1
The Patent Paradox Revisited:
An Empirical Study of Patenting in the US Semiconductor Industry, 1979-95
Bronwyn H. Hall* and Rosemarie Ham Ziedonis**
May 2000
Abstract
This paper examines the patenting behavior of firms in an industry characterized by rapid
technological change and cumulative innovation. Recent survey evidence suggests that
semiconductor firms do not rely heavily on patents to appropriate the returns to R&D, despite the
strengthening of US patent rights in the early 1980s. Yet the propensity of semiconductor firms to
patent has risen dramatically over the same period. This paper explores this apparent paradox by
conducting interviews with industry representatives and analyzing the patenting behavior of
approximately 100 US semiconductor firms during a period that spans the “pro-patent” shift in
the US legal environment. The results suggest that the strengthening of US patent rights spawned
“patent portfolio races” among capital-intensive firms, but also may have facilitated entry by
specialized design firms during this period.
JEL Codes: O34, O32, L63
Keywords: Patents, IPR, semiconductor industry, cumulative R&D
* Department of Economics, University of California, Berkeley; Nuffield College,
Oxford University; National Bureau of Economic Research; Institute for Fiscal
Studies, London.
**Wharton School of Business, University of Pennsylvania.
An earlier version of this paper was prepared for the January 1999 NBER Conference on “The
Patent System and Innovation,” sponsored by the Alfred P. Sloan Foundation. The Alfred P.
Sloan Foundation’s Competitive Semiconductor Manufacturing Grant to UC Berkeley and the US
Air Force Office of Scientific Research provided additional support for this study through
doctoral research grants to Ziedonis. We extend special thanks to the managers and intellectual
property attorneys who participated in our study and shared their time and insights with us. We
also thank Jerry Karls of Integrated Circuit Engineering, Inc. for sharing industry data and Jeff
Macher of UC Berkeley’s Haas School of Business for facilitating and participating in several of
our interviews. Finally, we gratefully acknowledge the helpful comments and suggestions we
received from two anonymous reviewers, Melissa Appleyard, Clair Brown, Wes Cohen, David
Hodges, Adam Jaffe, Jenny Lanjouw, Josh Lerner, Kristina Lybecker, Rob Merges, David
Mowery, Rob Porter (the editor), Cecil Quillen, Dennis Yao, Arvids Ziedonis and participants in
the Berkeley Innovation Seminar, the NBER “Patent System and Innovation” Conference, and the
STEP Board Conference on Intellectual Property Rights, Washington D.C., February 2-3, 2000.
1. Introduction
2
1. Introduction
In the early 1980s, important changes in the US legal environment ushered in an era
characterized by strong patent rights. Most notable among these changes was the 1982 formation
of a centralized appellate court, the Court of Appeals for the Federal Circuit (CAFC). Although
the CAFC is widely credited with unifying and strengthening the judicial treatment of patent
rights in the United States, the effects of the “pro-patent” court on the innovative activities of
firms remain unclear. For example, survey evidence suggests that firms in most industries have
not increased their reliance on patents for appropriating the returns to R&D over the decade of the
1980s (Cohen, Nelson, and Walsh 2000). Yet this period coincides with an unprecedented surge
in patenting in the United States unaccounted for by increases in R&D spending alone (Kortum
and Lerner 1998). If firms in most industries do not rely heavily on patents to profit from
innovation, then why are they patenting so aggressively?
This paper re-visits this “patent paradox” in the semiconductor industry, where the gap
between the relative ineffectiveness of patents (as reported in surveys) and their widespread use is
particularly striking. In two surveys on appropriability conducted in 1983 and 1994, (the “Yale”
and “Carnegie Mellon” surveys, respectively), R&D managers in semiconductors consistently
reported that patents were among the least effective mechanisms for appropriating returns to
R&D investments (Levin et al. 1987; Cohen et al. 2000).
1
Driven by a rapid pace of technological
change and short product life cycles, semiconductor firms tend to rely more heavily on lead time,
secrecy, and manufacturing or design capabilities than patents to recoup investments in R&D.
Nonetheless, the number of semiconductor-related patents issued in the United States has risen
sharply since the early 1980s, exceeding the overall increase examined by Kortum and Lerner
(1998).
2
Even more important, the propensity of semiconductor firms to patent has also risen
during this period. As shown in Figure 1, we find that patenting per million real R&D dollars in
the semiconductor industry has doubled between 1982 and 1992, from about 0.3 to 0.6.
3
During
the same period, the patent yield for manufacturing as a whole was fairly stagnant and that for
pharmaceuticals actually declined.
1
The 1994 Carnegie Mellon Survey on Industrial R&D in the U.S. Manufacturing Sector (Cohen et al. 2000)
updated and extended the influential “Yale” survey conducted in 1983 (Levin et al. 1987). Respondents in both surveys
were R&D lab managers in a variety of “focus industries.” Both surveys found that R&D managers in only a handful of
industries, including pharmaceuticals, chemicals, and (more recently) biotechnology and medical devices, considered
patents to be an effective mechanism by which to appropriate the returns to R&D. These results echo the findings of
Scherer et al. (1959), Taylor and Silberston (1973) and Mansfield (1986). As discussed below, the Carnegie Mellon
survey extends upon the Yale survey by asking questions on why firms seek patent protection.
2
The number of US patents issued in a narrowly-defined set of semiconductor patent classes more than doubled
between 1981 and 1984, while the number of US patents issued in all US patent classes rose by only 50 percent during
this period (USPTO 1995).
3
The semiconductor industry also provides an excellent setting within which to examine
the effects of stronger patent rights on firms engaged in rapidly advancing, “cumulative”
technologies (Scotchmer 1991; Merges and Nelson 1990). Much like multimedia or computer
firms, semiconductor firms often require access to a “thicket” of intellectual property rights in
order to advance the technology or to legally produce or sell their products. Given the rapid pace
of technological change in this industry, however, any new product or process is likely to overlap
with technologies developed, either in parallel or in the past, by an array of external parties
(Grindley and Teece 1997). Despite significant advancements in the theoretical literature on the
importance of strong patent rights on inducing investments in R&D when innovation is
cumulative (e.g., Scotchmer 1996; O’Donoghue et al 1998), there remains little systematic
evidence on how a shift toward stronger patent rights affects the innovative activities of firms in
the context of rapidly changing, cumulative technologies.
4
Our study casts new empirical light on
this issue by examining the patenting behavior of semiconductor firms during a period that spans
the “pro-patent” shift in the US legal environment.
In order to illuminate the factors underpinning the surge in patenting in this industry and
the effects, if any, of stronger US patent rights on the innovative activities of semiconductor
firms, we employed a combination of qualitative and quantitative research methods. First, we
conducted interviews with intellectual property managers and executives from several US
semiconductor firms—including manufacturers and specialized design firms. Although previous
studies have explored the motives for patenting in this industry,
5
we sought additional insights on
whether the increased patenting per R&D dollar we find in this industry appears to be related to
the strengthening of US patent rights in the 1980s or whether it seems to be driven by unrelated
technological or managerial factors (an important alternative hypothesis). These interviews also
enabled us to investigate the use of patents by semiconductor design firms, which specialize in
chip design and contract out the manufacture of their products to other firms. Many of these firms
entered the industry during the period associated with strong patent rights, but the importance of
patent rights to these firms was unclear.
3
The numbers in Figure 1 were compiled from several sources described in Section 4 below.
4
The most extensive treatment of these issues (Merges and Nelson 1990) relies on historical records to trace the
effects of conferring broad patent rights to inventors engaged in so-called “cumulative systems” technologies, such as
electronics, aircraft or automobiles. In settings where technological advance involves a diverse range of entities
involved in the “systems” architecture, the authors find that the issuance of broad, overlapping patents may lead to
mushrooming transactions costs and therefore slow the pace of innovation. See also discussion in Mazzoleni and
Nelson (1998).
5
For example, see Tilton (1971), Taylor and Silberston (1973), von Hippel (1988) and, more recently, Grindley
and Teece (1997) and Cohen et al. (2000).
4
Our quantitative analysis is based on a much larger sample of approximately 100 publicly
traded, US firms whose principal line of business is semiconductors and related devices
(SIC3674) and whose R&D expenditures are therefore primarily directed towards semiconductor-
related areas.
6
After compiling a detailed database of these firms’ patent portfolios from 1975 to
1998, we match these data with financial variables from Compustat (e.g., R&D and sales) that are
known determinants of patenting in general
7
and that capture additional variables identified in our
interviews. This methodology enables us to advance upon existing studies by constructing
reliable estimates of the patent propensities of individual firms during this twenty-year period
while keeping the broad technological area constant across firms.
8
Thus we are able to determine
whether the upsurge in patenting simply represents changes in the mix of firms in the industry
over time (i.e., the effects of entry and exit), changes in the economic behavior of firms, or both.
An unfortunate weakness of this approach is the exclusion from our analysis of large US
“systems” manufacturers (e.g., IBM, AT&T or Motorola) and non-US firms (e.g., Toshiba,
Samsung or Siemens) that are important patent owners and users of semiconductor technologies.
Because corporate R&D spending is reported for the entire portfolio of a firm’s R&D activities, it
is not possible to isolate the share of R&D expenditures directed toward semiconductor
technologies for these diversified firms.
Our central hypothesis is that the increase in the patent propensities of semiconductor
firms is driven by the “pro-patent” shift in the US legal environment in the 1980s. We
distinguish, however, between two potential effects of strengthened patent rights on the
intensified patenting we observe in this industry. First, the enhanced enforcement environment
could induce more aggressive patenting by firms most vulnerable to “hold up” in the new patent
regime. We label this our “strategic response” hypothesis. Because the costs and risks associated
with being “held up” by external patent owners appear to be especially salient for firms with large
sunk costs in complex manufacturing facilities, we test this hypothesis by examining the
patenting behavior of large-scale manufacturers and whether it has changed in the period of
strong US patent rights. Second, a shift toward stronger patent rights could facilitate vertical
6
As discussed in Section 4, we expanded our sample of firms specializing in semiconductor chip design by using
industry data provided to us by Integrated Circuits Engineering, Inc.
7
See Hausman, Hall, and Griliches (1984) and Hall, Griliches, and Hausman (1986), also discussed in Section 4.
8
In contrast, Kortum and Lerner (1998) examine the increase in patenting activity in particular classes (e.g.,
software and biotechnology) and then look at aggregate trends in R&D intensity across the US and other national
economies. As such, they are unable to determine whether the increased patenting activity in certain areas is simply due
to a simultaneous increase in R&D spending directed towards those areas (in which case, the propensity to patent in
such areas has remained unchanged). Using a different approach, Cohen et al. (2000) construct a “patent propensity”
measure based on the responses of R&D lab managers to questions in the Carnegie Mellon survey. Although the
authors are able to construct a useful estimate of firms’ propensity to patent at one point in time, they are unable to
track annual changes in this important variable for respondents over time.
5
specialization within the industry and lead to the emergence of ‘technology specialists’ (Merges
1998; Arora 1995). We examine this “specialization” hypothesis by assessing whether the surge
in patent propensities in this industry is explained by the emergence of more patent-intensive
design firms.
Several important findings emerge from our research. First, estimating firm-level
patenting behavior during 1979-95 reveals that the upsurge in patenting by US semiconductor
firms is indeed quite remarkable. Controlling for the changes in the mix of firms leaves the
upward trend shown in Figure 1 essentially unchanged. Second, our qualitative and quantitative
evidence suggests that the “pro-patent” shift has in fact contributed to intensified patenting in this
industry, but in ways not examined systematically in previous studies. On the one hand, we find
that large-scale manufacturers have invested far more aggressively in patents during the period
associated with strong US patent rights, even controlling for other known determinants of
patenting (consistent our “strategic response” hypothesis). Instead of being driven by a desire to
win strong legal rights to a standalone technological prize, these firms appear to be engaged in
“patent portfolio races” aimed at reducing concerns about being held up by external patent
owners and at negotiating access to external technologies on more favorable terms. On the other
hand, we find that firms entering the industry since 1982 patent more intensively than pre-1982
entrants. This is especially true of specialized design firms founded after 1982; in fact, we see a
considerable increase in entry by this type of firm (consistent with our “specialization”
hypothesis).
9
Our interviews suggest that stronger patent rights are especially critical to these
firms in attracting venture capital funds and securing proprietary rights in niche product markets.
Although these results highlight the multifaceted role that patents can play among firms
even within the same broad technological area, we find that the primary reason for intensified
patenting among our sample of semiconductor firms is more aggressive patenting by capital-
intensive firms (“strategic response”). The results of our econometric analysis corroborate views
commonly expressed by industry representatives that the 1980s “pro-patent” shift in the US legal
environment altered their firms’ incentives to patent. Nonetheless, in their analysis of the overall
surge in US patenting during this period, Kortum and Lerner (1998) pose two alternative
hypotheses that we also consider. First, Kortum and Lerner suggest that the surge in US
patenting could be attributed to more aggressive patenting by firms endowed with large patent
portfolios in the period preceding the CAFC’s formation. Using aggregate trends in the patenting
9
As discussed below, the emergence of specialized design firms into the industry was also facilitated by the
adoption of standard process technologies used in semiconductor manufacturing (see Macher et al. 1998). An
interesting question for future research is the extent to which this shift in process technology was itself facilitated by
the “pro-patent” changes in the legal environment.
