1
Carnegie Mellon University
H. JOHN HEINZ III COLLEGE School of Public Policy and Management
DISSERTATION By
Chirantan Chatterjee
Submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy In
Public Policy and Management
Title: Intellectual Property, Incentives for Innovation and
Welfare - Evidence from the Global Pharmaceutical Industry
June 21, 2011
Accepted by the Dissertation Committee:
____________________________________ __________
Professor Lee G. Branstetter, Chair Date
____________________________________ __________
Professor Serguey Braguinsky Date
_____________________________________ __________
Professor Matthew J. Higgins Date
Approved by the ________________________________ _______
Dean Ramayya Krishnan Date
2
Intellectual Property, Incentives for
Innovation and Welfare: Evidence from
the Global Pharmaceutical Industry
Dissertation Chair: Lee G. Branstetter
Copyright 2011
By
Chirantan Chatterjee
3
Abstract
This dissertation explores the relationship between Intellectual Property
(IP), incentives for firm level innovation and societal welfare investigating their
relationship in the context of empirical essays on the global pharmaceutical
industry. In doing so, it addresses two important public policy questions. First, the
importance of location for enhancing innovation productivity in the global
pharmaceutical industry. Second, recent debates in health care markets related to
Access versus Innovation and its linkages with a nation‟s optimal patent policy.
The question of whether IP incentivizes innovation is a long debated one in
the literature on economics of innovation and technological change. The first
chapter explores this fundamental question in an emerging market context, applying
a „private returns to R&D framework‟ to the Indian bio-pharmaceutical industry. In
a fundamental policy shift, India agreed to introduce product patents for
pharmaceuticals when it signed the WTO TRIPS treaty in 19951. This policy came
into effect through enabling legislation in 2000 and final implementation in 2005.
Using this policy shift as the setting for a natural experiment, the paper estimates its
impact by using data on a panel of 315 Indian pharmaceutical firms drawn from the
years 1990 to 2005. Private returns of a firm are measured using a hedonic stock
market valuation of the tangible total assets (A) and intangible inventive assets (K).
The findings indicate an economic and statistically significant increase in private
returns to inventive activity. However, this effect appears to be highly concentrated
in the most technologically progressive Indian firms. Subsequent investigations
1 World Trade Organization‟s Trade Related Intellectual Property Agreement
4
through firm-level field case studies, patent data analysis and discussions with
industry experts reveal that IP apart, economic liberalization in India since 1991 and
the Hatch-Waxman Act in the United States have had accompanying effects in
guiding the evolution of the industry.
During the period of our analysis, a substantial number of Indian bio-
pharmaceutical firms became export intensive, with enhanced access to Western
markets. This came about aided by a rationalized currency regime through an
economic reforms process in India. The 2nd
chapter explores how export
destinations and firm capabilities influence the extent of learning by exporting
(LBE) in Indian pharmaceutical firms that exported to a variety of both advanced
and emerging destinations between 1994 and 2007. Departing from previous
studies the paper explores if exports result in other gain besides improvements in
technical efficiency. We find that LBE is not restricted to technical efficiency gains
alone but also reduces costs of production. Furthermore, exporters also gain access
to other types of knowledge that improves R&D efficiency and the rate of new
product introductions. Interestingly these gains are more especially when firms
export to high income destinations (as evidenced from higher gains when firms
export to US rather than non-US destinations). Finally, results also indicate that the
gains are higher for more capable firms.
The third chapter connects the rise of the Indian bio-pharmaceutical
producers to the global value chain in the pharmaceutical industry. Specifically, it
explores the welfare effects of early generic entry in the United States during the
period 1997 and 2008. This is the period during which, with increasing frequency,
generic drug manufacturers in the United States (many from Israel, India, North
5
America, or European Union) have been able to challenge the monopoly status of
patent-protected drugs even before the patents expire. The legal foundation for
these challenges is found in Paragraph IV of the Hatch-Waxman Act. If successful,
these Paragraph IV challenges generally lead to large market share losses for
incumbents and sharp declines in average market prices. The 3rd
chapter estimates,
for the first time, the welfare effects of accelerated generic entry via these
challenges. Using aggregate brand level sales data between 1997 and 2008 for
hypertension drugs in the U.S. we estimate demand using a nested logit model in
order to back out cumulated consumer surplus, which we find to be approximately
$270 billion. We then undertake a counterfactual analysis, removing the stream of
Paragraph IV facilitated generic products, finding a corresponding cumulated
consumer surplus of $177 billion. This implies that gains flowing to consumers as a
result of this regulatory mechanism amount to around $92 billion or about $130 per
consumer in this market. These gains come at the expense to producers who lose,
approximately, $14 billion. This suggests that net short-term social gains stands at
around $78 billion. We also demonstrate significant cross-molecular substitution
within the market and discuss the possible appropriation of consumer rents by the
insurance industry. The findings from the 3rd
chapter have implications related to
innovation policy as it pertains to pharmaceutical markets around the world..
Key Words: Intellectual Property rights; incentives for innovation; private
returns to R&D; learning by exporting; demand estimation; regulation;
pharmaceuticals.
JEL Codes: F10, F20, I11, I38, O3, O47
6
To my paternal grandfather
Late Priyalal Chakraborty
7
Acknowledgements
“My debts are large, my failures great, my shame secret and heavy;
yet when I come to ask for my good, I quake in fear lest my prayer be granted.”
~ Prose 28 from Gitanjali by Bengali poet Rabindranath Tagore
I am immeasurably grateful to my chair, Lee Branstetter, for his unflinching
support and guidance throughout my doctoral studies. It is my great privilege to be
mentored by him in the Zvi Griliches tradition of economic thought. Going forward,
I am hoping to carry forward this rich legacy with humility. To Matthew Higgins,
my member of the committee from Georgia Institute of Technology, here is a high
five from Pittsburgh to Atlanta. Not only am I indebted to him for access to the data
in the final chapter of the dissertation and his many insights on the pharmaceutical
industry, I am also very thankful to him for opening up the Higgins household
during my research trips to Atlanta. Serguey Braguinsky, my 3rd
member of the
committee has been very kind with his support whenever I needed it and I am very
grateful to him as well. I would also like to acknowledge here the mentorship of
Ashish Arora during my initial years at the Heinz College.
This dissertation stands on the shoulders of giants, many of whom, from
academia or industry, have provided me invaluable inputs in my work. In particular,
I would like to take this opportunity to convey my immense gratitude to faculty
members and staff here at Heinz College, Carnegie Mellon for their support and
mentorship through all these years. In this regard, I am particularly reminded today
of all my professors in Pittsburgh‟s scholarly community – namely Shamena
Anwar, Edward Barr, Sourav Bhattacharya, Al Blumstein, Jonathan Caulkins,
8
Karen Clay, George Duncan, Maria Marta Ferreyra, Salavat Gabdrakhmanov,
Esther Gal-or, Martin Gaynor, Limor Golan, David Greenstreet, David Hounshell,
Steven Klepper, Brian Kovak, David Krackhardt, Ramayya Krishnan, Rob Lowe,
Irina Murtazashvili, Daniel Neill, Rema Padman, Denise Rousseau, Mel Stephens,
Michael Smith, Lowell Taylor, Rahul Telang, Francisco Veloso, William Vogt and
Ellerie Weber among others.
A scholarly journey, I believe, is always incomplete without the mentorship
of teachers and professors from one‟s past educational career. In this regard, I
would like to acknowledge my school teachers from St. Xavier‟s School, Durgapur,
specifically, Mr. Suvro Chatterjee, Ms Anita Mukherjee, Ms. Nandini Mukherjee,
Mr Parameshwaran, Ms. Malini Ramdas, Mr. Uday Ray, and Father Wavreil. From
my under graduate education in IIT Roorkee, I particularly would like to
acknowledge the guidance of Dr. Chander Mohan, Dr. K G Rangaraju, Dr. P K
Swamee and Dr. R S Tiwari. I am also grateful to Dr Biju Paul Abraham, Dr
Amitava Bose, Dr Leena Chatterjee, Dr Raghabendra Chattopadhyay, Dr. Sudip
Chaudhuri, Dr Rajneesh Das, Dr Nilanjan Ghosh, Dr Vidyanand Jha, Dr Alok Ray,
Dr Sougata Ray, Dr Runa Sarkar, Dr Anindya Sen and Dr Anup Sinha from IIM
Calcutta. Friends and professional colleagues from across countries, cities, time
zones and work settings in many small ways have played an integral part in this
journey and I am indebted to them for the many moments of intelligent, joyous, and
many a times, irrelevant and irreverent conversations from my days at IIT Roorkee,
IIM Calcutta, Times of India‟s Mumbai office and Carnegie Mellon University.
A doctoral thesis work, especially for a married graduate student, is a daily
tale of the pendulum‟s simple harmonic motion. There are the good highs and the
9
bad highs, not to forget the lows. In this regard, I have no words to express my
gratitude to Anubrata, my wife, who has been a source of unwavering support
during the thesis development work. She deserves a lot for taking in the rants of a
graduate student, teaching me the true tenets of patience and keeping me motivated
through all these years. Once upon a time, when her son ventured across the oceans,
a mother was sad – Maa, hopefully will be happy today, seeing the fruition of a
long endeavor. To my father, a young man at heart, I owe immeasurably for
learning about persistence and optimism, despite all adversities in life. I would also
like to thank my extended family, sister Deblina, brothers-in-law, Arindam and
Apratim, parents-in-law, uncles and aunts. Today I also remember my maternal and
paternal grandparents – without whose blessings, this thesis would not have been
completed.
I will cherish my graduate school days with the fondness of a side pillow for
a sleeping child. Especially, all those conversations with my friends in campus, who
came before me and left, and with those, who continue on their journey forward.
Here I am, and I wish you my best and look forward to all of you joining me and
others who have trodden the scholarly path. To quote Kris Kristofferson from The
Pilgrim: Chapter 33, “From the rockin' of the cradle to the rollin' of the hearse, The
goin' up was worth the comin' down”.
All usual disclaimers apply.
10
Contents
A. Introduction…………………………………………………………..15-19
B. Chapter 1: Fundamental Patent Reform and the Private Returns to
R&D: The Case of Indian Pharmaceuticals
1. Introduction…………………………………………………………22-23
2. Background………………..……………………………………......23-32
3. Theory on Private Returns to R&D………………………..…..…...32-37
3.1 Why Private Returns & Empirical Strategy…………....32-36
3.2 Identification Strategy with Period Dummies………….36-37
. 4. Data…………………………………………………………………38-40
4.1 Firm Data…………………………………………………..38
4.2 Knowledge Capital Data…………………….………….38-39
4.3 Other Data...................................................…...............40-41
5. Construction of Variables………………...………………………...41-42
6. Results and Discussion………………………..…..………………..43-48
6.1 Period trends in shadow …………………………... ......43-44
6.2 Trends in shadow with different depreciation.. ………..45-46
6.3 Changes in shadow in a non-linear model & σ≠1. ………...47
6.4 Shadow & previous literature..…………………………….48
7. Conclusion and Extensions…………………………………………49-52
8. References.………………………………………………………... 52-56
9. Appendix………………………………………………………..… 57-96
11
C. Chapter 2: Absorptive Capacity, Firm Capabilities & Destination in
Learning by Exporting in Indian Pharmaceutical Producers
1. Introduction………………………..………………………….…...99-101
2. Literature and Background………..….………..……………..…..101-105
3. Conceptual Framework and Hypotheses….….………………….106-109
3.1 Exports and Learning………………………………..106-107
3.2 Export Destination and Learning……………………107-108
3.3 Absorptive Capacity, Firm Capabilities & Learning..108-109
4. Data and Variable Construction........……….………………..…. 110-117
4.1 Firm Level Data…………...........……………………110-111
4.2 Product Market Data…….…………………………...111-112
4.3 Dependent Variables….…………………………………..113
4.4 Independent Variables……….………………………114-117
5. Empirical Model and Results…..……………………....………...117-128
6. Discussion, Extension & Conclusion…………………………… 129-131
7. References………………………………………………………. 129-135
8. Tables & Appendix…………………………………………........ 136-145
D. Chapter 3: Regulation and Welfare: Evidence from Paragraph IV
Generic Entry in the US Pharmaceutical Industry
1. Introduction…..............………………………………………….149-151
2. Regulatory environment and early generic entry………………...152-160
2.1 Hatch-Waxman and Paragraph IV challenges…….....152-159
2.2 Cross-molecular substitution ……………………… 159-160
3. Related Literature...…………..……………………………….... 160-163
12
4. Data and Methodology ……………….…………………….......163-177
4.1 Demand Estimation – Nested Logit Model……………....165-172
4.1.1 Quantities and prices…….………………… 169-170
4.1.2 Hypertension market & unconditional shares.170-171
4.1.3 Outside good share……………………………….171
4.1.4 Conditional share ……………………………171
4.1.5 Product Characteristics….……….…………..171-172
4.2 Instruments ….…..…………………………………….......173-174
4.3 Consumer Surplus (CS) & Counterfactual Analysis……....174-176
4.3.1 Real world……………….……………………..…174
4.3.2 Counterfactual world………….……………..175-176
4.4. Robustness - alternative nesting strategy ………………176-177
5. Empirical Results………………………………………………....177-186
5.1 Descriptive Statistics………….…………………………...177-178
5.2 Results………….…………………………………………179-186
5.2.1 Coefficient estimates………………..……….179-180
5.2.2 Welfare & counterfactual analysis…………..180-181
5.2.3 Robustness - alternative nesting strategies .........182
5.2.4 Welfare for whom? And at what cost? ……. .182-186
6. Discussion & Conclusion………………………………………..186-189
7. References….…………………………………………………...189-196
8. Appendix …..……………………………………………………197-198
9. Figures & Tables........................................................................199-211
E. Conclusion & Future Work..…..…………………………………..212-213
13
Figures & Tables
Chapter 12:
Figure 1: CAGR of R&D Expenditures of Indian Pharmaceutical Firms
Table 1: Intellectual Property Laws and Indian Pharmaceuticals
Table 2: Production of Bulk Drugs and Formulations
Table 3: Description of Variables
Table 4: Descriptive Statistics
Table 5: Private Returns to R & D activity – Shadow in Industry & sub-sets
Table 6: Shadow with approximation of constant returns to scale
Table 7: Shadow without approximation of constant returns to scale
Table 8: Coefficients of R & D in various economies and industries
Chapter 2:
Figure 1: An Integrated Conceptual Model
Figure 2: Results Summary
Table 1 – Cost function regressions
Table 2 – R&D regressions
Table 3 – New Product Regressions
Table 4 - Descriptive Statistics
Chapter 3:
Figure 1: Exclusivities and Innovation in Pharmaceuticals
Figure 2: ANDA Patent Certification Options for Generic Manufacturers
Figure 3: Rising Para-IV Certifications in the U.S. Pharmaceutical Market
2 Chapter 1 has some tables in Appendix II. These unnumbered tables are referred to where
appropriate, in the narrative of Appendix II.
14
Chapter 3 (Continued):
Figure 4: U.S. Revenues of Hypertension Drugs (in $ 000) 1997-2008
Figure 5: Discount Factor by Para-IV Generics in Quarter of Entry for Some
Representative Brand Name Products
Figure 6: Consumer Gains and Producer Loses in the U.S. Hypertension
Market
Figure 7: Distributions of Incumbents and Generic Entrants
Table 1: Cross Molecular Substitution Suggestions by Blue Cross Blue
Shield of Georgia
Table 2: Hypertension Molecules
Table 3: Paragraph IV Descriptives related to Hypertension Drug Products
Table 4: Overall Descriptives for Para-IV Generic Entry
Table 5: Demand Estimation: Nested Logit Specification
Table 6: Demand Estimation: Robustness Specification (Multinomial Logit)
Table 7: Drug Availability and Pricing
Table 8: Demand Estimation: Alternative Nesting Approach
Table 8.1: Demand Estimation: Robustness (International Prices and BLP-
Style Instruments)
Table 9: Comparison of Welfare Calculations
Table 10: Data Exclusivity Regime in Various Regions
15
A. Introduction
The relationship between intellectual property (IP) rights, incentives for
firm-level innovation and welfare is a well-documented rich literature within the
economics of innovation and technological change. Some of the earliest works
(Nordhaus 1969, Deardoff 1990, Chin and Grossman 1990, Scherer 1980) have
tried to understand how invention and patent protection could evolve dynamically
across the world with implications for global welfare. Of particular interest for
economists in this regard have been the effects of a global drive in harmonizing IP
and its concomitant effects in R&D intensive, yet socially important sectors like
pharmaceuticals. Many countries in the world today, pursuant to the WTO-TRIPs
agreement of 1995, have implemented stronger patent rights. This has generated
particular concern in terms of the responses of the domestic industry where stronger
patent rights were implemented. In contrast to historical precedents of domestic
firm responses to stronger IP in other nations like Italy, Japan or Korea, it has
seemed that in the last two decades, the Indian evolutionary story of its
biopharmaceutical sector has come as a counter-intuitive puzzle for scholars. This
dissertation investigates this puzzle in the first two chapters of the dissertation.
The first chapter provides empirical evidence on the relationship between IP
and firm-level incentives for innovation in the Indian bio-pharmaceutical sector. In
doing so, we adopt a „private returns to R&D‟ framework pioneered by Griliches
(1981). India, in a fundamental policy shift, agreed to introduce product patents for
pharmaceuticals when it signed the WTO TRIPS treaty in 1995. This policy came
into effect through enabling legislation in 2000 and final implementation in 2005.
16
The first chapter estimates the impact of this policy shift by using data on a panel of
315 Indian pharmaceutical firms drawn from the years 1990 to 2005. Private returns
of a firm are measured using a hedonic stock market valuation of the tangible total
assets (A) and intangible inventive assets (K). The findings indicate a statistically
significant increase in private returns to inventive activity. However, this effect
appears to be highly concentrated in the most technologically progressive Indian
firms. Subsequent investigations through firm-level field case studies, patent data
analysis and discussions with industry experts reveal that IP apart, economic
liberalization in India since 1991 and the Hatch-Waxman Act in the United States
have had an accompanying, carrot and stick effect in guiding the evolution of the
industry. In a nutshell, the dissertation‟s findings on the Indian bio-pharmaceutical
experience suggests that while stronger IP could potentially incentivize firm-level
innovation in a developing economy, it needs to be also accompanied by changes in
external and domestic markets, such that firms can market their ex-ante imitation
work, and also can profitably access export markets with greater felicity through an
economic reforms process.
The first chapter's exploration offers a natural segue to the work in the 2nd
chapter. Applying a lens from the perspective of the home economy, this chapter
investigates, whether internationalization benefits firms in developing economies by
offering them gains from exporting. Exploring this in an empirical setting of
learning by exporting (LBE), the 2nd
chapter documents some novel findings on a
sample of Indian bio-pharmaceutical producers between 1994 and 2007. During the
period of our analysis, a substantial number of Indian bio-pharmaceutical firms
became export intensive, with enhanced access to Western markets aided by a
17
rationalized currency regime through a macroeconomic reforms process in India.
Departing from previous studies, the investigation explores if, exports subsequently
resulted in other gain besides improvements in technical efficiency. We find that
LBE is not restricted to technical efficiency gains alone but also reduces costs of
production. Furthermore, exporters also gain access to other types of knowledge
that improves R&D efficiency and the rate of new product introductions.
Interestingly these gains are more especially when firms export to high income
destinations (as evidenced from higher gains when firms export to US rather than
non-US destinations). Finally, results also indicate that the gains are higher for
more capable firms.
The third chapter makes an attempt in integrating our investigations on IP
and incentives for innovation on the supply side and extends it to the demand side
in global pharmaceutical markets. In trying to do so, it particularly exploits the shift
in IP regulations pursuant to the implementation of the Hatch-Waxman Act in the
United States. The findings provide the first known evidence of short-run welfare
consequences of early entry by generic drug manufacturers (many from Israel,
India, North America, or European Union) in the United States through what is
known as the Paragraph IV mechanism outlined in the Hatch Waxman Act. If
successful, these Paragraph IV challenges generally lead to large market share
losses for incumbents and sharp declines in average market prices. Using aggregate
brand level sales data between 1997 and 2008 for hypertension drugs in the U.S. we
structurally model the utility of a consumer and estimate demand using a nested
logit model in order to back out cumulated consumer surplus, which we find to be
approximately $270 billion. We then undertake a counterfactual analysis, removing
18
the stream of Paragraph IV facilitated generic products, finding a corresponding
cumulated consumer surplus of $177 billion. This implies that gains flowing to
consumers as a result of this regulatory mechanism amount to around $92 billion or
about $130 per consumer in this market. These gains come at the expense to
producers who lose, approximately, $14 billion. This suggests that net short-term
social gains stands at around $78 billion. We also demonstrate significant cross-
molecular substitution within the market and discuss the possible appropriation of
consumer rents by the insurance industry. The findings from the 3rd chapter have
implications for innovation policy as it pertains to pharmaceutical markets in
developed and developing economies.
To summarize, this dissertation's findings have implications for both policy
makers and managers. On the one hand, it provides evidence of shifts in incentives
for firm-level innovation in developing economies with changes in IP regime, a
finding that could potentially guide the design of an optimal patent policy in the
context of both a developed and a developing economy. But as we point out, this
needs to be accompanied by associated events from both the external and domestic
market. The findings on learning by exporting point to mechanisms of gains from
exporting that could potentially accrue to firms from the developing world through
internationalization. Since these gains are conditioned, as we document, by ex-ante
capabilities of firms, the findings from the 2nd chapter also point to the crucial role
of development of absorptive capacity in firms from the emerging world ---
especially as they try to catch up to the global technological frontier. The 3rd
chapter
addresses the central issue of „Access versus Innovation‟ in pharmaceuticals and
healthcare markets of the developed world. Our findings on short run societal
19
welfare gains certainly needs to be balanced with future work on the long run
impact of relaxed IP regulations as facilitated through Hatch-Waxman Act in the
US. This would entail a careful investigation of the impact of early generic entry on
the long run incentives for innovation in pharmaceutical firms. The dissertation‟s
findings should also be generalizable to guide patent and international trade policy
in other developing economies, informing nations on how to develop socially
important R&D intensive sectors like pharmaceuticals for the betterment in the long
run of society around the world. Like most scholarly endeavors, this dissertation
intends to be motivated by its findings and pursue these open questions in future
work.
References
Deardoff Alan V., 1992: “Welfare Effects of Global Patent Protection", Economica,
59, 35-51.
Chin Judith C. and Grossman, Gene M., 1990: "Intellectual Property Rights and
North South Trade", in The Political Economy of International Trade: Essays in
Honor of Robert E. Baldwin (Ronald Jones and Anne Krueger edited), 90-107
Nordhaus, W.D., 1969: "Invention, Growth and Welfare: A Theoretical Treatment
of Technological Change", Cambridge, Mass: MIT Press.
Scherer, F. M. 1980: "The Economics of the Patent System", Chapter 16 in
Industrial Market Structure and Economic Performance, Houghton and Mifflin,
439-458.
20
B. Chapter 1: Fundamental Patent Reform
and the Private Returns to R&D - The
Case of Indian Pharmaceuticals3
3 This version co-written with Ashish Arora and Lee Branstetter was presented at the NBER Conference on Location in the Biopharmaceutical Industry in March 2008. An advanced version of
this paper containing updates from our field survey and a theoretical model, “Strong Medicine:
Patent Reform & the emergence of a Research-Driven Pharmaceutical Industry in India”, with
Ashish Arora, Lee Branstetter and Kamal Saggi is forthcoming in a NBER Book Volume on The
Location of Biopharmaceutical Activity, 2011, Iain M. Cockburn & Matthew J. Slaughter, editors.
21
Abstract4
Do private returns to inventive activity change when IPR regimes are
substantially strengthened in developing economies? Our paper investigates this
question by looking at the impact of patent reforms in India on India-based
pharmaceutical companies. In a fundamental policy shift, India agreed to introduce
product patents for pharmaceuticals when it signed the WTO TRIPS treaty in 1995.
This policy came into effect through enabling legislation in 2000 and final
implementation in 2005. We estimate the impact of this policy shift by using data
on a panel of 315 Indian pharmaceutical firms drawn from the years 1990 to 2005.
Private returns of a firm are measured using a hedonic stock market valuation of the
tangible total assets (A) and intangible inventive assets (K). The findings indicate
an economic and statistically significant increase in private returns to inventive
activity. However, this effect appears to be highly concentrated in the most
technologically progressive Indian firms.
Key Words: Patent reforms; private returns; intangible assets; market value
of innovation; Indian pharmaceuticals.
4 Acknowledgements: I would like to thank my committee members, Professor Ashish
Arora, Professor Lee Branstetter and Professor Shamena Anwar for their suggestions at various
stages in refining this research. Thanks are also due to seminar participants at the Strategy
Entrepreneurship and Technological Change Program in Carnegie Mellon, International Industrial
Organization Conference and NBER seminars for useful comments. Experts from Indian
pharmaceuticals along with Kevin Bodreau, Serguey Braguinsky, Marco Ceccagnoli, Prithwiraj
Chowdhury, Wilbur Chung, Matej Drev, Karen Eggleston, Susan Feinberg, Jeff Furman, Saibal
Ghosh, Matthew Higgins, Tarun Khanna, Steve Klepper, Kensuke Kobe, Frank Lichtenberg, Keith Maskus, Romel Mostafa, Anand Nandkumar, R Narayanswamy, Yi Qian, Subramanian Rangan,
Bhaven Sampat, Jayati Sarkar, F M Scherer, Lourdes Sosa, Scott Stern, Rik Santanu Sen, Francisco
Veloso, William B. Vogt, Jonathan Ying and Minyuan Zhao provided suggestions and useful
comments. All errors and omissions are mine.
22
1. Introduction
Do stronger laws related to intellectual property rights (IPR) and
appropriability regimes5 spur or hinder national innovative capacities? With that
debate continuing, researchers have also looked at associated issues trying to
understand how IPR laws in an economy affect firm-level changes in innovative
capacity. India provides a good experimental setting to add to that stream of
research. In the last three or four decades India as an emerging economy has seen
considerable changes in its regulatory structures and associated institutions,
especially those related to IPR. The country had a pro-product patent regime until
1972 having adopted the British Patents and Design Act of 1911 after independence
in 1947. The Indian Patent Act of 1970 implemented in 1972 shifted the country‟s
appropriability environment to a regime that was friendly towards process-patents.
These patent laws were changed after more than thirty years in 2004-2005. The
Patents Ordinance, subsequent to the 3rd
Amendment to the Patents Act followed up
on the Patents Amendment Bill 1995 and product patents were introduced in
pharmaceuticals, foods and chemicals, in India from 1st of January, 2005.
A recent paper (Chaudhuri, Goldberg, Zia 2006) argues that this
change in patent regime will cause potential adverse welfare effects in the Indian
pharmaceuticals industry. For the sub segment of the systemic anti-bacterial
segment fluoroquinolones, the authors argue that withdrawal from the domestic
markets due to stronger patents would imply a total annual welfare loss of $305 mil.
5 Appropriability regimes first were written about by Arrow (1962) and later on defined by Teece
(2000). They stand for the scope in which knowledge and innovations can be protected from
imitators. Loosely termed as, intellectual property regimes, laws related to appropriability have
tended to vary across a spectrum from being pro-process to pro-product across national economic
settings.
23
About 80% of that loss of welfare will be due to loss of consumer welfare, with
domestic producers loosing some $50 mil and foreign producers gaining $ 19.6 mil
in profits annually. We adopt a supply side approach to this issue and highlight the
increasing private returns to R & D for Indian drug firms with stronger patents. We
believe our findings provide new insights at three levels.
Firstly, temporal trends in private returns to R & D capture the
impact of patent reforms over time on the R & D activity of Indian pharmaceuticals.
The private returns to R & D increase monotonically and peak around 2005 the year
of stronger patents implementation. Secondly, one can expect older R & D (tuned
with the earlier process patent regime) to face obsolescence at a faster rate as
stronger patents are implemented. We try to econometrically capture this using
increasing depreciation rates in measuring R & D stocks which in turn implies new
R & D becoming more valuable. Such an exercise reveals increasing private returns
to R & D in the overall sample as well as in various subsets of the industry. That
early evidence of the markets positively valuing more recent R & D activities
should provide direction about shifting research capabilities of firms with stronger
patents in Indian pharmaceuticals. Finally, certain sections of the industry can be
expected to respond more favorably towards stronger patents. We develop an
understanding of that through industry level stratifications and investigate temporal
change in private returns of R & D in these various strata.
2. Background
The history of appropriability regimes in India dates back many
decades. India had adopted the British Patents and Design Act of 1911 after
independence in 1947 providing protection for drug patents (product patents) and
24
this remained in place until 1972. Such a regime stunted the growth of the domestic
pharmaceutical industry, as global pharma firms established their monopoly
presence in the domestic markets with pre-established global patents on drugs and
manufacturing processes (Chaudhuri, 2005). It was in 1970, that the Government of
India (GoI) initiated changes in the patent laws of the country bringing into effect
The Indian Patent Act of 1970. This was finally implemented in 1972. The
implementation of this Act (along with the Foreign Exchange Regulation Act of
1973 and the implementation of New Drug Policy in 1978) provided in the 1970s
an environment of comprehensive and concerted policy changes to nurture the
manufacturing and innovative capacities of the domestic industry, in particular that
of drugs and pharmaceuticals (Chaudhuri, 2005). These patent laws stayed in place
for more than thirty years. Following a series of amendments to the patent laws
beginning 1995, product patents in pharmaceuticals, foods and chemicals became
effective, in India from 1st of January, 2005. Between 1972 and 2005, processes but
not products were protected. The associated transitions for firms to a new regime
were not knee-jerk, as changes were foreseen ever since GoI had committed itself to
implement a globally consistent pro-product patents regime by signing the WTO –
TRIPS6 treaty in late 1994. It came with understandable hitches
7 taking more than a
decade to be implemented, after rounds of negotiations between various interest
6 The Uruguay round of the World Trade Organisation, featured the Trade Related Intellectual
Property Rights agreement between signatory countries covering five basic areas related to trading
systems and intellectual property laws, adequate protection for intellectual property rights,
enforcement of the rights in respective countries, dispute settlement related to IPR and transitional
arrangements related to the implementation of the new intellectual property regime. 7 The new set of Indian patent laws, its enforcement and transition has opened up debates on a variety of issues that have been investigated by researchers, especially those related to compulsory
licensing, importation clauses related to the site of invention, patentability criterion, post and pre-
grant opposition, data exclusivity and mailbox provisions for patents applied during the transition
period of 1995 to 2005 (Chaudhuri, 2005). Our study looks at the impact of this fundamental patent
reform on the Indian pharmaceuticals industry at an aggregate level over time.
25
groups ranging from domestic drug firms, small and large, the GoI, the World
Trade Organization or foreign drug firms operating in Indian markets (Refer Table
1 for a detailing of the IPR transition in India and its implications for Indian
pharmaceuticals collated from various sources).
In close to three decades of a pro-process patent regime, the Indian drugs
and pharmaceuticals industry has evolved, registering their presence vis-à-vis
foreign firms not only in home but also in some export markets mainly as generic
players. The industry also flourished on what is loosely termed as reverse
engineering. Essentially that meant firm competencies in imitative R & D and
manufacturing with favorable IPR policies playing a substantial enabling role for
firms to manufacture knock-off drugs from patented molecules by global firms. The
older set of patent laws starting from 1970 played a critical role in the development
of firm capability. Until 1972, the British Patents Act allowed a manufacturer to
patent all processes for a particular drug. The life of drug patents was also 14 years.
The 1970 Patents Act reduced the life of a patent to 5-7 years. Also, with the new
act, a manufacturer could patent only one method of production or process for a
drug product. Both of these steps ensured the tilting of scales towards bolstering the
manufacturing and R & D capabilities of domestic drugs and pharmaceuticals
industry rather than focusing on drug discovery (Chaudhuri, 2005). The results of
these policy changes showed up over time on more than one industry indicators. For
example, until early 1970 the multinational companies enjoyed 68% of the domestic
drug markets in India. Yet within a decade, by 1980, that share had come down to
50% - at par with that of domestic drug makers. That share for domestic drug
makers has only increased over subsequent years. By 2004, until when latest data
26
was available, domestic drugs and pharmaceutical firms in India had a 77% market
share in the home markets (Chaudhuri, 2005). The changes in market shares came
about as manufacturing capabilities of Indian pharmaceuticals witnessed
considerable improvements. Until the 1970s the country was dependent on imports
of essential bulk8 drugs to manufacture formulations. That meant foreign firms
ruling the roost as well as using their market power to maintain high drug prices.
After 1970 things changed. Domestic drug makers started picking up the gauntlet of
a looser patent term and relaxed patent laws related to manufacturing processes for
a drug product -- producing bulk drugs in larger quantities. Also with the growth of
the domestic bulk manufacturing industry, the formulation makers started spawning
(See Table 2). Thus around the late 1980s and 1990s the industry saw a dramatic
growth of not only old firms like Ranbaxy Ltd and Cipla Ltd, but also witnessed
new firms utilizing the entrepreneurial opportunity. With that, Indian drug makers
also started registering their presence in the global pharmaceutical markets. From
around $ 96 mil of exports in 1980-81, exports of Indian drugs and pharmaceuticals
had increased to $ 1910 mil in two decades by 2000-01 (adjusted with respective
year exchange rates, Chaudhuri, 2005). By 1988-89 India was already a net exporter
of drugs and pharmaceutical products (Source: Organisation of Pharmaceutical
Producers of India, 2004).
8 Bulk drug makers specialize in producing chemicals that go into the production of drug
formulations. Advanced intermediates or advanced bulk makers also focus on doing that at a further
specialized stage of production. Formulation makers are firms who make the final drug products in
various medicinal forms like tablets or capsules using the chemicals supplied by bulk makers and
adding excipients. Formulation makers could opt to be generic makers if they decide to use an off-patent going drug originally produced by another firm, and enter the market with a product of similar
chemical composition but a different brand name. All of these activities, bulk (also called active
pharmaceutical ingredients or APIs), advanced intermediates, or formulations could be one vertically
integrated operation within a pharmaceutical firm coupled with its R & D divisions who might
specialize not only in generic R & D but also engage in R & D for new drug molecules.
27
Table 1: Intellectual Property Laws and Indian Pharmaceuticals
Period IPR events in India Implications
From
Pre‟72 to
Post „72
British Patents and Design Act,
1911 - Patents Act 1970
• Pre 1972: A product and process patent regime; Life of
drug patents 14 years; One could patent all processes for
drug manufacturing.
• Post 1972: Product patent regime abolished, patent only
a method or a process, Life reduced to 5 – 7 years, for a
particular drug only one method or process patentable.
1994-
1995
Signing of the WTO TRIPs
treaty by India as a result of the
1986-1994 Uruguay Round of
negotiations
Dec 31st, 1994: The Patents (Amendment) Ordinance
allowing filing and handling of product patent
applications for pharmaceutical and agricultural
chemical products, as well as the granting of exclusive
marketing rights, EMRs on those products. The
Ordinance became effective on January 1, 1995.
The Patents Amendment Bill 1995 was introduced.
1996-
1997
Transition period • Indian Patent office keeps receiving product patent
applications.
Meanwhile disputes with US and EU at WTO related to
violation of product patents.
WTO asks GoI to complete institutional reform on new
IPR laws by April 1999.
1998 –
2001
India signs and ratifies Paris
convention and PCT
WTO reviews the TRIPs terms and grants an extension
to India beyond 2000 but before January 1st 2005 – the
new deadline to implement product patents.
2002-
2003
Breezy time of changes,
interest groups fighting out
granting of EMRs by IPO, City
High Courts putting up stay
orders.
Examples of disputes: Rejection of EMR for GSK‟s
Rosiglitazone and Hoffman La Roche‟s HIV drug
Squinavir, based on patent application having been filed
before 1995. Natco Pharma gets a stay order from
Chennai High Court on EMR for Novartis‟s cancer drug
Glivec – the Indian generic producers getting a safe
cushion against government enforcement.
Dec 2004
– 1st of
Jan‟
2005.
Amendments to Patents Act
before deadline of Jan 1st 2005
as set by WTO
• Product patent regime in place finally. From 1st of
Jan‟2005 a firm could also now file for a product patent
within India, and granted the same.
Source: Chaudhuri, 2005, Oxford, Analyst Reports, Thomson Scientific, World Wide Web.
28
Table 2: Production of Bulk Drugs and Formulations
As those changes unfolded, specific competencies could be identified with
firms in the Indian drugs and pharmaceuticals industry (Chaudhuri, 2005). Indian
drug makers excelled on four counts9. They were manufacturing bulks, advanced
intermediates, formulations and generics with globally acceptable standards in
imitative R & D and manufacturing processes, doing that at cheaper costs compared
to global drug firms. They were also proficient in producing drugs faster than their
global peers and managed to get them to export and domestic markets at lower
prices. These four counts, excellence in imitative manufacturing and R & D,
cheaper costs, lesser time10
and lower prices had already by 1995 ensured
considerable competitive advantages for Indian drug firms as they competed with
9 This section borrows from the book “WTO and India‟s Pharmaceuticals Industry” by Chaudhuri
(2005), Sampath (2005), and Fink (2001) along with various analyst reports from equity houses like Morgan Stanley, Mckinsey and Kotak Securities. 10 Analyst reports indicate that while for global firms, 30-40% levels of total costs are manufacturing
costs, for Indian drug firms this is at 20% levels. Lanjouw (1998) and Keayla (1998) have
investigated the dexterity of Indian drug firms to bring products into the market in increasingly
minimal amount of time.
Year Bulk Formulations
1974-75 900 4000
1979-80 2260 11500
1984-85 3650 18270
1989-90 6400 34200
1994-95 15180 79350
1999-2000 37770 158600
2003-2004 77790 276920
Source: Ministry of Chemicals and Fertilizers, Government of India Annual Report Various Issues,
Chaudhuri 2005; Figures in Indian Rupees million – at current prices.
29
global big pharmaceutical firms. By 199911
, 193 manufacturing plants in India –
41% of all complying plants from inspections in Japan, Taiwan, Korea and India -
complied with FDA regulations. In 2005, the largest number of 60 compliance
certificates of manufacturing facilities went to Indian drug firms – a number more
than drug firms in Israel, China, Italy, Taiwan, Spain and Hungary. The work
pressure for certification of drugs manufacturing facilities in India had reached such
levels that FDA, by 2006 was already contemplating the setting up of an India local
office in New Delhi, India. The prolific Indian drug makers also showed an average
time lag of around 3.5 years – this -- to pick drugs introduced in foreign markets
and introduce it into the domestic settings. This lag was as less as 2 years in some
cases like for Mebendazole which was introduced globally in 1974 and an Indian
firm introduced it domestically in just two years in 1976. Finally, cheaper drugs
from India are a common occurrence in the global pharmaceutical markets. An oft
cited example comes from the price of the triple drug combination of stavudine,
lamivudine and nevirapine – a combination in the category of ARVs or
antiretroviral drugs costing $10,000 in curing a disease like AIDS affecting more
than 7 million people globally. The United Nations along with the World Health
Organization promoted the Acceleration Access Initiative in tandem with firms like
Abbott, Boehringer, Bristol Myers, GSK, F Hoffman La Roche and Merck to steer
voluntary price controls of this ARV combination in developing countries. The
price decreased from $10000 to $ 931 in Jan‟2001 before Indian generics maker
Cipla Ltd. stepped in 2000 and introduced the combination in the global markets at
11 Based on data provided to us by the Food and Drug Administration, FDA, United States.
Compliance with FDA standards imply FDA inspectors visiting manufacturing facilities and
approving the firm‟s drug manufacturing conforming with their specifications related to
manufacturing.
30
$350. This was followed by three other Indian firms, Aurobindo, Ranbaxy and
Hetero entering the market with a combination priced at $201. The differential in
prices, some 25% of global prices, is a story not only for ARVs but also for many
other therapeutic areas like vaccines or anticancer drugs as well.
It is thus not without reason that the evolution of the Indian pharmaceutical
industry has captured the attention of the global pharmaceutical industry. A
Mckinsey report in 2005 suggested that while Indian pharmaceutical manufacturers
accounted for 2% of global pharmaceuticals markets, generating total output valued
at $ 8.8 billion - the industry grew at annual rates of around 21% during 1995-2005,
with predictions to reach an output level of $ 25bn by 2010. Analysts have also
predicted that the key opportunity for Indian pharmaceuticals lie in the area of
generics. The generics global markets is valued at $ 47-50 billion with estimated
sales from going-off patents being valued at $55-65 billion between 2003-2008 –
India being expected to capture 30% of that opportunity. The competitive position
of the Indian pharmaceuticals industry, especially its generic players however could
severely be affected with the implementation of the new patents regime. The
signing of the WTO TRIPS treaty would have meant a considerable change of
guard for the domestic drugs and pharmaceuticals industry as well as big pharma
operating in the Indian markets. This then we believe is an apt time to check how
inventive activities are being re-allocated in the light of a fundamental patent
reform for an industry which had reached a certain level of maturity in terms of its
R & D and manufacturing capabilities. The answer perhaps lies in understanding
how innovative capacities at firm level are adapting to this transition to a product
patent regime in 2005, that after a thirty-year long pro-process friendly
31
appropriability environment. That is the motivation of our research. Working on a
panel 315 Indian pharmaceutical firms from 1990 to 2005, we aim to test
empirically the effects of fundamental patent reform on the private returns to R &
D. GoI estimates point that in 2003, there were some 5877 drug manufacturing units
in India with some 300 in the organized sector being publicly traded on the bourses.
They are our focus of analysis, cumulatively accounting for more than 90% of the
market12
. For these larger firms from the more organized section of the industry, the
significant shift in the Indian patent system toward greater protection of product
innovation may induce important changes in firm strategy. It is possible that many
of these firms will seek to create new drugs, adopting a business model more like
that of established pharmaceutical firms in the West. To the extent that this is the
case, financial markets in India are likely to respond by changing the valuation of
firms‟ R&D investments. Our paper seeks to quantify the extent to which this has
occurred.
The following sections of the paper are organized thus. In the next section,
we outline the literature on market value of innovation which builds on the premise
of measuring the private returns to R & D and subsequently formalize our theory
for empirical estimation. Next, we describe our data and its felicity or limitations to
empirical investigations. We then describe the construction of our variables.
Finally, we discuss our results and conclude with the implications of our findings.
The paper also includes an appendix on the results of our investigations with other
proxies for inventive assets apart from R & D expenditure. Also included in the
12
This was shared to us in a conversation by Mr D G Shah, the Secretary General of the Indian
Pharmaceutical Alliance, the leading industry body for domestic drug firms in India.
32
appendix are a delineation of R & D in Indian drug firms, strategies for sample
stratification and a detailing of measurement of Tobin‟s Q. Further, we also
delineate a sensible heuristic to impute R & D expenditure for missing data and
outline the methodology used to create R & D stocks for the purposes of our
analysis. This apart, we conclude the appendix with a note on Indian stock markets
placing them in context with stock markets around the world.
3. Theory on Private Returns to R&D
3.1 Why private returns?
Both private firms and governments have a keen interest in measuring the
economic returns to innovative activities. This especially if one can place the
measurements in context with policy changes, aiding in interpretation of policy
effects in the economy. To measure the impact of changes in IPR regimes on
innovation investments a common approach would be to connect total factor
productivity or profit growth of a firm to measures of innovation investment
(Mairesse and Mohnen, 1995), yet this approach is not without difficulties (Hall,
1998). Firstly, the data may not cover a long enough period to enable a precise
measurement especially given the lags involved in spending on innovation and its
effects showing up on the bottom line of a firm. This could especially be true for
firms doing significant amounts of basic research or for certain industries which are
research intensive like pharmaceuticals. Secondly, even if one can wait, hiatuses of
the duration as above make measurement for the purposes of planning innovation
strategies not so useful, especially if that planning is with an aim to guide overall
firm strategies. Finally, a measurement strategy reliant on accounting data reported
by firms in their books may be subject to errors with a need for care in the
measurement and timing of associated inputs (Fisher and McGowan, 1983).
33
It is for this reason that researchers have turned to evaluating the private
returns to different types of innovation investment like R & D expenditure with firm
level performance measures derived from stock market data. This approach in
alignment with the total factor approach has in particular analyzed the relationship
between R & D expenditure as a proxy for knowledge capital and market value of a
firm as proxy for its performance measure. The literature in this area implicitly or
explicitly assumes that the stock market values the firm as a bundle of tangible and
intangible assets (Griliches [1981] and others). Thus the market value V of a firm is
a function of the set of assets that it is comprised of and looks thus:
.....)2,1(......), AAf2A,1(AV
-------------------------------------------- (1)
The functional form of f is not known and a critical assumption here is that
the markets are efficient, such that the market value of a firm truly is a reflection of
a value maximizing combination of the assets, as manifested through the functional
form. We can note here that should we assume that the firm is comprised of a single
asset A, with constant returns to scale and linear homogeneity of the profit function,
we will then arrive at a much accepted finding from literature in financial
economics; that the market value V of a firm is a multiple of the book value of the
asset A, with the multiplier equaling Tobin‟s Q.
Previous literature suggests that given the mystery of the functional
form of (1), (it is not easy to compute as well given its closed-form), one can
assume fairly ad-hoc approximations like linear or Cobb-Douglas (linear in logs).
As one proceeds with a linear combination of the assets, researchers might be
tempted to question themselves whether the assets of a firm can be treated as
34
additively separable – almost as if claiming that the firm is a sum of its parts and
those parts when unbundled out, could be sold off separately for the same price as
they are valued at when embedded in the firm. One can however defer such an
approximation as simply the first and most important terms in a more general
approximation of the true functional form of f (Hall, 1998). Following closely thus
such an approximation from literature, we set out the market value tiV , of a firm i at
a time t, with two kinds of assets, tiA , and tiK , . The tangible assets tiA , are measured
by the book value of total assets of the firm13
and the intangible assets tiK , , used as
proxy for knowledge capital, are measured by stocks of R & D expenditure of a
firm. The aim in such a setting is to capture the shadow of tiK , , the intangibles over
the tangible assets tiA , through the effect of tiK , / tiA , on the market value of the
firm. This effect is captured through the coefficient in the estimating equation, .
We can interpret the above as follows: the additive combination of both the tangible
and intangible assets assigns the firm a value – expectations of which leads a firm‟s
investor (assumed rational) price the firm i at time t hedonically at tiV , . The basic
market value equation of the firm thus assumes the form:
),*,(, tiKttiAtqtiV ------------------------------------------------------ (2)
where measures returns to scale and q is shadow of the measured market value of
the firm (as a combination of its tangibles and intangibles) over its real market
value, which in the long run should equal 114
. We can also impose the long-run
13 The construction of variables is outlined in subsequent sections and the appendix in the paper. 14
In effect this q is actually an estimate of the log of average Tobin‟s Q, where our LHS measures
market value of a firm, our RHS measures replacement cost of assets, as we explain in our appendix.
35
assumption on , claiming that it will be equal to 1 in the long run for a firm. The
basic market value equation of firm, taking natural logs on both sides and adjusting
thus reduces to:
ti,ε)ti,/Ati,(K*tβln(1tlnqti,Ati,Vln
----------------------------------------- (3)
We should note here that without the long run approximation of =1 Equation (2)
on expansion and adjustment would effectively look thus:
ti,ε))ti,/Ati,(Ktβln(1tσti,Alntσtlnqti,lnV ----------------- (4)
For the purposes of our empirical investigation, we use equation (3) as our primary
guiding specification since we find that the value of executing a few preliminary
regressions (using (4)) falling in the range of 1.03-1.08 which is close to 1 for our
purposes. The left hand side of (3), one can now note, is denoted by Tobin‟s Q for a
firm i at time t. The right – hand side of the equation is approximated for the term
within the logarithm thus: xx )1ln( for small values of x or here K/As. Hall and
others have previously suggested that an un-approximated form makes more sense
for K/A levels in the range of 15% and above, which can hardly be expected in our
case except for a couple odd firms in Indian drugs and pharmaceuticals. Thus the
estimating equation would assume the form:
ti,ε)ti,/Ati,(K*tβtlnqti,lnQ -------------------------------- (5)
36
Equation (5) is tested for K/As15
. While still at Equation (4), we do away with the
assumption of =1 and carry out our approximation. The reduced form for (4) is:
ti,ε))ti,/Ati,*(Ktβtσti,Alntσtlnqti,lnV ----------------------- (6)
Subtracting tiA , from both sides of (6) we get:
ti,ε))ti,/Ati,*(Ktβtσti,Aln*1)-tσtlnqti,lnQ ( --------------- (7)
We use pooled ordinary least squares to estimate Equation (7) including a full set of
year dummies for our estimations and introducing firm specific fixed-effects in the
pooled regressions. Further we check our estimation results with non-linear least
squares on an un-approximated specification of the form of (3) above. Our non-
linear estimation results include time dummies and first-differences to account for
firm specific unobserved heterogeneity.
3.2. Identification Strategy with Period Dummies
A key motivation for the paper is to investigate the time trends in the
shadow . This should capture the effects of a changing IPR regime on the private
returns to R & D undertaken by a firm. For our basic estimating Equation (5)
stands for the semi-elasticity of Q with respect to changes in K/As, or:
)8(
)
,
,(
),(ln
tiA
tiKtiQ
t
If the markets are efficient16
and investors are assumed to be rational one can expect
that stronger patents will give the firm with a higher K/A ratio better valuations in a
15 We also test for higher order terms of K/As and note that they are of no significance for the square
of R & D / Assets term, the coefficient estimates which would be square of ranging between -1.03
to -10.16.
37
stricter appropriability environment. While that might not be universally true (given
that the not only the quantum of research spends but also the quality of research will
also have an effect on valuation of a firm‟s R & D returns) we report our results
with suggestive interpretations and implications in the Indian context. The
movement of with time should provide researchers, if at all askance, evidence of
behavior of innovation investments with changes in patent regime, that measured
through private returns or market valuation of the firms. With this in mind, we
introduced period effects in our specification and check our results. We break our
period of analysis into three phases, noting the period 1990 – 1994 as a categorical
variable DA, 1995-1999 as categorical variable DB and 2000-2005 as categorical
variable DC. The basic estimating Equation (5) with interactions of these dummies
and the K/As would then look thus:
)9(
.....1
ti,ε DC*)
ti,/A
ti,(K*
2β
DB*)ti,
/Ati,
(K*βDA *)ti,
/Ati,
(K*0
βt
lnqti,
lnQ
Using only the interaction terms, we check for the movements of over our three
periods. In all our regressions we also use controls of log of sales and mean of
overall industry Q of the Indian industry as given by our dataset of 9976 firms,
using the same heuristic17
for measuring Q for the overall industry as the one we
use to measure for our pharmaceutical firms. The former control is used to offset
effects of firm size and the latter to offset the effect of rising market valuations for
the Indian industry on aggregate.
16 Indian stock markets are as real a mix of efficiency and inefficiency as any other stock markets around the world. Yet if foreign institutional investors and their keen interest in the Indian bourses
are to be believed we stand on a relatively safe ground with regards to measuring firm valuation for
our research. Our data utilizes Bombay Stock Exchange data and a note on the same is included in
the appendix. 17 A note on how we measure Tobin‟s Q is included in the appendix and in construction of variables.
38
4. Data
For the purposes of our research, we create a unique dataset collating firm
variables with inventive output data and peripheral data to use as controls in our
regressions. The details of our data are outlined below:
4.1 Firm Data
Our primary dataset comes from the Prowess database of the Centre for
Monitoring of Indian Economy, which gives a ready-made industry classification of
the firms. The Prowess database is similar to Compustat database for U.S.
companies providing information that incorporated companies are required to
disclose in their annual reports. Our study is conducted on a panel of 315 drugs and
pharmaceutical firms (National Industrial Classification 2423) from 1990 to 2005.
For these firms, the dataset also provides us annual data from 1990 to 2005, on
market capitalization of the firms at the Bombay Stock Exchange (BSE). This gives
us the market value of the common stock of a firm; we also collect data on
preferred stock for these firms. To capture the debt component of a firm‟s market
value, we collect data on borrowings and current liabilities; all of this comes from
the CMIE dataset. Further we collect data on the total assets of firms as a measure
of the tangible component in a firm‟s valuation. Our firm data also includes
information on ownership groups, R & D expenditures, exports, sales, profits and
age of the firm as measured from their year of incorporation. We validate all our
firm financial data from annual reports of firms and from the electronic data source,
EDIFAR, of the Securities and Exchange Board of India, Government of India.
4.2 Knowledge Capital data
For our independent variables, we collect information on inventive output
from CMIE and validate it from various public and commercial sources. Data on
39
inventive output is proxied with R & D expenditure of the firms from the CMIE
dataset. This is validated with information from annual reports for the firms and R
& D stocks are constructed at various literature specified depreciation rates. The
disclosure norms under the Indian Companies Act 1956 require companies to report
heads of expenditure accounting for more than 1% of turnover. Since R & D
expenditure in pharmaceutical firms in India are often less than 1%, the
management in firms quite frequently does not report it. Keeping this in mind, we
also construct an imputed R & D stock variable with R & D being imputed on a
firm by firm case basis, adopting a sensible benchmark for R & D intensity in the
industry.
K, the intangible assets were constructed using the R & D expenditure
reported by firms in their books. We use annual reportage on both the capital and
current account of firms and treat the additive combination as the total R & D
expenditure of the firms. These R & D expenditures were however levels of
investments in research assets, stocks were created with depreciation rates of 15%,
20%, 25%, 30%, and 35% following literature. We detail out issues in construction
of R & D stocks in the appendix. We also took care of the escalation of R & D
salaries in India in an era of globalization of R & D. Thus we depreciate R & D
expenditure on the current account (reporting current year salaries) more in
comparison to R & D expenditure on capital account (no salary related research
expenses). A combination of 15% - 30% and 20%-40%18
were the respective
18 CEO of Glenmark Pharmaceuticals Ltd, one of India‟s upcoming drug firms, Mr Glen Saldanha
confirmed our expectation about firm salaries for R & D scientists going up in Indian
pharmaceuticals in the last ten years. Professor Narayanswamy from IIM Bangalore, India shared his
insights on how to use firm accounting data on R & D stating that in most cases firms report salaries
on their current account and expense off other R & D related expenditure on their capital account.
40
depreciation rates used on the two kinds of R & D expenditure and details on that
are included in the appendix. Finally, we subject the same treatment to imputed R &
D expenditure to construct imputed R & D stocks at the various depreciation rates.
We report here coefficient of K/As with 15%, 25% and the 15%-30% combination
of depreciation rates for stocks arrived at from both treated and untreated R & D
expenditure. We also collect other inventive assets data proxying for K with counts
and stocks of Drug Master Files and Abbreviated New Drug Applications for our
set of firms from the Food and Drug Administration, United States. For our set of
firms we also used stocks of domestic patents starting from 1995, data on which is
collected from the Indian Patent Office as another measure of K. Details about other
Ks are specified in the appendix.
4.3 Other Data
The Food and Drug Administration, United States provides us ten year
approval data on manufacturing standards of pharmaceutical firms in India, Japan
Korea and Taiwan. This information is used to create a subset of the industry. We
also use log of sales as a control in our analysis. The CMIE dataset also provides us
information to construct overall Tobin‟s Q in the Indian industry from 1990 to
2005. We use that as an additional control in our regressions. Finally, we stratify
our samples19
based on data from firm websites and annual reports, governmental
reports of industrial R & D and analyst reports and using subsequently a sensible
Thus given higher salaries, a higher depreciation rate was applied on the current account reportage of
R & D expenditure. Rik Santanu Sen, faculty member in finance with Hong Kong University of
Science and Technology who has worked previously as an industry analyst with ICICI Bank, India finds such a usage of depreciation rates in measuring R & D expenditure for Indian firms sensible. 19 The various sample stratification strategies have been outlined in the appendix. The industry was
stratified into subsets of firms engaging in US patenting, firms identified by analysts as good
investment options, firms approved by FDA and modern firms. Results for shadow of K/A were
investigated for these broad sub-samples.
41
heuristics to classify our firms into industry subsets and checked our results for
them.
5. Construction of variables
For our 315 firms, Tobin‟s Q, as outlined in the appendix is constructed thus:
a. Tobin‟s Q of a firm = market value of a firm / replacement cost of its assets.
b. Market value of a firm = market value of its equity + market value of its debt.
c. Market value of firm equity = Market value of its common and preferred stock.
Market value of common stock = Outstanding shares * closing price (both
at BSE, year-end)
Market value of preferred stock = Preferred capital provided by the CMIE
data (Sarkar and Sarkar, 2005).
Market value of firm debt = It‟s borrowings with current liabilities and
provisions.
Replacement Costs of a firm‟s assets = Total assets of a firm in a year,
with misc. expenditure and intangible assets, both subtracted from it.
This heuristic for measuring Tobin‟s Q is also applied to the overall industry dataset
and an Industry Q is used as controls to offset the effects of rising bourses for the
entire Indian industry in our regressions. The dependent variable used was
logarithms of Tobin‟s Q for firms in our dataset using the above heuristic. Apart
from Industry Q, we also used log of sales as another control in the regressions.
When returns to scale is not assumed to be 1, log of assets, is used as one of the
independent variables, assets being measured with the same heuristic as for
measuring Q, given in (e) in the Tobin‟s Q heuristic above. Our K/As consisted of
two components. The numerator K as used to construct R & D stocks as discussed
42
above and the denominator A, the replacement costs of a firm‟s assets measured
with figures on total assets (misc. expenditure and intangible assets like advertising
expenditure reported by firms on this head subtracted)as delineated above. Table 3
and Table 4 outline the variables and descriptive statistics.
Table 3: Description of Variables
Table 4: Descriptive Statistics
Variable Obs Mean Std.
Dev.
Min Max
Q 2900 1.28 1.15 0.00 9.33
Log of Q 2899 -0.03 0.75 -5.46 2.23
Total Assets (Rs crore) 2903 104.90 269.93 0.00 4182.85
Sales (Rs crore) 2903 98.46 237.13 0.00 4275.30
Log of Sales 2798 3.14 2.00 -4.61 8.36
Industry Q 5119 1.18 0.36 0.00 1.66
Un-Treated R & D R&D Stocks at 15% depreciation /
Total Assets
2900 0.02 0.04 0.00 0.32
R&D Stocks at 25% depreciation /
Total Assets
2900 0.02 0.03 0.00 0.27
Treated R & D
R&D Stocks at 15% depreciation /
Total Assets
2900 0.02 0.04 0.00 0.33
R&D Stocks at 25% depreciation /
Total Assets
2900 0.02 0.04 0.00 0.29
Variable Name Description Q Tobin's Q of firm measured as per heuristic
ln Q log of Tobin's Q
A Total Assets measured as per heuristic
ln a log of Total Assets in a year in Rs crore
Sales Sales of firm in a year in Rs Crore
ln sales log of Sales of a firm in a year
Industry Q Mean of Tobin's Q of entire industry in CMIE database
K/A (untreated, 15%) Ratio of Un-treated R & D expenditure stocks at 15% depreciation and
Total Assets
K/A (untreated, 25%) Ratio of Un-treated R & D expenditure stocks at 25% depreciation and
Total Assets
K/A (untreated, 15%-
30%)
Ratio of Un-treated R & D expenditure stocks at 15%-30% combination
of depreciation and Total Assets
K/A (treated, 15%) Ratio of Treated R & D expenditure stocks at 15% depreciation and
Total Assets
K/A (treated, 25%) Ratio of Treated R & D expenditure stocks at 25% depreciation and
Total Assets
K/A (treated, 15%-30%) Ratio of Treated R & D expenditure stocks at 15%-30% combination of
depreciation and Total Assets
43
6. Results and Discussions
In this section we summarize our results. Firstly, the changes in the
shadow with time, reveal shifting private returns to R & D in Indian
pharmaceuticals. Alongside, the movement of the shadows in various industry
subsets provides us an initial sense of within industry segments benefiting most
from stronger patents. Secondly, changes in the private returns of R & D (or the
shadow) with increasing depreciation rates enable us to comment on the importance
of newer R & D in firms given stronger patents. Thirdly, we comment about the
effect on the shadows when we conduct robustness checks in our model (with a
non-linear functional form and non-constant returns to scale). We end our
discussion by placing our estimated shadows in context with previous literature on
the market value of innovation in other economies and industries. We use log of
sales and average of overall industry Q as controls in all our regressions with fixed
effects and time dummies.
6.1 Period trends in shadow - in industry subsets and overall sample
We use the specification in Equation (9) above to estimate the time effects
on the shadows for the entire industry as well as various subsets of the industry. Our
period dummies 1990-1994, 1995-1999 and 2000-2005 were interacted with the
K/As and the coefficients for the interactions provided us shadows relevant to
different periods (Table 5 is a snapshot of Table A – C in appendix I below). The
subsets are identified using a stratification approach as outlined in Appendix C.
During 1990-1994 the shadow is lowest at -4.334 for FDA approved firms. The sign
of the shadow is negative not only for FDA approved firms but even for modern
firms, or firms who were engaging in US patenting during 1990-1994.
44
Table 5: Private Returns to R & D activity – Shadow in Industry & sub-sets
Period Entire
Industry
Only
Modern
Firms
FDA
Firms
Firms
with US
Patenting
Firms with
no US
Patenting
Untreated R & D
1990-1994 0.94 -0.612 -4.334 -1.32 1.89
1995-1999 1.627 ** 1.393 0.8 1.18 1.69 *
2000-2005 2.02 ** 2.169 ** 2.444 ** 2.2 ** 1.63 **
Change 1.08 2.781 6.778 3.52 -0.26
Treated
R & D
1990-1994 0.946 -0.601 -4.367 -1.653 1.759
1995-1999 1.685 ** 1.485 * 1.891 1.702 1.653 *
2000-2005 1.791 ** 1.922 ** 2.185 ** 1.68 ** 1.52 **
Change 0.845 2.523 6.552 3.333 -0.239
So, during that period levels of Q decrease from previous levels with unit
investment in K/A by firms. It is positive for the entire industry and for firms not
engaging in US patenting. The shadow increases subsequently for all the subsets
and the entire industry except for firms with no US patents. The change of shadow
value from 1990-1994 to 2000-2005 is most pronounced for FDA approved firms
and analyst identified firms. Such a monotonic increase in the shadow for all firms
except the non-US patenting firms at large suggests that private returns to R & D
increasing as stronger patents got implemented. For those firms, who remain aloof
to US patenting the markets provide a decreasing private return to R & D (a
decreasing shadow) with time. We could reason here that with a tighter
appropriability, firms who are unable to shift their research capabilities to a regime
such that they could appropriate benefits from it (through knowledge assets like US
patents) are not being seen favorably by the market. One can note that significance
of the shadows increases near the last period as well.
45
6.2 Trends in Shadow with Change in depreciation rates
Table 6: Shadow with approximation of constant returns to scale
The shadow is the semi-elasticity of Q with respect to the K/As. That is
for a unit increase in K/A, Q would increase by from previous levels. As
depreciation rates to measure the Ks and therefore the K/As are increased from 15%
to 25% to the 15%-30% combination we would in effect be measuring R & D (our
Ks) in such a manner where new R & D is more valuable than older R & D. Thus
higher depreciation rates and their effects on the quantum of the shadow should
provide further evidence of obsolescence of R & D with a changing IPR. If one
hypothesizes that a changing appropriability regime is making previous R & D done
46
by firms more and more obsolete, the markets should react by providing higher
shadows (or higher private returns) with higher depreciation rates. Table 6 and
Table 7 shows that this is indeed the case for the approximated linear version of our
model with and without constant returns to scale. For example with σ=1, as in Table
6 and a 25% depreciation, Tobin‟s Q goes up by 2.087 from previous levels for a
unit increase in R & D / Assets investment which is higher than the 1.969 increase
in Q-ratio levels with 15% depreciation. This increase in the shadow holds true for
depreciation rates calibrated from 15% to 25% when σ≠1 (Table 7 below), though
not observed in an un-approximated nonlinear model.
Table 7: Shadow without approximation of constant returns to scale
47
6.3 Change in shadow with a non-linear model, treating R & D and σ≠1.
Tables 6 and 7 also show that the approximated linear form of our model
(specification as per Equation 5 and 7) leads to a decrease in the size of the
coefficient estimate for treated R & D expenditure only when a non-constant returns
to scale is assumed. As we assume σ≠1, the size of coefficients decreases in the
approximated linear specification but increases in the non-linear specification.
Treatment of R & D reduces the size of coefficient in the linear approximated form
with and without returns to scale and increases the shadow in the non-linear
specification. When we do away with the assumption of constant returns to scale
and introduce log of assets we find contrasting trends for the approximated (pooled
ols) and un-approximated (nlls) regressions. For pooled OLS, introduction of log
assets leads to a universal decrease in size of the coefficient estimates, from a range
of 1.783-2.151 to 1.719-2.071 across all depreciation rates, for both treated and un-
treated R & D. For a log-log non-linear specification, where we use non-linear least
squares with time effects and first differences, we find that the trend is reversed. In
other words, introduction of a non-constant returns to scale in an un-approximated
form results in an increase of the shadow, to a range of 1.939 – 2.16 (nonlinear and
σ≠1) from a range of 1.881-2.069 (non-linear and σ=1). We must note here that for
= 1 we arrive at the shadow directly from the regression coefficients whereas for
σ≠1 the shadow is computed from the coefficient of K/A which in this case would
be * and the coefficient for log of assets would give us -1 which would help
us compute and use it on the coefficient of K/A to arrive at ranges for .
48
6.4 Shadow and previous literature
Our primary results as indicated in Table 6 show that with constant returns
to scale, the coefficients or the shadow of R & D using time dummies and firm
specific fixed effects for both treated as well as un-treated R & D, ranged from
1.783 – 2.151. Without the constant returns to scale, (or in effect introducing log of
assets in our regressions) we arrive at a computed shadow range for Indian
pharmaceuticals in the range of 1.94-2.16, as indicated in Table 7. A look at
literature (Table 8) indicates that the our shadow falls within the values of the range
-2.3 in Japan to 11.96 in the United States the estimated shadow values in previous
studies of market value of innovation.
Table 8: Coefficients of R & D in various economies and industries
Study LHS Variable R & D Coefficient
(Standard Error)
Sample Characteristics
Griliches
(1981)
log q 1.23(0.54)-1.58(0.44) US, 157 firms, 1968-1974
Hall (1993
a & b)
log v & log q 3.1(0.08)-0.48(0.02) US 2400 & 3000 firms, 1973-
1991 & 1959-1991
Megna and
Klock (1993)
q 0.488 US Semiconductors, 11 firms,
1972-1990
Haneda
and Odagiri
(1998)
log q -2.3 Japan, 90 firms, 1981-1991
Blundell
et.al (1999)
log q 1.582 UK, 340 firms, 1972 -1982
Toivannen (2002)
log v 2.6-4.2 UK, 877 Firms, 1989-1995,
Hall and
Oriani (2004)
log q France: 0.28(0.08); Germany:
0.33(0.04); Italy: 0.01(0.12);UK:0.88(0.10)
France(51 firms), Germany (80
firms), UK (284 firms), Italy (49 firms) 1989-1998
Greenhalgh
and Rogers
(2005)
log v 3.703 UK, 347 firms, 1989-1999
Hall et.al (2005)
log q 1.376(0.069) US 4800 firms, 1965-1995
Our results Log Q 1. 783 (5.38) – 2.151(5.13) India 315 firms; 1990-2005
49
7. Conclusion & Extensions
The key contribution of this research lie in identifying monotonically
increasing private returns to R & D in Indian pharmaceuticals as stronger patents
got implemented in India post the WTO-TRIPs signing in Dec‟1994. Secondly, an
increase in the depreciation rates to measure the R & D stocks, leads us to a set up
of measuring knowledge stocks where newer R & D is valued at a higher level.
That process shows an increase in the shadow of the overall industry indicating the
markets positively valuing more recent R & D in the industry. We are also able to
identify firms with FDA approval or identified by analysts as good investment
options to be the ones benefiting most from stronger patents. This is revealed by the
highest quantum ff changes observed in their shadows during the time period of
investigation. The basic findings add to the literature on market value of innovation
with new results on market value of innovation from an industry in an emerging
economy, India. One can take this research forward by checking for the broad
trends in private returns to innovative activity on other proxies for intangible assets.
We carried out extensions as outlined in our appendix using ANDAs, DMFs and
Domestic patents as proxy for Ks. Apart from domestic patents, both ANDAs and
DMFs exhibit monotonically increasing private returns to R & D as measured
through their shadows.
An extension of this current work takes the investigation of this chapter
forward (Arora, Branstetter, Chatterjee, Saggi 2008). The follow-on work asks if a
shift to strong intellectual property rights induce higher levels of inventive activity
in developing countries. The conventional wisdom in economics has come to regard
this proposition with skepticism. Theoretical contributions, most recently that of
50
Grossman and Lai (2005), and a long line of empirical contributions have laid out
sound theoretical arguments and empirical evidence suggesting that this is unlikely
to be true in general. Qian‟s (2007) work focuses on pharmaceuticals and suggests
especially strong grounds for skepticism there. The experience of the Indian
pharmaceutical industry seems to challenge this received wisdom as we have
documented in this chapter. So the moot question then becomes if the conventional
analysis is wrong?
Yes and no. The follow-on paper points to two dimensions along which the
conventional analysis may be incomplete. First, if imitation and innovation are
strategic substitutes, then a domestic policy change could, in principle, push firms
to a greater level of research intensity by foreclosing the imitation option. Second, if
R&D collaboration between established producers and potential new entrants is
possible, then a domestic policy change could have a similar, related effect, by
pushing domestic firms into partnerships that would have not been incentive
compatible so long as the imitation option were open. Our empirical analysis
provides some evidence for both kinds of developments in the Indian
pharmaceutical industry, and we fully endorse the strong possibility that these two
dimensions could become more important over time. Indigenous industry analysts
and public statements by industry executives suggest this will be the case.20
However, a deeper and more comprehensive examination of the Indian
pharmaceutical industry‟s innovation surge to date points to important causes other
than TRIPs and the patent reform process TRIPs ratification triggered. The 1984
20
Appendix III documents evidence from some firm case studies collected on a field trip to India in
December 2008
51
Hatch-Waxman Act opened up a market for legal imitations in the world‟s largest
economy. The opening of the Indian economy in the early 1990s reinforced the
attractiveness of this external market and similar markets for generics opening up in
Western Europe and industrial East Asia, while enhancing the global
competitiveness of Indian producers. In a sense, TRIPs provided the stick -- the
imperative for Indian pharmaceutical firms to seek alternative market opportunities.
The generics market (and the related API/bulk drugs market) provided the carrot --
a new market opportunity that required some investment to profitably access, but
one for which Indian firms arguably had a comparative advantage even in the mid-
1990s. India‟s long transition period to the new regime, a decade of reform, may
have also played an important role, providing indigenous firms with continued
access to a protected domestic market while they invested in the capability to sell
abroad. Over the course of the 1990s and early 2000s, Indian firms were able to
upgrade their manufacturing and process R&D capability to the point where they
could succeed in a generics market that grew rapidly as a series of blockbuster
drugs went off-patent. Arora, et.al 2008 provide substantial evidence linking the
largest pieces of the expansion of R&D input and output to process innovations
focused on existing drugs rather efforts to develop new drugs or to engage in
collaborative R&D with Western pharmaceutical firms. The market‟s rising
valuation of R&D expenditure is largely unrelated to product development.
These considerations suggest important reasons why the Indian drug industry was
able to grow and develop even as other indigenous industries in developing
countries adopting stronger patent rights were unable to weather the changes in
their markets. To some extent, Indian firms were simply in the right place at the
52
right time. Other developing countries, such as Taiwan, South Korea, and Mexico,
were forced to adopt strong patents before the generics market had fully opened up,
and they were not given the lengthy transition period India managed to negotiate.
Other potential producers, such as Brazil and Argentina, were handicapped by
uncompetitive exchange rates and domestic macroeconomic turmoil that
undermined their efforts to explore foreign markets. These considerations also
reinforce the difficulties developing countries face in trying to shift from imitation
and incremental invention to more substantive product development. These barriers
to upgrading may be especially significant in the pharmaceutical industry, but they
surely exist to varying degrees across the product space. The next question then
becomes if that version of the Indian industrial development story is generalizable?
Is it true that most developing country industries that have successfully weathered
the imposition of strong IPR have done so by identifying an industry submarket in
which largely incremental, process-oriented R&D was sufficient to secure a
defensible global market position. One wonders if parallels can possibly be drawn
between the Indian pharmaceutical industry and the Korean semiconductor
industry. Further investigation of this possibility is the focus of ongoing research.
References
Achilladelis, B. and Antonakis, K., 2001: “The dynamics of technological
innovation: the case of the pharmaceutical industry”, Research Policy, 30, 535-588.
Arora, Ashish, 1995: “Licensing Tacit Knowledge: Intellectual Property Rights and
the Market for Know-How”, Economics of Innovation and New Technology, 4, 41–
59.
Arora, Ashish, 1997: “Patents, Licensing, and Market Structure in the Chemical
Industry”, Research Policy, 26, 391-403.
53
Ben-Zion, Uri., 1984: “The R & D and Investment Decision and its Relationship to
the Firm‟s Market Value: Some Preliminary Results”, in Griliches, Zvi (ed.), R &
D, Patents, and Productivity, Chicago: University of Chicago Press, 299-312.
Blundell, Richard, Griffith, Rachel and Reenen, Van John, 1999: “Market Share,
Market Value, and Innovation in a Panel of British Manufacturing Firms”, The
Review of Economic Studies, 66, 529-554.
Bond S., Harhoff D., and Van Reenen., J., 2002: “Corporate R&D and Productivity
in German and the United Kingdom,” IFS, Oxford University, University of
Munich, and London School of Economics: manuscript.
Bound J., Cummins C., Griliches, Z., Hall, B. H., and Jaffe, A., 1984: “Who Does
R&D and Who Patents?”, in Griliches, Z. (ed.), R&D, Patents, and Productivity,
Chicago: Chicago University Press.
Branstetter, L., 2001: “Are knowledge spillovers international or intranational in
scope?”, Journal of International Economics, 53(1), 53-79.
Chan L. K. C., Lakonishok J., and Sougiannis T., 2001: “The Stock Market
Valuation of Research and Development Expenditures”, Journal of Finance, 56,
2431-2458.
Chaudhuri, S, Goldberg K. Pinelopi and Jia, Panle, 2006: “Estimating the effects of
Global patent protection in pharmaceuticals: A Case study of Quinolones in India”,
American Economic Review, 2006.
Chaudhuri, S, 2005: “The WTO and India‟s Pharmaceuticals Industry”, Oxford
University Press, New Delhi, India.
Cockburn, I., and Griliches, Z., 1988: “Industry Effects and Appropriability
Measures in the Stock Market's Valuation of R&D and Patents.”, American
Economic Review, 78, 419-423.
Fink, C., 2000: “How Stronger Patent Protection in India Might Affect the Behavior
or Transnational Pharmaceutical Industries,” The World Bank, Working Paper no.
2352. Washington D.C.
Greenhalgh, C. and Rogers, M. 2004: “The Value of Innovation: The Interaction of
Competition, R&D and IP”, Department of Economics, Oxford University Working
Paper 192.
Griliches, Z., 1981: “Market Value, R&D and Patents”, Economics Letters, 7, 183-
187.
54
------- and Mairesse J., 1984: “Productivity and R&D at the Firm Level.” in
Griliches, Z. (ed.). R&D, Patents, and Productivity, Chicago: Chicago University
Press.
-------, 1990: “Patent Statistics as Economic Indicators: A Survey”, Journal of
Economic Literature, 28, 1661-1707.
Hall, B. H., 1993b: “The Stock Market's Valuation of R&D Investment During the
1980's.”, American Economic Review, 83, 259-264.
-------, 1996: “The Private and Social Returns to Research and Development.” In
Smith B., Barfield C., (Eds.). Technology, R&D and the Economy. AEI-Brookings
Institutions: Washington, DC.
-------, 2000: “Innovation and Market Value.” In Barrell, Ray, Geoffrey Mason, and
Mary O‟Mahoney (eds.), Productivity, Innovation and Economic Performance,
Cambridge: Cambridge University Press.
--------, Griliches Z., and Hausman J. A., 1986: “Patents and R&D: Is There a
Lag?.”, International Economic Review, 27, 265-83.
––----- and Rosemarie Ham Ziedonis, 2001: “The Patent Paradox Revisited: An
Empirical Study of Patenting in the U.S. Semiconductor Industry, 1979-1995,”
Rand Journal of Economics, 32(1), 101-128.
---------, Jaffe A. B., and Trajtenberg M., 2005: “Market Value and Patent
Citations,” Rand Journal of Economics, 36(1), 16-38.
---------, and Kim D., 2000: “Valuing Intangible Assets: The Stock Market Value of
R&D Revised”, Working Paper, University of California at Berkeley and Harvard
University, Berkeley, CA and Cambridge, MA.
---------, Mairesse J., Branstetter L., and Crepon B., 1999: “Does Cash Flow Cause
Investment and R&D: An Exploration Using Panel Data for French, Japanese, and
United States Scientific Firms.” In A.R. Thurik (Ed.), Innovation, Industry
Evolution, and Employment. Cambridge University Press: Cambridge.
Hausman, J.A., 1978: “Specification Tests in Econometrics.” Econometrica, 46,
1251-1272.
Hirschey M., 1982: “Intangible Capital Aspects of Advertising and R&D
Expenditures”, Journal of Industrial Economics, 30, 375-390.
Jaffe, A. B., 1986: “Technological Opportunity and Spillovers of R&D: Evidence
from Firms' Patents, Profits, and Market Value”, American Economic Review, 76,
984-1001.
55
Khanna, T and Palepu, K, 2000: “Is Group Affiliation Profitable in Emerging
Markets? An Analysis of Diversified Indian Business Groups”, The Journal of
Finance, 55, 867-891.
Lanjouw, O. J., 1998: “The Introduction of Product Patents in India: Heartless
Exploitation of the Poor and Suffering?”, Growth Center Discussion Paper, Yale
University, and NBER Working Paper no. 6366.
Lindenberg, B. Eric and Ross, A. Stephen, 1981: “Tobin‟s Q Ratio and Industrial
Organisation”, The Journal of Business, 54, 1-32.
Lerner, J. 2002:“150 Years of Patent Protection”, American Economic Review, 92,
221-225.
Lev B., and Sougiannis T., 1996: “The Capitalization, Amortization, and Value-
Relevance of R&D”, Journal of Business, Accounting & Economics, 21, 107-138.
Levin, R.C., Klevorick K A., Nelson, R.R., and Winter S.G., 1987: “Appropriating
the Returns from Industrial Research and Development”, Brookings Papers on
Economic Activity, 3, 873-820.
Megna, P and Klock M, 1993: “The Impact of Intangible Capital on Tobin‟s Q in
the Semiconductor Industry”, American Economic Review, 83, 265-269.
Morton, FMS, 2002: “Horizontal Integration Between Brand and Generic Firms in
the Pharmaceutical Industry”, Journal of Economics and Management Strategy, 11,
135-168.
Munari F. and Oriani R., 2002: “Privatization and R&D Performance: An Empirical
Analysis Based on Tobin‟s q”, FEEM Working Paper 63.2002: Milan.
Mulkay B., Hall B. H., and Mairesse J., 2000: “Firm Level Investment and R&D in
France and the United States: A Comparison”, National Bureau of Economic
Research Working Paper 8038: Cambridge, MA.
Romer, Paul, 1990: “Endogenous Technological Change,” Journal of Political
Economy, 98, 71-102.
Sakakibara, Mariko and Branstetter, Lee, 2001: “Do stronger patents induce more
innovation? Evidence from the 1988 Japanese patent law reforms”, The Rand
Journal of Economics, 32, 77-100.
Sampath, Gehl Padmashree, 2006: “Indian Pharma within Global Reach?”, UNU-
Merit Working Paper Series No 2006 – 031.
Sarkar, J and Sarkar, S, 2005: “Debt and Corporate Governance in Emerging
Economies: Evidence from India”, IGIDR India, Working Paper Series-2005-007.
56
Scotchmer, Suzanne, 1991: “Standing on the Shoulders of Giants: Cumulative
Innovation and the Patent,” Journal of Economic Perspectives, 5, 29-41.
Teece,J. D., 2002: “Managing Intellectual Capital: Organizational, Strategic and
Policy Dimensions”, Oxford University Press.
Teece,J. D., 1986: “Profiting from technological innovations”, Research Policy, 15,
285- 306.
Thomadakis, S.B., 1977: “A model of market power, valuation, and the firm‟s
returns”, Bell Journal of Economics, 17, 150-162.
Tobin, J., 1969: “A general equilibrium approach to monetary theory”, Journal of
Money Credit and Banking, 1, 15-29.
Toivanen O., Stoneman P., and Bosworth D., 2002: “Innovation and Market value
of UK Firms, 1989-1995”, Oxford Bulletin of Economics and Statistics, 64, 39-61.
Trajtenberg, Manuel, 1990: “A Penny for Your Quotes: Patent Citations and the
Value of Innovations,” RAND Journal of Economics, 21(1), 172-187.
57
I. Appendix of Detailed Results
A. Shadow and sample stratification – period trends
Table A.
58
Table B.
59
Table C.
60
II. Appendix on associated research
A. Other Inventive Assets or proxies for Ks
We extended our investigation of the impact of patent reforms in Indian
pharmaceuticals by measuring inventive activity with other proxies, checking the
movement of private returns for them. These proxies included Drug Master Files
(DMFs) and Abbreviated New Drug Applications (ANDA) filed by our set of firms
with the Food and Drug Administration, United States. Our third proxy for
inventive assets was stocks of domestic patent counts procured from the EKASWA
database from the Indian Patent Office.
A “drug” refers to a specific chemical entity that may be called by its generic or
brand (proprietary) name. It may be manufactured by either the NDA or the New
Drug Application holder or one of many ANDA holders. A firm decides what forms
and concentrations to produce, a different (A) NDA being required for each. With
the Indian pharmaceutical industry focusing mainly on the generics markets in the
last 10-15 years, ANDA filings are frequent and would be a realistic proxy to
measure innovation and research output. We use stocks of only ANDA counts as
well as stocks of ANDA concentration counts to proxy for innovation output
following Morton (2002). The stocks are created using a similar range of
depreciation rates as used for R & D and results are reported at depreciation rates of
15% 25% and 35%. A drug master file or a DMF is on the other hand an optional
submission by the firm to the FDA of information, usually about the chemistry,
manufacturing and controls (and also other information) of a drug product or a
component of a drug product. This is filed to permit the FDA to review this
61
information in support of a third party‟s submission. The FDA currently classifies
DMFs into four types Type I to Type IV depending on whether the information in
the DMF pertains to the manufacturing plant or clinical and toxicological
information, drug substance (products, intermediates or materials used in their
manufacturing), packaging or excipient information (Shaw, 2003). Our rationale in
using DMFs as a measure for inventive assets is vindicated by industry reports.
These cite that in the DMF filings space, India‟s share in global filings has
increased from 2.4% in „91, to around 35% in ‟04. Further it is only second to the
United States according to 2005 data in the DMF filings space. We also check our
results on stocks of domestic patent counts, information on which is procured from
the EKASWA database made available by the Indian Patent Office. The EKASWA
database provides us domestic patent filing information starting from 1995.
Summary statistics of these four proxies of inventive assets are reported with their
K/As at 15% depreciation rates in the table below.
Summary Statistics of Other Inventive Assets
Variable Obs Mean Std. Dev. Min Max
ANDA count 5112 0.021518 0.407241 0 15
ANDA Concentration count 5112 0.0446009 0.904329 0 35
DMF count 5106 0.1329808 1.008748 0 29
Domestic Patent count 5120 0.1408203 1.168836 0 40
ANDA Stocks/Total Assets 2900 0.0007893 0.012671 0 0.34785
ANDA Concentration Stocks/ Total
Assets
2900 0.0008282 0.012694 0 0.34785
DMF Stocks/ Total Assets 2900 0.0023699 0.01 0 0.141498
Domestic Patent Stocks/ Total Assets 2900 0.0025559 0.016521 0 0.344439
62
The results of our analysis are indicated below in the table on the next page, these
are our R & D results at various depreciation rates. We argue that while ANDAs
might surely face obsolescence, DMFs might still not be subject to obsolescence
while measuring inventive output of a firm given its voluntary nature in filing for a
firm. Given that DMFs were used in our baseline regression not only for
depreciated stocks but also for its counts (100% depreciation figures in the table
below). In summary following broad observations could be made based on the
results. For ANDAs, usage of stocks of only counts or even concentration counts
show only minor changes in the quantum of the shadows as seen in the estimation
coefficients of K/As. It seems that during the period 1990 – 1994 ANDAs were
valued negatively by the markets the quantum of the shadow ranging between -30.1
to -32.2. These negative shadows by 1995 to 2000 had decreased in its amount but
still ranged between -8.8 to -12.0 during that period. As the implementation date for
product patents neared, the markets seem to have started viewing ANDAs
positively, their shadows becoming positive and ranging from 0.05 to 1.7. The
broad monotonic increase in private returns to inventive activity with
implementation of stronger patents is thus equally applicable when one proxies
intangibles with the ANDAs. For DMFs, shadows ranged from -6.1 to -2.4 during
1990-1995 but had increased monotonically by 2000-2005 ranging from 1.9 to 2.3.
This trend persists whether we take into account the non-obsolescence nature of
DMFs as indicated in figures with depreciation rates of 100%, where we used pure
DMF counts as Ks in our regressions. For domestic patents, the shadow during the
period 1990-1995 is 0 because no information on domestic information is available
during that period. However uniformly across various rates of depreciation the
63
shadow which ranged from 0.4 to 0.5 during 1995-2000 fell during 2000 to 2005
ranging then from -1.05 to -0.95 during that period.
Table on Shadows of Other Inventive Assets
B. R & D expenditures in Indian Pharmaceuticals
For our sample firms, mean R & D expenditures was Rs 2.42 crore (approx
$0.52 million) with maximum R & D expenditure reported by Ranbaxy
Laboratories in 2004 of Rs 399.67 crore ($ 85 million). The mean R & D
expenditure has been rising over the years. In 1990 the mean of R & D expenditure
was Rs 0.2 crore and it rose to Rs 0.8 crore in 1995 and to Rs 1.5 crore in 2000 and
finally to Rs 9.3 crore in 2005. The standard deviation of R & D expenditures was
Rs 14.7 crore ($3.5 million). R & D expenditures were also studied on a firm-by-
firm basis. Observations which were not zeros or missing for any year for a firm
were taken and 173 firms were studied.
64
Figure 1.
Compounded Annual Growth Rate of Total R & D expenditures of Indian
pharmaceutical firms
-100.0
-50.0
0.0
50.0
100.0
150.0
200.0
250.0
300.0
firmid
Numeric firm Identifier
CAG
R
cagr
This sample of firms as in Figure 1 above showed a compounded annual growth
rate (CAGR) of R & D expenditure ranging from de-growth of 50% to a growth of
248% for the period of analysis as shown in Exhibit 1 above. 8 firms had a CAGR
of greater than 100%, 10 firms showed a CAGR between 50-100%, 36 firms had a
CAGR between 25-50% while 54 firms showed a CAGR greater than zero and less
than 25%. In all 173 such firms were studied and the top high-growth firms are
listed in table below.
Table on Indian Pharmaceutical Firms in terms of R & D expenditure
FIRM Name CAGR of R & D
Expenditures (%)
Period for
CAGR
Unimark Remedies Ltd 248.8 1999-2002
Matrix Laboratories Ltd. 169.3 2001-2004
Glenmark Pharmaceuticals Ltd. 79.5 1996-2005
Cipla Ltd. 74.6 1993-2005
Nicholas Piramal India Ltd. 65.3 1994-2005
Morepen Laboratories Ltd. 50.6 1995-2002
Orchid Chemicals & Pharmaceuticals Ltd. 47.8 1994-2005
Vorin Laboratories Ltd. [Merged] 46.5 1996-2001
Shasun Chemicals & Drugs Ltd. 46.3 1993-2005
Dishman Pharmaceuticals & Chemicals Ltd. 45.8 2003-2005
Dr. Reddy'S Laboratories Ltd. 45.6 1992-2005
Indoco Remedies Ltd. 45.3 1998-2005
Torrent Pharmaceuticals Ltd. 29.6 1994-2005
65
We also identified firms who reported R & D expenditures in excess of $ 1
million anytime during our period of analysis. A dozen firms as in Table above
reported R & D expenditure in excess of $ 1 million anytime during our period of
analysis and most of this increase occurred during 1999-2005. The compounded
annual growth rate (CAGR) of R & D expenditures for this elite group of 12 firms
shows an increasing trend of investment in R & D expenditures. CAGR for these
firms ranged from 24.8% for Aurobindo Pharma to 79.5% for Glenmark
Pharmaceuticals. This rate was computed from the year when data was reported and
was not zero or missing. The CAGR reveals and upward movement of R & D
expenditures. Yet they might be misleading when one places the quantum of R & D
expenditures of this elite group of Indian pharmaceutical firms with global
pharmaceutical majors. Global pharma majors were already spending in R & D
back during the 1950-1960s (Landau, Achilladelis and Scriabine [2005]) as much
amounts as what Indian pharmaceuticals are spending in R & D today. Even in
2005, R & D expenditures of the elite Indian firms, ranging between $ 11 million to
$ 84 million, were just about 1-10% of what global pharma majors were spending in
R & D (in excess of $ 1000 million) by the mid 1990s. Globally though, with
escalating costs of drug R & D, researchers are looking at the leverage of a drug
firm to re-invest its sales earnings or total assets into R & D for future growth.
It is on this count that Indian drug firms are catching up with their R & D
expenditure as % of sales reaching levels of global Pharma. Today, global Pharma
is spending some 15%-20% of their sales in R & D expenditure. Indian firms like
Dr Reddy‟s Laboratories are touching those R & D intensity levels having spent
some 17% of their sales in R & D in 2005.
66
Table on Firms with R & D expenditures in excess of $ 1 million
T
The intensity level has risen from less than 0.5% levels even ten years back
to somewhere between 3 % - 7% in 2005, with individual firms like Dr Reddys,
Ranbaxy, or Glenmark spending in excess of 10% of sales or more in the last two or
three years. Our overall industry dataset shows a mean of intensity of R & D
expenditure as a percentage of sales at 0.12% in 1990, increasing to 0.84% in 1995,
0.9% in 2000 and reaching 2.7% levels by 2005.
FIRM Name Total R & D
expenditure in 2005 (in Rs crore,
in braces dollar
equivalents)
CAGR for
total R & D expenditure
YEAR
computed from
Mean
Intensity (R & D
expenditure
as % of sales)
Aurobindo Pharma 54.31 ($ 11.4 mil)
24.8 2000 2.4
Cadila Healthcare 103.2
($ 21.7 mil)
40.8 1996 4.7
Cipla 98.38 ($ 20.7 mil)
74.6 1993 3.0
Dr Reddy‟s
Laboratories
297.79
($ 62.5 mil)
45.6 1992 5.0
Glenmark Pharmaceuticals
48.68 ($ 10.2 mil)
79.5 1996 5.3
Lupin 83.61
($ 17.6 mil)
48.7 1993 4.0
Nicholas Piramal 108.44
($ 22.8 mil)
65.3 1997 2.4
Orchid Chemicals &
Pharmaceuticals
52.21
($ 10.9 mil)
47.80 1994 4.1
Ranbaxy Laboratories 399.66 ($ 84 mil) 30.9 1993 4.7
Sun Pharmaceutical Industries
115.98 ($ 24.4 mil)
31.1 1994 5.8
Torrent
Pharmaceuticals
67.32
($ 14.1 mil)
29.6 1994 6.4
Wockhardt 69.28 ($ 14.5 mil)
32.6 1992 5.6
67
Along side intensity levels, some firms are showing signs of indulging in R & D
geared towards drug discovery. Indian firms are also forging strategic alliances with
global pharma firms. At the forefront of such alliances are generic makers, like Par
Pharma and Watson striking R & D alliances with their Indian counterparts like Dr
Reddy‟s or Cipla Ltd. In the new drug discovery regime, a few licensing deals have
already been chalked out – Dr Reddy‟s has for example licensed out its DRF 2593,
DRF 2725, diabetic molecules to Novo Nordisk and DRF 4158 metabolic disorder
related molecule to Novartis AG. Ranbaxy has inked a deal with Bayer AG for its
Cipro XR antibiotic molecule and with Schwarz Pharma for its RBx 2258 molecule.
These deals are not one-off, with other drug makers in India, like Torrent and
Glenmark catching up in licensing new molecules to global firms like Novartis AG
and Forest labs respectively. Indian drug makers also now realize the importance of
owning knowledge assets, Indian filings at the FDA for Abbreviated New Drug
Application counts or for Drug Master Files to manufacture generics increasing in
the last few years. We find that this evolution in R & D capabilities is most keenly
evidenced in firms who engage in US patenting activity, or those firms who were
picked by analysts in the Indian equity markets. Motivated by these developments
at the industry subset levels, we also try to understand the movement of R & D
expenditures in Indian pharmaceuticals at the various industry strata identified. As
shown in Table on R & D expenditures in various industry subsets below, firms
with US patenting activity reported a mean of total R & D expenditure of Rs 0.1
crore in 1990, this figure increasing to Rs 3.6 crore in 1995, Rs 8.1 crore in 2000
and Rs 48 crore in 2005. For these firms intensity of R & D expenditure rose from
0.8% of sales in 1990 to 1.8% in 1995, 2.6% in 2000 and 6% in 2005. For firms
68
picked by analysts, mean of total R & D expenditure was Rs 0.4 crore in 1990,
increasing to Rs 3.0 crore in 1995, Rs 8.6 crore in 2000 and Rs 40.9 crore in 2005.
The intensity stood at 0.2% of sales in 1990, rising to 1.6% in 1995, 2.3% in 2000
and further to 4.8% by 2005. The movement of R & D expenditure and its intensity
for other industry subsets are also noted. The table in the next page shows that there
is a rising trend in terms of both absolute expenditure and intensity of R & D.
Table on R & D expenditures in industry subsets of Indian pharmaceuticals
1990 1995 2000 2005 Overall
Only US Patenting firms (47 firms)
Mean of Total R & D expenditure
(in Rs crore)
0.1 3.6 8.1 48.1 11.9
Mean of Total R & D Expenditure
as % of sales
0.1 1.8 2.6 5.9 2.2
Firms Picked by Analysts (41 firms)
Mean of Total R & D expenditure
(in Rs crore)
0.4 3.0 8.6 40.9 12.6
Mean of Total R & D Expenditure
as % of sales
0.2 1.6 2.3 4.8 2.3
Firms approved by FDA(42 firms)
Mean of Total R & D expenditure
(in Rs crore)
0.2 3.5 8.1 40.4 12.5
Mean of Total R & D Expenditure
As % of sales
0.1 1.5 1.7 4.8 2.0
Modern Firms (145 firms)
Mean of Total R & D expenditure
(in Rs crore)
0.2 1.6 3.0 16.9 4.8
Mean of Total R & D Expenditure
As % of sales
0.1 1.0 1.0 4.6 1.7
Entire Industry (315 firms)
Mean of Total R & D expenditure
(in Rs crore)
0.2 0.8 1.5 9.3 2.4
Mean of Total R & D Expenditure
as % of sales
0.1 0.8 0.9 2.7 1.2
C. Sample Stratification
To understand the dynamics within the industry and effects of a
changing IPR regime on that, we adopted a sample stratification strategy as outlined
below. Our basic stratified samples included dividing up the drug firms in our
dataset as modern or non-modern firms, firms‟ whose manufacturing facility(s) has
69
been certified by the FDA at any time as per the data provided to us by FDA, also
firms who were identified as front-runners in the Indian drugs and pharmaceuticals
space by analysts in the Indian financial markets and finally firms who had or had
not filed for a US patent with the US patent office. We also carried out
stratifications based on their ownership group, i.e. either domestic or foreign firms
and outline here the same.
Modern or non-modern drug firms in the Indian setting
We found the Indian pharmaceuticals industry to be quite varied in terms of
the horizontal spread of the industry as per firm size as well as the vertical depth
depending on a firm‟s business focus. Trade association figures suggest that there
are some 10000 units in the industry with some 300 in the organized sector. One
can ascertain this from the number of interest groups that are currently present in
the industry – there are the bulk drug manufacturers associations, and the
pharmaceuticals exports association, association of subsidiaries of foreign
pharmaceutical manufacturers in India as also the association for domestic
formulations makers or small and medium drug entities. The spread and depth in
the industry reflects also in a firm‟s positioning related to R & D and manufacturing
capabilities. We do understand that a classification process in such a scenario while
by no means conclusive still might aid in understanding consolidation dynamics
within the industry. This especially with changes in IPR and the interest of the
researcher in separating the wheat from the chaff. Our research based on talking to
a few firms, looking at their websites, consulting analyst reports on the industry or
going through the firms‟ annual reports suggested that the Indian drugs industry
could be at the first level broadly identified in terms of the products they were
70
finally manufacturing. So, in many cases firms were just bulk or active
pharmaceutical ingredients makers – some other times they were a little more than
just that; bulk drugs entity producing what the industry calls as advanced
intermediates which are almost finished dosage forms except that a formulations
company picks up an advanced intermediate and gives it a final shape adding
excipients, assigning a brand and selling them as tablets, capsules or any other final
medicinal form. Yet other times, firms were found to be specializing in the
production of new drug delivery systems, or were just pure capsule makers viz.
Empty Hard Gelatine Capsules made by Bharti Healthcare from northern India. For
this research we stayed out of Indian drug firms (and there were about 10 of them in
our dataset) who specialized in the traditional medicines space like Ayurveda or
Herbal formulations space.
The primary source for classifying the product portfolio of firms was the
CMIE – Prowess data base we were using for our research. After eliminating firms
in the traditional medicines space and those with spurious data we arrived at a
consolidated sample of 315 drug firms for analysis in the Indian context. From the
dataset we then tried in assessing the product portfolio or the nature of
manufactured products for a firm. Such an approach was not without
inconsistencies. For example, capacities for manufacturing always didn‟t
correspond to the product types that were being sold by any given firm. We used
checks not only on the „capacity for manufacture‟ fields on the database but also on
the „product on sales‟. Some other times, the years for which data were cited ended
much before 1999 or 2001. And then again for some firms in the database there was
absolutely no observation at all. For both of the above situation we have validated
71
the observations from the database by visiting the concerned firm‟s website and
looking at the respective firm‟s latest product portfolio where available. Finally,
many a times, bulk, formulations, intermediates, medical products or NDDS were
observed not in a consistent language across all firms for all years. We applied our
judgment in such a case as to what constituted for coming under the broad category
of bulk, formulations, intermediated, medical products or NDDS. Finally, as was
mentioned above, the judgment factor could have induced a bias in the
classification process, but for that the researcher has only confirmed his
understanding of a firm‟s positioning by visiting the websites of the concerned firm
in question. Most often a bulk maker or a formulations maker could easily be
identified by visiting the Products section of the respective firms‟ website – though
it must be said quite a few firms didn‟t have any website at all – and then again
there were a few instances where integration was judged based on the range of
products that were being manufactured and sold by the firm.
Broadly speaking thus, the heuristic for classification went like this.
i. For any firm X one accessed the products manufactured section on
the CMIE database. Checked that with the products sold by the same firm from
their website.
ii. That provided the categories under which capacities were installed in
the firm cited for the last few years – say for example 2001-2005.
iii. We benchmark our data for the year 2000. 2001 data was taken as
indicative of the firms‟ positioning in 2000 and so was 1999 data. However if the
data citing ended much before 2000 say for example, due to reasons of mergers,
acquisition or winding up of operations by the firm – with such a year in question
72
being 1996 for example – 1996 then was noted as the ending year product category
with a note in the database of the year by which the category was noted.
iv. If the data on the CMIE database started from 2001 or ended in 1999
– unlike say specifically the company‟s position in 2000 – the assumption was
checked by visiting the Products section of the firms‟ website. We thus ended up
identifying our firms as belonging to one of the following categories: bulk only
makers (94), makers of only formulations (98), bulk and formulations with
evidences of integrated operations (52), and bulk and formulations without any
evidence of integration (28). The rest, 43 firms were classified as others who could
not be identified as belonging to one among the above set of drug makers. It might
be worthwhile to mention here that our dataset had 6 biotech firms, 13 firms who
specialized in medical products or diagnostic kits and 11 specializing in new drug
delivery systems.
v. Having classified firms as per their product portfolio we moved on to
collecting data on the nature of R & D being done by the firms. 10 of the firms were
focusing on research purely on Active Pharmaceutical Ingredients, 7 of them had
added biotech research capabilities to API research capabilities, 31 were
specializing in R & D on both bulk and formulations, 8 of them seem to be
conducting drugs R & D in advanced intermediates apart from API research
capabilities, 18 specialized also on research related to novel drug delivery systems,
23 were specializing on research only in formulations, and likewise.
vi. We then matched our data on R & D with that of the product
portfolio data for firms. This was done along with three other associated checks to
arrive at a broad classification in the industry comprising of modern or non-modern
73
drug firms. Firstly we checked our R & D classification data with analyst reports on
strategic alliances with foreign drug firms for a domestic firm either in Europe or in
the United States. Next, we also checked if the firm had a full-fledged R & D center
or were claiming to conduct R & D in their manufacturing facilities, consigning the
latter to be practically doing no real R & D. Finally, we also confirmed the presence
of a GoI21
approved R & D facility with the firm. We used the above three
classifications and identified firms as being a modern drug firm or a non-modern
drug firm if it had a strategic R & D alliance OR had a separate R & D center OR
had a government approved R & D facility while keeping in mind the matching of
the R & D portfolio with the product portfolio of the firm.
We thus identified, one can perhaps say with a liberal classification system,
(we tighten the screws in subsequent stratification strategy), at 145 modern drug
firms and 170 non-modern drug firms in our dataset. For the modern drug firms
over the years, total R & D expenditure at the mean was Rs 4.8 crore (approx $ 1.02
mil), that being about 2.3% of mean sales which was around Rs 172 crore (approx $
36.6 mil).
Firms whose facilities were certified by FDA any time
Another stratification strategy used the FDA certification data we collected
from the Food and Drug Administration, USA. The compelling reason behind such
a stratification stemmed from the importance given to Indian drug manufacturers by
FDA. We found it worthwhile to check that subset in the industry which is trying to
get this FDA certification in manufacturing so that they could compete with their
21
The Department of Science and Technology of the Indian Government approves industrial R & D
facility of firms based in India.
74
products at the global pharma markets. The data provided to us details on
certification of drug manufacturing facilities in India, Taiwan, Korea and Japan, by
firm name, their manufacturing facility, and date of certification with information
on whether the standards were in line with FDA benchmarks. We used the date of
inspection to assign if the certification happened in a particular year as in our panel.
For India, 193 drug manufacturing plants –41% of all complying plants from
inspections in Japan, Taiwan, Korea and India – were in compliance with FDA
regulations. This was a number that matched with increasing numbers of FDA
approved facilities as available from other sources22
. Out of the overall data, 8
facilities, in various years for different or same firms were not granted approvals
despite FDA inspections. We used the FDA data, to create a dummy for firms,
assuming that if a firm in any year had a FDA certification it would be coded as 1
and otherwise zero. There were 672 such observations, during which an FDA
inspection in a year, for a particular firm resulted in a compliance certification. We
report our movement of the shadow for this set of 42 firms – who had a mean total
R & D expenditure of Rs 12.5 crore (approx $ 2.7 mil ), some 3.7 % of mean sales
of Rs 342 crore (approx $ 72.8 mil ).
Firms identified by analysts as good investment options
For the purpose of stratifying our sample, we also consulted analyst reports.
There were firms that were picked by analysts as front runners in the drugs and
pharmaceutical industry, in terms of their potential to be good investment options.
The first was a research report by Morgan and Stanley‟s global equity research
22 In 2005, the largest number of 60 compliance certificates of manufacturing facilities went to
Indian drug firms – a number more than drug firms in Israel, China, Italy, Taiwan, Spain and
Hungary, cited a Morgan Stanley, 2004 analyst report.
75
division, published in December 2004, which identified firms from as the report
was titled the “Emerging Indian Pharmaceutical Industry”. The second also was
from Morgan and Stanley but this time coming from a field trip report by its
Asia/Pacific equity research division, written again in 2004. The final list of firms
came from Kotak Institutional Securities, an equity house based out of Mumbai, in
which Goldman Sachs had a 25% stake for its Indian operations until 2006. The
consolidated list of 41 firms favored by analysts had a mean total R & D
expenditure of Rs 12.6 crore (approx $2.8 mil), some 3.5% of mean sales of Rs
359.2 crore (approx $76.4 mil).
Firms with at least one US Patent filed and granted as per the US
Patent Office
The sample was also stratified on the basis of data collected from the US
Patent office, US Patents being identified by assignee name and/or inventor country
filed by drug companies based out of India. There were 47 such firms, who had at
least one US Patent filed at the USPTO. For these set of firms with US patenting
history, the earliest patent filed by an Indian firm dated back to 1974, average
patents filed by an Indian firm were 0.076, and the maximum filed by a particular
firm in a particular year was by Ranbaxy Ltd -- 11 patents in 2002. The
complementary set of firms was also identified as a stratum for investigation. It
might be an aside to mention here but there were 281 firms with domestic
ownership, 34 with foreign ownership and 288 firms who didn‟t face a merger or
acquisition in our dataset.
Characterization of sub-sets in Indian Drugs and Pharmaceuticals Industry
76
D. Measuring Tobin’s Q for Indian drug firms
Our study hinges critically on the measurement of Tobin‟s Q. When one
looks at literature from financial economics, one finds a basic definition of Tobin‟s
Q starting from an entire stream of literature pioneered by James Tobin in his
seminal paper in 1969 titled, “A general equilibrium approach to monetary theory”.
Researchers have used Tobin‟s Q as a measure of firm performance subsequently in
various areas of research, whether to understand ownership issues or the impact of
mergers and acquisitions in an industry. The idea of Tobin‟s Q is based on the fact
that a firm needs to be valued not only by the markets but also in terms of
replacement cost of its assets. Thus computationally, Q has been a ratio of the
Characterization Entire
Industry
Only
Modern
Firms
Analyst
Firms
FDA
Firms
Firms
with US
Patenting
Firms
with no
US
Patenting No of firms 315 145 41 42 47 268
Un-treated R & D
Mean R & D expenditure
2.42 4.76 12.62 12.47 11.96 0.41
Max R & D
expenditure
399.66 399.66 399.66 399.66 399.66 28.65
Treated R& D
Mean R & D
expenditure
1.58 3.39 10.69 9.69 9.26 0.26
Max R & D expenditure
639.20 639.2 639.2 639.2 639.2 28.9
Mean Sales 98.46 172.95 359.16 341.76 336.05 48.43
Max Sales 4275.30 4275.3 4275.3 4275.3 4275.3 1501.21
Mean Assets 104.90 177.84 389.27 377.96 361.98 50.76
Max Assets 4182.85 4182.85 4182.85 4182.85 4182.85 2322.55
Un-Treated R & D
R & D expenditure/Sales at
the mean (%)
2.46 2.75 3.51 3.65 3.56 0.85
Treated R & D
R & D
expenditure/Sales at the mean (%)
1.60 1.96 2.98 2.84 2.76 0.53
*Absolute numbers are in 10 million Indian Rupees
77
market value of a firm to its replacement cost of assets. There are certain subtleties
associated with such an aggregate definition of how to compute Tobin‟s Q. Firstly,
accounting standards differ in various economies, be that in European, US or Asian
context, thus accounting reportage has to be carefully handled to use them as data
for economic analysis. Secondly, the numerator, market value of the firm is an
aggregate of the market value of not only the equity portion of a firm‟s valuation
but also the debt portion. Measuring market value of equity could be not so straight-
forward given that a firm‟s equity capital could consist of not only its common
stock but also its preferred stock. Measuring market value of debt also requires
approximation for in most economies arriving at that with an underdeveloped bond
or debt trading markets might be difficult. To counter all these problems, the
literature in financial economics has over the years arrived at various contextual
algorithms or approximations to measure Tobin‟s Q. (Lewellen and Badrinath,
1997, Thomadakis, 1977 and Perfect and Wiles, 1994). The literature on market
valuation of innovation, started by Griliches, Hall and others broadly adopts the
above algorithms modifying them in a contextual setting as required by accounting
standards in a particular economy. We follow closely both the financial economics
literature thus, along with literature on market valuation of innovation to construct
our Tobin‟s q, falling back when required to adapt ourselves to the measurement of
Q in the Indian context borrowing from previous studies on measuring Tobin‟s Q in
research related to India (Khanna and Palepu, 1990 and Sarkar and Sarkar, 2005).
We outline here thus our broad steps to construct Tobin‟s Q from the CMIE data.
Firstly we look at the numerator, to compute the market value of Q ratio. For
78
measuring market value of debt, we use the following accounting data to measure
market value of debt:
1. Total borrowings & Total liabilities
2. Total borrowings along with current liabilities and provisions.
3. Debt to Equity ratio * Net Worth
All of the above are available in the CMIE dataset for a firm in a particular year,
giving us four options for measuring market value of debt. For market value of
equity, we use the preferred capital of a firm in a year, and add it up to the market
capitalization of a firm for a particular year for the common stock of the firm.
Market capitalization is arrived at by using the number of outstanding shares of a
firm in the Bombay Stock Exchange multiplied with its closing price on the last day
of a financial year. Thus along with four ways of measuring debt, and the market
value of equity and preferred capital, we arrive at four possible ways of measuring
the numerator, the market value of a firm. Financial economics literature has
several strands of debate on how to appropriately measure the replacement cost of
assets. We use the total assets figures of a firm with four variations in the Indian
context from the CMIE data. These four ways of measuring replacement costs of
assets were:
Total assets subtracting Misc. Expenditure not written off from Total assets.
(The second term, industry sources whom we also consulted pointed out was not
really measuring the real assets of the firm. We confirmed that with our dataset
for it included data on items like voluntary retirement expenses and thought it
wise to subtract from total assets.)
79
Total assets with Misc. Expenditure not written off and Intangible Assets both
subtracted from it. (If R & D is less than 1% of total sales, Indian Companies
Act 1956 doesn‟t make it mandatory to report R & D expenditure separately.
Our industry discussions pointed out thus that, firms many a times club R & D
expenditure with advertising expenditure and report it under the head of
intangible assets leaving no separate field for advertising expenditure; we thus
felt it wise to check with intangible assets being additionally subtracted from
Total assets).
We also used figures on total assets in a stand alone way to measure Tobin‟s Q.
Finally, the CMIE dataset also provides us average of previous and a current
year data; we used the average total assets to measure for replacement cost of
assets.
Four possible measures of replacement cost of assets and four possible
measures of market value, gives us sixteen different combinations to measure
Tobin‟s Q. We checked for the pair- wise correlations for all these 16 different
measures of Tobin‟s Q. Using correlation figures along with literature and our
judgment on what could be the most adroitly yet comprehensively measurement
strategy for Q, we arrived at the following:
a. Tobin‟s Q = market value / replacement cost of assets.
b. Market value = market value of equity + market value of debt.
c. Market value of equity = Market value of common stock and
preferred stock:
80
d. Market value of common stock = Outstanding shares * closing price
(both at BSE, year-end) and Market value of preferred stock = Preferred
capital provided by the CMIE data (Sarkar and Sarkar (2005))
e. Market value of debt = Total borrowings and current liabilities and
provisions.
f. Replacement Costs of Assets = Total assets of a firm in a year, with
misc. expenditure not written off and intangible assets, both subtracted from
it.
Additionally, we also use the same heuristic to measure Q for computing
overall industry Q from all the firms, made available by the CMIE-Prowess dataset.
We use the mean of overall industry Q as a control on the right – hand side of our
regressions throughout our regressions to offset overall industry effects on the set of
drug firms.
E. Heuristic for imputation of R & D expenditure
While we discuss about the nature of distribution of R & D expenditure in a
separate section of the appendix, as has been discussed, R & D reportage is a matter
of concern in the Indian context. The disclosure norms under the Indian Companies
Act 1956 require companies to report heads of expenditure accounting for more
than 1% of turnover. Since R & D expenditure in pharmaceutical firms in India are
often less than 1% the management could quite frequently not report it, clubbing it
with advertising expenditure while reporting it under intangible assets in the
balance sheet. Further our R & D expenditures came both from capital account as
well as current account. R&D expenditures on capital account takes into
consideration R&D spending related to capital assets such as buildings, plant and
81
equipment, laboratories, etc. R&D expenditure on revenue or current account means
R&D spending related to current year's R&D activities such as R&D staff salaries,
materials used in R&D projects, etc. We used an additive combination of R & D
expenditure on both the capital and current account to measure total R & D
expenditure for a firm in a year. These expenditure figures were then used to
construct R & D stocks at various literature specified depreciation rates – as
outlined in another section of the appendix. The same method was applied to the
imputed R & D expenditures. Imputed R & D expenditures were created using a
sensible heuristic of R & D intensity, especially for firms who don‟t report R & D
expenditure either or on both the capital and current account during in-between
years. In all 54 firms were treated for missing R & D observations in in-between
years with before and after observations of R & D expenditure being present. Firm
specific approaches were followed, though very broadly, we use average intensities
(R & D expenditure on sales) to impute data. Thus for Cipla Ltd, the Indian firm
which created waves bringing in its low-cost AIDS drugs, R & D is consistently
reported starting from 1993 until 2005 with observations missing in 1996 and 2003.
Annual reports or SEBI database didn‟t prove of much help for imputing data for
Cipla, because it doesn‟t mention R & D expenditure separately in the annual
reports of those years – presumably they having fallen below reportable
percentages. However one cannot simply do away with R & D expenditure, in those
years more so, given previous and subsequent period data existing. Thus for Cipla
we take the average intensity of R & D expenditure, 3% of sales, as benchmark
intensity, and impute data of R & D expenditure in 1996 and 2003 on current
account and capital account. In some other cases, when CMIE failed to capture
82
data, we used data available from firm websites and ten-year financial summaries to
impute data. This for example was the treatment strategy for R & D expenditure of
the Indian subsidiary of Abbott Ltd which has had a 61 year presence in Indian
markets. As mentioned, we apportion the imputed amount on current or capital
account relying on previous year and subsequent year trends of shares of current or
capital account on total R & D expenditure and if that is unavailable, we allocate
them equally among the two accounts. This heuristic for treatment of R & D took
into account 2217 missing observations in total untreated R & D expenditure in our
panel. The mean of treated total R & D expenditure was Rs 1.6 crore (approx $ 0.34
mil) – about 37% lower than the mean of un-treated total R & D expenditure at Rs
2.4 crore (approx $ 0.51 mil ), with standard deviations of former being around Rs
14.3 – Rs 14.7 crore ($ 3.1 mil approx). The correlation between treated and un-
treated total R & D expenditure was more than 99.5%.
F. Creation of R & D stocks
R & D expenditure by firms is undertaken as an investment intended to create a
knowledge stock. This stream of thought has given rise to an entire body of
literature related to endogenous growth theory. Essentially, there is a need for
creation of R & D stocks, treating R & D expenditure in the same way as ordinary
investment in tangible assets. This more so, because R & D conducted today, and
thereby generating knowledge stocks in the same or subsequent periods, might not
be as valuable as R & D tomorrow and its associated knowledge stocks generated.
This means much like in investments in tangible assets, even in investments in R &
D, an intangible asset or knowledge capital, one needs to assume an appropriate
depreciation rate. To identify an appropriate depreciation rate, researchers have
83
adopted various approaches, mostly using a perpetual inventory model and using
standard depreciation rates of 15-20%. A recent paper (Hall, 2006) highlights this
problem acutely, and tries also to look at the lag structure of R & D in generating
returns and if that is influenced, in the case of private returns, by the markets
themselves, rather than the other way around. For our purposes we create our R &
D stocks both on treated and un-treated R & D expenditure
thus: tiRtiKtiK ,1,, where for firm i, K represents the stock of R & D
expenditure in year t and t-1 and R represents the flow of R & D expenditure in a
particular year t. Researchers (Griliches and Mairesse, 1984; Hall, 1990; Jaffe,
1986; Hall 1993; Blundell et al., 1999; Hall and Oriani, 2006) have used a perpetual
inventory method to compute the stocks using depreciation rates in the range of 15-
20%. We adopt a slightly modified approach. Firstly, for our starting year t=1990,
we assume that the flow of R & D expenditure is also the stock of R & D for a firm.
That is tiRtiK ,, for t=1990. Given that, we apply literature specified rates, as
well as depreciation rates in increments of 5% and check our results. Thus we use
depreciation rates of 15%, 20%, 25%, 30% and 35%, reporting results with only
15% and 25% depreciation rates. Further, conversations with some industry insiders
suggests that with globalization of R & D (setting up of centers for drug R & D by
global pharmaceutical firms like AstraZeneca in India) it could be possible that R &
D salaries (rather than R & D capital) accounted for a substantial chunk of total R &
D expenditures of these firms. In other words this also might have meant that R &
D salaries might have depreciated more in comparison to R & D assets. Thus for
our R & D expenditure we apply depreciation rates of 15% on R & D expenditure
84
reported on capital account (spending on capital assets and machinery) and 30% on
the current account (which takes into account salaries) and do the same with a 20%-
40% combination on capital and current account and check our results. We report
results on 15%-30% combination of depreciation rates in this paper, the other
combination giving us no significant variations.
G. The Bombay Stock Exchange
Efficient markets are a prerequisite for carrying out this analysis. We stay close
to the literature on market value of innovation that was formulated adding to
findings from other studies in various economies and industries in the developed
world. Concerns about its applicability to the context of an emerging economy like
India are natural, in this section we try to take care of them. Our study focuses on
the Bombay Stock Exchange (BSE), which established in 1875 is not only among
the most prominent stock exchanges around the world, but also the oldest in Asia.
In 2006-07, stock markets like NYSE bought in ownership stakes in the Bombay
Stock Exchange. The confidence on the Indian financial system and specifically the
BSE bourses can also be evidenced from the high interest of foreign institutional
investors in the last decade. With some 3500 firms listed, BSE as of Oct‟2006 had a
total market capitalization of Rs 33.4 trillion (US $ 730 billion). If we consider our
dataset of publicly traded drug and pharmaceutical firms on the BSE, 2004 and
2005 data indicates that it constituted about 4% of that overall market capitalization
of BSE. Three pharmaceutical companies, Cipla Ltd, Dr Reddy‟s Laboratories and
Ranbaxy are part of the elite 30 stocks comprising the most popular index of the
BSE – known as the SENSEX that is widely represented by the largest firms from
other sectors in the Indian economy. In recent times, Indian stock markets are
85
witnessing quite some high-returns to investments perhaps portraying the
confidence of not only institutional but also retail investors. A recent study from the
Wharton Business School on “Financing Firms in India” (Allen, Chakrabarti, De
et.al 2006) points out that a one dollar investment at the BSE – India in 1992, would
have yielded close to a four and a half dollars, almost 15 years on, in 2006. That
was the highest return witnessed from any prominent stock exchange in the world –
SSE China gave around 1.8 dollars back, FTSE – London would have given 2.1
dollars back, Nikkei in Japan would have returned back the dollar, while S & P 500
would have given back three dollars during that period. That study also looks at the
Indian financial markets and places it at par with other English origin, French
origin, German origin or Scandinavian origin financial markets in terms of the
banking and market sector‟s activity, size and efficiency. (Refer Table 3-B page 56
of Allen, Chakrabarti, and De et.al, 2006). Finally, Bombay Stock Exchange is
competitively placed in the hierarchy of global bourses (Refer Table below).
Global stock exchanges
Stock
Exchange
Total Market Cap
(US $ billion)
Concentration
(%)
Turnover Velocity
(%)
NYSE 12, 707, 578.3 55.8% 89.8%
Tokyo SE 3,557, 674.4 56.9% 97.1%
NASDAQ 3,532,912.0 59.3% 124.8%
London SE 2,441,261.4 68.8 115.0
BSE, Mumbai 386,321.1 89.2 43.1
Source: www.fipv.com, website of international organization of stock exchanges; Allen, Chakrabarti, De et.al 2006. Concentration is share of total yearly turnover of an exchange, from
turnover of top 5% firms in terms of market cap. Turnover velocity is the total turnover for the
year expressed as a percentage of the total market capitalization.
86
III. Appendix on Firm Case Studies & Findings from India Trip23
Introduction
This appendix documents summary of conversations with firms, the
interviews being part of a field-trip undertaken in India during December 2008. Our
overall goal was to understand the evolution of firm capabilities in the Indian
pharmaceutical industry. The 14 firms ( details available on request) covered in the
trip were selected to represent the wide spectrum of capabilities in the industry. We
interviewed 5 firms (Suven Labs, GVK Biosciences, Aurigene Technologies,
Advinus Therapeutics and Jubilant Organosys) specializing in contract and/or
collaborative drug discovery research with Western partners. 5 more were Indian
generic firms, 2 among which Ranbaxy and Dr Reddy‟s were early leaders in drug
discovery in the industry, setting up their R & D centers in early 1990s. The other 3
among these five, Glenmark, Sun Pharmaceuticals and Zydus Cadila entered drug
discovery in the last decade with different business models to fund NCE (New
Chemical Entity) and NBE (New Biological Entity) research. We also spoke to 4
other firms, 1 of which was a MNC with a 50 year presence in India, Novartis‟s India
subsidiary; 2 firms focusing on India‟s growing domestic markets, USV Limited and
Emcure Pharmaceuticals. The other remaining firm in the sample was Avesthagen, a
fledgling biotech entity in the industry. The field trip was undertaken in a tumultuous
time in India‟s recent history; a fortnight after the Mumbai terror attacks riddled and
shocked the country‟s financial center, leaving the world dismayed. Over a span of
23 The India Field Trip was done in December 2008 and was supported by the Alfred P. Sloan
Foundation. We are immensely indebted to industry participants for providing us their valuable time
and inputs in the conversations. A more detailed field report is available from Chirantan Chatterjee
on request.
87
12 odd days we crisscrossed 5 Indian cities of Mumbai, Hyderabad, Bangalore, Delhi
and Ahmedabad talking to the 14 Indian firms and getting a sense of
entrepreneurship, firm capabilities and industry evolution to get a sense of the
industry‟s responses to macro and micro level changes in the last decade.
Sample Selection & Broad Issues
Our overall goal was to understand firm capabilities and in this section we
document our pre-trip preparation. Mr. D G Shah of the Indian Pharmaceutical
Alliance, a premier industry body was extremely helpful to us in arranging the
interviews. We wrote directly to the CEO of each firm and coordinated with the
CEO‟s office to set up 2-3 hourly interview slots with their top management
including their R&D managers.
Some specific questions around which we focused our conversations were:
a. Was stronger IP really an incentive for the more technologically
progressive Indian firm (and also others) to ramp up its R & D efforts and
investments? Could drug discovery have been undertaken by the modern firms as a
natural progression (from the previous weak IP regime enabled complementary
capabilities) while being agnostic to TRIPs? Did the changes in patent laws that
underwent several amendments from the original 1970 version in 1999, 2002, and
2005 influence firm strategy? What kind of an impact did the Indian economic
reforms since 1991 have on the Indian pharmaceutical firm‟s (and the industry‟s)
strategic direction?
b. While stronger IP was officially approved to be in place since 2005
as per WTO-TRIPs, India has taken an inordinately long time to get its act going in
88
terms of setting up its patent laws and administration. Might it then be true that this
delay in getting its act right in establishing the nation‟s IP laws deterred MNC entry
in setting up R & D centers within the country?
c. How is the industry sub-segmenting out in the face of the new IP
regime and global pharmaceutical changes? Do we see more
acquisitions/mergers/exits? How are the firms generating cash if they are at all
interested in doing drug discovery?
d. Are some firms able to better exploit the new business environment
to fit in to the global pharmaceutical innovation value chain? What are the prospects
of witnessing the first Indian blockbuster drug and the first Indian big
pharmaceutical firm going forward? Finally, what is the role of the Hatch-Waxman
in the US and generic market opportunities amidst all these changes in the
industry‟s capabilities?
Summary of Findings
Our discussion from here on documents the findings based on the firm-case
interviews. First, we take a broad look at how entrepreneurial orientation has guided
firm strategy and its R & D capabilities. Next, we document the overall response
from our field-trip participants on the broad response of their particular firm (and if
possible of the industry) to WTO-TRIPs, the 1991 opening of the Indian economy
and the Hatch-Waxman Act of 1984 opening US generic markets. Our discussions
also generated some interesting hypotheses, some of which we believe could be
tested in an econometric setting with future acquired data.
89
Entrepreneurship & Firm Strategy
Our 14 firms had very different entrepreneurial antecedents and our meeting
with these firms clearly indicated that firm-strategy to a large extent in these firms
wase a legacy of entrepreneurial orientation. The sample had a wide distribution. In
Mumbai for example, our first interviewed firm Glenmark had a dynamic CEO, Mr
Glenn Saldanha (son of the original founder) with past US degrees and exposure to
Western pharmaceutical firms. Mr Saldanha‟s vision was clearly visible in the
firm‟s strategy, he having guided Glenmark out of a pre-1995 history of selling
domestic formulations into being a NCE research driven firm with a separate
generics entity. At the other end of the spectrum was Mr Dilip Shanghvi, CEO of
Sun Pharmaceuticals who started his firm as a seller of Active Pharmaceutical
Ingredients (APIs) on cycles in Kolkata, India but today has guided Sun Pharma to
being one of the most aggressive generic entrants in the US market. Mr Shanghvi
clearly gave us a sense that he rather be conservative, than have high hopes on
becoming an innovative behemoth in global pharmaceuticals. Accordingly, Sun
Pharma has also spun off its R & D into a subsidiary to de-risk investor
expectations and R & D results from firm performance. At USV Limited, again in
Mumbai, we met an entrepreneurial couple, Mr and Mrs. Prashant Tewari, both
returnee migrants to India after an US education, who took up the family business
from their previous generation ever since return. The Tewaris gave us an impression
of being far more conservative than what one would expect them to be, given their
exposures to international settings. USV has been focusing strongly on its
marketing strength to open up the bottom of the pyramid markets within India. Our
travel from Mumbai to Hyderabad set us up with Mr Venkat Jasti at Suven Labs, a
90
contract research firm. Mr Jasti, who came back from US again after a long stint as
a pharmacist in the State of New York, seemed pretty clear about where he wants to
take his firm, focusing on the markets for innovation in global pharmaceuticals,
citing Eli Lilly as a strong partner for its contract and collaborative research. We
could not meet Dr Anji Reddy, founder of Dr Reddy‟s (DRL), but instead we got a
strong sense that DRL was a scientifically driven firm, having been one among the
earliest in the industry to have taken to a scientific drug discovery approach.
Moving over to Bangalore, we met Dr Rashmi Barbhaiya, an ex-scientist with a
large Western pharmaceutical firm, who left his job in the US after close to two
decades, came back and started Advinus with funding from the Tata Group. At
Bangalore again, we met Aurigene and Avesthagen, two more scientific contract
research firms, the latter having an interesting story of scientist-entrepreneurship
with its CEO Dr Billoo Morawalla Patel setting up the firm in the labs of the
National Center of Biological Sciences during her research career. In sum, the
entire spectrum was indicative clearly of differing firm directions conditional on the
entrepreneur‟s assessment of new opportunities in line with the classical
Schumpeterian world (Schumpeter 1934, 1951).
WTO-TRIPs, Indian Economic Reforms and Hatch-Waxman Act & US
Generic Markets
One key motivation for our field trip was to understand the extent of firm-
level impact of WTO-TRIPs, the opening of the Indian economy with the 1991 and
onwards reforms process and the role of the Hatch-Waxman Act; this apart from
91
getting an aggregated sense of these three exogenous shocks on the industry‟s
future strategic direction.
Firm’s response to WTO-TRIPS, Indian Reforms, Hatch-Waxman Act
FIRM NAME WTO TRIPS
INDIAN
ECONOMIC
REFORMS
HATCH
WAXMAN
ACT'1984
Glenmark Pharma √ √ √
Sun Pharma × × √
Novartis √ - Not Satisfied √ -
Emcure
Pharmaceuticals - √ √
USV Ltd. × - Not Sure √ √
Suven Labs × - -
GVK Biosciences √ - -
Dr Reddy’s × √ √
Aurigene Technologies × - -
Advinus Therapeutics √ - -
Ranbaxy √ √ √
Jubilant Organosys √× √ √
Zydus Cadila × √ √
√ - Impacted positively; × - No Impact, same business model with or without TRIPs; - Not
discussed, no comments or not applicable. √× - Impacted positively, probably in industry image
terms, but at the firm-level unaffected.
Again, we received a wide distribution of responses. In brief, WTO-TRIPs
and its implementation in India seemed to have not played the most influential role as
we thought it would have. Sure, signs of a stronger domestic IP environment gave
courage to firms like Glenmark to embark on NCE research, but ala Grossman and
Lai (2005), they seemed to be well aware that should they succeed in drug discovery,
more than domestic patenting what might matter are the markets for new products in
Western markets. Others like Sun Pharma, Dr. Reddy‟s, and Ranbaxy dead panned
that with or without TRIPs the firm would have invested in drug discovery – and
infact they did. Mr Dilip Shanghvi, Sun Pharma‟s CEO was infact provocative in
stating that „I often think about this, if without TRIPs my business model would have
been the same‟ – perhaps indicating that while WTO-TRIPs did play an overall role
92
in improving India‟s business-environment image to the world (Ranbaxy and GVK
managers noted this too), firms were not sure about how to interpret it for their own
strategic direction. Some other firms infact used the new IP scenario to their
advantage though indirectly, stating that they would do non-infringing research and
thereby win over Western clients for collaborative or contract research. We got a
sense of this in our interviews with Jubilant, Aurigene and Suven Labs. Finally,
Zydus Cadila‟s Jay Kothari was of the view that more than TRIPs, it was the
shrinking pipeline and reduced R & D productivity of Big Pharma that was impacting
the sector‟s prospects as an outsourcing destination for R & D.
For Indian Economic Reforms, the view was unanimous. Firms confirmed our
previous economic intuition that indeed, a rationalized currency regime and easy
access to foreign exchange, and more generally opening of the economy, aided firms
to think big and use economic decision making to exploit opportunities in the global
markets. This came about not only by increasing the competitiveness of
pharmaceutical products from India, but also with increased Outward and Inward
Foreign Direct Investment, and easier access to capital equipment that helped firms to
increase their productivity.24
Finally, apart from Suven, GVK Bio, Aurigene and
Advinus (pure contract/collaborative R & D entities), all our firm respondents
responded by saying that they treated the growing generics opportunity in the US
markets with the Hatch-Waxman Act to be of utmost importance. Some like Ranbaxy
asserted that learning to surmount USFDA regulations have taken time, though since
late 2003 onwards, this has infact been overcome by most first time and subsequent
24 In a related piece of research, we show econometrically, how firms are engaging in ex-ante
investments in capital equipments before entering export markets, the scale of investments being
more pronounced for entry into advanced markets like in US than in less advanced ones.
93
entrants. Further, they wanted to also look at the world generic market opportunities
in European Union and Latin America. Finally, generics cash seems to be playing a
key role in firm NCE research – at places like Ranbaxy and Zydus Cadila they
termed it as a bridging strategy - to use generic cash for funding NCE research, while
(and this is not exclusive of employing a bridging strategy) at places like Dr Reddy‟s
and Sun Pharma, generics research is being viewed to be exclusive and non-intrusive
from drug discovery – the result, spinning out of subsidiaries focusing on pure drug
discovery.
Going Forward for Indian Bio-pharmaceuticals
The firm discussions also generated several new propositions that Indian
firms are prospecting for their evolution going forward. In terms of their frequency of
occurrence, either motivated through our questions or even without, two key areas
looked to be top on the agenda of the firms we interviewed. The first was related to
the supply of human capital to do NCE research and drug discovery. Firms had a
favorable impression of public research facilities in India and even of the domestic
scientific pool but certainly were of the view that a slingshot from an exogenous
source was necessary to catapult the industry to global prominence. Related to this
discussion, one key area that emerged was the mindset of Indian R & D personnel,
being oriented towards chemistry rather than biology which firms like Advinus
believed could be a key deterrent to basic drug discovery research. Dr Rashmi
Bharbhaiya from Advinus has infact already started a drug discovery school in house
in his Pune collaborative research center where one teaches young Indian chemists
recruited to do what Dr Bharbhaiya described as think biology while doing
94
chemistry.The head of structural biology at GVK Biosciences, seconded Dr
Bharbhaiya‟s view – Dr J B Gupta, infact went one step further to state that, one
needed to tap into the global pool of Western trained bio-pharmaceutical scientists of
Indian origin and design suitable schemes to attract and retain them in domestic
Indian firms. He further noted that such human capital is especially essential in a
supervisory role to monitor work done by down-the-line R & D personnel in the
team. The second issue top of mind on firms was that, global pharmaceutical‟s model
of doing R & D, with high costs and time will progressively become untenable. We
heard several anecdotes on how Indian firms could offer a new model of
development to basic research in pharmaceuticals and locate themselves between the
continuum of large pharmaceutical firms and biotech companies of the West aided by
items like lesser iterations on New Chemical Entities, use of medium throughput
screens and nimbler decision making in R & D. While this was not clear if it will
indeed hold out in the long run, one might want to test this hypothesis going ahead.
We also touched on the following with firms during our discussion:
The role of the entrepreneurial background in guiding firm‟s strategic
direction. Firms also were open to comparisons of Indian pharmaceutical with its
Information Technology industry, citing that perhaps the former was still trying to
catapult itself into the product markets domain rather than remaining a mere services
supplier to global markets. Further, there was also a discussion on the role of rising
wages, if at all, for the different kind of firms within the industry, as also
comparisons with the Chinese and Korean industry‟s comparative advantages.
95
We also discussed the selective geographical out-licensing strategy of firms
like Zydus Cadila and Glenmark when they have a new product developed in-house.
It looks like that they do out-license their NCEs but try to retain US rights for the
same – perhaps motivated by size of the US pharmaceutical markets.
Several firms are already getting into benchmarking themselves with models
of Western Pharmaceutical firms. Glenmark for example believes that its goal was to
be a Pfizer in the cardiovascular therapeutic area while Ranbaxy looked at its Daiichi
acquisition similarly to Novartis‟s position of a generics arm within an innovator.
Firms wishing to attract the best of talent also mentioned to us that they are
using ESOPs (Employee Stock Option Plans) to attract and retain the best of human
capital in R & D. Jubilant along with GVK Bio also cited that in terms of domestic
supply of human capital in R & D, Ranbaxy and Dr Reddy‟s seem to be spurring a
healthy network effect, ex-scientists from these two firms carrying forward R & D
programs in the new generation firms.
A few Indian managers also discussed with us reasons for MNCs delaying
setting up R & D centers in India, this despite the country‟s transition to a stronger
IP. India it sems is being punished for an inordinate delay in gettings it IP
administration in place, some others cited that Big Pharma was especially reticent in
loosing control over their R & D programs, and finally, some shared that Contract
Research and Collaborative Research firms are being used as pilot projects before
seriously offshoring drug discovery R & D to India. This might be an interesting
issue to investigate in the literature on location of innovation.
96
Fiinally, the business model of new generation contract research
organizations, the CROs, like Advinus and Aurigene was an issue that provoked
much interest. We discussed on the nature of projects outsourced by Western firms to
Indian CROs and if at all Indian CROs had capabilities to work on unvalidated
targets or more complex therapeutic markets like Central Nervous System related
drugs. The answers were ambiguous. Questions were raised on how these decisions
are being made by Big Pharma and what are the Indian firms‟ comparative
advantages in the same. We discussed also on IP sharing mechanisms in such
partnerships and harped on any limits to growth for Indian CRO players. At an
overall level, this might be worth investigating especially if one wants to understand
where Indian CROs could fit, in between the Western big pharmaceutical firm and
the Western biotechnology start-up.
We hope to address many of these issues in future research going forward.
97
C. Chapter 2: Absorptive Capacity, Firm
Capabilities & Destination in Learning by
Exporting in Indian Pharmaceutical
Producers25
25 This is the earlier version of a paper co-written with Anand Nandkumar from Indian School of
Business Hyderabad. It builds on Chirantan Chatterjee‟s 2nd paper at Heinz College in 2008 titled,
“No Such Thing as a Free Lunch: Investment, Technological Upgrading, and Exports in Indian
Pharmaceuticals”. An updated version of this current work is now going to be presented at the
Strategic Management Society‟s annual conference at Miami in November 2011. This version was
awarded a Best Paper in Emerging Economies Track at the annual summer gathering of Academy of
International Business in Rio Di Janeiro, Brazil in 2010. It was also nominated for the Temple/AIB
best paper award. Accepted also for the Academy of Management‟s annual gathering in 2010 in
Montreal under the International Management track, this version is currently being updated with instruments to control for sample selection and is being prepared for journal submission. We
acknowledge very useful feedback from Ashish Arora, Paul Beamish, Lee Branstetter, Charles
Dhanaraj, Nicolai Foss, Adam Fremeth, Marty Gaynor, Guy Holburn, Samuel Kleiner, Kwok Leung,
Romel Mostafa, Lilach Nachum, Torben Pederson, Brian Richter, Chad Syverson and Mark
Zbaracki.
98
Absorptive Capacity, Firm Capabilities &
Destination in Learning by Exporting: New Evidence from
Indian Pharmaceutical Producers
Abstract
This article explores how export destinations and firm capabilities influence the
extent of learning by exporting (LBE) using a novel sample of Indian
pharmaceutical firms that exported to a variety of both advanced and emerging
destinations between 1994 and 2007. Departing from previous studies we
investigate if exports result in other gains besides improvements in technical
efficiency. We find that LBE is not restricted to technical efficiency gains alone.
Exporters also gain access to other types of knowledge that improves R&D
efficiency and the rate of new product introductions. Interestingly these gains are
more especially when firms export to high income destinations as evidenced from
higher gains accrued by firms when exporting to US versus non-US destinations.
Results also indicate that the gains are higher for more able firms. Our findings
have several managerial and policy implications and these are discussed.
JEL codes: F10, F20, O47
Key words: Learning by exporting; absorptive capacity; firm capability; Indian
pharmaceuticals
99
1. Introduction
Exports have been acknowledged to be a key contributor to economic
growth of nations. They are also an important source of competitive advantage for
firms. It has been argued that a critical, but perhaps an unintended consequence of
exporting is that they also confer upon firms, learning benefits that make exporters
more productive and by implication also more profitable than non-exporting firms.
Since knowledge tends to be spatially bounded (c.f. Kogut, 1991; Nelson, 1993),
participation in international markets is likely to facilitate acquisition of valuable
information that firms would otherwise not be privy to (see for instance Grossman
and Helpman, 1991 and 1993). Despite such theoretical consensus, empirical
evidence on whether or not firms learn by exporting is a mixed bag for a variety of
reasons. This paper is an attempt to explain some of the diverse empirical results in
the literature using a unique sample comprising of Indian pharmaceutical producers
that exported to a range of foreign destinations from 1994-2007. Our empirical
results suggest that learning by exporting critically depends on the destination
exported to as well as, the firm‟s ability to learn and apply the “new” knowledge.
Moreover, technical efficiency is not the only type of learning that accrues from
exporting. Firms also acquire new product ideas apart from better ways of
conducting R&D.
Thus, this paper contributes to a growing body of scholarly work that
empirically investigates whether exporting confer on exporters, positive
externalities from exporting. Our work contributes to this growing literature and is
also novel along three dimensions. First, we explore whether exporting results in
better R&D efficiency to exporters and whether exporting in itself is a source of
100
new product ideas. We argue that cheaper ways to manufacture products need not
be the only type of learning that arises from exporting. Exporting firms may also
learn new ways of conducting R&D especially when they operate in more
developed markets. Further, catering to the needs of more sophisticated buyers
might provide exporters with new ideas, which along with the enhanced R&D
knowledge might eventually translate into full fledged product lines. All of these
will eventually contribute to enhancing the competitive advantage of exporting
firms.
Second, we investigate the extent to which learning by exporting depends on
capabilities of firms developed ex-ante. We explore the extent to which R&D,
manufacturing and marketing capabilities influence the extent of learning from
exporting. The learning by exporting hypothesis critically hinges upon the ability of
firms to acquire and assimilate new knowledge. Third, we explore if the extent of
learning benefits from exporting depends upon the destination exported to. We
explore how the extent of learning by exporting (LBE) also depends upon the level
of sophistication of agents at the export destination.
We test our intuitions using a novel dataset of Indian pharmaceutical firms
that exported to a variety of foreign destinations between 1994 and 2007. To this
end, we draw on annual firm-level financial information along with firm-level
product market data. In addition, we use within and between variation in export
history that vary by US and non-US destinations and tease out if exports to the US
are more beneficial to an Indian exporter. As we will explain later, access to
detailed product market data provides us with the ability to tease out learning
effects from merely differences in market power enjoyed by firms. This although is
101
incidental to this paper is nonetheless an important contribution to the empirical
literature.
This draft is organized as follows. The following section provides a brief
overview of the literature. In section 3, we explain our conceptual framework and
hypothesis. Section 4 outlines the data and variables. Section 5 contains the
empirical approach and the results. We conclude in Section 6.
2. Literature and background
Our work draws from a rich literature both in economics and strategy. The
economics literature that relates to LBE thus far has not dealt with how firm
heterogeneity or more precisely, how differences in firm capabilities influence the
extent of LBE. Moreover, most empirical work in this stream estimate Total Factor
Productivity (TFP) regressions which by implication attempts to identify
productivity shifts, or technical efficiency gains from exporting. While the strategy
literature does focus on firm capabilities insofar as entry decisions are concerned,
we are unaware of studies that explicitly explores how firm capabilities influence
the extent of LBE. Also several studies in both these strands fail to separate out
differences in market power from the intended effect, productivity gains from
exporting. Also few studies have systematically explored if the extent LBE varies
by destinations. As outlined above, we explore whether LBE gains vary by
destination and ex-ante firm capabilities. Moreover we also develop several novel
measures to identify LBE effects rather than merely looking for productivity
differences. An added contribution is that we propose a method to separate out
market power effects from the LBE effects.
102
Across strategy and economics research, learning benefits from exporting
has been a topic that has attracted attention over the last few decades. Despite this,
our reading of the literature suggests that there appears to be no consensus on
whether exports confer learning benefits to firms. The results are inconclusive not
only by the context investigated – i.e. whether the focal firm is located in the
developed world vis-à-vis emerging economies - but also by the destination
exported to. While several scholars provide evidence of improvements in
productivity that arise on account of exports (Clerides, Lach, and Tybout 1998;
Bernard and Jensen 1999; Bernard and Wagner 1997; Delgado, Farinas and Ruano
2002; Bernard and Jensen 2004; Trefler 2007) others provide results that do not
support the hypothesis that exports confer learning benefits (Aw, Chung, and
Roberts ,2000; Baldwin and Gu, 2003; Van Biesebroeck, 2004; Lileeva 2004;
Hallward-Driemeier, Iarossi, and Sokoloff , 2005; Fernandes and Isgut, 2006; Park,
Yang, Shi, and Jiang, 2006; Aw, Roberts, and Winston, 2007 and De Loecker,
2009).
A number of conjectures have emerged for the lack of consensus of
the learning by exporting hypothesis. Scholars have typically relied on Total Factor
Productivity (TFP) regressions to explore the influence of exporting. This approach
relies on being able to accurately estimating productivity shifts that arise from
exporting. As has been often pointed out, TFP approach to identifying LBE effects
suffers from a variety of problems. First, there may be other gains from exporting
such as learning new ways to conduct R&D that may not show up in the TFP
regressions. Also not controlling for firm capabilities may not only introduce bias
but also neglect the fact that not all firms gain from exporting -- some may lack the
103
capabilities to absorb new knowledge or translate this learning into tangible gains.
Accordingly in our empirical analysis we not only explore if exporting firms
become more efficient, but also explore if exporters gain other types of knowledge
on account of exporting. In particular we test for whether or exports increase R&D
productivity and also whether exporters are more likely to have higher rates of
product introductions. In so doing, we attempt to ride past the TFP approach, which
probably is nothing more than a “black-box” to researchers trying to unearth
managerial implications of exporting. This we believe enables us to better identify
the mechanisms through which firms learn by exporting.
Another explanation advanced for the lack of empirical evidence for the
learning through exporting hypothesis is that the TFP regressions, (the most often
used technique to tease out learning effects) may not be adequate in contexts in
which firms enjoy substantial market power. On account of data constraints
researchers that rely on TFP regressions typically use sales revenues as proxy for
outputs. By implication, price variations are essentially treated as output variations
or productivity differences. While in a competitive market in which all firms realize
the same prices both in domestic as well as, in export markets this is perhaps less of
an issue. However, in contexts in which firms exhibit price setting behavior, TFP
regressions are inadequate to identify changes in productivity on account of LBE
because market power differences are likely to be confounded with productivity
changes. Such a situation is particularly salient in a context such as ours, wherein
many firms enjoy substantial market power in domestic markets. In cases where
export markets are more competitive, exports could then actually lower measured
TFP. Starting with Hall‟s famous paper in 1987, this literature then shows how such
104
results per se would be inadequate to conclusively test the learning by exporting
hypothesis. A variety of scholars have proposed alternatives that suit their contexts
(c.f. (Klette and Griliches 1996; De Loecker 2009; Katayama and Tybout 2009;
Foster, Haltiwanger and Syverson 2008) We depart from the literature and
investigate how a firm‟s average cost of production changes with exports in order to
identify the effects of learning by exporting. This approach is natural to our context
-- most of our firms are multi product firms producing a variety of outputs and we
observe the quantites sold by every firm in each product market. Using this
information we construct an output index and estimate cost function regressions.
The learning by exporting phenomenon is however more nuanced than just
the methodological modifications that we are offering above. Borrowing on both
strategy and economics research, we also explore how differences in firm abilities
along with destination of exports condition firm-level learning from exporting. Past
work has documented that one of the keys to assimilating new knowledge is the
absorptive capacity of firms – the capability to “evaluate and utilize outside
knowledge” which is a function of “prior related knowledge” (Cohen and Levinthal,
1990). In what to our knowledge is a first-such exercise in this literature, we
provide new evidence of how variation in absorptive capacity of exporters along
with manufacturing and marketing capabilities translates into differential gains from
exporting. In so doing, our work provides new building blocks in related literature
where scholars have investigated the extent to which absorptive capacity of firms
influences the capability of domestic firms to learn from FDI investments (Girma,
2005) or how absorptive capacity influences export participation (Harris and Li,
105
2008; Campa and Guillen, 1999), or the role of absorptive capacity in international
joint ventures (Lane Salk and Lyles, 2001).
Finally, we also explore if destination of export conditions LBE. Building
on De Loecker, (2007), who provides evidence of Slovenian manufacturing firms
learning more by exporting to high income countries, relative to exporting to low
income countries, we explore if exporting to high income destinations results in
greater learning benefits.
To summarize, ours is certainly not the first paper that tries to unearth
whether exporting results in learning benefits. Our paper however provider deeper
insights into how exports translate into enhanced competitive advantage for
exporters. This, we accomplish by not only looking at different types of learning
that can accrue from exporting (such as costs, R&D productivity and new product
introduction26
) but also in conjunction with firm capabilities (which we argue is
captured in absorptive capacity and capability measures) as well as, export
destination, we attempt in offering a more nuanced view of firm level gains27
from
exporting activity – especially through new outcome variables like in domestic
markets.
26 Throughout this paper, we use new product introductions in a manner as outlined in the variable
construction of Section 4 of this paper. 27 At this point, let us also humbly acknowledge that that we are mindful of limitations in the current draft of the paper. Future versions of this work will address selection-effects in learning-by-
exporting behavior with propensity scores and other matched sampling techniques employing
methods documented in recent literature in this area, for example see Girma Kneller and Pisu,
(2007).
106
3. Conceptual Framework and Hypotheses
In this section we motivate our empirical analysis by laying out the
hypothesis that we empirically test later in this paper.28
Several scholars in the
international trade and management literature have suggested that exporting
promotes growth in general and in particular exports leads to productivity
improvements to exporters.
3.1 Exports and learning:
Since knowledge is spatially bounded (c.f. Kogut, 1991; Nelson, 1993), one
of the unintended consequences that arises to firms from gaining access to
international markets is the acquisition of valuable information that firms would
otherwise not be privy to. Similar to previous work, we too argues that exporters are
likely to gain access to information that would enable them to become more
technically efficient, this translating them into becoming more productive
(Grossman and Helpman 1991 & 1993).
However, manufacturing may not be the only type of knowledge that
exporting firms may gain by exporting. We conjecture that exporters may also
benefit from other types of technological spillovers which include knowledge that
enable exporters to conduct R&D in a more efficient manner. Moreover, some of
this R&D knowledge could be in the nature of both process improvements that
eventually results in cost reductions and ultimately enhanced productivity while
some may also be in the nature of knowledge that relate to product innovations
which, would also lead to introduction of newer products. As our literature survey
highlighted, a preponderance of papers that relate to learning by exporting focus
28 We are currently developing a model that rationalizes these hypotheses.
107
only on productivity benefits that accrue from exporting barring a few that show
that exporters can also be privy to valuable R&D knowledge as well (Salomon &
Shaver 2005). We test not only for whether exporters gain access to intermediate
R&D knowledge that enables them to conduct R&D more efficiently, but we also
test whether exporters are in general more technically efficient as well as whether
exporters have higher rates of product introductions relative to non-exporters. We
formally state these intuitions as hypotheses below:
H1a: Exports enhance technically efficiency; exporting firms are more productive
on account of having a lower average cost of production relative to non-exporting
firms.
H1b: Exports enhance R&D efficiency; exporting firms are more efficient in
conducting R&D relative to non-exporting firms.
H1c: Exports increase the number of new product introductions; Exporting firms
have a higher rate of product introduction relative to non exporting firms.
3.2 Export destination and learning:
We also argue that exports to high income destinations are likely to confer
greater learning benefits than others. High income countries are likely to have larger
markets as also greater level of consumer expectations of quality from suppliers of
products – all of which might promote greater firm specialization (Adam Smith
1863). This implies that firms operating in such markets might just be more
technologically efficient and more sophisticated on account of having larger scale
and further having surmounted advanced market regulatory burden and consumer
quality expectations. Larger markets also imply that firms located in such
destinations are just likely to do more R&D on an average (Cohen and Klepper,
108
1992, 1996). Hence firms exporting to high income destinations are likely to be
exposed to possibly more sophisticated knowledge. Thus all else equal, exporters
selling in high income destination are likely to gain access to better quality R&D
knowledge, especially relative to exporters that export to other destinations. Also
such better R&D knowledge is likely to lead to both relatively higher efficiency as
well as higher rate of new product introductions. Formally, this is tested out in the
following manner:
H2a: Exports to high income destinations relative to exports to other destinations
enhance technically efficiency.
H2b: Exports to high income destinations, relative to exports to other destinations
enhance R&D efficiency.
H2c: Exports to high income destination, relative to exports to other destinations
increases the number of new product introductions by exporters.
3.3 Absorptive capacity, firm capabilities and learning:
All else equal, the learning benefit that accrues from exporting should also
depend on the focal firm‟s ability to assimilate and utilize knowledge. Thus a firm
with better absorptive capacity should learn more by exporting. Thus exporting
should enhance R&D efficiency especially in firms with higher absorptive capacity
than those with lesser absorptive capacities. Moreover, how a firm utilizes such
knowledge should depend both on its manufacturing as well as its manufacturing
and complementary capability. This implies that firms with better manufacturing
capability should all else equal, see greater cost reductions from exporting relative
to firms with relative inferior manufacturing ability. Also, both relative differences
in manufacturing and complementary capability should matter for how much
109
exports increase the number of new product introductions. We formally state this
intuition below (also see Figure 1 for an integrated portrait of the conceptual
model):
H3a: Exports increase technical efficiency more for firms with superior
manufacturing ability than for others.
H3b: Exports increase R&D efficiency more for firms with superior R&D ability
than for others.
H3c: Exports increase new product introduction, more for firms with superior
manufacturing ability than for others.
H3d: Exports increase new product introductions, more for firms with superior
complementary capability than for others.
Figure 1: An Integrated Conceptual Model
110
4. Data & Variable Construction
The dataset for the purposes of this paper comes from a number of sources.
4.1 Firm level data
We gathered firm data from the Prowess database provided by Center for
Monitoring of Indian Economy (CMIE). The Prowess database is similar to
Compustat database for U.S. companies providing information that incorporated
companies are required to disclose in their annual reports. From this data source, we
constructed a panel of firms relating to years 1990 or date of incorporation
whichever is earlier and this is done through 2007. The CMIE data is widely
regarded as the best available data source for Indian firms and has been used by
many scholars both in the economics and strategy literature (Khanna & Palepu
2000, Bertrand Mehta & Mullainathan 2002, Fisman and Khanna 2004).
Nonetheless the data source is not devoid of issues, some of which may have
implications for results presented in this paper. As we explain in the following
pages on our variable construction we will also briefly touch upon some of the
constraints imposed by this data source wherever appropriate.
We augmented this dataset with the following: for each firm we added the
number of Indian patents (grants and applications) from Indian patent gazette, as
well as the number of US patents (grants and applications) from the US Patent
Office (from US.PTO patent BIB and grant DVDs). Moreover, for each firm-year
combination we also identified the number of Abbreviated New Drug Applications
(ANDA) as well as Drug Mater Files (DMF) filed with the U.S. Food and Drug
111
Administration from their web site (www.fda.gov)29
. As we will explain in detail
later, we also use the Annual Survey of Indian industries (ASI), which contains
industry level labor statistics from years 1994 through 2007.30
4.2 Product market data
Our source for product data is the IMS data base, which provides detailed
information relating to the prices and units sold of about 436 pharmaceutical
products produced and sold in India. This data is collated by IMS using sales data
collected from about 3500 stockists (wholesalers) that cater to about 55,000
retailers across India. As we will describe later in this section, the IMS database
constitutes our main source from which we construct our output Index, the proxy
for output that we use in the cost regressions. We augmented the IMS data source
with the number of US patents, New Drug Applications (NDAs) for the focal
product market for a specific year and also identified if the focal product was an
Over the Counter (OTC) product.
Before describing our variables in detail, we briefly describe the Indian
pharmaceutical industry. Growing at unprecedented levels in recent years, the
pharmaceutical industry in India represented just 1% of the $550 billion global
29 It should be noted at this point that in the pharmaceutical industry context, firms usually file a
New Drug Application for a new chemical entity it might have discovered once it enters the approval
process of the USFDA. USFDA allows generic entry through filing and approval of Abbreviated
New Drug Application (for previously approved NDAs). The DMFs are indicative of bulk-suppliers
intent to supply raw materials, either to branded or generic producers of drug products. In the Indian
industry context, almost all the firms are generic producers or bulk drug suppliers, unlike the
Western world, where one might find a Pfizer filing 5 NDAs in a year for approval. Thus ANDAs
and DMFs are the best indicative proxies for regulatory approval available for Indian pharmaceutical
firms – and in absence of demand side US sales data – help us best to identify our destination
dummies in US and non-US markets. 30 Following previous literature on Indian manufacturing (Basant and Fikkert 1996, Hasan Mitra and
Ramaswamy 2007), we used the code 304 for drugs and pharmaceuticals till ASI 1997-1998 as per
the NIC 87 3 digit level. For ASI data from 1998-1999 till 2004-2005, the 4 digit industry code 2423
was used to extract total emoluments and yearly man day employees data for Indian drugs and
pharmaceuticals firms.
112
pharmaceutical industry in 2005 (KPMG, 2006). However, its share is increasing at
10% every year compared to 7 percent annual growth for the world market overall.
In terms of volume, Indian domestic sector represents just 8 percent of the global
industry total by volume, and in 2005 occupied fourth place worldwide. Exports by
Indian pharmaceutical firms have also been growing at a unprecedented rate -- drug
exports by Indian firms have been growing at an annual rate of about 30% (Indian
Government National Pharmaceuticals Policy, January 2006) from $ 87.9 million in
1990 to close to $ 3.72 billion in 2004. There are broadly two sectors within the
industry. The “organized” sector comprises of about 250 firms that account close to
70% of the pharmaceutical products produced and sold in India. The “unorganized”
sector comprises of over 20,000 firms catering to 75% of the local demand. A key
driver of export growth is the cheap cost of production – in 2005 KPMG estimated
that the average cost of pharmaceutical production was almost 50% lower than that
of western firms.
Our dataset consists of a panel of 111 drug producers (from the “organized”
sector) of the Indian pharmaceutical industry (National Industrial Classification
2423) and this panel runs between 1994 and 2007. The nature of our data mandates
that we restrict our analysis to firms that produce at least one pharmaceutical
product in the domestic market (producers, henceforth); the IMS data does not
contain products manufactured by bulk drug producers. We also exclude 3
producers that do not report salaries and wages for most years in the sample. In all,
this leaves us with 946 firm year observations consisting of 108 firms in our
sample.
113
4.3 Dependent variables
Cost (lnCit): The fact that we observe domestic output only,
constrains us to use a proxy that reflects the cost of production for domestic output.
We do not observe this in usual firm-level data. As a result we had to construct a
proxy adopting an elaborate procedure that uses the domestic sales and export
earnings to back out the total costs that relate to domestic output. (details outlined in
Appendix A and Appendix B of the paper). In short the method uses total domestic
sales and export revenue to apportion total costs of production to domestic output
and export output. In regressions we use the natural log of this variable as a
dependent variable.
R&D output: We use the number of Indian patent applications
(lnTotIndPatAppsit) filed by a firm in a year as our proxy for R&D output. From the
Indian gazette publication we collected the number of Indian patent applications
using a keyword match of firm name in the Prowess data with patent owner‟s
(assignee) name in the patent. Since the distribution of this measure is skewed we
use the natural log of this measure as one of our dependent variables.
New product introductions: From IMS data for each firm year
combination, we calculated the number of new products introduced by that firm in
that year. From this number we deducted the number of products that were retired
in that year to get the number of net product introductions in that year
(NetNewProductsit) in the domestic market.
114
4.4 Independent variables
Factor price of labor (plt): We use the cost of labor per man hour as
our proxy for the factor price of labor. In order to construct this variable we used
the RBI statistical handbook to calculate the average number of man hours per firm
per year. For each year, the RBI statistical handbook lists the total number man
hours by industry, as well as the total number of employees by industry. Using
these two values we first calculated the average number of man-hours per employee
by dividing the total man hours by the total number of employees for the
pharmaceutical industry. We then divided the deflated (by the whole sale price
Index or WPI) salaries and wages of the focal firm by the number of man hours per
employee. This gives us the average cost of labor per man-hour, pl.
Factor price of capital (pkt): We use the cost of fixed assets per man
hour as our proxy for the factor price of capital. We constructed this variable as
follows: We first calculated the cost of capital for a focal firm by applying the long
term interest rates as reported by RBI on the deflated (by WPI) gross fixed assets
reported by the focal firm. We then divided the cost of capital by the number of
man hours per employee (calculated using the RBI statistical bulletin), to arrive at
our proxy for the factor price of capital, pk.
We are mindful of the fact that the literature typically calculates cost of
capital using the focal firm‟s reported investments in new plant and equipment after
applying suitable depreciation rates. Unfortunately, in our case since Indian
accounting standards do not require a compulsory reporting of new plant and
equipment data, we resort to using gross fixed assets, deflated with the wholesale
price index, as our measure of capital stocks. It is plausible that this measure is
115
susceptible to differing accounting practices of firms. In the future drafts, we intend
exploring the robustness of this measure.
Factor price of material (pmt): We use the cost of material per man
hour as our proxy for the factor price of material. We constructed this measure
dividing the deflated material cost reported by the focal firm by the number of man-
hours per employee calculated using the RBI statistical bulletin.
Output Index (Qit): We constructed our proxy for total output of a
firm using the following procedure. This procedure takes into account the relative
complexity in manufacturing a product and weights it appropriately.
Lag exports (lagexportsit) =1 if the focal firm‟s export earnings in the
previous year exceeded 0. We use this measure as a proxy for export intensity in
regressions in which we do not distinguish between export destinations.
US export history (USExpHistit): In regressions where we do
distinguish between export destinations, we use the number of years from 1990 that
the focal firm had exported to the US and argue that this is a reasonable proxy to
measure export intensity to the US. We identified a firm as exporting to the US if
that had at least 1 abbreviated new drug application (ANDA) or an approved Drug
Master File (DMF) in that year and a non-zero firm-year export earnings
observation. Arguably, this measure cannot identify whether a firm continued to or
stopped exporting to the US once they began exporting to the US. If anything this
would imply a measurement error in this variable which would bias our estimates
towards zero. However regressions in which we used the number of ANDAs or
DMF instead of USExpHistit yield qualitatively similar results (not reported here
but available on request).
116
NonUS export history (OthExpHistit): We use a measure that
captures the number of year the focal firm had exported to other non-US
destinations in regressions in which, we distinguish between destinations. We
identified a firm as not exporting to US if it had no approved ANDAs or DMFs for
that year but did have a non-zero export-earnings observation Like with our
measure of US export history, this measure also suffers from not being precisely
able to identify if the focal firm continued to export to non-US destinations in the
subsequent years. Once again, this is likely to result in measurement error that is
likely to bias the estimates towards zero.
R&D over sales (lnRDSalesit): As the name suggests, we construct
this variable using reported R&D expenditure, divided by total sales, both the
numerator and denominator deflated by the whole sale price index (WPI) retrieved
from Reserve Bank of India website. In R&D regressions, we use the natural log of
this measure as a independent variable.
Manufacturing capability (salwageoversalesit): Our proxy for
manufacturing ability is the proportion of salaries and wages over sales, both
deflated by the WPI. We argue that a lower value of this ratio implies higher
manufacturing ability.
Absorptive capacity (lnlagTotalIndPatsit): Consistent with the
literature, our proxy for absorptive capacity is the lag number of total Indian patents
held by a firm (Zahra & George, 2002). We use the natural log of this measure in
our regressions. Since we do not have access to citations data, we were unable to
weigh these patents by the number of forward citations. However, using the number
of forward citation weighted stock of US patents yield us qualitatively similar
117
results (results can be reported on request). We believe that this is the right measure
of capturing realized absorptive capacity since very few firms in our sample own
US patents and therefore the stock of Indian patents is a practical proxy for
absorptive capacity in these sample firms.
Marketing capability (markoversalesit): We use the marketing
expenditure per Indian rupee of sale as our proxy for a firm‟s marketing ability.
From the prowess dataset, we divide the deflated marketing expenditure reported by
firms over deflated sales to construct this measure. Not all firms (32 firms, 280
observations in all are the deviant firms) report their marketing expenses.
Prior manufacturing experience (lnlagTotalOutputit): As the name
suggests we use the (log of) lagged total output to control for manufacturing
experience of a firm. We constructed this variable by aggregating the focal firm‟s
total output (output index in our case) of a firm since 1994.
5. Empirical Model and Results
In this section, we discuss the results from our empirical analysis.
Empirical Model
Our goal is to examine how exports influence exporting firm‟s technical
efficiency, R&D efficiency and the number of new product introductions. As
alluded to earlier, since the standard method used in the literature, namely TFP
regressions is plagued by a variety of issues we estimate the cost functions and test
if exporting firms have a lower average cost of production.
118
Exports and technical efficiency
We adopt a procedure outlined in Berndt 1991, to derive our estimation
equations for the cost function regressions. Suppose a firm i's total output in a year
t, Qit is a function of three inputs, labor (Lit), capital (Kit) and material (Mit).
Further,
let us suppose that the production technology is given by .
In order to allow for differential productivity between exporters and non-exporters
as well as differential firm abilities, we modify the production function as
where Ait is a firm specific efficiency component that
depends on that firm‟s export intensity, manufacturing ability and stock of
knowledge. Given this production technology, if pj is the price of output that a firm
i realizes for product j the corresponding cost minimization problem for the firm
that sells N products is just
(1)
Suppose plt, pkt and pmt are the factor prices of labor, capital and materials in
period t. If the returns to scale parameter, , and kit a
constant which is equal to it can be shown that the
corresponding cost function for this technology is
just . Taking logs, the cost function of a firm i in
period t is just
(2)
119
where uit is the iid error term. We impose the homogeneity (of degree 1) assumption
by setting in (2) above. This results in the following estimating
equation
(3) -
where
. As we will explain later, in line with our hypotheses we use several different
proxies for ait in order to tease out how export intensity influence costs. We use the
output index described in the data section as a proxy for Qit and the total costs that
relates to domestic output as a proxy for Cit. Under column 2 of table 1, we report
the results of this specification. This specification suggests that r is marginally
above 1. In column 3, of the same table, we explore the influence of exports on cost
of production of domestic output. To this end, we let and if
, we estimate the following regression specification
(3a)
Results of this specification shown under column 3 of table 1 suggest that
firms that exported in the previous year are more technically efficient. From
specification 2, firms that exported in the previous year are likely to have about 5-
6% lower cost of domestic output. Interestingly r in this specification is much closer
to 1, suggesting that the cost function exhibits constant returns to scale. This result
supports H1a that suggested that exporting firms must be more technically efficient
than their non-exporting counterparts. In column 4, by letting
120
, , we explore for any destination
specification learning effects by estimating
(4)
Our results support H2a. Exporting to US markets appear to be more
beneficial relative to exporting to other export markets. While a one year of
exporting to US is associated with about a 12% decrease in total cost of domestic
output, a similar increase is associated with only a 8% decrease in total cost of
domestic output. Exporting to the US thus is associated with about a 3% lower cost
of domestic production (std. err – 0.01; p-val 0.01). All other coefficients appear to
be similar to those of the earlier specifications.
A key concern in specifications 1 through 3 of table 1, is that the difference
in capability of firms that are also time varying could plausibly lead to biased
estimates of influence of exports. Stated otherwise, it is plausible that our estimates
are biased due to accounting for firm specific time varying differences that
influence total output. In specification 4, we additionally control for firm ability by
using laggedoutputit as an additional regressor. More specifically, we estimate:
---- (5)
While our other coefficients are largely unchanged, exports to US continue
to be more beneficial even in this specification. While a one year of exporting to US
is associated with about a 9% decrease in total cost of domestic output, a similar
increase is associated with only a 5% decrease in total cost of domestic output.
Exporting to the US thus is associated with about a 4% lower cost of domestic
121
production (std. err – 0.01; p-val 0.01). In specification 5 we further control for
differences in manufacturing ability using salwagesoversalesit. That is in effect an
investigation of whether firm-pedigree, that is firms that are more capable learn
more by exporting. To this end, in specification 5, we also use two interacted
variables -- salwagesoversalesit*USexphistit and salwagesoversalesit*Othexphistit
and estimate:
(6)
<Insert Table 1 about here >
Results of specification 6 are shown in column 6 of table 1. Once again,
results suggest that export to the US is more potent in reducing cost of production
relative to exporting to other destinations. While an additional year of exporting to
US is associated with about an 8% decrease in total cost of domestic output, a
similar increase is associated with only a 1% decrease in total cost of domestic
output. Moreover, firms with better manufacturing capability have a lower cost of
production on an average. Further more capable firms are likely to benefit more
from exporting – a firm that has a 10% point lower labor cost (salary and wages) as
a proportion of sales has a 3.6% lower cost of domestic output for every additional
year of exports to the US and a 1.7% lower cost of domestic output for every
additional year of exports to other destinations. These results support H3a, which
stated that firms with superior manufacturing capability, should all else equal,
122
become even more technically efficient relative to a firm with inferior
manufacturing capability.
Exports and R&D efficiency
Our next level of analysis involves looking at an alternative mechanism of
learning from exporting -- using Indpatappsit as a dependent variable, we now test
how exports influence R&D efficiency. To this end we estimate R&D production
functions, results of which are shown in table 2. As with cost regressions, we start
with a base specification, in which we only include RDsalesit and time dummies.
Additionally we also control here for another firm-pedigree relative variable, that
for realized absorptive capacity (in the spirit of Zahra and George 2002) using
lnlagTotalIndPatst. Specifically if Kt denotes time dummies and αi time fixed
effects, we estimate
Our regressions control for firm fixed effects. Not surprisingly results
suggest that firms with higher R&D outlays also have more patent applications.
Also firms with better absorptive capacity are also likely to be more productive. In
specification 2, we additionally include a lagexportsit and then test H1b by
estimating:
The results of this specification are shown under column 2 of table 2. The
coefficient on the lag export dummy is insignificant. Moreover, the coefficient on
the lag export dummy is also small. However, the point estimates suggest that a
123
firm that exported in the previous year is likely to have higher number of Indian
patent applications – exporters in the previous year are likely to have 1% higher
patent applications. We thus do not find overwhelming support for H1b. In
specification 3 we make two modifications to specification 2. First instead of using
a single dummy variable we use USexpHistit and OthExpHistit. We thus explicitly
distinguish between export destinations to test H2b. with these changes out
estimating equation is as follows:
Results of this specification are shown under specification 3 of table 2. We
note here that the point estimates of both USexpHistit and OthExpHistit are positive
but, only the coefficient of USexpHistit is significant. Our results suggest that a firm
with an additional year of US export experience is likely to have about 10% higher
patent applications. The point estimate of OthExpHistit however suggests that an
additional year‟s export to other non-US destinations increases the patent
application rate by only 1%. Interestingly as H2b conjectured, the difference is
positive and significant – exporting to the US is more beneficial than exporting to
any other destination. Exporting to the US translates to a 10% additional patent
application rate (std. err 0.02; p-val-0.00). Other coefficients are largely unchanged
when compared to our previous specifications.
To test H3b, we employ specification 4, wherein we additionally interact our
proxy for absorptive capacity lagTotIndPatsit with USexpHistit and OthExpHistit.
The new estimating equation then is:
124
Results of this specification are shown in specification 4 of table 2. While
the point estimates of both interactions are positive, only the interaction with
is significant. From specification 4, an
additional year of exporting to the US is likely to increase the number of patent
application by 8% more for a firm that had 1% higher stock of Indian patents. A
similar increase however is associated with about a statistically insignificant
increase of only 2%. Even in this specification exporting to the US appears to be
more beneficial than exporting to other destinations. While an additional year of
exporting is associated with about a 9% increase in the propensity to apply for
Indian patents, a similar increase is associated with only a statistically insignificant
1% increase in patent application propensity. As in the case of our earlier results the
difference however is significant, suggesting that exporting to the US is more potent
in increasing patent applications relative to exporting to any other destination. Thus,
we find support for both H2b and H3b, while we only find partial support for H1b.
< Insert Table 2 about here >
Exports and new product introductions:
We now test our hypotheses that relates to the number of new product
introductions. We use NetNewProductsit as a dependent variable and explore how
export activity affects the number of new product introductions. Results of these
regressions are shown in table 3. We start with a base specification that includes
125
only lagexportsit and firm fixed effects and time dummies. Using Kt to denote time
dummies and αi time fixed effects, in column 2 of table 3, we estimate
As suggested by H1c, firms that exported in the previous year have a higher
rate of product introductions. Our results suggest that an export in the previous year
is associated with about a 9% increase in the number of new product introductions.
In specification 2, we distinguish between exports to US and to other countries
using USExpHistit and OthExpHistit. We thus estimate:
(12)
Results of this specification suggest that an additional year of exporting to
US is associated with about a 17% increased new product introductions, whereas
exporting to other destinations is associated with only a 9% more product
introductions. Thus a marginal extra year of exports to US is associated with about
a 8% more product introductions (Std err. 0.04; p-value 0.04). This result supports
H2c. In specification 3, we further control for the focal firms‟s manufacturing and
marketing capabilities as well as previous production experience and find that our
results are largely similar to that of specification 2. Specifically we estimate,
…………………………………………….(13)
Results of this specification also suggest that exports to US are associated
with higher number of product introductions. Moreover as expected, better
marketing and manufacturing capabilities are associated with higher number of
126
product introductions. Further previous manufacturing experience is also associated
with higher number of new product introductions. In specification 4, we test H3c.
To this end, we interact salwageoversalesit, and markoversalesit with USExpHistit
and OthExpHistit. We thus estimate
The interaction term of US export history with our proxy for manufacturing
capability is positive and significant. However the interaction of non-US export
history term with manufacturing ability is positive but not significant. While
marketing per se is associated with higher number of product introductions, the
interaction terms are very small and insignificant. This suggests that most
marketing knowledge required to be more profitable may in fact be local and
exporting may not be a source of marketing knowledge that can be profitably
applied in an Indian context. Thus while our results support H3c we do not find
support for H3d.
< Insert Table 3 about here>
127
Figure 2 – Results Summary: New Mechanism of Firm-Learning by Exporting
To summarize our results, Figure 2 indicates that at all the 3 levels
investigated, this analysis provides some fresh insights and support to the
hypotheses investigated. Firm pedigree as captured by its marketing and
manufacturing capability along with absorptive capacity plays an influential role in
ensuring that exporters can appropriate the benefits from exporting – be that
through cost reductions, enhancing R&D efficiency or introducing new products in
domestic markets. In addition, the pedigree dimension is also influenced by the
destination of exports, in general the H2 series of hypotheses were validated,
indicating that exports to the US relative to non-US destinations, enhanced
technical efficiency, improved R&D efficiency and enabled firms to launch new
128
product introductions in domestic markets. Off course, in all of the above outcome
variables, exporting per se turned out to be influential for firm-level learning,
whether one was looking at cost reductions or aiming to measure new product
launches in domestic markets. Having said that, we don‟t find conclusive evidence
to support H1b, i.e. exporting per se, didn‟t enhance R&D efficiency at the very
broad firm-level. Why might that be so? Here is an explanation. To the extent that
our R&D productivity regression was using as an input measure the amount of
R&D spent as a fraction of sales, given the generic nature of our firms, one could
expect that most of the outputs of R&D got translated into ANDAs and DMFs or at-
best some process-improvements in the form of New Drug Delivery Systems that
were only sometimes patented at the US Patent Office. This and also anecdotal
evidences suggest that exporting firms can generate more cash from even US-
generic market sales which could be pumped back into the R&D process, but
whether that should be measurable in the form of Indian patents or through ANDA
and DMF atocks or even US Patent stocks remain debatable.
We are working on ironing out robustness checks of the R&D regressions
using US Patent Application Stocks on the LHS (this despite its sparse presence in
only a select few technologically progressive Indian drug firms) and hope that this
alternative measure might stand up to the scrutiny to validate H1b. At this point,
even at the cost of repetition, we would again like to acknowledge the limitations of
our work in controlling for selection and commit to report in future versions robust
results with propensity scores to address that issue.
129
6. Discussion, Extensions & Conclusion
Our work provides new results on learning by exporting from an emerging
economy context, India and that too in an R&D intensive industry, pharmaceuticals.
In this paper, we make several contributions to the literature.
First, steering clear of the market power issues that plague identification of
learning benefits from exporting we identify how exporting benefits firm‟s
technical efficiency. This method enables us to understand the influence of
exporting without the interference of market power issues. This to our knowledge is
the first such attempt with costs in this literature and is essentially made possible
since in our context as we also observe output of firms. We find evidence that
exports decrease average cost of production and influences technical efficiency.
Second, we also identify other mechanisms of learning by exporting. We
investigate if exporting promotes R&D efficiency as well as, the new products
introductions. We also additionally investigate how the exporter‟s manufacturing
and marketing capability influence all three facets of learning. We find that
absorptive capacity of a firm moderates the amount of learning from exporting –
firms with higher absorptive capacity appear to be learning more R&D related
knowledge than those with relatively lower absorptive capacity. Moreover, firms
with better manufacturing capability appear to benefit more from exporting; Better
firms not only become more technically efficient, but they also appear to have more
new product introductions. However differences in marketing ability r do not
appear to differentially benefit exporters.
130
Finally, we also find strong evidence that learning from exporting depends
on the destination to which exported. More specifically, we find evidence that
suggests that exporting to developed countries all else equal results in higher
amounts of learning perhaps due to presence of more sophisticated agents in
developed countries.
As with other work, our paper has many limitations. Our first two
limitations are essentially imposed by the nature of the data. First, not all firms
report data on marketing and our results are based on firms that report marketing
expenses. Second, some of our measures such as export history may introduce bias
in our estimates; however we do find evidence of exporting being beneficial and
any variable that alleviates this error such only find a higher impact from exporting.
Finally we are also mindful of endogeneity issue that is typical in this literature31
–
exporting firms may be capable and this in fact may result in tem entering export
markets especially the more sophisticated markets. In the future drafts of this paper,
we hope to address some limitations by undertaking additional robustness checks.
Finally, we also plan to validate our empirical results with interviews with members
in the industry.
To validate our research findings we also take our hypotheses to executives
in Indian pharmaceutical firms. The Vice President Corporate Strategy of a leading
Indian firm buys our notion that US-exports actually results in more gains than from
other-exports, he notes: “It is especially important since it gives cash opportunities
to firms with ambitions and capabilities to fund drug discovery research.”
31 We are currently developing instruments to address this issue
131
This clearly indicates how advanced market opportunities like those in the
US are important to emerging economy firms. Head of Pharmaceutical Research at
one of the top 2 Indian drug firms, and himself a scientist who returned to India
after working in a large US pharmaceutical firm, also offers insights that further
strengthens our capability arguments:
“Indian firms took time to learn advanced market regulations and as
economic reforms dawned in India it became easier for them. However, the early
players in export markets especially to the US could use their previous experience
of US-presence as a bridging strategy to engage in all kinds of profit-enhancing
behaviors.”
Finally, CEO of a domestic Indian player agrees that presence in export
markets, especially moderated by destinations and capability of firms, could
actually influence “new product introduction capabilities” in the industry. In sum,
combining our findings with on-field anecdotal observations, we are confident that
our paper not only provides insights into whether or exporting results in learning
but also on which types of knowledge is gained from exporting and how a firm‟s
capabilities and ability to assimilate knowledge matter for learning by exporting to
result in enhanced competitive advantage. This, to our knowledge is serious
substantial contribution to the literature on learning by exporting. That said, like
most scholarly work, much more remains to be done going forward.
132
References
Alvarez, R. 2005. Exporting and performance: evidence from Chilean plants. The
Canadian Journal of Economics/Revue canadienne d'Economique 38, no. 4: 1384–
1400.
Arrow, K. J. 1962. The economic implications of learning by doing. The review of
economic studies 29, no. 3: 155–173.
Aw, Bee Yan, Sukkyun Chung, and Mark J. Roberts, 2000. Productivity and
turnover in the export market: Micro evidence from Taiwan (china) and the
Republic of Korea. World Bank Economic Review 14(1):65–90.
Aw, Bee Yan, Mark J. Roberts, and Tor Winston. 2007. Export market
participation, investments in R&D and worker training and the evolution of firm
productivity. World Economy 14(1):83–104.
Baldwin, J. R, and W. Gu. 2003. Export-market participation and productivity
performance in Canadian manufacturing. The Canadian Journal of
Economics/Revue canadienne d'Economique 36, no. 3: 634–657.
Bernard, Andrew B. and J. Bradford Jensen. 1999. Exceptional exporter
performance: Cause, effect, or both? Journal of International Economics 47(1):1–
25.
Bernard, Andrew B. and J. Bradford Jensen. 2004. “Exporting and productivity in
the U.S”. Oxford Review of Economic Policy 20(3):343–357.
Bernard, Andrew B. and Joachim Wagner. 1997. Exports and success in German
manufacturing. Weltwirtschaftliches Archive 133(1):134–157.
Berndt, ER. 1991. The Practice of Econometrics. Addison-Wesley.
Bertrand M, Mehta P. & Mullainathan. 2002. Ferreting Out Tunneling: An
Application to Indian Business Groups. Quarterly Journal of Economics. 117(1)
pg:121-148.
Bigsten, A., P. Collier, S. Dercon, M. Fafchamps, B. Gauthier, J. W Gunning, A.
Oduro, et al. 2004. Do African manufacturing firms learn from exporting? Journal
of Development Studies 40, no. 3: 115–141.
Blalock, G., and P. J Gertler. 2004. Learning from exporting revisited in a less
developed setting. Journal of Development Economics 75, no. 2: 397–416.
Branstetter, L., 2001. Are knowledge spillovers international or intranational in
scope?. Journal of International Economics, 53(1), 53-79.
133
Campa JM and Guillen MF, 1999. “The Internalization of Exports: Firm and
Location-Specific Factors in a Middle-Income Country.” Management Science.
Castellani, D. 2002. Export behavior and productivity growth: evidence from Italian
manufacturing firms. Review of World Economics 138, no. 4: 605–628.
Clerides, S. K, S. Lach, and J. R Tybout. 1998. Is Learning by Exporting
Important? Micro-Dynamic Evidence from Colombia, Mexico, and Morocco*.
Quarterly Journal of Economics 113, no. 3: 903–947.
Cohen WM and Klepper S, 1992. The anatomy of industry R&D distributions. The
American Economic Review. 82(4), pg 773-799.
Cohen WM and Klepper S, 1996. A reprise of Size and R&D. The Economic
Journal, 106(437), pg 925-951
Cohen WM and Levinthal DA, 1990. Absorptive Capacity: A new perspective on
learning and innovation. Administrative Science Quarterly. 35(1).
De Loecker J, 2007. Do exports generate higher productivity? Evidence from
Slovenia?. Journal of International Economics. 73(1). Pg 69-98.
De Loecker J & Warzynski, S. 2009. Markups and Firm-level Export Status. NBER
Working Paper # 15198.
Delgado, Miguel A., Jose C. Fariñas, and Sonia Ruano. 2002. Firm productivity and
export markets: A non-parametric approach. Journal of International Economics
57(2):397–422.
Fernandes, Ana M. and Alberto E. Isgut. 2006. Learning-by-exporting effects: Are
they for real?
Mimeo, The World Bank and Institute for Competitiveness & Prosperity.
Fisman R and Khanna T, 2004. Facilitating Development: The role of Business
Groups. World Development.
Foster L, Haltiwanger J and Syverson C, 2008. Reallocation, Firm Turnover, and
Efficiency: Selection on Productivity or profitability. American Economic Review.
98(1). pg 394-425.
Girma, S, Kneller, R and Pisu M. 2005. Exports versus FDI: an empirical test.
Review of World Economics.
Greenaway, D., and R. Kneller. 2007. Firm heterogeneity, exporting and foreign
direct investment. Economic Journal 117, no. 517: F134–F161.
134
Grossman GM and Helpman E. 1991. Trade, Knowledge spillovers and growth.
NBER working paper 3485.
Iacovone, Leonardo and Javorcik, S. Beata, 2008: "Getting Ready: Preparation for
Exporting," Working Paper, World Bank.
Isgut, A. E. 2001. What's different about exporters? Evidence from Colombian
manufacturing. Journal of Development Studies 37, no. 5: 57–82.
Katayama, H., Lu, S. and Tybout, J. (2003). “Why Plant-Level Productivity Studies
are Often Misleading, and an Alternative Approach to Inference, NBER WP.
Keller, 2002. Geographic Localization of international technology diffusion.
American Economic Review. 92(1).
Khanna, Tarun and Palepu, Krishna. 2000. Is group affiliation profitable in
emerging markets?
An analysis of diversified Indian business groups. Journal of Finance, 55(2), April,
867- 893.
Klette, T.J. and Griliches, Z (1996). “The Inconsistency of Common Scale
Estimators When Output Prices are Unobserved and Endogenous”, Journal of
Applied Econometrics, vol 11 (4), 343-361.
Lane, P. J., Salk, J. E., & Lyles, M. A. 2001. Absorptive capacity, learning, and
performance in international joint ventures. Strategic Management Journal, 22:
1139–1161.
Hallward-Driemeier, Mary, Giuseppe Iarossi, and Kenneth L. Sokoloff. 2005.
Exports and manufacturing productivity in East Asia: A comparative analysis with
firm-level data. Mimeo, The World Bank and UCLA.
Levinthal, D. A. and J. G. March (1993). 'The myopia of learning', Strategic
Management Journal, Winter Special Issue, 14, pp. 95-112.
Meller, P. 1995. Chilean export growth, 1970-90: An assessment. Manufacturing
for export in the developing world: problems and possibilities 1970: 21.
Kogut, B., & Zander, U. 1992. Knowledge of the firm, combinative capabilities,
and the replication of technology. Organization Science, 3: 383-397.
Lileeva, Alla & Trefler, Daniel, 2007. Improved Access to Foreign Markets Raises
Plant Level
Productivity... for some Plants, NBER Working Paper.
Nelson, R., & Winter, S. 1982. An evolutionary theory of economic change.
Cambridge, MA: Belknap Press of Harvard University Press.
135
Park, Albert, Dean Yang, Xinzheng Shi, and Yuan Jiang. 2006. Exporting and firm
performance:
Chinese exporters and the Asian financial crisis. Mimeo, University of Michigan.
Smith A, McCulloch JR, 1863. An Inquiry into the Nature and Causes of the
Wealth of Nations.
Van Biesebroeck, J. 2005. Exporting raises productivity in sub-Saharan African
manufacturing firms. Journal of International economics 67, no. 2: 373–391.
Zahra AS and George G, 2002. Absorptive Capacity: A Review,
Reconceptualization and Extension. Academy of Management Review 27(2). 185-
203.
136
Table 1 – Cost function regressions, dependent variable, lnCit
BASE REGS.
WITH
EXPORTS
WITH EXP.
HISTORY
WITH EXP.
HISTORY
& LBD32
WITH EXP.
HISTORY &
LBD, CAP
pl 1.62 ***
1.59 ***
0.94 ***
0.82 ***
0.88 ***
(0.13) (0.13) (0.08)
(0.04)
(0.09)
lnQit 0.71 * 0.88
** 0.86
*** 0.87
*** 1.02
***
(0.41) (0.42) (0.40)
(0.41)
(0.35)
pk 0.22 ***
0.24 ***
0.27 ***
0.19 ***
0.21 ***
(0.04) (0.04) (0.04) (0.08)
(0.05)
USExpHistit
-0.12 ***
-0.09 **
-0.08 ***
(0.00) (0.01)
(0.01)
OthExpHistit
-0.08 ***
-0.05 ***
-0.01
(0.01) (0.01) (0.03)
Lagexportsit
-0.06
*
(0.03)
Lag lntotal outputit
-0.06 ***
-0.08 ***
(0.01) (0.01)
salwageoversalesit
-0.26
**
(0.11)
salwageoversalesitXUSExpHistit
-0.36
*
(0.18)
salwageoversalesitXOthExpHistit
-0.17
***
(0.04)
Constant 16.02 ***
16.48 ***
15.08 ***
13.21 ***
14.17 ***
(0.43) (0.47) (0.30) (0.26) (0.32)
N 946 946 946 946 946
# firms 108 108 108 108 108
Time dummies(12) Y Y Y Y Y
R2 overall 0.89 0.89 0.78 0.74 0.69
R2 within 0.45 0.45 0.52 0.51 0.59
R2 between 0.95 0.94 0.56 0.55 0.51
Test for CRS 1.42 * 1.14
** 1.16
*** 1.14
*** 0.98
**
(0.81) (0.54) (0.36)
(0.25)
(0.47)
USExpHistit - OthExpHistit
-0.03 ***
-0.05 ***
-0.07 **
(0.01) (0.01) (0.03) Notes: Standard errors in parenthesis. * Significant at 90% level. ** Significant at 95% level. *** Significance at 99%
level. All specification include 12 time dummies.
32
LBD – refers to controls for „learning by doing‟ in all our regression specification and tables using
the variable Lag lntotal outputit
137
Table 2 – R&D regressions, Fixed Effects, dependent variable lnIndpatappsit
Notes: Standard errors in parenthesis. * Significant at 90% level. ** Significant at 95% level.
*** Significance at 99% level. All specification include 12 time dummies.
BASE REGS.
WITH
EXPORTS
WITH EXP.
HISTORY
WITH EXP.
HISTORY &
LBD, CAP
lnLagRDsalesit 0.06 ***
0.08 ***
0.05 **
0.04 **
(0.02) (0.02) (0.02)
(0.02)
lnlagIndpatsit 0.30 ***
0.41 ***
0.23 ***
0.19 ***
(0.03) (0.02) (0.03)
(0.04)
USExpHistit
0.11 ***
0.08 ***
(0.03) (0.03)
OthUSExpHistit
0.01 0.01
(0.02) (0.02)
Lagexportsit
0.01
(0.03)
lnlagIndpatsit
XUSExpHistit
0.08 *
(0.04)
lnlagIndpatsit
XOthExpHistit
0.02
(0.03)
Constant -0.06 -0.02
-0.07
*** -0.07
***
(0.07) (0.09) (0.08) (0.08)
N 946 946 946 946
# firms 108 108 108 108
R2 overall 0.51 0.52 0.51 0.51
R2 within 0.26 0.25 0.26 0.27
R2 between 0.78 0.78 0.76 0.77
USExpHistit - OthExpHistit
0.10 ***
0.08 **
(0.02) (0.03)
138
Table 3 – New Product Regressions, dependent variable lnNetNewProdIntit
EXPORTS HISTORY
HISTORY
AND CAP.
WITH EXP.
HISTORY &
LBD, CAP
Lagexportit 0.09 ***
(0.01)
USExpHistit
0.17
*** 0.04
*** 0.13
***
(0.03) (0.01)
(0.05)
OthExpHistit
0.09 ***
0.01 0.02
(0.02) (0.01) (0.04)
markoversalesit
0.06 **
0.07 **
(0.02) (0.03)
salwageoversalesit
0.11
0.08
**
(0.04) (0.04)
lnlagTotalOutputit
0.03 ***
0.03 **
(0.01)
(0.01)
markoversalesit*USExpHistit
0.01
(0.02)
markoversalesit*OthExpHistit
0.01
(0.03)
salwageoversalesit*USExpHistit
0.06 ***
(0.02)
salwageoversalesit*OthExpHistit
0.03
(0.02)
Constant 3.52 ***
3.52 ***
3.32 ***
3.33 ***
(0.02) (0.02) (0.11) (0.10)
N 946 946 666 666
# firms 108 108 76 76
Time dummies & FE (12) Y Y Y Y
R2 overall 0.03 0.03 0.30 0.31
R2 within 0.02 0.03 0.20 0.30
R2 between 0.05 0.06 0.31 0.30
Test USExport - OtherExport=0
0.08 **
0.03 ***
0.11 **
(0.04) (0.01) (0.06) Notes: Standard errors in parenthesis. * Significant at 90% level. ** Significant at 95% level. *** Significance
at 99% level. All specification include 12 time dummies.
139
Table 4: Descriptive Statistics of the Variables
Variable N Mean Source of
variation
Std. Dev
lnCit 946 9.84 Firm, year 1.42
lnIndPatAppsit 946 0.29 Firm, year 0.79
lnNetNewProdIntit 946 3.64 Firm, year 0.86
lnRipt 201722 2.27 Product, firm,
year
3.62
lnNDAcountit 201722 1.48 Product, year 1.64
lnUSPatCountit 201722 2.65 Product, year 3.59
OTCit 201722 0.02 Product, year 0.13
pl 946 0.21 Firm, year 0.37
pk 946 0.74 Firm, year 3.14
pm 946 0.86 Firm, year 3.73
Qit 946 11.78 Firm, year 39.18
lagexportsit 946 0.53 Firm, year 0.49
USExpHistit 946 0.58 Firm, year 1.67
OthExpHistit 946 3.85 Firm, year 4.12
RDSalesit 946 4.34 Firm, year 3.77
salwageoversalesit 946 0.07 Firm, year 0.09
lnlagTotalIndPatsit 946 0.48 Firm, year 1.12
markoversalesit 666 0.36 Firm, year 0.45
140
Appendix
A. Backing out Costs for domestic output as one of our dependent variable
We first start by aggregating the (deflated) total salaries and wages (deflated
by wholesale price index), material expenses (deflated by raw material deflator) and
cost of capital. We calculate the cost of capital for a firm year by applying the
appropriate long term interest rate provided by the Reserve Bank of India ( RBI,
henceforth and available at www.rbi.org.in) for that year on the total fixed assets
reported by the focal firm. We use this as our proxy for the total cost of output (both
domestic and export) produced by a firm. In order to separate out the material cost
that relates to domestic quantity from total material cost, we start with decomposing
the total cost incurred by a firm i at time t, Pit as:
,
where xit represents the average cost per unit of output with qitD representing
the domestic output and qitw representing the export output of firm i in a year t.
Since we only observe export earnings and total domestic sales (total sales – export
sales) we write the above expression in terms of domestic and export sales. That we
further decompose Pit as
=
Since is domestic sales (Dit) and is export sales (Eit), taking
logs we regress
141
where εit represents an iid error term. We finally compute the share of cost
related to domestic sales, is just:
where K is the smearing factor (Duan, 1983). All our cost regressions use Cit
as a proxy for total cost of domestic output. The natural log of this measure is used
as one of our dependent variables.
B. Using Demand Side Data for a Novel Domestic Cost Measure
This section in the appendix outlines our methodological contribution to the
learning by exporting literature. As we have argued, using domestic costs was the
approach we adopted, but further, obtaining predicted domestic costs in themselves
required us to gather some extra variables that pertained to the pharmaceutical
industry and adopt an estimation framework, both of which are outlined below.
First, we note the variable construction and additional data collection procedure:
Product market variables: We first describe all the variables that we
use to construct one of our key independent variables, out index (Qit) that we use to
back out our domestic cost measures.
a. Total revenues (lnRipt): For each product market firm pair, IMS data
reports the gross revenues realized by the focal firm in that product market for that
year. As we will describe later, we use the natural log this variable as a dependent
variable in the regression that we use to construct the output index (Qit)
b. lnUSpatcountpt : Represents the natural log of the number of US
patents for a particular product market p at a point in time t. For each product
market year combination we added the number of citation weighted granted patents
from the US.PTO patent DVD. In order to do so, we used the services of a bio-
142
chemist who identified the chemical name(s) for each product market. Using these
identified chemical names, we did a key word match on the abstract and claims
using the US.PTO DVD to identify the US patents for that product market. We then
forward-citation weighted these patents using the procedure described in
Trajtenberg, (1999).
c. lnNDACountit: Represents the natural log of number of NDAs for
the focal product market at time t. From the US FDA web site (www.fda.gov), we
collected the number of NDAs for a particular product market. Once again we used
the identified chemical names and matched these chemical names with the “key
ingredient” field of NDAs.
d. OTCp=1 if the product was an OTC drug. From the list of Over the
Counter (OTC) products available at the FDA web site, we matched whether a
product was an OTC product. Once again we matched the chemical name with the
active ingrient of a OTC product and created a dummy=1 if the product was an
OTC product.
e. ndaNApt =1 if the focal product did not have any NDA for that year.
Of a total of 436 product markets, we could not match NDA for 32 markets.
f. USpatNApt=1 if we did find a match in the US.PTO database for the
focal product market at time t.
We now describe the key dependent variables we use in our empirical
analysis.
Factor price of labor (plt): We use the cost of labor per man hour as
our proxy for the factor price of labor. In order to construct this variable we used
143
the RBI statistical handbook to calculate the average number of man hours per firm
per year. For each year, the RBI statistical handbook lists the total number man
hours by industry, as well as the total number of employees by industry. Using
these two values we first calculated the average number of man-hours per employee
by dividing the total man hours by the total number of employees for the
pharmaceutical industry. We then divided the deflated (by the whole sale price
Index or WPI) salaries and wages of the focal firm by the number of man hours per
employee. This gives us the average cost of labor per man-hour, pl.
Factor price of capital (pkt): We use the cost of fixed assets per man
hour as our proxy for the factor price of capital. We constructed this variable as
follows: We first calculated the cost of capital for a focal firm by applying the long
term interest rates as reported by RBI on the deflated (by WPI) gross fixed assets
reported by the focal firm. We then divided the cost of capital by the number of
man hours per employee (calculated using the RBI statistical bulletin), to arrive at
our proxy for the factor price of capital, pk.
We are mindful of the fact that the literature typically calculates cost of
capital using the focal firm‟s reported investments in new plant and equipment after
applying suitable depreciation rates. Unfortunately, in our case since Indian
accounting standards do not require a compulsory reporting of new plant and
equipment data, we resort to using gross fixed assets, deflated with the wholesale
price index, as our measure of capital stocks. It is plausible that this measure is
susceptible to differing accounting practices of firms. In the future drafts, we intend
exploring the robustness of this measure.
144
Factor price of material (pmt): We use the cost of material per man
hour as our proxy for the factor price of material. We constructed this measure
dividing the deflated material cost reported by the focal firm by the number of man-
hours per employee calculated using the RBI statistical bulletin.
Output Index (Qit): We constructed our proxy for total output of a
firm using the following procedure. This procedure takes into account the relative
complexity in manufacturing a product and weights it appropriately. Recall, that the
IMS data contains data on total units sold of each product manufactured by a firm in
a year, as well as the total revenues that a firm earned in that market. Using these
we start by writing a firm‟s total revenue by selling a product j in time t as function
of three components: marginal costs, markup and quantity. Thus if γijt is the price
plus the markup realized by firm i through selling product j, Rit its total gross
revenues in that product market and qijt the quantity sold by that firm in that market,
Taking logs this is just,
We now put a structure on lnqit to weight different products manufactured
by the focal firm that denote the relative complexity to manufacture that product.
More specifically we assume that and if
we estimate
is the producer specific scaling factor.
145
We further assume that,
and
The above regression is based on the intuition that products that have many
US patents should on an average be easier to manufacture because of information
being available on the underlying technology to manufacture that product. Further
the number of NDAs is intended to account for the relative complexity to
manufacture the product. Similarly since OTC products are typically medication
with lower dosages, these products may be relatively easier (and more stable) to
manufacture. Using these estimates, we construct our proxy for total quantity or the
output index by summing up the predicted values for each market for that producer.
Stated otherwise, we compute (Duan, 1983). The
output for a firm is then got by aggregating over the quantities for that firm. This is
just .
146
D. Chapter 3: Regulation and Welfare:
Evidence from Paragraph IV Generic
Entry in the US Pharmaceutical Industry33
33 This version co-written with Lee Branstetter and Matthew J Higgins has recently been updated.
The new version has been accepted for the NBER Summer Institute to be presented in July 2011.
147
Regulation and Welfare: Evidence from Paragraph IV Generic
Entry in the US Pharmaceutical Industry†
†Acknowledgements: We thank Tamer Abdelgawad, John Asker, Abhijit Banerjee, Antonio Bento ,
Garrick Blalock, Serguey Braguinsky, Lawton Burns, Kitt Carpenter, Iain Cockburn, Susan Cohen,
Chris Conolon, Prithwiraj Chowdhury, Esther Duflo, Antara Dutta, Susan Feinberg, Maryann
Feldman, Maria Marta Ferreyra , John Gardner, Marty Gaynor, Ron Goettler, Dana Goldman, David
Greenstreet, Bart Hamilton, Guy Holburn, Tarun Khanna, Steven Klepper, Brian Kovak, Vineet
Kumar, William Lesser, Jura Liaukonyte, Richard Manning, Romel Mostafa, Anand Nandkumar,
Rohini Pande, Tomas Philipson, Ivan Png, Fu Qiang, Seth Richards, John Romley, Matt Shum,
Steven Sklivas , John Solow, Lowell Taylor, Andrew Tepperman, Jerry Thursby, Kenneth Train,
Ellerie Weber, and Mark Zbaracki as well as seminar participants at Emory, Cornell, Carnegie
Mellon, Pittsburgh, IIM Bangalore, IIM Ahmedabad, Charles River Associations and Precision
Health Economics for valuable comments and discussions. We thank Susan Jack, Margaret Warner
and Robert Anderson for guidance in using the National Health Interview Survey. Programming
assistance by Nachiket Sahoo, Anubrata Banerjee and Trupti Natu is appreciated. We are grateful for the time Dr. Terry Simon and Dr. Ronald Severtis spent explaining to us the treatment of
hypertension. We also thank Alexandra Kondo and IMS Health Incorporated for their generous
support and access to their data. The statements, findings, conclusions, views, and opinions
contained and expressed herein are not necessarily those of IMS Health Incorporated or any of its
affiliated or subsidiary entities. The statements, findings, conclusions, views, and opinions contained
and expressed in this article are based in part on data obtained under license from the following IMS
Health Incorporated or affiliate information service(s): IMS Midas™, IMS Lifecycle™, IMS
National Disease and Therapeutic Index™, IMS National Prescription Audit™, June 1997 to
December 2008, IMS Health Incorporated or its affiliates. All Rights Reserved. Higgins
acknowledges funding from The Imlay Professorship and Pfizer, Inc. and Chatterjee acknowledges
funding from NSF SCISIP Grant #0830233. Authors are listed alphabetically and the usual disclaimers apply.
148
Abstract
With increasing frequency, generic drug manufacturers in the United States
are able to challenge the monopoly status of patent-protected drugs even before the
patents expire. The legal foundation for these challenges is found in Paragraph IV
of the Hatch-Waxman Act. If successful, these Paragraph IV challenges generally
lead to large market share losses for incumbents and sharp declines in average
market prices. This paper estimates, for the first time, the welfare effects of
accelerated generic entry via these challenges. Using aggregate brand level sales
data between 1997 and 2008 for hypertension drugs in the U.S. we estimate demand
using a nested logit model in order to back out cumulated consumer surplus, which
we find to be approximately $270 billion. We then undertake a counterfactual
analysis, removing the stream of Paragraph IV facilitated generic products, finding
a corresponding cumulated consumer surplus of $177 billion. This implies that
gains flowing to consumers as a result of this regulatory mechanism amount to
around $92 billion or about $130 per consumer in this market. These gains come at
the expense to producers who lose, approximately, $14 billion. This suggests that
net short-term social gains stands at around $78 billion. We also demonstrate
significant cross-molecular substitution within the market and discuss the possible
appropriation of consumer rents by the insurance industry. Policy and innovation
implications are also discussed.
Key words: demand estimation; regulation; pharmaceuticals; intellectual property
JEL codes: I11, I38, O3
149
1. Introduction
The Drug Price Competition and Patent Term Restoration Act,
informally known as the “Hatch Waxman Act” was designed to balance access to
pharmaceutical products while at the same time incentivizing pharmaceutical
innovation in the United States. An important provision in Hatch-Waxman allows
generic manufacturers to attempt to enter the market before patents protecting
original branded products have expired. Using this mechanism, known as a
Paragraph IV (Para-IV) certification, generic manufacturers attempt to enter patent
protected markets either by claiming non-infringement or invalidity of the product‟s
patent. Even though the Para-IV certification has been available to generic
manufacturers since the passage of Hatch-Waxman in 1984, for reasons we will
discuss, the number of challenges did not grow significantly until the late 1990s.
Using unique and novel data we quantify, for the first time, the welfare effects of
this accelerated generic entry as a result of Para-IV certifications.
For our analysis we focus on the hypertension (HTN) market in the
U.S. over the time period 1997 to 2008. The hypertension market is significant in
terms of pharmaceutical revenues and in terms of disease prevalence. Our findings
on welfare gains build on and extend a rich literature on discrete choice estimation
of demand (e.g., Trajtenberg 1989; Berry, 1994; Stern, 1996; Petrin 2002;
Cleanthous, 2002; Dutta, 2011). The empirical framework starts by modeling the
utility of a consumer as a function of observed and unobserved product
characteristics. The final outcome is a nested logit demand model that reduces to a
standard shares specification (Berry, 1994) where shares are a function of drug
prices and product characteristics. Coefficient estimates from the demand
150
regressions are then used to estimate consumer welfare. The implementation of a
counterfactual analysis allows estimation of the gains flowing to the consumer from
Para-IV entry. By the end of 2008, cumulated consumer surplus amounted to
approximately $270 billion in the HTN market. In a counterfactual world, over the
same period without the stream of Para-IV generic products, cumulated consumer
surplus was $177 billion. This implies that the gains flowing to the consumer
because of Para-IV entry amounted to $92 billion, which translates into
approximately $133 per representative customer in this market. Placed in context
of the annual cost of treating hypertension, which the American Heart Association
estimated to be $60 billion in 2007, these are significant gains from early generic
entry. 34
Furthermore, over the same period we document the loss to producers to
be approximately $14 billion. This leaves us with substantial net social gains
arising from accelerated entry, at least in the short run.35
We also document significant cross-molecular substitution in this
market. Cross-molecular substitution occurs when patients move from one branded
product to the generic of another branded product. While this movement is
facilitated by physicians, we also present some anecdotal evidence suggesting that
insurance companies encourage this shift among their customers. The implications
of cross-molecular substitution are profound; a branded product‟s intellectual
property (IP) protection, within a market, is only as strong as their weakest branded
competitor‟s IP. For this reason, cross-molecular substitution may have important
34 See http://www.nmanet.org/images/uploads/Publications/OC34.pdf 35 The presence of an insurance industry that mediates transactions between drug sellers and drug
consumers poses some complications for us in terms of our ability to interpret our estimates as
“social gains.” While a full consideration of these issues is beyond the scope of the current paper,
we discuss some of their implications in Section 5.
151
implications in terms of how incumbent pharmaceutical firms allocate their R&D
resources across therapeutic disease programs, especially for rare or pediatric
diseases.
Hatch-Waxman was designed to balance access with innovation.
We demonstrate that the Para-IV regulatory challenge has been an effective
mechanism at providing early access to consumers but at what cost? Are the short
term consumer gains impacting the long term incentive to innovate or influencing
the type of drugs that are developed? With rising costs of drug discovery and
longer development times (DiMasi et al, 2003; 2007) researchers have
contemplated on how best to design an optimal patent regime that will effectively
balance access to cheaper drugs for the consumer without compromising the
incentives for innovation (Knowles, 2010; Higgins and Graham, 2009). A full
welfare analysis of the balance struck by Hatch-Waxman will remain incomplete
until we have analyzed its impact on innovation; ongoing research seeks to
undertake this kind of analysis, but until then we make an important contribution
estimating the first-order welfare effect in the short run from accelerated generic
entry due to Para-IV.
The paper proceeds as follows. Section 2 offers a brief discussion of the
regulatory environment in which pharmaceutical firms operate. Section 3 discusses
related literature. Our data and methodology are presented in Section 4. Section 5
presents results and we discuss the implications of our work and conclude in
Section 6.
152
2. Regulatory environment and early generic entry
2.1 Hatch-Waxman and Paragraph IV challenges
The current regulatory environment faced by pharmaceutical
companies in the United States can be traced to the passage of Hatch-Waxman in
1984. One of the hallmarks of the legislation is its purported trade-off: it allows
expedited Food and Drug Administration (FDA) approval for generic entry while
extending the life of pharmaceutical patents in order to compensate innovators who
lost time on their “patent clocks” waiting for FDA approval (Grabowski, 2007).
This balance was deemed necessary to equalize two conflicting policy objectives:
giving pharmaceutical firms incentives to conduct drug research while
simultaneously improving consumer welfare by enabling generic firms to quickly
bring copies to market (Federal Trade Commission (FTC), 2002).
When a pharmaceutical company submits a New Drug Application
(NDA) to the FDA for approval they are required, by law, to identify all relevant
patented technologies necessary to create the drug; these patents are subsequently
listed in the FDA Orange Book. Upon approval of a drug, the FDA will restore
patent term to the pharmaceutical firm for time used by the FDA in the approval
process (Grabowski, 2007).36
In addition, the FDA will also grant each new
approved product regulatory protection lasting for five years (“data exclusivity”)
which runs concurrently with patent protection.37
During this data exclusivity
period, regardless of the status of the underlying patent(s), no generic entry may
36 There are limits to this. Pharmaceutical firms cannot receive a patent extension of more than five
years, nor are they entitled to patent extensions that give them effective patent life (post approval) of
greater than 14 years. 37
Orphan drugs receive 7 years of data exclusivity; reformulations receive 3 years of data
exclusivity and pediatric indications receive an additional 6 months of data exclusivity.
153
occur. At the conclusion of data exclusivity branded products are protected only by
their patents; this period running from the cessation of data exclusivity to the
expiration of the patent(s) is commonly referred to as “market exclusivity” (See
Figure 1).
Prior to the passage of Hatch-Waxman, generic manufacturers
seeking to sell their products in the U.S. market had to demonstrate the safety and
efficacy of their products by putting them through clinical trials. While the
outcome of these trials lacked the uncertainty involved in the trials of an innovative
new drug, the time and expense involved were a significant disincentive for
generics manufacturers to put products on the market, since they could not charge a
premium price to offset the costs of clinical trials. Prior to Hatch-Waxman, it is
estimated that more than 150 products existed without any patent protection and
without any generic entry (Mossinghoff, 1999). While Hatch-Waxman did not
lessen the burden of the clinical trials process for branded pharmaceutical
companies seeking approval for new drugs, it essentially eliminated the requirement
for separate clinical trials for generic manufacturers. This was made possible since
generic manufacturers could simply demonstrate “bioequivalence” with branded
products by showing that the active ingredient in their product diffused into the
human bloodstream in a manner similar to the original product.38
38 In theory, generics should be perfect substitutes for branded drugs since they are bioequivalent.
Cleanthous (2002) shows that the data do not support this relationship and suggests this is the result
of „spurious product differentiation‟, which he defines as arising “…when consumers perceive
physically identical products to differ in quality.” Recent evidence, however, may suggest that
consumer perceptions have merit and while drugs may be bioequivalent, they may indeed differ in quality. Several articles appeared in the April 17, 2007 edition of the prestigious journal Neurology
discussing the high incidence of break-through seizures with generic anti-epileptic. Insurance
companies such as Blue Cross Blue Shield of Georgia allow pediatric customers to stay on branded
anti-epileptic medications even though generics are available. Differences across generics for the
same brand have also been reported. We are not suggesting all generics have problems but it
154
Hatch-Waxman provides four pathways (or “Paragraphs”) a generic
firm may follow in order to gain entry into a market. The process starts with the
filing of an Abbreviated New Drug Application (ANDA) by a generic manufacturer
with one of the four Paragraph certifications. A Paragraph I certification is one for
which the originator firm has not filed patent information for its branded product.
Paragraph II certification relates to when the branded product‟s patent has already
expired (i.e., the end of market exclusivity), and Paragraph III certification relates
to cases when the generic manufacturer notes that the patent on the branded product
will expire on a certain date and that it seeks to enter only after patent expiry or end
of market exclusivity. The fourth certification, Paragraph IV, argues that the
generic manufacturer does not infringe on a branded product‟s patents or that those
patents are invalid. More importantly, however, a Paragraph IV certification can be
acted on by the FDA after the conclusion of data exclusivity anytime during the
market exclusivity window (See Figure 2).39
This suggests that, if successful, these
challenges can significantly decrease the effective patent life of branded products
bringing generics to the market earlier than otherwise would be the case (Higgins
and Graham, 2010; Grabowski and Kyle, 2007).
When a generic manufacturer files an ANDA with a Para-IV
certification, the generic manufacturer is obligated to notify the incumbent. Upon
receipt the pharmaceutical firm has two options: (1) do nothing or (2) sue the
generic manufacturer (for patent infringement) within 45 days. If the
pharmaceutical firm chooses not to file suit and does nothing then the FDA is
appears in some instances where the therapeutic window is very narrow these perceptions may have
some merit. 39
Generic manufacturers may file a Para-IV certification up to one year prior to the end of data
exclusivity but the FDA may not act on it until the conclusion of data exclusivity.
155
entitled to approve the generic version of the branded product. If, however, the
pharmaceutical firm chooses to file suit then that action automatically triggers a 30-
month stay on FDA action. During this stay the FDA is unable to take any action
on the ANDA unless there is a first court ruling. If the court rules in favor of the
pharmaceutical firm, the Para-IV certification fails and the FDA is unable to
authorize generic entry until the branded product‟s patents expire (i.e., end of
market exclusivity). On the other hand, upon a first court ruling in favor of the
generic manufacturer the FDA may approve the ANDA. The first generic
challenger is awarded 180-days of exclusivity.40
During this 180-day exclusivity
period the first challenger is the sole generic provider and occupies a position of
duopoly with the original branded provider. This period of exclusivity for a
challenger was deliberately placed in Hatch-Waxman as an incentive for generics
manufacturers to challenge “weak” patents. As competition in the generics market
has intensified, generics manufacturers have increasingly sought these periods of
180 day exclusivity and the high profits they bring. After 180 days, other generic
entrants are allowed to freely enter the market.
Even though Para-IV certifications have been available to generic
manufactures since the passage of Hatch-Waxman, the number of challenges
remained low until the late 1990s. The acceleration of challenges since then can be
tied to a series of court decisions, changes in FDA policy, and passage of the
40 While the FDA may approve an ANDA upon a first court ruling in favor of the generic
manufacturer if that ruling is overturned on appeal the generic manufacturer exposes itself to
damages for lost branded revenue. This is a gamble many generics manufacturers are willing to
take, in part because the number of cases in which the first ruling is reversed on appeal is limited.
156
Medicare Act of 2003 (See Figure 3).41
Over time, the ability of pharmaceutical
firms to delay generic entry has been limited in important ways, dramatically
intensifying competition and accelerating generic entry. In the early years, after
passage of Hatch-Waxman, pharmaceutical firms were allowed to appeal initial
judgments against the validity of their patents (or findings of noninfringement), and
the FDA could not approve generic entry until all appeals had been exhausted -- a
time consuming process that often held generic manufacturers at bay until patents
expired or were about to expire. In more recent years, the FDA has approved entry
as soon as courts issue a first ruling. Throughout the 1990s, incumbents often
followed a practice of taking out additional patents after an initial Para-IV filing and
invoking non-concurrent 30-month stays for each patent allegedly infringed.42
In
more recent years, due to changes in the Medicare Act of 2003, pharmaceutical
firms have been limited to one 30-month stay per product. Generic manufacturers
have also been given greater leeway in their use of the 180-day exclusivity period
granted to first-filers under the law. Finally, recent court rulings have made it
easier to demonstrate patent invalidity and harder to demonstrate infringement.
In the early 1990s ANDA applications with Para-IV certifications
accounted for about 10-20% of all generic entry, however, by the end of the 2000s
they accounted for more than 40% (Higgins and Graham, 2009; Berndt et al., 2007).
Teva Pharmaceuticals is the most prolific filer of Para-IV challenges although there
has been increased activity in recent years by Indian generic manufacturers. In
41 In 1998, Mova v. Shalala, changed the interpretation of 180-day exclusivity. In 2000, the FDA started allowing generic entry following a first favorable court decision irrespective of a final court
ruling. In 2003, the FDA started giving 180-day exclusivity to multiple applicants filing on the same
first day. In 2006, KSR v. Teleflex lead to a new standard of obviousness for patents and in 2007,
Medimmune v. Sun Pharma, a new standard for non-infringement was established. 42 The interested reader can see Bulow (2004) for a more complete discussion.
157
2007, Teva‟s Annual Report claimed that 92 of their 160 ANDA filings were Para-
IV challenges putting at risk over $100 billion in branded sales.43
Because the costs
of challenging a patent are a relatively small -- $5 million (American Intellectual
Property Law Association, 2007) -- compared with the large, average, potential
payoff of $60 million for the 180-day generic exclusivity period (Federal Trade
Commission (FTC), 2009), generic manufacturers have begun to engage in
“prospecting” by filing numerous ANDAs with Para-IV certifications. Evidence
suggests that these challenges are occurring earlier in a branded products lifecycle
(Panattoni, 2011; Saha et al., 2006; Grabowski, 2004; Scherer, 2001) and now
affect smaller market drugs with yearly sales under $100 million; “blockbusters”
are not the only target (Grabowski and Kyle, 2007). Moreover, in a recent event
study, Panattoni (2011), reports cumulative market losses of slightly over $1 billion
when a pharmaceutical firm losses a Para-IV case in court.
Studying the outcomes of the 104 ANDA applications with Para-IV
certifications, the FTC found that 75 of these applications had resulted in litigation
(FTC, 2002). Of the 53 cases that had been resolved, 22 (42%) were resolved in the
generic manufacturer‟s favor thus allowing a generic product introduction prior to
the expiration of the branded product‟s patents. As we have already noted,
regulatory and legal changes have increased the ability of generic manufacturers to
mount successful challenges. In an effort to counteract the increase in Para-IV
certifications (more than 230 were filed in 2010) pharmaceutical firms have turned
to authorized generic agreements (FTC, 2009). 44
That is, pharmaceutical firms
43
U.S. SEC, Form 20-F, (December 31, 2007); http://www.tevapharm.com/pdf/teva20f2007.pdf. 44 http://money.cnn.com/2011/03/11/news/companies/big_pharma_lawsuits.fortune/index.htm
158
exercise their right as patent owners to license their technology to a generic
manufacturer, and that generic producer then enters the market at a low price. This
means that a successful Para-IV challenger will confront a triopoly, not a duopoly,
and will earn lower profits. Pharmaceutical firms may have hoped that the use (or
threat of use) of authorized generics producers might have reduced the incentives
for Para-IV challenges and, therefore, their incidence. However, the evidence
suggests otherwise (Berndt et al., 2007a; Berndt et al., 2007b). Evidence also exists
that pharmaceutical firms have engaged in out-of-court settlements to delay entry
(FTC, 2002; Bulow, 2004) and maintain patent monopolies. This practice has
attracted the attention of antitrust authorities, however, the Supreme Court recently
refused to overturn the practice.45
Regardless of any efforts by pharmaceutical
firms, the intensity of Para-IV certifications has remained high.
To summarize, over the years of our sample period (the late 1990s-
2008), a series of legal and procedural changes turned Para-IV certifications from a
rare occurrence to a first-order challenge to the profitability of the branded
pharmaceutical industry. We will think of these changes as a kind of gradually
unfolding natural experiment. Collectively, these changes substantially reduced the
cost of a Para-IV certification and dramatically increased their likelihood of
success. Thus, the measured increase in Para-IV certifications will be viewed
largely as a response to the exogenous (at least from the point of view of any one
generic challenger) policy shift. In addition to presuming that the overall increase
in incidence of Para-IV challenges was largely exogenous, we also posit that there
45 On March 7, 2011, the Supreme Court rejected a challenge to a 1997 settlement in which Bayer
AG paid Barr Pharmaceuticals to drop an early bid to market an antibiotic.
http://online.wsj.com/article/SB10001424052748703386704576186713514684524.html
159
was a strong element of exogeneity in determining which branded products were
ultimately challenged.46
The patents protecting most products hit with a Para-IV
certification were written before such challenges became prevalent (Knowles,
2010). The weakness of some of these patents in court proceedings was evident, in
many cases, only in hindsight.
2.2 Cross-molecular substitution
Most prescription health plans in the U.S. allow for the use of
branded products until generics become available. In most cases patients will be
given the generic by retail pharmacies unless the prescription is written “Dispense
as Written” or if the patient explicitly asks for a branded drug (in which case there
is usually a much higher co-payment). More recently, however, insurance firms
have begun to engage in “cross-molecular” substitution. For example, let‟s assume
there are 3 branded products in a particular market, Drug A, Drug B and Drug C,
sold by three different pharmaceutical firms and that Drug B just lost a Para-IV
challenge in court and a generic, Generic B, has entered the market. Cross-
molecular substitution exists when insurance companies attempt to encourage
patients taking Drug A or Drug C to switch to Generic B. While insurance firms
cannot force patients to move they can entice them with lower (or no) copayments
for Generic B. Table 1 provides an example extracted from a letter Blue Cross Blue
Shield of Georgia (BCBSGA) sent to patients suggesting generic alternatives to
different branded products across several therapeutic categories. In the letter
BCBSGA offered to pay for the generic co-pay for a period of three months.
46 Evidence exists that both large market (Panattoni, 2011) and small market (Grabowski and Kyle,
2007) have been targeted. Recent work by Hemphill and Sampat (2010) make a first attempt to
explore characteristics of patents in order to understand the likelihood of generic challenge.
160
The implications of cross-molecular substitution are significant. For
pharmaceutical firms with branded products, cross-molecular substitution implies
that their drug‟s market protection is only as strong as the weakest drug in a
particular market. In the above example, the transition from Drug A and Drug C to
Generic B is occurring irrespective of the underlying IP protection or exclusivity
periods for those drugs. This activity has obvious welfare implications; the gains to
consumers are potential larger since patients from Drug A and Drug C can
potentially benefit but the producer loss will also be larger because the incumbents
that market Drug A and Drug C will lose revenues. The extent of these impacts will
vary across therapeutic categories as some drugs are more easily substitutable. For
example, we would expect higher substitutability in markets such as hypertension
and allergy and lower substitutability in markets such as depression and epilepsy.
3. Related literature
This paper draws upon the literature of welfare estimation and
demand modeling dating back to Trajtenberg‟s (1989) pioneering work on the CT
scanner industry. Trajtenberg, in turn, built upon an even older literature relating
discrete consumer choices to product characteristics (Lancaster, 1966; McFadden,
1973; Lancaster, 1977). This approach has proven extremely useful in studying
product choice in the presence of significant product differentiation (Anderson et
al., 1992), and it has been applied to a range of industries, including , automobiles
(Petrin, 2002 and Goldberg 1995), digital goods (Brynjolfsson and Smith, 2003),
personal computers (Bresnahan et al., 1997), Medicare HMOs (Town and Liu,
2003) and banking (Dick, 2008). Methodological improvements have also
permitted researchers to utilize these models using aggregate data in the absence of
161
consumer level data (Berry, 1994; Berry et al., 1995 and 1999). Given their
theoretical underpinnings it is also worthwhile to note that discrete choice models
have been applied to understand a variety of economic situations beyond studying
the social value of new goods. For example, Nevo (2000, 2001) examines price
competition and mergers in the ready-to-eat cereal industry and Davis (2000)
studies spatial competition in movie theaters, while some other papers have used
this framework to understand gains from trade and globalization (Broda and
Weinstein 2006; Clerides 2008).
These approaches have also been applied to the pharmaceutical
industry, although researchers have had to confront some additional challenges
unique to this sector. Products are highly differentiated and the choice of a
particular product is often made by agents of the consumer (i.e., doctors) versus the
consumers themselves. Consumer choice can also be influenced by the presence (or
not) of insurance coverage. Notwithstanding these difficulties, Ellickson et al.
(2001) explore patient welfare from new drugs and find that gains are contingent
upon compliance and the role of the physician. Stern (1996) employs a discrete-
choice framework to estimate demand to evaluate patterns of substitutability
between branded and generic drugs. Ellison, et al. (1997) models demand in order
to compute substitution elasticities between branded and generic antibacterial drugs.
More recently, studying the U.S. anti-depressant market Cleanthous (2002) finds
large welfare gains from drug innovation; Bokhari and Fournier (2009) report
welfare gains due to first time generic entry; and, Dutta (2011) analyzes the welfare
impacts of stronger intellectual property (IP) in India. Other complementary work
has focused on the enhancements to social welfare through reductions in mortality,
162
morbidity and total medical expenditures Lichtenberg (1996a, 1996b, 1998, 2001,
and 2005).
In addition to work that quantifies welfare effects from product
innovation using discrete choice models, this paper also relates to other work
focusing on various dimensions and implications of generic entry (Caves et al.,
1991). Saha et al. (2006) report the dramatic rise in generic introductions since the
passage of Hatch-Waxman while Reiffen and Ward (2005) show that the cost to
obtain generic drug approval has decreased. Time to market for generics after
branded product patent expiration has also declined substantially, from
approximately three years prior to Hatch-Waxman to only one to three weeks
(Congressional Budget Office, 1998).47
Other research has focused on entry
decisions by generic manufacturers (Morton, 1999, 2000; Grabowski and Vernon,
1992, 1996; Berndt et al., 2003, 2007; Frank and Salkever, 1997; Hurwitz and
Caves, 1988; Hudson, 2000; Appelt, 2010), prices (Danzon and Chao, 2000a,
2000b), price controls (Kyle, 2007; Danzon et al., 2005; Lanjouw, 2005) and entry
costs (Djankov et al., 2002).
Researchers have also recognized the potential that accelerated
generic entry can have on the incentives to develop new drugs. Terming the
phenomenon „Napsterization of Pharmaceuticals‟, Hughes et al. (2002) show in
theoretical work that providing greater access to a current stock of prescription
drugs yields large benefits to existing consumers. However, this access comes at a
cost in terms of lost consumer benefits from reductions in the flow of future new
47
Berndt and Aitken (2010) provide an excellent summary of the increase of generic competition in
the U.S. over the last decade.
163
drugs. This trend could be problematic since newer drugs have been shown to be
positively related to life expectancy (Lichtenberg, 2005) and in lowering non-drug
medical spending of all types (Lichtenberg, 2001). This potential problem has been
discussed by recent researchers (Grabowski and Kyle, 2007; Higgins and Graham,
2009; Graham and Higgins, 2010; Knowles, 2010; Panattoni, 2011) and the extant
literature provides ample evidence suggesting that acceleration of generic entry has
undermined incentives for R&D. What is still lacking in the literature, however, is
firm econometric evidence demonstrating that generic entry has actually led to a
decline (or changes) in R&D investment that can be plausibly ascribed to the
generic entry itself. Our focus on estimating the first-order impact from accelerated
generic entry, with some suggestive short run estimates of producer loses on the
supply side provides the link in the literature that will allow researchers to more
accurately analyze the impact on R&D decisions.48
4. Data and methodology
Previous research in this area has struggled with data limitations.
We are fortunate to have access to a range of unique and comprehensive data sets
that provide us with powerful leverage over some of the econometric challenges we
confront. First, data from Parry Ashford Publications (www.paragraphfour.com)
allows us to identify each Para-IV certification dating back to 2003.49
This data
provides full drug level information about the challenge and outcome which we can
link to our other data resources. For data prior to 2003 we filed a Freedom of
48 This is the focus of on-going research. 49 Para-IV certification data only became publicly available in 2003. In order to supplement the
data prior to 2003 we filed a FOIA request with the FDA. Other researchers (Berndt et al., 2007)
have used survey data in order to capture pre-2003 activity. In a recent study, Panattoni (2011)
collected data from District Court decisions.
164
Information Act (FOIA) request with the FDA. Next, we need demand side
information for the drugs (and markets) where challenges occurred. For this we
turned to the IMS MIDAS™ database which provides sales (quantity) and revenue
information for every product sold by every firm, across all therapeutic disease
categories, branded and generic, in the U.S. This database also provides
information on dosages and expected market exclusivity expiry dates. All branded
products are also listed in the FDA Orange Book, and this database provides an
alternative source for approval, data exclusivity and patent expiry dates. Our final
time period covers 1997 to 2008 and unlike the prior literature (Dutta, 2011;
Cleanthous, 2002) that use annual data, we instead utilize quarterly data. This
choice is necessary so we can more accurately track entry (and subsequent entry
after 180-day exclusivity periods granted to first-filers) by generics.
A key aspect of our methodological framework is the definition of
our market and how it relates to the measurement of the outside good. We choose
to focus on the U.S. hypertension market as our initial research setting. The market
is large; for example, in 2007 the American Heart Association estimated the burden
of hypertension on the healthcare budget to be close to $60 billion (See Figure 4).
This category of disease is medically significant too, with prevalence around 26%-
29% of the population in 2008.50
We consulted experts at the Center for Disease
Control (CDC) to help retrieve disease prevalence statistics for hypertension from
the National Health Interview Survey (NHIS), which we discuss more fully below
in Section 4.1.2. Para-IV certifications have also been active in this market.
Finally, discussions with physicians suggest a relatively high degree of substitution
50 See http://www.cdc.gov/nchs/data/databriefs/db03.pdf.
165
across different drugs, allowing for cross-molecular substitution. Inspection of
product level data confirms the reality of extensive cross-molecular substitution in
this market.
4.1 Demand estimation – nested logit model
Our potential market is all prospective U.S. hypertension patients
who might consume one of a number of drugs spread across five specific drug
categories used broadly to treat hypertension (See Table 2).51
Our product level
data are organized in a taxonomy known as the anatomical therapeutic chemical
(ATC) classification system, and we will refer henceforth to ATC codes and
categories. As identified in the table, we assign chemically distinct products to
different categories. A drug containing a single active ingredient is treated
differently from a drug that combines multiple active ingredients; each chemically
distinct category will be classified as a “molecule” (even though some substances
may combine multiple active ingredients). We follow Berry (1994), Stern (1996)
and Dutta (2011) and model demand with a nested logit model. Formally, let M be
the number of molecules (including combinations) being sold in the hypertension
market, , ,….. . Let be the set of all products that contain molecule .
Consumer i‟s utility from consuming firm j‟s product in molecule m is given by
(excluding a time subscript):
(1)
51 The ATCs used to define the market are: (1) C7A0 BETA-BLOCKING AGENTS, PLAIN; (2)
C7B1 BETA-BLOCKERS IN COMBINATION WITH HYPOT/DIURETICS; (3) C8A0 CALCIUM ANTAGONIST, PLAIN; (4) C9A0 ACE INHIBITORS, PLAIN; (5) C9B1 ACE
INHIBITORS USED IN COMBINATION+A-HYP/DIURETICS; and, (6) C9B3 ACE INH
COMB+CALC ANTAG. Angiotensins and their combinations with calcium antagonists or diuretics
(C9C0+C9D1+C9D3) have not faced patent expiration or Para-IV challenges and were thus
excluded from the sample.
166
All products in a molecule (that is, all products that possess the same active
ingredient or same combination of active ingredients) are presumed to be closer
substitutes than chemically different products associated with other molecules. In
the language of discrete choice models, we think of them as belonging to the same
“nest.” Our approach also allows for the possibility of substitution across
molecules. This was based on direct conversations with physicians, who maintain
that it is consistent with their prescribing behavior and consistent with published
prescription guidelines.52
As an example, in Table 1, BCBSGA is attempting to
induce patients to switch from branded drugs made with the molecule nifedipine to
generic drugs made with the molecule amlodipine.
In Equation 1, captures the correlation of consumer choices between
molecules (or the degree of substitution across molecules) and the are i.i.d.
extreme value. is common to all drugs that belong to the molecule m and
Cardell (1997) proves that since the are i.i.d. extreme value so is
[ . As approaches 1, the within molecule correlation of utility
levels converges to 1 and indicates perfect substitutability of the molecules. In
other words, if approaches 1, products within a molecule are no more close
substitutes for one another than products in a different molecule. If, on the other
hand, approaches 0, then there is no substitution across molecules. An important
test of the validity of our estimation procedure will be an estimated value for that
lies between 0 and 1. is the mean utility of the product, is invariant across
consumers (i.e., patients) and is a function of observed product characteristics ( ),
52 See http://www.nhlbi.nih.gov/guidelines/hypertension/phycard.pdf
167
price ( and unobserved product characteristics ( that could affect utility.
Formally, the mean utility can be specified as follows:
(2)
Following Berry (1994), market shares in this framework are given by:
(3)
where is the probability of choosing a product j, given a choice of a molecule m
and is:
] (4)
The probability of choosing molecule m is given by:
(5)
Thus, the market share of product j in molecule m is given by:
(6)
Finally, the share of the outside good in this model is given by:
(7)
Berry (1994) then demonstrates that this can be substituted and worked out into a
linear estimable specification:
(8)
The expression in Equation (8) gives us the mean utility from Equation (2) along
with a conditional share term, , which captures within group correlation.
Depending on the nature of the product characteristics vector, the expected sign of
the elements in the coefficient vector can either be positive or negative. For
example, if the focal product characteristic is the number of contraindications for a
168
drug then one might reasonably expect a negative coefficient; its usage may be
more limited because of the larger number of contraindications. On the other hand,
if the focal product characteristic is advertising then we might reasonably expect a
positive coefficient. The sign of the price coefficient, α, is expected to be negative
and should be between 0 and 1 assuming a large enough market such that the
conditional shares on the right-hand side (RHS) are different from the unconditional
shares on the left-hand side (LHS). Without the conditional within-group term,
on the RHS, Equation (8) would reduce to a simple multinomial logit
specification, which we include as a robustness check.
Our approach carries with it the important benefits of simplicity and
ease of estimation, and recent related work using discrete choice models to measure
the welfare impact of changes in the monopoly power of pharmaceutical producers
(e.g., Dutta, 2011) has demonstrated its usefulness in that context. Of course, these
benefits come at a cost. Unlike the more complicated approach of Berry,
Levinsohn, and Pakes (1995) -- which we will denote hereafter as BLP -- we do not
simultaneously estimate the core parameters of a discrete choice demand model
with those of a fully specified supply side model, nor do we allow for an interaction
between consumer and product characteristics, either at the individual level or at the
market level. Our particular distributional assumptions and the "nests" we impose
on the data necessarily restrict the substitution possibilities across products in ways
that BLP are able to avoid. While we explore the empirical implications of
alternative nesting structures for our core conclusions, we are unable to estimate the
differential reactions of different kinds of consumers to product characteristics with
our current approach. These are important considerations, and the eventual
169
combination of our rich data on product level sales and characteristics with
extensive data on the demographic characteristics of consumers using these drugs is
the focus of ongoing research.53
However, the effective combination of these data
presents its own challenges, and, given the current absence of any credible estimate
of the welfare impact of Para-IV entry on U.S. drug consumers, our current
approach would seem to offer an expedient path to an informative, if incomplete,
estimate of the welfare impact that could be compared with the revenue losses of
incumbent producers.
Equation (8) includes variables on the right hand side that could be
endogenous and/or correlated with omitted variables in the error term. These
concerns extend to the conditional share term capturing substitutability, ,
and price, We include data on advertising expenditures for particular products,
and this variable could also be endogenous. Therefore, we consider the use of
instrumental variables, which we discuss below.
4.1.1 Quantities and prices
As we indicated above, our unit of observation is molecule-firm-
brand-quarter, which means that if two branded firms and a generic firm are each
selling chemically identical products, we still treat each of the three products as
distinct, and we track sales on a quarterly basis. IMS MIDAS™ provides sales data
in Standard Units (SU). The SU measurement is designed to equate different
dosage forms (e.g., tablets, capsules, liquid) into comparable patient dosages.
53 The National Health Interview Survey (NHIS) is an annual, large-sample, comprehensive survey of Americans' health status, and the data collected include information on the incidence of diseases
like hypertension and the consumption of particular categories of drugs, as well as important
demographic information on the survey respondents. In principle, these data on the distribution of
consumer characteristics within the drug consuming population could be integrated into our analysis,
as in Cleanthous (2002).
170
Products that have multiple presentations of a molecule sold by a single firm are
aggregated together. Revenues are reported over the same period and are converted
into real dollars using a base year 2000 GDP deflator. For each quarter observation,
wholesale price is defined as revenues divided by SU sales.
4.1.2 Hypertension market and unconditional shares
We consulted experts at the CDC to help retrieve and construct
disease prevalence statistics for hypertension from the National Health Information
Survey (NHIS). Specifically, for children and adults, we retrieve counts of the
number of U.S. residents taking the NHIS who answered that they ever had
hypertension or the related conditions of high blood pressure, a heart condition, or
coronary heart disease. CDC recommended weightings were then applied in order
to back out national estimates of hypertension prevalence.54
To create a correspondence between sales data (pill-level) and
prescription-level data or the number of patients consuming hypertension drugs, we
impose an assumption on the length of treatment. For this we assume chronic
intake of hypertension drugs (i.e., patients stay on the drug throughout a year) and
combine this with prescription-level data on average treatment days from the IMS
National Prescription Audit™ (NPA) and the IMS National Disease and
Therapeutic Index™ (NDTI), supplemented with information from medical
references. We take our estimates of the number of patients with hypertension and
multiply this by 90 (days in a quarter) and then by 4 (four quarters in a year). This
multiplication yields the potential market size for hypertension drugs. We therefore
54
For more information on the CDC weighting recommendation and methodology please
see:ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2009/srvydesc.pdf
171
define the unconditional share for a particular quarter, , as brand level sales
divided by our estimate of the potential market size, in SU.
4.1.3 Outside good share
The approach in Section 4.1.2 also allows us to create the outside
good measure, or the number of potential patients with hypertension that are not
actively receiving treatment. In any given quarter, we can sum hypertension drug
sales across all five relevant ATC categories and compare that with the overall
potential market derived from the CDC NHIS data above, the difference is our
measure of the outside good.55
4.1.4 Conditional share
Equation (8) includes the term which captures within group
correlation, σ, or, as we argue, provides an average measure of cross-molecular
substitution. Construction of the numerator is defined as branded product-level
sales for each quarter. The denominator is the summation of all sales for a specific
molecule (branded and generic) in each quarter. Using the calcium channel blocker
Norvasc and Q1:2000 as an example, the numerator would be Norvasc sales in
Q1:2000 divided by all the sum of products (branded and generic) that contain the
molecule amlodipine.
4.1.5 Product Characteristics
Formal modeling of demand in this setting occurs in a manner such
55 We experiment with the outside good measure in various ways, since its reasonable computation might have a bearing on the results. This involved imposing different assumptions on the length of
treatment, imposing treatment card information, and reconciling all the measures of outside good
shares with epidemiological estimates of U.S. patients who are aware of they being hypertensive but
are not getting treated (for a variety of reasons). Our findings remain robust to these various
measures and are available from the authors upon request.
172
that the utility of a consumer from consuming a drug product, in a certain time
period, is a function of product characteristics, observed and unobserved.
Following the extant literature (e.g., Stern, 1996; Cleanthous, 2002; Dutta, 2011),
we include information on drug side effects. More specifically, we use drug-label
information, cross-checked with medical references, to gather the number of
contraindications (i.e., circumstances under which the drug cannot be safely taken)
for each drug. Advertising expenditures on a product are also a significant
determinant of sales and the literature points to biases in demand estimates without
advertising (Moul, 2006). As a result, we incorporate product-specific detailing
information from IMS MIDAS™. Detailing data is comprised of three
components: (1) direct journal advertising, (2) direct mail advertising and (3) direct
interactions between drug representatives and physicians. We aggregate three
forms of detailing together at the product level, converting all financial variables to
real, year 2000, dollars. We do not possess quarterly data on direct-to-consumer
advertising (DTC) but since physicians are acting as agents on behalf of patients
(consumers), it can be argued that detailing is a critical component of a
pharmaceutical firm‟s advertising strategy. Rosenthal et al. (2003) supports this
notion and demonstrates that total promotion expenditures were approximately 14%
of sales. Of this 14%, DTC comprised 2.2% while total physician promotion
accounted for 11.8%. Finally, because there are important product attributes that
are difficult to measure consistently across products we include product-specific
fixed effects. These fixed effects will vary across products but not over time.
173
4.2 Instruments
A clear econometric challenge arises with respect to price, the level of
advertising, and the conditional share term. All of these variables are potentially
correlated with the unobserved product characteristics, , in Equation (2), a
problem that is well recognized in the demand estimation literature (e.g., Berry,
1994; Berry et al., 1995, 1999; Nevo, 2000). Consistent with other studies (e.g.,
Stern, 1995; Cleanthous, 2003; Dick, 2008; Dutta, 2011) we use two competition
related instrumental variables and one related to product design that are less likely
to be correlated with the error term.
Our first instrument is the count of dosage levels in which a product is
available, Form Numbers. FDA data allows us to track changes in these numbers
over different molecules. As Stern (1996) points out “Unless consumers value
products sold by a particular firm because it is a multiproduct firm, measures of
multiproduct ownership will be correlated with price and advertising, but be
uncorrelated with unobserved quality.” (p.18) The second instrument we employ is
the Time Since Generic Entry and is coded as 0 for all observations before entry
into a molecule market. After entry a counter starts and increases by one unit for
each quarter that elapses. Our final instrument is the Number of Firms (branded
and generics) selling products across the market. The second and third instruments
relate to the level of competition in the market, and under the assumption that entry
is exogenous, should be uncorrelated with unobserved product characteristics. For
174
all instruments we also used lagged values. The results of the standard over-
identification tests are reported below.56
4.3 Consumer surplus (CS) and counterfactual analysis
4.3.1 Real world
Given the coefficient estimates on price and product characteristics,
quarterly consumer surplus in the nested logit model outlined above follows Small
and Rosen (1981) and Train (2003) and is given by:
*P*ln [1+ ] (9)
Where P captures prevalence information and is converted into daily-
doses using our previously discussed length of treatment assumption (see Section
4.1.2). Within the bracketed term, 1 represents the utility derived by the average
consumer from consuming the outside good; the remaining expression comes from
the indirect utility derived by the average consumer, through coefficient estimates
applied on realizations of price and product characteristics (Train, 2003).57
The
double summation term implies that the indirect utility in each quarter is first
summed across all products (brands, generics and Para-IV generics) within a
molecule and then across all molecules. The entire expression is then logged,
multiplied by P and (where is the coefficient on price and is defined as the
marginal utility of income) in order to express consumer surplus in dollar terms.
56 See Appendix 1 for a discussion on additional instruments currently being explored. 57 It is important to note that in our specification not only includes the product characteristic
variables, number of contraindications, detailing, but also price, and further, the constant term and
the coefficient on product fixed effects that are generated in the regressions.
175
4.3.2 Counterfactual world
In order to understand the impact of Para-IV early generic entry we
need to establish the counterfactual which will allow us to determine the consumer
surplus generated had there not been entry before patent(s) expiry (i.e., end of
market exclusivity) by the generic entrants. In constructing the counterfactual
consumer surplus series we first drop the Para-IV generic entrants (and the
subsequent follow-on generic entrants that entered as a result of the Para-IV action
after the 180-day exclusivity period) but retain the branded pharmaceuticals along
with the generics that entered after patent expiry or the end of market exclusivity.
We need to make several assumptions relating to pharmaceutical firm
action in this counterfactual world.58
First, we assume a pharmaceutical firm would
not launch a product reformulation or me-too drug since they would end up
cannibalizing sales of their existing product. Second, we assume that the
pharmaceutical firm follows pre-entry trends in terms of price and advertising until
the end of market exclusivity, after which they follow normal post-entry trends in
the face of normal (non Para-IV) generic entry. As a consequence of these
assumptions we are able to impute price and advertising from the quarter of Para-IV
generic entry until patent expiry of the incumbent‟s product. We have tested the
accuracy of our counterfactual predictions by using early sample data to estimate
late sample prices, quantities, and advertising in markets where generic entry never
58 In the real world incumbents could possibly launch “me-toos” or reformulations (Graham and
Higgins, 2010), authorize their own generics and/or modify price or advertising in an attempt to
create customer loyalty (Bhattacharya and Vogt, 2003). However, once a generic has entered, in
most cases, prescription plans move patients to them unless they are willing to pay either a higher
co-pay or, in some cases, full retail price.
176
occurred. We find that our predictions track actual values fairly closely.59
We
repeat the calculation outlined in Equation (9) to create a series by
quarter with the difference between the two consumer surplus series giving us
welfare going to consumers as follows:
- (10)
In order to get a sense of the plausibility of these CS estimates we
need to reconcile them with the average patient‟s annual drug expenses, estimates
of surplus in hypertension in other economies, and producer surplus figures. We
discuss all these issues below.
4.4. Robustness - alternative nesting strategy
While our exposition, and our results, suggest a high degree of cross-
molecular substitution in the hypertension market, it is useful to consider the
robustness of our results to alternative characterizations of the market. One logical
way to do this is to segregate hypertension drugs into broad categories based on the
biochemical pathway through which the drug treats the underlying disease. Broadly
speaking, three major categories of hypertension drugs coexist during our sample
period: beta blockers, ACE inhibitors, and calcium channel blockers. We will not
go into the biochemical details of the different ways in which these drugs treat
hypertension, but the differences are substantial enough that each of these
categories is independently recognized in the ATC classification system.
59
We thank Brian Kovak for this suggestion. Results of this test will be incorporated in later
versions of the paper and are available from the authors upon request.
177
We consider two substantially different ways to nest hypertension
drugs in the product space. We could imagine that consumers choose molecules but
not treatment categories, and model consumer choice as being one in which
consumers are just as likely to substitute a beta blocker for an ACE inhibitor as they
are to substitute between ACE inhibitors. This is the approach described above.
An alternative is to model consumers as first choosing a treatment category – ACE
inhibitors, beta blockers, or calcium channel blockers – then choosing molecules
within (but not outside) these categories. This second approach allows for
significant cross molecular substitution within categories, but zero substitution
across them. The reality of prescription and consumption behavior lies somewhere
between these sharply different modeling approaches. If, however, we can
demonstrate that estimated consumer gains are broadly similar, regardless of which
approach we take, then that would engender greater confidence in our results and
suggest that they are not simply an artifact of a particular approach to demand
modeling. As it turns out, our estimates, which we discuss below, do appear to be
impressively robust to this alternative nesting strategy.
5. Empirical results
5.1 Descriptive statistics
Table 3 presents descriptive statistics for the hypertension market and
the products we focus on in this study. The total number of products (including all
dosage forms) varied between 147 and 175 during our focal 47 quarters starting
from the second quarter of 1997 to the fourth quarter of 2008.60
Over the same time
frame the number of generic products varied between 41 and 69 and the number of
60 This time frame was chosen due to limitations in the IMS data.
178
firms varied between 54 and 73, with a marked increase towards the end of the
panel. Average real price per SU was approximately $0.90 and average sales were
approximately 16 million SUs each quarter. While product-level variation exists,
there are, on average, three contraindications per product. Finally, aggregate
quarterly detailing, on average, was $1.67 million for branded products and zero for
generic products.
As Table 4 demonstrates, on average, challenged products had $600
million of sales per quarter over our sample period. Average wholesale prices per
SU (for all versions, branded and generics) ranged between $0.40 and $4.44, but
decreased, on average, by 26.8% after Para-IV entry. It is interesting to note that
after entry, the originator increased price by about 28%; this is not unexpected since
the remaining brand consumers will have a more inelastic demand (Bhattacharya
and Vogt, 2003). In the quarter of entry we find that the discount factor offered by
the generic entrant is approximately 38% in comparison to branded product prices,
however it varies widely depending on the type of drug. For example, in one case
the discount factor offered in the quarter of entry was just over 97%. Figure 5
illustrates the variability in these discount rates.
Another key feature in the data is the intensity of entry that follows a
Para-IV challenge (i.e., other follow-on generic firms that enter once the entrant
firm‟s 180-day exclusivity period ends). Across therapeutic markets there are, on
average, 11 subsequent generic entrants. This subsequent follow-on entry, more so
than the first (Para-IV) generic entrant, drives prices down and compresses
pharmaceutical firm revenue. For example, during the first year after Para-IV entry,
179
average branded pharmaceutical product revenues eroded by approximately 52%
and collapsing by over 78% by the year after subsequent follow-on generic entry.
5.2 Results
5.2.1 Coefficient estimates
The nested logit demand regression results are reported in Table 5 and for
robustness purposes multinomial regression results are reported in Table 6.61
Recall
that the dependent variable, defined in Equation 8, is the difference between the
logs of the conditional share and share of the outside good. Focusing on the nested
logit results, Model 1 (Table 5) presents results utilizing instrumental variables.
Several items are worth noting. First, the sign on the coefficient Price is negative (-
0.303) and statistically significant. When we compare this coefficient estimate for
hypertension with those obtained from recent work on other therapeutic markets,
our results lie within the range of results obtained by those authors. For example,
Stern (1996) uses aggregate U.S. data across four therapeutic categories from 1978
to 1991 and finds a price coefficient ranging between -0.91 and -10.95. Cleanthous
(2002) uses aggregate data on depression from 1980 to 2001 and finds a price
coefficient of -1.93. A more recent comparison is with Dutta (2011) who uses a
panel of drugs across various therapeutic markets to estimate pharmaceutical
demand in India. For anti-hypertensive drugs she reports a price coefficient of -
0.07, perhaps indicative of how elasticity of demand varies across different
countries.
61
When σ approaches 1 (perfect substitutability) ln(sj|m) becomes 0 and the nested logit simplifies to
a multinomial logit.
180
The results also point to strong evidence of cross-molecular
substitution; the coefficient on the within-group term ln(sj|m) is positive (0.497) and
significant. This indicates a fair amount of substitutability across molecules within
hypertension markets. Discussions with physicians and drug substitution guidelines
issued by a major insurer also attest to a high degree of substitution across
molecules. In future work, we hope to compare analysis of therapeutic categories
with fairly high levels of cross-molecular substitution with other markets
characterized by lower levels of cross-molecular substitution such as epilepsy. The
overall impact of Para-IV entry on consumers and producers is likely to be much
greater in treatment areas with high levels of cross-molecular substitution. In terms
of the signs on the other variables, our results seem to follow economic intuition.
Shares are decreasing in number of contraindications for drugs, which is to be
expected. Finally, the coefficient on the log of advertising, Lnadver, is positive and
significant providing evidence of its critical role in influencing drug demand.
5.2.2 Welfare implications and counterfactual analysis
Welfare calculations are derived using Equation (10). Cumulative
consumer surplus generated from 1997 to 2008 was approximately $270 billion.
This includes all generic entry with all choice sets available to the consumer.
Average quarterly consumer surplus generated from this calculation is
approximately $5.7 billion or about $23 billion per year. When we remove the
Para-IV facilitated generic entry from the sample and generate consumer surplus in
a counterfactual world the figure drops to approximately $177 billion. This implies
that early Para-IV generic challenges created a cumulative consumer surplus of $92
billion over our time period in the U.S. hypertension market (See Figure 6). Hatch-
181
Waxman set out to strike a balance between access and innovation; for the first time
we have been able to quantify the benefits to consumers as a result of the early
entry via Para-IV challenges. These results should not be surprising since prior
work has already demonstrated the gains to consumers as a result of normal generic
entry.
The average quarterly consumer surplus in the counterfactual world is
about $3.8 billion ($5.7 billion in the real world), suggesting that in each quarter of
our analysis, Para-IV entry delivered gains to the consumer of just over $2 billion or
$8 billion annually. These are large flows that can be explained by lower prices and
the expansion in the consumer choice set (See Table 7). To illustrate the size of
these gains for expositional purposes, we can re-express them in terms of cost
savings per-dose of medication.62
If we assume that, on average, 60 million
consumers in the U.S. are annually suffering from hypertension and that all of them
took medication for this condition (a counterfactual, but illustrative assumption),
this translates to annual consumption of 21.6 billion pills (60 million * 360),
assuming chronic intake. Given annual consumer surplus gains of $8 billion, this
amounts to per pill consumer gain of approximately $0.37 which translates to a
yearly savings of $133 per consumer. Given the high incidence of prescription drug
coverage in the U.S., a question remains however as to whether the full welfare
gains flow through to consumers or whether some of that value is appropriated by
the insurance industry, an issue we raise below.
62
This back-of-the-envelope calculation is offered purely for expositional purposes, to help readers
comprehend the magnitude of estimated gains.
182
5.2.3 Robustness - alternative nesting strategies
In Section 4.4 we discussed alternative ways of partitioning the
product space and modeling consumer choice across potential substitutes in the
hypertension market. In the main results stressed in this paper, we “nest” or cluster
products into groups with the same active ingredient(s), but we do not impose any
additional structure on consumer choices, and we allow for relatively unconstrained
substitution across the major subcategories of hypertension drugs, such as ACE
inhibitors, calcium channel blockers, beta blockers, and their combinations.
An alternative approach is to allow for substitution across molecules
within a subcategory (i.e., ACE inhibitors and drugs containing ACE inhibitor
compounds), but not across subcategories. We treat the subcategories as distinct
markets. Discussions with prescribing physicians suggest that the reality is
somewhere between the assumptions we made in the previous section and the
assumption we make here. Fortunately, we can demonstrate that we get broadly
similar estimates of consumer surplus gains, regardless of which set of assumptions
we make. Table 8 reports the coefficients we obtain for the parameters of our
indirect utility function when we estimate demand by subcategories. Table 9
reports the difference between real and counterfactual, consumer and producer
surplus when we calculate both measures by subcategories. The results are not that
different from those described above.
5.2.4 Welfare for whom? And at what cost?
A number of issues arise when considering these estimates of
consumer surplus. First, how do these gains compare with producer surplus losses,
and are their net social gains? Such a comparison requires an estimate of what
183
producer prices and quantities would have been in a counterfactual world without
Para-IV generic challenges. While we are experimenting with a formal supply side
model, as in Dutta (2011), we find that we are able to predict incumbent sales and
price outcomes with surprising accuracy, simply by extrapolating the trends found
in early sample data for particular products to their later periods.63
In many cases,
patent expiry lies beyond our sample period, so we can simply predict
counterfactual producer behavior as a continuation of pre-Para-IV entry trends that
extends to the end of our sample. In other cases, patent expiry occurs before the
end of the sample. In that case, we take the price and quantity declines observed
after the initial 180-day exclusivity period, but re-date these to the expiration of the
patent rather than the successful Para-IV challenge.
Of course, calculating producer surplus requires that we subtract costs
from revenue. Rather than use a formal model to derive an estimate of marginal
cost, we take the late sample generic price as a measure of marginal cost.64
Conversations with industry insiders confirmed that significant generic entry tends
to drive prices close to marginal cost, limiting the profits even for generic
producers. We also subtract product-specific advertising expenditures from
producer surplus. In calculating the counterfactual, we presume that advertising
expenditures follow early sample trends. In the real world, producers tend to cut off
advertising almost entirely after generic entry, and we use real world expenditure
63 We actually applied this procedure to markets that either never saw generic entry or saw it very
late in the sample, and found that our “out of sample” predictions were surprisingly accurate for price and quantity. 64 To be precise, for each molecule we looked for the lowest late sample generic price and took that
as our estimate of marginal cost for all producers in the molecule. In two cases of combination
molecules, there was no generic entry, so we took the average of the lowest generic price for the
element molecules in that combination, and used that average as our estimate for marginal cost.
184
levels in computing producer surplus for the actual market history we observe.
While we have total R&D spending for publicly listed pharmaceutical firms, we do
not possess data on R&D spending for creation or improvement of particular
products. As such, we assume that branded pharmaceutical firms spend as much on
R&D as on advertising, and we deduct this from revenues to create a final measure
of “producer surplus.”
Generics producers earn profits in the real world that they would not
have earned in the counterfactual world, but the intensity of generic competition
compresses these profit flows very rapidly after the 180-day exclusivity period
ends. Cumulatively, by the end of our sample in 2008, total losses to all producers
because of Para-IV entry amount to approximately $14 billion or approximately
about 15% of the gains that flow to the consumer (See Figure 6). This suggests that
the social gains are quite large – cumulating to nearly $78 billion over our sample
period. While more in-depth work needs to be completed on the supply-side, these
figures represent the first attempt that we are aware of to quantify the loss to
producers and net social gains resulting from Para-IV entry under Hatch-Waxman.
But the extent to which these social gains are being realized by the
consumer is open to question. Most American consumers do not purchase
prescription drugs directly; the drugs they consume are prescribed by physicians
and they pay a fraction of the retail price (the co-pay) at the point of purchase, with
the rest being covered by prescription drug insurance. In this paper, we abstract
from the reality that there are institutions mediating between the ultimate drug
consumers and the sellers, and we assume that the measured price declines and
generic entry really do translate into “consumer surplus.” In a world in which
185
doctors and insurance firms really do act in the best interests of their patients and
policyholders, it would not matter whether consumers participate directly or
indirectly in this market. Drug price declines would, in such a world, be passed
through to consumers, either in the form of lower co-pays, lower premiums or both.
For our purposes, we will maintain the assumption that insurance firms work this
way.
In future work, though, a much more complete and thorough
consideration of the degree to which insurance firms appropriate the gains
generated by pharmaceutical price declines is certainly warranted. Inspection of
actual co-pay data from IMS surveys revealed the existence of a number of products
in which the insurance company realized very significant declines in real drug cost,
but did not adjust patient co-pays by the same amount, in percentage change terms.
Furthermore, while originator products‟ prices typically rise modestly after generic
entry, the co-pays required of consumers who choose these products often go up by
much more than the price, in percentage change terms. As an illustrative example,
we observed the case of a branded drug, Altace, which encountered very strong
generic competition toward the end of the sample. 65
Using co-pay data, we
constructed a “pass-through” coefficient that measured the percent change in the
branded drug‟s price after generic entry divided by the percent change in co-pay for
patients who elected to continue to purchase the branded drug. This coefficient was
about .55, well below one, demonstrating that the percent change in price was much
smaller than the percent change in patient co-pay. On the other hand, the price of
65
We find that the data behaves in a strikingly similar manner in other brands like Coreg and
Norvasc.
186
generic versions of the drug plummeted. However, patient co-pays fell only
modestly. The ratio of the percent change in price over the pre- and post-entry
periods to the percent change in co-pays was well above one, implying that co-pays
fell by far less than the wholesale price for patients switching to generic versions.
Finally, even if the measured price declines generate increased
welfare for consumers in the short run, the prospect of limited profitability from
future drug development could lead to either a significant decline in pharmaceutical
R&D or a change in the type of drugs being developed.66
If this, in turn, leads to a
significant slowdown in the rate at which new effective drugs are introduced, then
consumer welfare could decline in the long run, even if the short-run gains from
increased access are large. For example, work by Lichtenberg (2005) has
demonstrated that newer drugs were positively related to life expectancy. In this
paper, we do nothing to estimate the losses associated with foregone future
innovation. However, a full welfare analysis of the balance struck by Hatch-
Waxman will remain incomplete until our estimates of short-run gains can be
compared to appropriately discounted (potential) losses induced by an R&D
slowdown. Conducting such an analysis is the focus of ongoing research efforts by
the authors.
6. Discussion & Conclusion
This paper estimates the impact of early generic entry facilitated via
Para-IV certifications in the U.S. hypertension market between 1997 and 2008.
While Para-IV certifications have been legally possible since the passage of the
Hatch-Waxman Act in 1984, a series of legal and institutional barriers kept the
66 Especially in orphan drugs or drugs for rare pediatric diseases with low patient population base.
187
number of challenges quite low until the late 1990s. Since then, a series of court
cases and procedural changes have significantly lowered the costs and raised the
success probabilities of these challenges. We view these changes as constituting a
slowly unfolding natural experiment. Using unusually rich data, we estimate a
discrete choice demand model, and use this model to back out the first known
estimates of impact of these challenges on social welfare. We calculate welfare
gains to the consumer of approximately $92 billion over our time period. Our
regression results also point to substantial cross-molecular substitution within the
hypertension market, a feature which amplifies the impact of early generic entry on
consumers and producers. The gains to the consumer are an order of magnitude
larger than estimated losses incurred by producers ($14 billion). Placed in context
with the cost of treating hypertension in the United States, the gains appear to be
quite large. Given the scope of Para-IV certifications in other large drug markets,
our results suggest that Hatch-Waxman has generated substantial (short-term) net
benefits.
However, our estimates come with important caveats. In the long run,
diminished profits for pharmaceutical innovators could lead to lower consumer
welfare as the level of R&D and the pace of product development decline in the
future. It may be the case that we are trading gains today at the cost of future drug
development; an issue we are currently exploring in follow-up work. Industry
representatives‟ point to the relatively short period of data exclusivity provided for
under Hatch-Waxman, and note that other major pharmaceutical markets like
European Union, Japan or Canada provide longer data exclusivity periods (See
Table 10). There is no comparable Para-IV mechanism in these other markets.
188
Recently, US Congress passed legislation mandating a 12 year data exclusivity
period for large molecule (biotechnology) drugs. Our results point to significant
loss to producers as a result of Para-IV entry in the small molecule hypertension
market. If incentives, on the margin, push producers to shift R&D away from small
molecule drugs then policy makers may need to consider altering current data
exclusivity periods for these drugs.
While we presume that the “consumer surplus” generated by increased
generic availability and price declines actually goes to consumers, it is at least
possible that some of this downstream surplus is appropriated by insurance firms.
Using proprietary co-payment data from IMS we intend to bring this possibility into
our analysis in future work. If it is the case that the insurance industry is
appropriating a significant portion of the consumer surplus generated from these
regulatory actions, then our work calls for a much deeper understanding of the
distortion that the insurance industry is causing in the pharmaceutical market.
Our current efforts to estimate prices and quantities in a counterfactual
world without generic entry rely on extrapolation of early sample trends rather than
a formal supply side model that endogenizes pricing decisions. As it turns out,
these simple extrapolations appear to result in a high degree of predictive accuracy.
Nevertheless, we are currently seeking to augment our results herein with a more
sophisticated supply side model, such as that utilized by Dutta (2011).
Finally, the hypertension market is just one of many important
therapeutic categories in the U.S. pharmaceutical industry. As our research agenda
proceeds, we intend to supplement the results described herein with results from
other markets, many of which are plausibly characterized by significantly different
189
levels of cross-molecular substitution (e.g., depression, epilepsy and GERD). Only
when we have studied a much larger fraction of the total pharmaceutical industry
will we be in a position to render even partial judgment on the efficacy of Hatch-
Waxman. As is usually the case in economic research, much more remains to be
done.
References
Acemoglu, D. and Linn, J., 2004. Market size in innovation: Theory and evidence
from the pharmaceutical industry. The Quarterly Journal of Economics 119, 1049–
1090.
Appelt, S., 2010. Authorized generic entry prior to patent expiry: Reassessing
incentives for independent generic entry. Unpublished working paper, Munich
Graduate School of Economics, University of Munich.
Anderson, S.,de Palma, A., Thisse, J., 1992.Discrete Choice Theory of Product
Differentiation. Cambridge, Massachusetts: Harvard University Press.
Berndt, E., Cockburn, I., Griliches, Z., 1996. Pharmaceutical innovations and
market dynamics: Tracking effects on price indexes for antidepressant drugs.
Brookings Papers on Economic Activity: Microeconomics2, 133-188.
Berndt, E.,Bir, A., Busch, S., Frank, R., Normand, S., 2002. The medical treatment
of depression, 1991-1996: Productive inefficiency, expected outcome variations,
and price indexes. Journal of Health Economics21(3), 373-396.
Berndt, E., 2002. Pharmaceuticals in US health care determinants of quantity and
price. Journal of Economic Perspectives 16, 45-66.
Berndt, E., Pindyck, R., Azoulay, P., 2003. Consumption externalities and diffusion
in pharmaceutical markets: antiulcer drugs. Journal of Industrial Economics 51(2),
243-270.
Berndt, E., Mortimer, R., Bhattacharjya, A., Parece, A., Tuttle, E., 2007. Authorized
generic drugs, price competition and consumers‟ welfare. Health Affairs 26(3), 790-
799.
Berndt, E., Bui, L., Reiley, D., Urban, G., 1995. Information, marketing and
pricing in the U.S. anti-ulcer drug market. American Economic Review 85(2), 100-
105.
190
Berndt, E., Aitken, M., 2010. Brand loyalty, generic entry, and price competition in
pharmaceuticals in the quarter century after the 1984 Hatch-Waxman legislation.
NBER Working Paper No. 16431.
Berndt, E., Newhouse, J., 2010. Pricing and reimbursement in U.S. pharmaceutical
markets. NBER Working Paper No. 16297.
Berry, S., 1994. Estimating discrete choice models of product differentiation.
RAND Journal of Economics 25(2), 242-262.
Berry, S., Levinsohn, J., Pakes, A., 1995. Automobile prices in equilibrium.
Econometrica 63(4), 841-890.
Berry, S., Levinsohn, J., Pakes, A., 1999. Differentiated products demand systems
from a combination of micro and macro data: the new car market. NBER Working
Paper No. 6481.
Berry, S., Waldfogel, J., 1999. Free entry and social inefficiency in radio
broadcasting. RAND Journal of Economics 30(3), 397-420.
Bhattacharya, J., Vogt, W.B., 2003. A simple model of pharmaceutical price
dynamics. Journal of Law and Economics 46, 599-626.
Bokhari, F., Fournier, G., 2009. Entry in the ADHD drugs market: welfare impact
of generics and me-toos. Unpublished working paper, Florida State University.
Branstetter, L. G., Sakakibara, M. 2001. Do stronger patents induce more
innovation? Evidence from the 1988 Japanese patent law reforms. RAND Journal
of Economics 32(1), 77─100.
Bresnahan, T.F., Gordon, R.G., 1997. The Economics of New Goods. University of
Chicago Press, Chicago, IL.
Bresnahan, T., Reiss, P.,1991. Entry and competition in concentrated markets.
Journal of Political Economy 99(5), 977-1009.
Bresnahan, T., Stern, S., Trajtenberg, M., 1997. Market segmentation and the
sources of rents from innovation: personal computers in the late 1980‟s. RAND
Journal of Economics 28, 17-44.
Broda, C., Weinstein, D., 2006. Globalization and the gains from variety. Quarterly
Journal of Economics 121(2), 541-585.
Brynjolfsson, E., Kemerer, C., 1996. Network externalities in microcomputer
software: An econometric analysis of the spreadsheet market. Management Science
42, 1627–1647.
191
Brynjolfsson, E., Hu, Y., Smith., M., 2003. Consumer surplus in the digital
economy. Management Science 49(11), 2003.
Bulow, J., 2004. The gaming of pharmaceutical patents. In Innovation, Policy and
the Economy, vol .4. Jaffe, Stern, Lerner (eds.), MIT Press: Cambridge, MA.
Cardell, S., 1997. Variance components structures for the extreme value and logistic
distributions with applications to models of heterogeneity. Econometric Theory 13,
185-213.
Caves, R., Whinston, M., Hurwitz, M., 1991. Patent expiration, entry and
competition in the U.S. pharmaceutical industry. Brookings Papers on Economic
Activity: Microeconomics, 1-48.
Cleanthous, P., 2002. Patient welfare implications of innovation in the U.S.
antidepressant market. Unpublished working paper, New York University.
Clerides, S., 2008. Gains from trade in used goods: evidence from automobiles.
Journal of International Economics 76, 322–336.
Cockburn, I., 2006. Is the pharmaceutical industry in a productivity crisis? In
Innovation Policy and the Economy, Vol. 7, A. Jaffe, J. Lerner, and S. Stern, (eds),
MIT Press, Cambridge, MA.
Congressional Budget Office, 1998. How increased competition from generic drugs
has affected process and returns in the pharmaceutical industry. Congressional
Budget Office, The Congress of the United Sates, http://www.cbo.gov.
Danzon, P., Chao, L., 2000a. Cross-national price differences for pharmaceuticals:
how large and why? Journal of Health Economics 19, 159-195.
Danzon, P., Chao, L., 2000b. Does regulation drive out competition in the
pharmaceutical markets? Journal of Law and Economics, 311-357.
Danzon, P., Wang, Y., Wang, L., 2005. The impact of price regulation on the
launch of delay of new drugs. Health Economics 14(3), 269-292.
Davis, S.J., Murphy, K.M., Topel, R.H., 2004. Entry, pricing and product design in
an initially monopolized market. Journal of Political Economy 112(1), 188-225.
Dick, A., 2008. Demand estimation and consumer welfare in the banking industry.
Journal of Banking and Finance 32(8), 1661-1676.
DiMasi, J., Hansen, R., Grabowski, H., 2003. The price of innovation: new
estimation of drug development costs. Journal of Health Economics 22, 151-185.
192
DiMasi, J., Grabowski, H., 2007. The cost of biopharmaceutical R&D: Is biotech
different? Managerial and Decision Economics 28, 469–479.
Djankov, S., LaPorta, R., Lopez-de-Silanes, F., Schleifer, A., 2002. The regulation
of entry. The Quarterly Journal of Economics 117(1), 1-37.
Dutta, A., 2011. From free entry to patent protection: Welfare implications for the
Indian pharmaceutical industry. Review of Economics and Statistics 93(1), 160-
178.
Ellickson, P., Stern, S., Trajtenberg, M., 2001. Patient welfare and patient
compliance: an empirical framework for measuring the benefits from
pharmaceutical innovation. In Medical Care Output and Productivity, Ernst Berndt
and David Cutler (eds), University of Chicago Press: Chicago, IL.
Ellison, S., Cockburn, I., Griliches, Z., Hausman, J., 1997. Characteristics of
demand for pharmaceutical products: an examination of four cephalosporins.
RAND Journal of Economics 28(3), 426-446.
Federal Trade Commission, 2002. Generic entry prior to patent expiration: FTC
study. U.S. Government Printing Office, Washington, D.C., July.
Federal Trade Commission, 2009. Authorized Generics: An Interim Report. FTC
study. U.S. Government Printing Office, Washington, D.C., June.
Ferrandiz, J.G., 1999. The impact of generic goods in the pharmaceutical industry.
Health Economics 8, 599-612.
Frank, R., Salkever, D., 1997. Generic entry and the pricing of pharmaceuticals.
Journal of Economics and Management Strategy 6(1), 75-90.
Gallant, R., Hong, H., Khwaja, A., 2009. Estimating a dynamic oligopolistic game
with serially correlated unobserved production costs. Unpublished working paper,
Duke University.
Goldberg, P., 1995. Product differentiation and oligopoly in international markets:
the case of the automobile industry. Econometrica 63, 891–951.
Grabowski, H., 2004. Are the economics of pharmaceutical research and
development changing? Pharmacoeconomics 22, 15-24.
Grabowski, H.G., 2007. Data exclusivity for new biological entities. Unpublished
working paper, Duke University, Department of Economics.
http://www.econ.duke.edu/Papers/PDF/DataExclusivityWorkingPaper.pdf
Grabowski, H., 2007. Competition between generic and branded drugs. In Sloan, F.,
Hsieh, C. (Eds.), Pharmaceutical Innovation: Incentives, Competition, and Cost-
193
Benefit Analysis in International Perspective. Cambridge University Press:
Cambridge, MA.
Grabowski H, Vernon J., 1992. Brand loyalty, entry, and price competition in
pharmaceuticals after the1984 act. Journal of Law and Economics 35(2), 331–350.
Grabowski, H., Vernon, J., 1996. Longer patents for increased generic competition:
The Waxman
Hatch Act after one decade. Pharmaco Economics 10, 110–23.
Graham, S., Higgins, M.J., 2010. The effect of Paragraph IV patent challenges on
“new” drug introductions. Working paper, College of Management, Georgia
Institute of Technology.
Griliches, Z., Cockburn., I., 1994. Generics and new goods in pharmaceutical price
indexes. American Economic Review 84, 1213–1232.
Hausman, J., Hall, B., Griliches, Z., 1984. Econometric models for count data with
an application to the patents-R&D relationship. Econometrica 52(4), 909-938.
Hemphill, C., Sampat, B., 2010. When do generics challenge drug patents?
Columbia Law and Economics Working Paper No. 379.
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1640512
Henderson, R., I. Cockburn. 1996. Scale, scope, and spillovers: The determinants of
research productivity in drug discovery. Rand Journal of Economics, 27, 32-59.
Higgins, M., Graham, S., 2009. Balancing innovation and access: patent challenges
tip the scales. Science 326, 370-372.
Hudson, J., 2000. Generic take-up in the pharmaceutical market following patent
expiry – a multi-country study. International Review of Law and Economics 20,
205-221.
Hughes, J., Moore, M., Snyder, E., 2002. Napsterizing pharmaceuticals: Access,
innovation and consumer welfare. NBER Working Paper No. 9229.
Hurwitz, M., Caves, R., 1988. Persuasion or information? Promotion and the shares
of brand name and generic pharmaceuticals. Journal of Law and Economics 31,
299-320.
Knowles, S., 2010. Fixing the legal framework for pharmaceutical research.
Science 327, 1083-1084.
Kyle, M., 2007. Pharmaceutical price controls and entry strategies. Review of
Economics and Statistics 89, 88-99.
194
Kyle, M., Grabowski, H., 2007. Generic Competition and Market Exclusivity
Periods in Pharmaceuticals. Managerial and Decision Economics 28, 491-502.
Lancaster, K., 1966. A new approach to consumer theory. Journal of Political
Economy 74, 132-157.
Lancaster, K., 1977. Consumer Demand: A New Approach. Columbia University
Press. New York, NY.
Lanjouw, J., 2005. Patents, price controls and access to new drugs: how policy
affects global market entry. NBER Working Paper No. 11321.
Lichtenberg, F., 1996a. Do (more and better) drugs keep people out of hospitals?
AER Papers and Proceedings 86(2), 384-388.
Lichtenberg, F., 1996b. The effect of pharmaceutical utilization and innovation on
hospitalization and mortality. NBER Working Paper No. 5418.
Lichtenberg, F., 1998a. Pharmaceutical innovation, mortality reduction and
economic growth. NBER Working Paper No. 6569.
Lichtenberg, F., 2001. Are the benefits of newer drugs worth their costs? Evidence
from the 1996 MEPS. Health Affairs 20(5), 241-251.
Lichtenberg, F., 2005. The impact of new drug launches on longevity: Evidence
from longitudinal, disease-level data from 52 countries, 1982-2001. International
Journal of Health Care Finance and Economics 5, 47-73.
McFadden, D., 1973. Conditional logit analysis of qualitative choice behavior,” In
Frontiers of Econometrics, P. Zarembka (eds.), Academic Press, New York, NY.
Mossinghoff, G., 1999. Overview of Hatch-Waxman act and its impact on the drug
development process. Food and Drug Law Journal 54(2), 187-194.
Moul, C. C., 2006. Consequences of omitting advertising in demand estimation: an
application to theatrical movies. Mimeo, Washington University in St. Louis.
Murray, F., Stern, S., 2006. Do formal intellectual property rights hinder the free
flow of scientific knowledge? An empirical test of the anti-commons hypothesis.
Journal of Economic Behavior and Organization 63(4), 648-687.
Nevo, A., 2000a. A practitioner‟s guide to estimation of random-coefficients logit
models of demand. Journal of Economics and Management Strategy 9(4), 513-548.
Nevo, A., 2000b. Mergers with differentiated products: the case of the ready-to-eat
cereal industry. Rand Journal of Economics, 31(3), 395-421.
195
Nevo, A., 2001. Measuring market power in the ready-to-eat cereal industry.
Econometrica, 69(2), 307-342.
Nordhaus, W.D.. 1969. Invention, Growth, and Welfare: A Theoretical Treatment
of Technological Change. MIT Press, Cambridge, MA.
Nordhaus, W., 2002. The health of nations: the contribution of improved health to
living standards. In The Economic Value of Medical Research. Kevin Murphy and
Robert Topel (eds.), University of Chicago Press, Chicago, IL.
Panattoni, L., 2011. The effect of Paragraph IV decisions and generic entry before
patent expiration on brand pharmaceutical firms. Journal of Health Economics 30,
126-145.
Petrin, A., 2002. Quantifying the benefits of new products: the case of the minivan.
Journal of Political Economy 110(4), 705-729.
PhRMA, 2001. Pharmaceutical industry profile 2001: research for the millennium.
Pharmaceutical Research and Manufacturers of America (Washington, DC).
Reiffen, D., Ward, M., 2005. Generic drug industry dynamics. Review of
Economics and Statistics 87, 37-49.
Rosenthal, M., Berndt, E., Donohue, J., Epstein, A., Frank, R., 2003. Demand
effects of recent changes in prescription drug promotion. In Frontiers in Health
Policy Research, v. 6, David M. Cutler and Alan M. Garber (eds.), MIT Press,
Cambridge, MA.
Scherer, F.M., 2001. The link between gross profitability and pharmaceutical R&D
spending. Health Affairs 20, 216-220.
Saha, A., Grabowski, H., Birnbaum, H., Greenberg, P., Bizen, O., 2006. Generic
competition in the US pharmaceutical industry. International Journal of the
Economics of Business 13, 15-38
Scott Morton, F., 1999. Entry decisions in the generic pharmaceutical industry.
RAND Journal of Economics 30(3), 421-440.
Scott Morton, F., 2000. Barriers to entry, brand advertising, and generic entry in the
US pharmaceutical industry. International Journal of Industrial Organization 18(7),
1085-1104.
Small, K., Rosen, H., 1981. Applied welfare economics with discrete choice
models. Econometrica 49(1), 105-130.
Stern, S., 1996. Market definition and returns to innovation: Substitution patterns in
pharmaceutical markets. Working paper, MIT Sloan School of Management.
196
Suslow, V., 1992. Are there better ways to spell relief: a hedonic pricing analysis of
ulcer drugs. Unpublished working paper, School of Business Administration,
University of Michigan.
Town, R., Liu, S., 2003. The welfare impact of Medicare HMOs. RAND Journal of
Economics 34(4), 719-736.
Train, K., 2003. Discrete Choice Methods with Simulation, Cambridge University
Press, Cambridge, UK.
Trajtenberg, M., 1989. The welfare analysis of product innovations, with an
application computes tomography scanners. Journal of Political Economy 97(2),
444-479.
Williams, H., 2010. Intellectual property rights and innovation: evidence from the
human genome. NBER Working Paper No. 16213.
Wooldridge, J., 1999. Distribution-free estimation of some nonlinear panel data
models. Journal of Econometrics 90(1), 77-97.
197
7. Appendix 1: Instruments - International Prices and Competitor
Characteristics
Instruments from International Prices
Debated extensively in the literature on discrete choice estimation of
demand (Nevo and Whinston, 2010; Angrist and Pischke, 2010), much emphasis is
placed on precise identification of coefficient estimates because of endogeneity of
prices, some of the product characteristics and the within group share term. In our
context, endogeneity is a key concern for prices, log of stocks of advertising and the
within group share term. It is important to understand here that the source of
endogeneity could come from demand shocks for a particular brand that could be
correlated to all the above mentioned regressors. Since the challenge lies in
identifying instruments that would be uncorrelated with the demand shocks, the
literature has suggested that the prices of the same brand in other geographic
contexts might be unrelated to the prices of the brand in the focal market of interest
(Nevo 2001). Accordingly, exploiting the international nature of our pharmaceutical
demand data from IMS MIDAS™, we use novel instruments from international
prices of pharmaceutical drugs as instruments for their U.S. prices. For U.S. prices
of brands, we take for each molecule, branded price in Japan, if that molecule was
being sold in Japan and use that as an instrument for the U.S. price. If the molecule
was not sold in Japan, we turn to the United Kingdom (U.K.) market and retrieve
branded prices in U.K. as instrument for the prices of U.S. brands. For the prices of
generics here in the U.S., we instrument using a similar approach, arguing that
average prices of generics here in the U.S. would be strongly correlated with
198
average prices of generics in Japan (if the molecule is available) or in the U.K. (if
the molecule was not being sold in Japan but in the U.K. instead).67
Instruments from Competitor Characteristics
Other instruments used for precise identification come from competitor
characteristics in line with the suggestion in the BLP literature. We have 6
therapeutic sub-markets of interest in hypertension where Para-IV entry occurred,
notably: C7A0 Beta Blocking Agents, Plain; C7B1 Beta-Blockers in Combination
with diuretics; C8A0 Calcium Antagonist Plain; C9A0 Ace Inhibitors Plain and
C9B1 and C9B3 Ace Inhibitors Combinations. Within each of these ATC4
therapeutic sub-markets, we turn to competitor characteristics for each focal brand
to instrument for its log of stocks of advertising (lnsadver) and the log of the within
group share term (lnsj|m). For each ATC4, there was always more than 1 branded or
generic observation that was competing with the focal brand or generic in our
dataset. For generic observations advertising is 0 in each quarter, its stocks are
therefore 0 too, and therefore within each such ATC4 therapeutic sub-markets, the
lnsadver variable for generics was a 0 in our dataset. Thus for generic observations,
the instrument from competitor generics in the same ATC4 was 0. This same
instrument for brands was the competitor mean stock of advertising from other
brands within the same ATC4 in the same quarter. We employ this approach also to
instrument for the focal brand‟s ln(sj|m) observation and use competitors‟ mean
ln(sj|m) in each quarter (for brands and generics) within each ATC4.
67
We are very grateful to Lowell Taylor for suggesting us to use „international price‟ as instrument
for the U.S. prices in the pharmaceutical context.
199
Figure 1: Exclusivities and Innovation in Pharmaceuticals
NDA Approval Date Expiry of NDA Approval Date by FDA 1
st Patent Expiry on NDA
Data Exclusivity Market Exclusivity
Figure 2: ANDA Patent Certification Options for Generic Manufacturers
ANDA CERTIFICATION OPTIONS FOR GENERIC ENTRANTS
Paragraph III – Patent has not expired
but will expire on a particular day
Paragraph II – Patent
Has Expired
Paragraph I – Required Patent
Information has not been filed
Paragraph IV – Patent is invalid or is
Non-infringed by generic applicant
FDA may approve ANDA
immediately; one or more generic
applicants may enter FDA may approve ANDA
immediately; one or more generic
applicants may enter
FDA may approve ANDA
effective on the date, the
patent expires; one or more
generic applicants may
enter at that time
Generic Applicant provides notice
to the patent holder and NDA
filer; entry of the first-filer may or
may not occur
Source: FTC Study 2002(http://www.ftc.gov/os/2002/07/genericdrugstudy.pdf)
200
Figure 3: Rising Para-IV Certifications in the U.S. Pharmaceutical Market
Figure 4: U.S. Revenues of Hypertension Drugs (in $ 000) 1997-2008
Source: IMS Midas, figures are in $ 000, X-axis is 47 quarters, Q21997-Q42008.
201
Figure 5: Discount Factor by Para-IV Generics in Quarter of Entry for
Some Representative Brand Name Products
Source: IMS MIDAS™. Note: Pre-entry average originator price is normalized to 100. Originator price in quarter of entry on this scale is computed and the exercise is repeated with generic price from Para-IV entrants. The discount factor is subsequently reported.
202
Figure 6: Consumer Gains and Producer Loses in the U.S. Hypertension
Market
Note: This figure comes from calculating the quarterly consumer and producer surplus in the real
and counterfactual world using Equation 10. The final cumulated gains and losses in $ billion by
2008 are reported here.
203
Figure 7: Distributions of Incumbents and Generic Entrants
204
Table 1: Cross Molecular Substitution Suggestions by Blue Cross Blue Shield of Georgia
DRUG CLASS BRAND NAME
GENERIC SUGGESTED THROUGH
GENERICSELECT PROGRAM
Angiotensin Receptor
Blocker/Angiotensin Converting Enzyme
Inhibitor
Cozaar, Diovan/HCT, Hyzaar, Altace (2.5, 5,
or 10 mg), Atacand/HCT, Avapro, Avalide,
Benicar/HCT, Micardis/HCT, Teveten
Benazepril, Enalapril, Enalapril HCTZ, Lisinopril,
Lisinopril HCTZ
Antidepressant
Cymbalta, Effexor XR, Lexapro, Prozac (2 mg/
5 ml solution), Effexor, Luvox/CR, Paxil/CR,
Pexeva, Celexa, Zoloft, Prozac (non solution
formulations) Citalopram, Sertraline, Fluoxetine
Calcium Channel Blocker
Sular, Adalat CC, Cardene/SR, Norvasc,
Plendil, Procardia XL Amlodipine
Statin
Lipitor, Crestor, Pravachol, Zocor, Lescol,
Lescol XL, Vytorin, Mevacor Simvastatin, Lovastatin
Triptan
Maxalt, Maxalt – MLT, Zomig, Zomig –
ZMT, Amerge, Axert, Frova, Imitrex, Relpax Sumatriptan Tablets (Limit to 9 tabs/rolling 30 days)
Note: This comes from a consumer communication issued by Blue Cross Blue Shield in the state
of Georgia, United States. As is evident, insurance firms are looking out for substitution
opportunities across molecules.
205
Table 2: Hypertension Molecules Molecule** Did It Face a Para-IV
Entry? (Y-Yes, N-No)
Quarter of Para-IV
Entry#
Patent Expiration***
AMLODIP BES/BENAZ (6) Y Q22007 Dec-17
AMLODIPINE (3) Y Q12007 Mar-07
ATENOLOL (1) N
ATENOLOL/CHLORTHAL (2) N
BENAZEPRIL (4) N
BENAZEPRIL/HCTZ (5) N
CAPTOPRIL (4) N
CAPTOPRIL/HCTZ (5) N
CARVEDILOL (1) Y Q32007 Jun-07
DILTIAZEM (3) Y Q21999 Jul-05
ENALAPRIL (4) N
ENALAPRIL MAL/HCTZ (5) N
ENALAPRIL/FELODIPINE (6) N
FELODIPINE (3) Y Q42004 Oct-07
FOSINOPRIL (4) Y Q42003 Jul-09
FOSINOPRIL/HCTZ (5) Y Q42004 Jul-09
ISRADIPINE (3) N
LABETALOL (1) N
LISINOPRIL (4) N
LISINOPRIL/HCTZ (5) N
METOPROLOL SUCCIN (1) Y Q42006 Sep-10
METOPROLOL TART (1) N
METOPROLOL/HCTZ (2) Y Q32004 Sep-10
MOEXIPRIL (4) Y Q22003 Feb-07
MOEXIPRIL HCL/HCTZ (5) Y Q12007 Feb-07
NADOLOL (1) N
NADOLOL/BENDROFLUM (2) N
NICARDIPINE (3) N
NIFEDIPINE (3) Y Q21997
PERINDOPRIL (4) N
PINDOLOL (1) N
QUINAPRIL (4) Y Q42004 Feb-07
QUINAPRIL HCL/HCTZ (5) Y Q22004 Feb-07
RAMIPRIL (4) Y Q42007 Oct-10
TIMOLOL (1) N
TRANDOLAPRIL (4) Y Q12007 Oct-10
TRANDOLAPRIL/VERAPAMIL N
VERAPAMIL (3) Y Q32007 Jun-07 Note: This table lists all the molecules across the three broad hypertension drug products that we examine – namely Ace Inhibitors and their variants, Beta Blockers and their variants, and Calcium Antagonists and their variants. The molecules are coded according to the category they belonged to and these numbers are noted in brackets in the first column: Beta Blocking Agents – 1; Beta Blockers + Diuretics – 2; Calcium Antagonist Plain – 3; ACE Inhibitors – 4; ACE Inhibitors + Diuretics – 5; ACE Inhibitors + Calcium – 6. Hypertension drugs witnessed Para-IV entry from the very first quarter of our analysis. Patent expiration month and year are retrieved from paragraphfour.com, and cross checked with FDA Orange Book data and IMS Patent Expiry information where available (or was required) and are for the key patent protecting the drug product.
206
Table 3: Paragraph IV Descriptives related to Hypertension Drug Products
Variable Mean Std. Dev. Min Max
Price per SU 0.927 3.953 0.004 116.410
SUs sold per quarter („000) 16343.7 41617.5 1.0 474276.0
Revenues („000 real USD) 9466.1 39181.3 0.9 630432.9
No of Contraindications 3.8 2.2 0.0 10.0
Advertising (detailing) per quarter
per brand („000 real USD) 1678.7 3145.9 0.0 19983.1
No of brands per quarter 166.5 4.6 157.0 175.0
No of Generic Brands per quarter 50.9 8.8 41.0 69.0
No of firms per quarter 59.3 4.9 54.0 73.0
Note: These numbers are for disaggregate data. For our analysis in order to avoid issues with trivial market shares we aggregate all generics generating prices for them weighed by sales. Incumbent brands are retained with no modifications.
Table 4: Overall Descriptives for Para-IV Generic Entry
Average
Std
Deviation Median Minimum Maximum
ATC Market Size ('000 USD) 622123.2 617686.3 337706.4 9610.6 2699945.0
Average Price Per SU ($) 31.8 90.9 2.9 0.4 444.5
Decrease/Increase In Normalized
P After Entry - All -26.4 37.2 -34.6 -77.9 68.8
Originator Increases/Decreases In
Price After Entry (Normalized) 28.3 52.0 24.0 -61.0 232.4
Discount Factor (%) 38.4 37.5 28.4 2.3 97.5
No Of Entrants 11.0 7.0 10.0 1.0 24.0
Mean Revenue Of Originator (1
Year Before 1st Entry - '000 USD) 166270.0 210101.0 84063.5 2077.0 1087622.0
Mean Revenue Of Originator (1
Year After 1st Entry - '000 USD) 66307.8 97310.5 31633.5 1133.0 479468.0
Revenue Erosion Of Originator
(%) 52.7 30.2 59.5 -42.8 93.9
Revenue Erosion From
Subsequent Entry (%) 78.9 68.0 89.0 -218.6 408.2
Note: These numbers are across therapeutic markets for 55 cases of Para-IV entry.
207
Table 5: Demand Estimation: Nested Logit Specification
Nested Logit – IV Nested Logit - OLS
VARIABLES Diff Diff
Price -0.303*** -0.0182***
(0.0563) (0.000863)
Lnsj|m (Conditional Share Term) 0.497*** 0.981***
(0.135) (0.00708)
Lnadver (Log of
advertising/promotions/detailing) 0.433*** 0.0455***
(0.115) (0.00637)
No of Contraindications -0.189 0.0695***
(0.153) (0.00888)
Constant -4.652*** -2.706***
(1.544) (0.0837)
Product Fixed Effects Yes Yes
Observations 3,745 3,813
R-squared 0.711 0.973
Hansen J Statistic 2.436
Chi-sq(1) P-val 0.11859
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
F-First Stage (Price) = 12.62; F-First Stage (lnsjm)= 314.59; F-First Stage
(lnadver)= 201.85
Instruments used: Form number, number of firms in quarter, time since generic
entry and their lags
Dependent Variable: diff = lnsj-lns0
Note: These estimates come from implementing Equation 8 in our regressions.
208
Table 6: Demand Estimation: Robustness Specification (Multinomial Logit)
Multinomial Logit -
IV
Multinomial Logit –
OLS
VARIABLES Diff Diff
Price -0.394*** -0.00772**
(0.0706) (0.00301)
Lnadver (Log of
advertising/promotions/detailing)
0.828*** 0.370***
(0.0401) (0.0143)
No of Contraindications 0.363*** -0.403***
(0.0407) (0.0347)
Constant -10.21*** -6.487***
(0.415) (0.345)
Product Fixed Effects Yes Yes
Observations 3,745 3,813
R-squared 0.388 0.803
Hansen J Statistic 7.798
Chi-sq(1) P-val 0.02026
Robust standard errors in parentheses *** p<0.01, **
p<0.05, * p<0.1
F-First Stage (Price)=12.62; F-First Stage (lnadver)=201.85
Instruments used: Form number, number of firms in quarter, time since generic entry
and their lags
Dependent Variable: diff = lnsj-lns0
Note: These results are for a simple multinomial logit specification, where we adopt a specification similar to Equation 8 but without the conditional share term, ln(sj|m), on the right hand side.
209
Table 7: Drug Availability and Pricing
Choice
Sets*
Average Price
Per SU ($)
Choice
Sets
Average Price Per SU
($)
1st Quarter 68 1.52 64 1.56
23rd
Quarter 78 1.65 71 1.73
47th
Quarter 92 2.23 68 2.98
TIME
REAL WORLD COUNTERFACTUAL WORLD
*This counts number of brands having non-zero sales. Note: This table documents why
one should see consumer gains with increasing choice sets and price shifts in the real and
counterfactual world.
Table 8: Demand Estimation: Alternative Nesting Approach
Calcium Antagonist Beta Blockers Ace Inhibitors
VARIABLES Diff Diff Diff
Price -1.164 -0.276** -9.075***
(0.917) (0.133) (1.485)
ln(sj|m) (conditional share term) 0.722*** 0.739*** 0.483***
(0.163) (0.0505) (0.179)
lnadver (Log of detailing) 0.445*** 0.348*** 0.184
(0.111) (0.0979) (0.154)
No of Contraindications -0.166 -0.381* -7.738***
(0.199) (0.203) (1.632)
Constant -2.384*** -2.702 36.98***
(0.533) (2.089) (8.395)
Product Fixed Effects Yes Yes Yes
Hansen J-Stats 0.077 2.456 0.654
Chi Sq P-Value 0.78206 0.11709 0.41865
First Stage Fs 45.7;72.5;54.3 114.9;1293.3;654.7 121.7;135.8;485.9
Observations 1,312 1,024 1382
R-squared 0.890 0.971 0.407
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Instruments used: Form number, number of firms in quarter, time since generic entry and their lags
Dependent Variable: Diff = ln(sjm) – ln(s0)
Note: These estimates come from implementing Equation 8 in our regressions but this was done sub-
market by sub-market.
210
Table 8.1: Demand Estimation: Robustness (International Prices & BLP-Style
Instruments)
Nested Logit – IV
VARIABLES Diff
Price -0.512***
(0.0998)
Lnsj|m (Conditional Share
Term) 0.595***
(0.0525)
Log of stock of advertising
@ 80% persistence of prior
period advertising (Stern
1996)
0.257***
(0.0896)
No of Contraindications -0.251**
(0.103)
Product Fixed Effects Yes
Observations 3,745
Hansen J Statistic 0.119
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
F-First Stage (Price) = 15.27; F-First Stage (lnsjm)= 512.52; F-First Stage
(lnadver)= 133.59
Instruments used: Japanese and UK prices of brands and generics within each
molecule, competitor log of stock of advertising, competitor Lnsj|m in each 4-
digit therapeutic market, & time since generic entry.
Dependent Variable: diff = lnsj-lns0
Note: The method for constructing the instruments for this table are outlined in Appendix I.
211
Table 9: Comparison of Welfare Calculations
Cumulated CS Gains Cumulated PS Gains
Cumulated Social
Gains
Aggregate Nesting 92.2 -14.2 78.0
Market by Market Gains and then Aggregation
Calcium 67.7 -7.1 60.5
Beta 0.8 -2.1 -1.3
Ace 7.7 -4.7 3.0
Total 76.1 -13.8 62.3
All numbers are in USD billion
Note: These figures come from calculating the quarterly consumer and producer surplus in the real
and counterfactual world using Equation 10 and computing their differences, sub-market by sub-
market. The final cumulated gains and losses to the consumer and producer are reported here within
each sub-market and the aggregate number is compared to the gains estimates found from the
aggregate nesting approach.
Table 10: Data Exclusivity Regime in Various Regions
Region Former Present
European Union 6 to 10 10+1
Canada 5 8+ 1/2
Japan 6 8+2
United States 5 5
Data Exclusivity
Regime (years)
Source: Higgins and Graham (2009)
212
E. Conclusion & Future Work
The baseline results of the three essays in this dissertation
provides new empirical intuition supported by econometric evidence on the
relationship between intellectual property (IP) rights and incentives for firm-level
innovation in developing economies, understanding in developing economy firms,
learning by exporting, and ascertaining the relationship between shifts in IP
regulation and welfare in a developed world pharmaceutical market. Like most
scholarly endeavors, each of these chapters is a work in progress and this section
hopes to outline in brief future efforts.
The findings of the first chapter point to the merit of a future
investigation where one might want to understand if hitherto exploitative endeavors
by imitative pharmaceutical firms in India are going to result in explorative
outcomes. To translate into industry parlance, might process imitation be able to
provide complementarities that Indian bio-pharmaceutical firms can use to engage
in basic, product and drug discovery research going forward? The second chapter
offers some intuition that learning gains accrue to these firms from exporting, and
more so by exporting to high-income destinations. However, if one can control for
sample selection in the empirical framework, might those learning gains still hold
true? Efforts are currently underway to address these using appropriate instruments
and propensity score matching techniques. The third chapter documents the welfare
effects in the short run because of early generic entry in a developed world
pharmaceutical market like the United States. The natural follow-on question to this
finding is a subsequent investigation on the long run incentives for innovation in
213
innovator pharmaceutical firms as conditioned by early generic entry in therapeutic
markets in the United States.
Apart from these possible future investigations, this dissertation
also points to two important public policy issues on both the supply and demand
side in global pharmaceutical markets. What might be the role of globalization of
innovation in R&D intensive sectors like pharmaceuticals? Will globalization of
innovation be able to bring about a changed nature of technological change (Arora
& Gambardella 1994) in the context of the pharmaceutical industry and enhance
questions about its research productivity that has cropped up in the last few years?
The investigation in this dissertation also has a natural segue to questions on the
demand side. What have been the effects of stronger IP and a global harmonization
of property rights for global welfare in pharmaceutical markets? In developing
economies like India per se, might a stronger IP regime result in its own Access
versus Innovation tussle with a public policy debate on rising domestic drug prices
in the post-2005 era? How will policy makers and scholars address that issue going
forward? Issues around biologics and large molecule drugs also deserve attention in
this regard. We intend to follow up on these questions in the coming years and
contribute to scholarly work going forward.