effects.
Results of Specification 4 suggests that our results hold up even when we control for unobserved time
effects that are qualitative, similar to those of Specification 2. In Specification 5, we tested if stronger
patents increase multinational patenting more than domestic patents (Result 2b) and whether the
difference between multinational patenting and domestic patenting is more in disembodied industries
relative to embodied industries (Result 4b). To this end, we ran two regressions: one for which the
number of patents held by foreign corporations was the dependent variable and the other in which the
number of patents held by Indian corporations (both in log) was the dependent variable. Results of
Specification 5 suggests that doubling the number of IP cases increases patenting activity by foreign
corporations by 12%, whereas a similar increase augments domestic patenting activity only by 2%. This
is in support of Result 2b, which suggests that an increase in the strength of patents increases
multinational patenting more than other factors (difference 10%; std. err 0.6; p-value – 0.09). Moreover,
the foreign assignees patent an order of magnitude more than domestic entities in disembodied industries
relative to embodied industries. Specification 5 suggests that a standard deviation increase in both the
proportion of US patent assignments along with doubling the number of IP cases increases foreign
patenting by 5%, whereas a similar increase amplifies domestic patenting by only 1.8% (difference 3.2%;
std. err. 0.01; p-value-0.00). This is in support of Result 4b.
Entry
Next, we explored how entry changed after the patent reform in Table 5. Our hypotheses suggested that
stronger patents should increase entry in the product markets (Result 1) and more so in disembodied
markets relative to embodied markets (Result 4a). Also, we had also hypothesized that disembodied
markets should on an average be more competitive than embodied markets.
Recall that we utilized a proxy, the total number of entrants in an industry using the total number
of active units in an industry (in natural log), in a year from the ASI data. As before, we started with using
the reform dummy to test the effect of reform in Specification 1 and also included the proportion of U.S
patents that were assigned in an industry: our proxy to measure the extent to which an industry is
disembodied. In addition, we also controlled for industry effects using 33 industry dummies. In
Specification 2, we tested Result 4a by interacting the proportion of US patent assignments with the
reform dummy. In Specification 3, we used an alternative proxy for the strength of patents: the number of
foreign Indian patents (in log) as our dependent variable of interest. We had shown earlier that the
numbers of foreign patents are higher when patents are strong. Since the reform dummy only varies by
year, we employed the number of foreign Indian patents, which varies both by industry and year.
However, since the number of foreign patents may also respond to demand for technology, and is
potentially endogenous, we instrumented the number foreign Indian patents with the number of patents
filed by the number of US patents held by non-US corporations in an industry, in a year. Our results are
qualitatively similar even when we used the reform dummy instead of foreign Indian patents. In
Specification 4, we tested Result 4a by interacting the number of foreign Indian patents with the
proportion of US patent assignments in an industry. Since this variable might also be endogenous, we
used the interaction term of the number of US patents held by non-US corporations with the proportion of
US patent assignments in an industry as an instrument. In Specifications 3 and 4, we controlled for time
effects using fiscal year dummies. Since our dependent variable is aggregated by fiscal year, we preferred
to implement fiscal year dummies instead of time dummies to control for unobserved time variations.
Results of specification 1 suggest that the reform increased the amount of entry by about 27% on
average. Further, an industry with 1 standard deviation higher assignments is likely to see a 10.1% higher
entry, suggesting that disembodied industries in general have higher entry rates. Specification 2 suggests
that entry rates were higher in industries which had a greater proportion of US patent assignments.
Results suggest that in industries with one standard deviation higher proportion of US patent assignments,
the increase in entry was about 3.7% more. Specification 3 suggests that entry rates are higher in
industries in which there is more patenting by foreign assignees—a 10% increase in foreign patenting is
associated with about a 4.1% increase in entry. Also, industries with one standard deviation higher
assignments are likely to see about 15% higher entry. Specification 4 suggests that entry is even higher in
industries with both a higher number of foreign Indian patents and a higher proportion of U.S patent
assignments; in industries with one standard deviation higher proportion of patent assignments, a 10%
increase in the number of foreign Indian patents increases entry by about 11.1% more. All of these are in
support of Results 1 and 4a.
License revenues
Next, we explored how the industry license revenues changed after the patent reform in Table 6. Our
hypotheses suggested that license revenues should increase with stronger patents (Result 3) and more so
in disembodied markets relative to embodied markets (Result 4c). Also, we had hypothesized that the
license payments should be higher in disembodied markets relative to embodied markets.
Using our proxy for license revenues, we assessed these hypotheses. As before, we started with
using the reform dummy to test the effect of reform in Specification 1 along with the proportion of U.S
patents that were assigned in an industry and 33 industry dummies. In Specification 2, we tested Result 4c
by interacting the proportion of US patent assignments with the reform dummy. In Specification 3, we
assessed the number of foreign Indian patents (in log) and instrumented it with the number of patents filed
by the number of US patents held by non-US corporations in an industry, in a year. In Specification 4, we
test Result 4c by interacting the number of foreign Indian patents with the proportion of US patent
assignments in an industry and instrumented this interaction term with another interaction term of the
number of US patents held by non-US corporations with the proportion of US patent assignments in an
industry in a year.
Results of Specification 1 suggest that the reform increased industry level license payments by
about 1.09 times on average. Further, an industry with 1 standard deviation higher assignments is likely to
see a 24% increase in license payments, indicating that disembodied industries in general have higher
license payments as well. Specification 2 suggests that the increase in industry level license payments
with stronger patents was higher in industries that had a greater proportion of US patent assignments.
Results propose that in industries with one standard deviation higher proportion of US patent
assignments, the increase in patent activity was about 29% more. Specification 3 suggests that license
payments were higher in industries in which there is more patenting by foreign assignees; a 10% increase
in foreign patenting is associated with about a 7.3% increase in license payments. Specification 4 suggests
that license payments are even higher in industries with both a higher number foreign Indian patents and a
higher proportion of U.S patent assignments. In industries with one standard deviation higher proportion
of patent assignments, a 10% increase in the number of foreign Indian patents increases entry by about
17% more. All of these are in support of Results 3 and 4c.
Discussion
In this paper, we explored how the recently concluded patent reforms in India, which strengthened
patents, affected entry and innovative activity in India. There is a lack of both theoretical and empirical
consensus on how strengthening patents influences competition. While one stream of the literature argues
that stronger patents which increase concentration in the technology market will also likely increase
product market concentration (Nordhaus 1969), another body of literature asserts that stronger patents
will likely facilitate licensing, which by encouraging entry, will lead to less concentration in product
markets (Teece 1986; Arora et al. 2001). Thus in essence, through a stylized model, we hypothesized that
whether stronger patents increase concentration in downstream markets depends on the extent to which it
alters the technology strategy of firms. When capabilities are fragmented such that only a few firms can
produce proprietary technology of their own, stronger patents may promote specialization and licensing
which, in turn, will likely lead to more entry and less concentration. In our case, since a preponderance of
domestic firms do not have the ability to produce technology of their own, while multinationals typically
possess that capability, stronger patents increases incentives for specialization, resulting in multinationals
specializing as technology suppliers and domestic firms specializing in commercializing technology.
Our empirical results showed that stronger patents in India increased multinational patenting to
more than domestic patenting, which reflected both the ability to produce proprietary technology and the
intent to license out technology. Moreover, stronger patents increased entry and licensing as evidenced by
higher entry and industry license payments after reform. All of these effects were also more pronounced
in disembodied industries, in which technology can be separated with ease relative to embodied
industries. Therefore, the broader strategy literature suggests stronger patents can create opportunities for
licensing and facilitates MFT (Arora et al. 2001); Teece 1986; (Gans et al. 2002). In essence, our results
speak to the broader technology strategy literature that identifies a set of conditions under which stronger
patents can actually facilitate entry rather than retard it. By identifying conditions under which stronger
patents facilitate licensing, our results in part, also explain the mixed evidence in the empirical literature
by highlighting that the effect of stronger patents on competition is contingent upon the extent to which
upstream and downstream capabilities are concentrated or fragmented in that industry.
Our results thus have significant managerial implications. An obvious implication of our results is
that for firms with significant downstream capability, stronger patents is not likely to be a threat, but
rather beneficial because with stronger patents, such firms are likely to have better access to cutting edge
technology. For firms with ability to produce proprietary technology, stronger patents are likely to
increase the number of options to monetize technology; by facilitating out-licensing, firms can just out-
license their technology to domestic firms rather than embed the technology into a product.
As with most work, ours also has limitations. Our main limitation is imposed on us by the nature
of the data. First, our results are based on proxies to measure the strength of patents. However, our results
appear robust to different measures of the strength of patents, which provides us with confidence in our
results. Also, our results are based on data from a single country. However, whether our results are driven
by the idiosyncratic country context that this study is based on at this point is unclear. Nonetheless, we
believe the paper does contribute in a novel way to understanding the possible effects of strengthening
patent protection in a country that hitherto had a weak IPR by especially highlighting the conditions under
which stronger patents can increase competition rather than decrease it.
References (incomplete) Anand, B. N., and T. Khanna. 2000a. “The Structure of Licensing Contracts.” The Journal of Industrial
Economics 48 (1): 103–135.
———. 2000b. “Do Firms Learn to Create Value? The Case of Alliances.” Strategic Management
Journal 21 (3): 295–315.
Arora, A., and M. Ceccagnoli. 2006. “Patent Protection, Complementary Assets, and Firms’ Incentives
for Technology Licensing.” Management Science 52 (2): 293–308.
Arora, A., M. Ceccagnoli, and W. M. Cohen. 2008. “R&D and the Patent Premium.” International
Journal of Industrial Organization 26 (5): 1153–1179.
Arora, A., A. Fosfuri, and A. Gambardella. 2001. “Markets for Technology and Their Implications for
Corporate Strategy.” Industrial and Corporate Change 10 (2): 419–451.
Arora, A., and A. Gambardella. 1994. “The Changing Technology of Technological Change: General and
Abstract Knowledge and the Division of Innovative Labour.” Research Policy 23 (5): 523–532.
Arrow, K. 1962. “Economic Welfare and the Allocation of Resources for Invention.” In The Rate and
Direction of Inventive Activity: Economic and Social Factors, 609–626. Nber.
http://www.nber.org/chapters/c2144.pdf.
Arrow, K. J. 1962. “The Economic Implications of Learning by Doing.” The Review of Economic Studies:
155–173.
Bertrand, M., P. Mehta, and S. Mullainathan. 2000. Ferreting Out Tunneling: An Application to Indian
Business Groups. National Bureau of Economic Research. http://www.nber.org/papers/w7952.
Cassiman, B., and R. Veugelers. 2002. “R&D Cooperation and Spillovers: Some Empirical Evidence
from Belgium.” The American Economic Review 92 (4): 1169–1184.
———. 2006. “In Search of Complementarity in Innovation Strategy: Internal R&D and External
Knowledge Acquisition.” Management Science 52 (1): 68–82.
Chari, Murali D. R., and P. David. 2012. “Sustaining Superior Performance in an Emerging Economy:
An Empirical Test in the Indian Context.” Strategic Management Journal 33 (2): 217–229.
doi:10.1002/smj.949.
Chen, Y., and T. Puttitanun. 2005. “Intellectual Property Rights and Innovation in Developing
Countries.” Journal of Development Economics 78 (2): 474–493.
Eaton, J., and S. Kortum. 1996. “Trade in Ideas Patenting and Productivity in the OECD.” Journal of
International Economics 40 (3–4) (May): 251–278. doi:10.1016/0022-1996(95)01407-1.
Evenson, Robert E., and S. Kanwar. 2001. “Does Intellectual Property Protection Spur Technological
Change?” SSRN eLibrary (June). http://papers.ssrn.com/sol3/papers.cfm?abstract_id=275322.
Fink, C., and K. E. Maskus. 2005. Intellectual Property and Development: Lessons from Recent
Economic Research. World Bank. http://books.google.co.in/books?hl=en&lr=&id=c--
C5LYu0_YC&oi=fnd&pg=PR5&dq=Fink+(2005)+patents+strength+german&ots=WJeLUc7s8Z
&sig=h2Cx6s0kj4dlB7ppkAwSQpc2j1w.
Fisman, R., and T. Khanna. 2004. “Facilitating Development: The Role of Business Groups.” World
Development 32 (4): 609–628.
Gambardella, A., and M. Giarratana. 2008. “General Technologies, Product Market Fragmentation, and
Markets for Technology: Evidence from the Software Security Industry.” Unpublished
Manuscript. http://www2.druid.dk/conferences/viewpaper.php?id=2875&cf=28.
Gambardella, A., and S. Torrisi. 1998. “Does Technological Convergence Imply Convergence in
Markets? Evidence from the Electronics Industry.” Research Policy 27 (5): 445–463.
Gans, J. S, D. H. Hsu, and S. Stern. 2002. “When Does Start-up Innovation Spur the Gale of Creative
Destruction?” RAND Journal of Economics 33: 571–586.
Gilbert, R., and C. Shapiro. 1990. “Optimal Patent Length and Breadth.” The RAND Journal of
Economics 21 (1) (April 1): 106–112. doi:10.2307/2555497.
���������A�BC�D�E�FE�F����E���E��
Spec. 1 Spec. 2 Spec. 3 Spec. 4 Spec. 5
All patents
All
patents
All
patents All patents
Foreign
patents
Indian
patents
Reform dummy 0.67 ***
0.55 ***
0.51 ***
(0.06) (0.10) (0.16)
Log total IP cases 0.12 **
0.12 *
0.02 *
(0.05) (0.06) (0.01)
Proportion of US patent assignments 0.11 ***
0.10 ***
0.12 ***
0.09 **
0.09 **
0.06
(0.02) (0.02) (0.03) (0.04) (0.04) (0.06)
Prop. Assignments X reform 0.03 ***
0.02 **
(0.01) (0.01)
Prop. Assignments X Log total cases 0.02 ***
0.03 ***
0.01
(0.00) (0.00) (0.01)
GDP per person employed 2.65 ***
(0.24)
Log engineering enrollments 0.49 ***
(0.10)
Constant -0.51 ***
-0.49 ***
-1.62 *
-1.20 ***
-0.96 ***
-0.49 ***
(0.14) (0.14) (0.91) -(0.28) (0.49) -(0.19)
N 924 924 924 924 924 924
Adj. R2 0.91 0.91 0.94 0.91 0.96 0.89
Industry fixed effects(33) Y Y Y Y Y Y
Year fixed effects (28) N N N Y Y Y Notes:* Sig. at 10% level; ** Sig. at 5% level; ***Sig. at 1% level. Standard errors in parentheses
�
���������A�BC��FE����
Spec. 1 Spec. 2 Spec. 3 Spec. 4
Reform dummy 0.27 ***
0.21 ***
(0.04)
(0.06)
Log foreign assignee Indian patents
0.41 ����
0.35 ����
(0.04)
�(0.05)
�
Proportion of US patent assignments 0.06 ***
0.06 ***
0.09 ����
0.07 ����
(0.01)
(0.01)
(0.01) �
(0.02) �
Prop. Assignments X reform
0.02 **
� �
(0.01)
� �
Prop. Assignments X for. Assignee Indian patents �
0.06 ����
�(0.01)
�
Constant 11.79 ***
11.81 ***
6.25 ����
6.16 ����
(0.09) -(0.09) (0.11) (0.15)
N 924 924 924 924
Adj. R2 0.93 0.93 0.94
Centered R2 0.91
Industry fixed effects(33) Y Y Y ��
Year fixed effects (28) N N N Y
F-statistic 266.14
Stock and Yogo (2005) critical values 16.38 19.93 Notes:* Sig. at 10% level; ** Sig. at 5% level; ***Sig. at 1% level. Standard errors in parentheses
���������A�BC��F���E�������������F���D����FE��
Spec. 1 Spec. 2 Spec. 3 Spec. 4
Reform dummy 1.09 ***
1.67 ***
(0.25)
(0.35)
Log foreign assignee Indian patents
0.73 ����
1.02 ����
(0.18)
�(0.24)
�
Proportion of US patent assignments 0.13 ***
0.10 ***
0.16 ����
0.10 ���
(0.01)
(0.01)
(0.02) �
(0.05) �
Prop. Assignments X reform
0.18 ***
� �
(0.07)
� �
Prop. Assignments X for. Assignee Indian patents �
0.09 ���
�(0.05)
�
Log(sales) 0.22 ***
0.21 ***
0.20 ����
0.18 ����
(0.02)
(0.02)
(0.02) �
(0.03) �
Constant 1.90 **
1.28 ��
1.04 ��
-1.41 ����
(0.73) (0.73) (0.71) (0.46)
N 462 462 462 462
Adj. R2 0.84 0.85
Centered R2 0.89 0.90
Industry fixed effects(23) Y Y Y ��
Year fixed effects (13) N N Y Y
F-statistic 28.65 15.30, 8.22
Stock and Yogo (2005) critical values 16.38 19.93 Notes:* Sig. at 10% level; ** Sig. at 5% level; ***Sig. at 1% level. Standard errors in parentheses
� �
�
Figures:
Figure 1 Equilibrium
�
Figure 2 Patenting activity over time
�
M
pp*p**
M*
M**
�
����
����
����
����
����
����
ABC A�� A�C AA� AAC ���� ���C ���
DEF��������������
���������������
�E���