The Quantity, Price, and Variety Response
of U.S. Exports to Stronger Patent Protection
Olena Ivus∗
October 2010
Abstract
Using detailed product data from 1990 to 2006, I estimate the effect of strengthening patent rights (PRs) in developing countries on U.S. export variety, prices and
quantities. For over 20 years, the highly contentious debate over the TRIPs agreement centered on these issues. I find that the strengthening of PRs under TRIPs
expanded U.S. export variety. New products added $366 million to U.S. exports
into 24 developing countries. This effect is strong in complex product industries
which rely most on PRs. Existing exports contracted by $136 million, as quantities
of already exported products fell and their unit prices rose.
Keywords: Trade; International law; Intellectual property rights
JEL classification: F10, K33, O34
∗
Queen’s School of Business, Queen’s University, Goodes Hall, 143 Union Street, Kingston, ON K7L
3N6, Canada. E-mail address: [email protected]. Phone: 613-533-2373. The work on this
paper began at and was partly funded by the Institute of Economic Research, Hitotsubashi University
while I was a Visiting Researcher. M. Scott Taylor provided valuable insight and comments which
have led to significant improvements to the paper. I would also like to thank seminar participants
at Hitotsubashi University, Kyoto University, Osaka University, and Soka University (Japan) for their
helpful comments. Feedback from seminar participants at the World Trade Organization and the Fall
2010 Midwest International Trade Meetings is also gratefully acknowledged.
1
Introduction
Intellectual property rights (IPRs) were designated as ‘trade-related’ in 1994, when the
Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPs) was ratified. Despite the passage of time since this designation, it is unclear how IPRs are related
to trade. Do stronger IPRs affect the variety of traded goods, their quantities, or their
prices? If yes, how are they affected? For over 20 years, these questions have been at the
centre of the highly contentious debate over strengthening IPRs in developing countries.
This study analyzes the trade impact of patent rights (PRs) in terms of variety, price
and quantity components. I estimate the response of U.S. exports to the strengthening
of PRs in 64 developing countries over the 1990-2000 period. Specifically, I answer the
following questions: (i) Have strengthened PRs changed the variety of U.S. products
exported or the value of products exported already?; and (ii) How have strengthened
PRs affected the prices and quantities of products already exported? These issues carry
direct implications for developing countries’ access to high-tech products and technological advancement, and also their roles in the global market. Yet despite these important
ramifications, they have been left unexamined to date.
This paper considers the above questions using highly detailed data on U.S. exports
organized by 10-digit product categories.1 More than 7,600 categories were exported into
the 64 developing countries in my sample over the 1990-2006 period, and the data allow
me to analyze how the individual components of exports were affected. I first decompose
U.S. exports into the variety, price and quantity components. I then attempt to estimate
the causal effect of the strengthening of PRs in developing countries on each component.
To plausibly identify the impact of PRs, three possible econometric problems are
addressed. First is the difficulty of quantifying PRs, as such measure must take into
1
Feenstra et al. (2002) data on U.S. exports organized by 10-digit HS classification codes is employed.
As examples, the Pharmaceuticals, Medicinal Chemicals, and Botanical Products industry contains 202
product categories, such as aminoglycoside antibiotics, insulin and its salt, sulfonamides used as drugs,
etc., and the Special Purpose Machinery industry contains 469 product categories, such as air humidifiers
and dehumidifiers, brewery machinery, snowplows, part of guided missiles, etc.
1
account not only the content of legislation, but also the degree of enforcement. In those
developing countries where patent legislation is poorly enforced, the actual strength of
PRs could be systematically overstated.2 Second, a wide range of domestic factors may
influence countries’ imports and their implementation of patent laws.3 This concern of
confounding factors is highly relevant for developing countries, where reforms to PRs are
typically not clearly defined and often tied up with other trade-related market reforms.4
Third, the decision to strengthen PRs could be driven by trade itself: technological
information received through imports of high-tech products, for example, could help a
country to build its own innovative capacity and thus motivate stronger PRs.
In the face of these problems, I implement an instrumental variable approach in which
colonial origin explains changes in developing countries’ PRs over the 1990-2000 period.
To control for unobserved measures of exports correlated with colonial origin, I measure
the outcome variable as the growth rate of an export component in a given industry
relative to the growth rate in the group of industries with the lowest patent effectiveness.
I find that the strengthening of PRs in developing countries under the TRIPs agreement increased the annual value of U.S. exports in industries that rely most heavily on
patent protection by 4.7%. This increase in exports was entirely driven by an expansion
in product variety. As a result of new products exported, the annual value of U.S. exports increased by 7.4%. Existing exports contracted by 2.8%, as the quantities of already
exported products fell and their unit prices rose. The unit price increase and quantity
reduction of products already exported are observed across a wide range of industries,
while the variety expansion is confined to industries with complex product technologies,
such as Special Purpose Machinery, Office and Computing Machinery, and Autoparts.
2
Developing countries’ judicial systems may vary in their independence, efficiency in dealing with
violations, and susceptibility to institutional corruption. These factors are hard to identify and quantify,
lending to a potential systematic overstatement of the actual strength of PRs in a given country.
3
For example, competition policy, innovative capacity, openness to trade, economic integration, level
of development.
4
PRs changes could be tied with other domestic or global policy movements affecting trade. TRIPs,
for example, was bundled together with other multilateral agreements into a single package, and these
changes also may have affected exports.
2
The results are robust to changes in country sample, time period, and industry grouping.
In terms of the dollar value, new products exported in response to the strengthening
of PRs under TRIPs added $366 million to U.S. patent-relying exports into 24 developing
countries. Accounting for the contraction in existing exports reduces the impact to $230
million. This amount explains 48% of the total increase in U.S. patent-relying exports
into 24 patent-reforming countries over the 1990-2000 period.
The empirical strategy is based on a simple observation that the time pattern of PRs
changes is strongly correlated with British or French colonial origin of a developing country.5 Prior to 1990, developing countries formerly colonized by Britain or France (colonies)
implemented most of the strengthening of PRs, while those not colonized by Britain or
France (non-colonies) were less willing to enforce patent laws. This pattern changed in
the 1990s, when non-colonies strengthened their PRs relatively more than colonies. The
strengthening of PRs in non-colonies was externally imposed rather than internally motivated by import-related factors. The external imposition came in the form of the global
movement towards stronger IPRs, which culminated in the ratification of TRIPs in 1994.
The TRIPs agreement prescribes minimum standards for domestic protection of IPRs on
par with developed countries and also requires effective enforcement.6 When the Uruguay
Round ended, acceptance of TRIPs was a condition for WTO membership. If exclusion
from the WTO was not incentive enough to ratify TRIPs, any unwilling country would
likely have also found itself under pressure and sanctions from the U.S., and of course
without recourse to the WTO’s systems of protection. Faced with such consequences,
many non-colonies ratified TRIPs despite internal opposition and unresolved debate.7
5
This approach follows Ivus (2010). PRs protection is measured by the Ginarte and Park (1997)
index which is available for each five-year time period from 1960 to 2005. See Section 2 for details.
6
Members’ compliance with their obligations under TRIPS is closely monitored by the TRIPs Council
(comprised of all WTO Members), and intellectual property disputes are resolved by the WTO’s Dispute
Settlement Body. The inclusion of TRIPs on the agenda of the Uruguay Round is discussed thoroughly
in Maskus (2000) and Stegemann (2000).
7
External pressure only partially accounts for the strengthening of PRs in developing countries in
the 1990s. For those developing countries with their own innovative capacity, strengthening PRs could
have been motivated by their inherent interest in protecting domestic innovations (Maskus, 2000). This,
however, should mainly concern the newly industrialized countries (NICs), which advanced from imi-
3
The changes of PRs and colonial origin are strongly correlated, suggesting that colonial origin is a relevant instrument.8 The instrument however is unlikely excluded. Colonization could have been determined by a wide range of factors related to trade (e.g.,
geographical proximity to the colonizing power), and these other factors must be isolated
from effects properly attributable to PRs. I hope to meet the exclusion restriction by first
measuring each export component in growth rates, and then comparing growth across
industries on the basis of industry classification by patent effectiveness.9 Specifically, I
identify the comparison industry group, which is composed of industries with the lowest
patent effectiveness, and measure growth in a given industry relative to growth in the
comparison industry group.
With the changes outlined above, the exclusion restriction requires that colonial origin
has no effect on the relative growth in U.S. export components, other than its effect
through changes in PRs. It would fail if non-colonies’ changes in PRs were determined
partly as a function of U.S. relative export growth. U.S. pressure to strengthen PRs
could have been primarily directed on non-colonies if these countries were seen to provide
unique growth prospects for U.S. exports in patent-relying industries (systematically
different from growth prospects in industries with the lowest patent effectiveness). This
scenario seems unlikely, especially since countries with lucrative markets are excluded
from the analysis.10 Nevertheless, to address this concern I identify developing countries
which were actively targeted for intervention by the USTR and then check the sensitivity
of my results to exclusion of these countries. The results remain much the same.
The existing theoretical literature provided valuable insights into the relationship between Southern IPRs and Northern export variety, prices, and quantities. Stronger IPRs
encourage Northern firms to develop new technologies (Diwan and Rodrik, 1991) and
tating foreign technologies to innovating. These countries (i.e., Brazil, China, India, Malaysia, Mexico,
Philippines, South Africa, and Thailand) are not considered here.
8
The first-stage regression results are provided in Section 2.
9
The classification of industries by patent effectiveness is documented in Cohen et al. (2000).
10
Developed countries and the NICs are not considered here. See Section 2 for details.
4
export a wider range of goods (Ivus, 2011). Prices may rise if the monopolistic power
of innovators rises (Chin and Grossman, 1990; Deardorff, 1992; Maskus and Penubarti,
1997) or innovation for quality improvements falls (Glass and Saggi, 2002);11 but may fall
if innovators employ the best practice research technologies (Taylor, 1994). Quantities
may rise if demand for Northern products rises (Maskus and Penubarti, 1997), production reallocates to the North (Helpman, 1993), or Northern firms free resources from
masquing their technologies to compensate for lax PRs (Taylor, 1993); but may fall if
newly exported goods dilute the market share away from existing exports (Ivus, 2011).
The methodology of export decomposition into the individual components follows
Feenstra and Kee (2004) and Hummels and Klenow (2005). As such I owe much to
these papers. Variety is measured by the extensive margin of exports, where product categories are weighted by their importance in U.S. exports. Existing exports are measured
by the intensive margin, which is further divided into price and quantity components. The
decomposition utilizes cross-country variation in exports, which is akin to my empirical
strategy of comparing U.S. export growth across non-colonies and colonies.
The empirical strategy used here is related to Ivus (2010), however in Ivus (2010) no
distinction between new and already exported products was made, and this distinction is
critical for the results. As such, while Ivus (2010) concluded that stronger PRs increased
the value of developed countries’ exports by increasing the quantity of exports, rather
than their price, this paper clarifies that the positive impact on export quantity is driven
entirely by new products exported. The expansion in export variety is strong enough
that it more than offsets the contraction in existing exports.
This is the first paper to estimate the impact of PRs on export variety, prices, and
quantities. Whether the adoption of stronger PRs results in an expansion of the range of
exported goods, as opposed to resulting in merely higher prices and lower quantities of
11
In the presence of FDI, stronger IPRs may encourage the introduction of new varieties (Lai, 1998),
but discourage quality improvements in existing products (Glass and Wu, 2007). In the presence of
licensing, stronger IPRs can increase innovation (Yang and Maskus, 2001).
5
existing exports, are crucial considerations from the perspective of developing countries.
There is a substantial empirical literature that has considered whether IPRs are related
to trade,12 but the prediction on aggregate exports may hide important differences across
individual export components. Our lack of understanding of how individual export components respond to stronger PRs continues to cause tension during international trade
policy negotiations despite more than 20 years of attention devoted to this issue.
Following this introduction, Section 2 describes the data on PRs changes in developing
countries, U.S. exports, and industry patent effectiveness. I decompose exports into the
variety, price, and quantity components in Section 3, and outline my empirical strategy in
Section 4. The results are presented in Section 5, their sensitivity is examined in Section
6, and the paper is concluded in Section 7.
2
2.1
The Data
Patent rights
I examine the changes of PRs in 64 developing countries which imported U.S. commodities during the 1990-2006 period, and for which data on the strength of PRs are available.
These countries are classified as developing economies in the IMF’s World Economic Outlook for the year 2009.13 I exclude from the analysis the newly industrialized countries
(NICs), such as Brazil, China, India, Malaysia, Mexico, Philippines, South Africa, and
Thailand as in these countries, specific measures to enhance IPRs protection were likely
pushed forward by U.S. pressure, exerted through the USTR on a bilateral level since
1989.14 The USTR’s targeting of countries for unilateral trade pressure may be deter12
See, for example, Ferrantino (1993), Maskus and Penubarti (1997), Fink and Braga (1999), Smith
(1999), Rafiquzzaman (2002), Co (2004), Falvey et al. (2009), and Ivus (2010).
13
IMF World Economic Outlook, October 2009 is available at: http://www.imf.org/external/pubs/
ft/weo/2009/02/weodata/groups.htm. Trinidad and Tobago is excluded from the analysis since it is
classified as developed economy by the World Bank.
14
Countries which are rigourously targeted for the USTR’s intervention pursuant to Section 301 are
placed on the 301 Report under the “Priority Watch List” category (which emphasizes increased atten-
6
mined partly as a function of export growth in patent-relying industries. An important
concern in this respect is that reverse causality of export growth affecting PRs could have
influenced the estimated impact of those PRs, producing an endogeneity bias.
The strength of countries’ patent protection is measured by the index of patent rights
documented in Ginarte and Park (1997) and Park (2008). The index spans from 1960
to 2005, and is broken into five-year increments. It covers five measures of patent laws:
patent coverage, membership in international treaties, duration of protection, method of
enforcement, and restrictions on patent rights. For each of these five measures, a country
is assigned a score between zero and one depending on the share of conditions satisfied.
The final index is an unweighted sum of the five scores.
Table 1 enumerates the changes in PRs implemented in the 1990s. In the first column,
the first-stage regression results are presented. The next regression was estimated:
∆P Rj = a0 + aN C + ej ,
(1)
where ∆P Rj ≡ [ln (1 + P Rj,2000 ) − ln (1 + P Rj,1990 )]/(2000 − 1990) is the average annual
change in PRs over the 1990-2000 period, and N C is the non-colony dummy variable
which equals one if a developing country is a non-colony and zero otherwise. The list of
countries (provided in the Appendix A) contains 24 non-colonies and 40 colonies.
It is apparent from the first column that the coefficient on N C is positive (.041)
and highly statistically significant. Thus on average, non-colonies increased their PRs
more than colonies after the year 1990. These changes were a consequence of the global
movement towards stronger IPRs, which culminated in the ratification of TRIPs. The
results also indicate that colonial origin is relevant for explaining variation in PRs changes.
tion in the area of concern) or the “Priority Foreign Country” category (which triggers an additional
investigation, at the end of which sanctions are possible). Before the TRIPs agreement was ratified,
the USTR primarily targeted developed countries and the NICs. Out of the 64 developing countries
considered here, only two appeared on the “Priority Watch List” before 1994: Argentina (non-colony)
and Egypt (colony). As I show in Section 6.3, the results remain similar when these two countries are
excluded from the analysis.
7
The colonial origin instrument is not weak since the F statistic of 32 exceeds its critical
value of 10 (Stock et al., 2002). The magnitudes of estimates and their significance are
not driven by aggregation over time. The coefficients on N C are also positive and highly
statistically significant for the 1990-1995 and 1995-2000 periods, which is shown in the
last two columns.
Table 1: The changes in PRs
Non-colony (N C)
Constant
N of observations
R2
F (1, 62)
P rob > F
1990-2000
.041∗∗∗ (.007)
.017∗∗∗ (.002)
64
.42
32.01
.000
1990-1995
.038∗∗∗ (.012)
.016∗∗∗ (.003)
64
.20
10.14
.002
1995-2000
.043∗∗∗ (.012)
.018∗∗∗ (.004)
64
.22
12.56
.001
Note: ∗∗∗ , ∗∗ , and ∗ denote 1%, 5%, and 10% significance level. Robust standard errors are in parentheses. The outcome variable is the average annual log change in the PRs index over a given period. For
1990-2000, for example, the outcome variable is computed as [ln(1+PR2000 )−ln(1+PR1990 )]/(2000-1990).
For the 2000-2005 period, there is no evidence against the hypothesis that the changes
of PRs are the same for non-colonies and colonies. This could be because TRIPs required
that most obligations with respect to IPRs in developing countries be implemented by the
year 2000.15 In reviewing developing countries’ progress in 2000, the USTR concluded
that “the vast majority of developing countries have made a serious effort to comply
with their TRIPs obligations.”16 This was especially true for non-colonies, which had
to increase their PRs substantially more to meet the TRIPs standards. Given that noncolonies updated their PRs substantially in the 1990s, the relative constancy of their PRs
during the 2000-2005 period is not surprising.
15
Longer transition periods were given for the protection of pharmaceuticals and agriculture chemicals
in some developing countries. Also for the least developed countries, the deadline for the implementation
of TRIPs was 2006. Most of the least developed countries are in the colony group.
16
Source: 2000 Special 301 Report, Office of the USTR, p.6.
8
2.2
U.S. Exports
U.S. exports to each of 64 developing countries are organized by 10-digit HS codes, with
each code representing a specific product category.17 The data are annual and include
information on the value, quantity, and units of exports, as well as descriptive information
about each category. I measure prices as unit-values, which I construct using the data
on values and quantities. For about 20% of HS codes, quantity data are missing. These
categories are thus considered for the analysis of the variety component only.
For each HS code, the units of exports do not vary across destination countries in
a given year. The units do, however, vary for some codes across years and comparing
unit-values for a given country over time requires disregarding those codes that cannot
be converted. For the purposes of this paper, the unit-value of exports into a developing country relative to the average unit-value of exports into all developing countries is
compared within a category over time.
Over the 1990-2006 period, the U.S. exported to the group of 64 developing countries
in 7,622 manufacturing categories. These categories have varied over the years, with the
invention of new products and discontinuance of others.18 Overall, the number of exported
categories has been steadily rising since 1990, as shown in Figure 1. The upward trend
is likely a consequence of U.S. progress in fostering product innovations. It is notable,
however, that a similar pattern does not hold when countries are grouped by colonial
origin. This is illustrated in Figure 2, where the number of categories exported into noncolonies relative to colonies is plotted over time. It is apparent that non-colonies’ relative
number has been rising during the first two years, then began falling in the 1992-1994
period, continued falling until 2002, and thereafter exhibited no discernable pattern.
17
The data are documented in Feenstra et al. (2002) and can be found at www.internationaldata.org.
A change in the composition of exports is driven by a change in the set of categories, rather than by
a change in the HS code used for the identical category. To make sure that a newly introduced HS code
corresponds with a newly developed category, I check the full alphabetic description about each category
and revise the data if the codes for the same category have simply changed over time. In each such case,
I used the latest code and substituted it in place of all other codes previously used for that category.
18
9
6200
6000
5800
Number of categories
5600
5400
1990
1995
2000
2005
year
Figure 1: Number of categories
A cursory review of the data thus suggests that the relative importance of non-colonies
in the number of categories sent from the U.S. changed over the 1990-2006 period. This
could be driven by a wide range of factors common to countries in one group, but not the
other. What is interesting however, is that the pattern shown in Figure 2 closely resembles
the pattern of changes over time in the relative progress of non-colonies in strengthening
their PRs. On average, non-colonies initiated changes to their patent regimes in the early
1990s. At about that time, their relative number of categories in U.S. exports started
declining. The period over which the decline is observed largely overlaps the period
during which non-colonies strengthened their PRs the most. Further, from 2000 to 2005,
the PRs data did not provide evidence that non-colonies strengthened their PRs more
than colonies. Likewise, non-colonies’ relative number of categories stopped declining in
or around 2002, following which there is no remarkable pattern across the groups.
While it appears that the number of categories exported to developing countries and
the reforms of their PRs are negatively associated, such a conclusion is potentially misleading. As an example, assume that when non-colonies were strengthening their PRs,
colonies were implementing other types of domestic policy reforms to attract a wide range
of U.S. products. In such circumstances, policy changes in both country groups might
10
1.22
1.2
1.18
1.16
1.14
Non-colonies' relative number of categories
1.12
1990
1995
2000
2005
year
Figure 2: Non-colonies’ relative number of categories
well be increasing the number of categories exported from the U.S., but if the increase was
more pronounced for colonies, then non-colonies’ relative number of categories would fall.
This example illustrates the importance of controlling for unobserved measures of exports
potentially correlated with colonial origin. For this purpose, cross-industry variation in
exports can be utilized, which I discuss below.
2.3
Industries
For each HS code, the data include the corresponding codes by SITC Revision 3. I used
this information to obtain data by 4-digit ISIC codes Revision 3 and then identified 31
manufacturing industries which conform to the industry definitions used in Cohen et al.
(2000).19 These industries are listed in the Appendix B, where the data on patent effectiveness documented in Cohen et al. (2000) are also provided.20 Patent effectiveness is
19
The correspondence table between SITC Rev.3 and ISIC Rev.3 codes is available at http://ec.
europa.eu/eurostat/ramon. In the correspondence, there are instances when a single SITC code is
matched with several ISIC codes. To find a unique correspondence, I relied on the additional information
available in my export data, such as SIC and NAICS classification codes and categories’ description.
20
Cohen et al. (2000) reports the results of the Carnegie Mellon Survey administered to 1478 R&D
labs in the U.S. manufacturing sector in 1994. Two of the 33 industries defined in Cohen et al. (2000) are
not included here because of no export data (i.e., Semiconductors and Related Equipment; and Search
and Navigation Equipment).
11
defined as the mean percentage of product innovations for which patent protection has
been effective in securing the “firm’s competitive advantage from those innovations.” Protecting product innovations from imitation is the primary motive for obtaining patents.21
To utilize cross-industry variation in exports, I identify the comparison industry group.
This group is composed of six industries that have the lowest patent effectiveness and
are not covered by other categories of IPRs, namely Basic Metals; Non-metallic Mineral
Products; Electronic Valves, Tubes, and Other Electronic Components; Basic Iron and
Steel; Electric Motors, Generators and Transformers; and Communications Equipment.22
Between 1990 and 2006, the number of categories exported in 31 manufacturing industries from the U.S. increased by 20.8% for colonies and 16.4% for non-colonies. Noncolonies’ poor relative performance in attracting a wide range of categories suggests that
reforms of PRs in a developing country and the variety of products it imports from the
U.S. are negatively related. Importantly, this negative relationship fails to hold when
industries are differentiated by patent effectiveness. As such relative to the comparison
industry group, the number of categories exported in the remaining 25 industries rose by
5.4% for non-colonies and fell by 4.1% for colonies.
3
Export Decomposition
To decompose U.S. exports into the variety, price, and quantity components, I adopt the
methodology developed by Feenstra and Kee (2004) and Hummels and Klenow (2005).
The methodology has two advantages. First, it measures export variety by the extensive
margin (EM), which weights categories by their value in total U.S. exports. Compared
21
Blind et al. (2006) reviewed recent survey studies on motives to patent and concluded that in all of
the studies examined, prevention of imitation is the main motive behind patenting.
22
This approach follows Ivus (2010), where the same industries comprised the comparison group. In
Ivus (2010), the classification of industries in the export data was different from that used in Cohen et al.
(2000) and hence, the industry names did not match. Three industries (i.e., Textiles, Food Products, and
Publishing and Printing) have the lowest patent effectiveness in Cohen et al. (2000) but are covered by
other categories of IPRs (such as copyrights, plant variety rights, geographical indications, trademarks)
which are correlated with patent rights.
12
to the simple count of the number of categories, the EM is a more refined measure of
variety, as it accounts for categories’ importance in U.S. exports. Secondly, the decomposition utilizes cross-country variation in exports. This is akin to my empirical strategy
of comparing U.S. export growth across non-colonies and colonies. This approach is also
suitable for the export data utilized here, where the units of categories do not vary across
destination countries in a given year, but they do vary for some categories across years.
Let Xijt represent the nominal value of U.S. exports of a category i into a destination
country j at year t. The set of all categories i exported within an industry k to country j at
k
. The decomposition requires choosing a consistent “reference country
year t is given by Ijt
group” which contains the widest possible set of categories. Let this reference group
include all 64 developing countries. Then U.S. exports into the reference group is given
P
by the total value of U.S. exports into the 64 developing countries, that is Xit ≡ j Xijt .
The entire set of categories exported from the U.S. each year is It ≡ ∪j Ijt .
For each industry k, country j’s share of total U.S. exports into the developing countries
can be decomposed into the extensive margin (EM) and the intensive margin (IM) as:
P
k
i∈Ijt
P
i∈It
Xijt
Xit
=
k
Xjt
≡ EMjtk IMjtk ,
Xt
(2)
where the respective margins are defined by:
P
k
i∈Ijt
EMjtk ≡ P
i∈It
P
Xit
Xit
=
k
i∈Ijt
Xt
P
Xit
and
k
i∈Ijt
IMjtk ≡ P
k
i∈Ijt
Xijt
Xit
=P
k
Xjt
k
i∈Ijt
Xit
.
(3)
EMjtk measures the variety of exports in industry k to country j. It equals the total
value of U.S. exports summed over the set of categories exported in k to j relative to the
total value of U.S. exports summed over the entire set of categories. Alternatively, it is the
share of categories exported in k to j in total U.S. exports. This share depends on the set
of categories exported and not on export value. As such, the EM varies across developing
countries because of the difference in the variety (and not the value) of exports.
13
IMjtk equals U.S. exports into j relative to U.S. exports into the reference group with
both nominal values taken over the set of categories exported in k to j. As such, the set
of categories exported to each country does not explain country difference in the IM.
As an example of export decomposition consider the Basic Chemicals industry, where
non-colonies’ share of U.S. exports was 6.75 times larger than colonies’ share in 1990.
The difference is partially explained by a larger number of categories shipped into noncolonies from the U.S.23 Also, the categories shipped into non-colonies were of a higher
value in total U.S. exports in the Basic Chemicals industry. As a result, the EM of U.S.
exports was 2.44 times greater for non-colonies than colonies. The remaining difference
in export shares was due to the IM, which was 2.77 times greater for non-colonies.
The IM can further be decomposed into the price index Pjtk and the implicit quantity
index Qkjt as IMjtk = Pjtk Qkjt . The price index is measured as a weighted average of the
price ratios:
Pjtk
≡
Y pijt wijt
pit
k
i∈Ijt
where pijt = Xijt /qijt and pit = Xit /qit =
P
j
Xijt /
,
(4)
P
j
qijt are individual prices (unit
values) in country j and the reference group respectively. The weights wijt are computed
as the logarithmic means of sijt and sit as follows:
(sijt − sit )/(ln sijt − ln sit )
,
i∈I k ((sijt − sit )/(ln sijt − ln sit ))
wijt ≡ P
(5)
jt
P
where sijt ≡ Xijt / i∈I k Xijt is the share of category i in k’s exports to j and sit ≡
jt
P
Xit / i∈I k Xit is the share of category i in k’s exports to the reference group.24
jt
23
Non-colonies received 549 categories of basic chemicals in 1990, which was 1.44 times more than
colonies.
24
Feenstra and Kee (2004) develop the exact price index given by πjt = Pjt (λjt /λt )−1/(σ−1) . The
“conventional” price index Pjt assumes the set of categories is the same across the two countries. This
index is defined in (4), where the product is taken over the common set of categories. The term λjt /λt
reflects cross-country differences in categories. Hummels and Klenow (2005) show that this term is
k
equivalent to EMjt
in (3).
14
4
Empirical Strategy
I relate the changes in developing countries’ PRs to the growth rates of U.S. export
components. The average annual growth rate of the variety component over the 19902000 period, for example, is approximated by the log change as follows:
k
k
∆EMjk ≡ (ln EMj,2000
− ln EMj,1990
)/(2000 − 1990).
(6)
When growth is measured in log changes, the change in country j’s share of total
U.S. exports, ∆(Xjk /X), is additively decomposed into the variety, price, and quantity
components as:
∆(Xjk /X) = ∆EMjk + ∆Pjk + ∆Qkj ,
(7)
where ∆Pjk and ∆Qkj are defined similar to (6). To estimate the impact of PRs changes
on each component of exports in (7), the next model could be postulated:
∆Yjk = b0 + bk ∆P Rj + S k + ukj ,
(8)
where ∆Yjk is the outcome variable (i.e., ∆EMjk , ∆Pjk , and ∆Qkj ). The coefficient bk is
industry subscripted, since patent effectiveness varies across industries.25 The vector S k
is the set of industry fixed effects, which is important to include since the PRs index does
not vary across industries. Industry fixed effects will pick up the effects of technological
progress, changes in industry structure, shifts in output mix, industry specific shocks,
etc. Last, b0 is the constant term and ukj is the error term.
Isolating exogenous to exports variation in PRs changes in (8) is highly problematic
for three reasons. First, the strength of PRs may be measured with systematic errors.
Second, a wide range of domestic factors may confound the impact of PRs (i.e., influence
countries’ trade and their patent reforms). Third, U.S. exports may be causing changes
25
The coefficient bk as a function of patent effectiveness is specified in Section 6.1.
15
in PRs, but not the reverse. In the face of these problems, I instrument ∆P Rj by
colonial origin. This procedure is reliable provided the colonial origin instrument is (i)
not weak; and (ii) excluded. The first criterion was evaluated in Section 2.1, where from
the first-stage regression results I concluded that the instrument is not weak. The second
criterion requires that colonial origin has no effect on the outcome variable, other than
its effect through changes in PRs. In other words, the average value of ukj in (8) must not
depend on colonial origin. The exclusion restriction will fail if colonial origin is related to
unobserved measures of exports, potentially embedded in ukj . Denoting the set of these
measures by Sj , I rewrite the error term as ukj = Sj + εkj . Now (8) changes to:
∆Yjk = b0 + bk ∆P Rj + S k + Sj + εkj ,
(9)
where the vector Sj is the set of factors common to all industries within a country.
I remove Sj from (9), and thus hope to meet the exclusion restriction, by analyzing
export growth in relative terms. Specifically, for each country I evaluate the growth rate
of an export component in industry k relative to the growth rate of that component in
the comparison industry group.
Let the extensive and intensive margins in the comparison group be defined as:
P
∗
i∈Ijt
EMjt∗ ≡ P
i∈It
P
Xit
Xit
∗
i∈Ijt
IMjt∗ ≡ P
and
∗
i∈Ijt
Xijt
Xit
,
(10)
∗
indicates that the margins are measured over the full set of categories exported
where Ijt
within six industries comprising the comparison group (∗ ). Also, let the price and quantity
indices of the IM in the comparison group be defined as:
Pjt∗
Y
≡
Pjth
λhjt
and
∗
h∈Ijt
where λhjt ≡
P
h
i∈Ijt
Xijt /
P
∗
i∈Ijt
Q∗jt
IMjt∗
≡
,
Pjt∗
(11)
Xijt is the export share of an industry h in the compari-
16
son group. Now the relative growth in the variety, price, and quantity component is given
by (∆EMjk − ∆EMj∗ ), (∆Pjk − ∆Pj∗ ), and (∆Qkj − ∆Q∗j ) respectively.
To rewrite (9) in terms of the relative growth rates, I first specify the model for the
comparison industry group as follows:
∆Yj∗ = b∗0 + b∗ ∆P Rj + Sj + ε∗j ,
(12)
where ∆Yj∗ varies by country only.26 I then subtract (12) from (9) to obtain the secondstage regression:
∆Yjk − ∆Yj∗ = β0 + β∆P Rj + S k + kj ,
(13)
where β0 = b0 − b∗0 , β = bk − b∗ , and kj = εkj − ε∗j . Notably, the set of factors common to
industries within a country, Sj , is removed. The exclusion restriction now requires that
colonial origin of a developing country does not directly determine the relative growth
rate of U.S. export components. Under this key assumption, the differential effect of
strengthening PRs on U.S. export components, β, is identified.
5
Regression results
Table 2 presents the estimation results for the variety component. To illustrate the
importance of addressing potential endogeneity of PRs to exports, I estimated both (8)
and (13) at this stage. The results are displayed in Panel A and Panel B respectively.
In Panel A, ∆EMjk is the outcome variable. The data are pooled across all 31 industries, and OLS regression is run. Standard errors are robust and clustered by country.27
The results indicate that the coefficient on ∆P Rj is negative, -.030, and statistically
insignificant. This estimate, however, is misleading if PRs changes are endogenous to
The set S k does not appear in (12) because ∆Yj∗ does not vary by industry.
The clustering serves to account for the within-country correlation among observations. It is particularly important given that the PRs index varies across countries but not industries.
26
27
17
exports (in which case the OLS estimator is inconsistent).28
In Panel B, ∆EMjk − ∆EMj∗ is the outcome variable. Here the concern of potential
endogeneity between PRs changes and exports is addressed by measuring the outcome
variable in terms of relative growth rates. The data are pooled across 25 industries, with
the other six industries included in the comparison group. It is apparent that both OLS
and 2SLS estimator produce similar results. The OLS estimate of the coefficient on ∆P Rj
is .801, and the 2SLS estimate is .954. Both estimates are highly statistically significant.
The 2SLS estimate implies that for each 1% increase in the developing countries’ index of
PRs, the variety of U.S. exports in patent-relying industries increased by 0.945% (relative
to the comparison industry group).
Table 2: The variety component
Panel A: EM industry growth
PRs changes
Constant
Industry fixed effects (n=30)
Number of observations
R2
OLS
-.030
(.11)
-.081∗∗∗ (.01)
Yes
1662
0.23
Panel B: EM relative growth
PRs changes
Constant
Industry fixed effects (n=24)
Number of observations
R2
OLS
∗∗∗
.801
-.220∗∗∗
Yes
1357
0.16
(.24)
(.02)
2SLS
.954∗∗∗ (.14)
-.225∗∗∗ (.02)
Yes
1357
0.16
Note: ∗∗∗ denote 1% significance level. Robust standard errors clustered by country (64 clusters) are
in parentheses. In Panel A, EM industry growth (∆EMjk ) is the outcome variable, computed for each industry k as the average annual log change in the EM. In Panel B, EM relative growth (∆EMjk − ∆EMj∗ )
is the outcome variable, computed as the average annual log change in the EM of industry k relative
to the average annual log change in the EM of the comparison group. All averages are taken over the
1990-2000 period.
28
Whether the endogenous regressor is in fact exogenous can be tested after 2SLS estimation, provided
the instrument is excluded. Colonial origin, however, cannot be treated as excluded instrument in (8).
18
Importantly, the sign of the coefficient on PRs changes differs across Panel A and
Panel B. This difference could have resulted from systematic measurement errors in PRs,
confounding factors, or reverse causality from exports to PRs. Neither of these three
potential problems is addressed in (8), and this could have caused an endogeneity bias
severe enough that the direction of the impact was reversed.
Table 3 reports the price and quantity results. ∆Pjk − ∆Pj∗ is the outcome variable in
Panel A, and ∆Qkj − ∆Q∗j is the outcome variable in Panel B. In each model, the data
are pooled across 25 industries, and robust standard errors are clustered by country.
Table 3: The price and quantity components
Panel A: Unit price relative growth
PRs changes
Constant
Industry fixed effects (n=24)
Number of observations
R2
Tests of endogeneity (H0 : PRs changes are
Robust chi2(1)
Robust F(1,1088)
OLS
2SLS
9.015∗ (5.1) 12.947∗∗∗ (1.5)
-.414∗
(.25) -.578∗∗∗
(.19)
Yes
Yes
1112
1112
0.07
0.06
exogenous)
14.94 (p-value=.00)
17.07 (p-value=.00)
Panel B: Quantity relative growth
OLS
2SLS
PRs changes
-8.533 (5.2) -12.705∗∗∗ (1.5)
Constant
.346
(.24) .521∗∗∗
(.19)
Industry fixed effects (n=24)
Yes
Yes
Number of observations
1112
1112
R2
0.06
0.05
Test of endogeneity (H0 : PRs changes are exogenous)
Robust chi2(1)
16.14 (p-value=.00)
Robust F(1,1088)
18.33 (p-value=.00)
Note: ∗∗∗ and ∗ denote 1% and 10% significance level. Robust standard errors clustered by country
are in parentheses. Each model has 59 clusters, since the quantity data in the comparison group are
missing for 5 countries (Burkina Faso, Central Afr. Rep., Malawi, Sudan, and Togo). In Panel A, Unit
price relative growth (∆Pjk − ∆Pj∗ ) is the outcome variable, computed as the average annual log change
in the price index in industry k relative to the average annual log change in the price index in the comparison group. In Panel B, Quantity relative growth (∆Qkj − ∆Q∗j ) is the outcome variable, computed
as the average annual log change in the quantity index in industry k relative to the average annual log
change in the quantity index in the comparison group. All averages are taken over the 1990-2000 period.
19
In Table 3, OLS and 2SLS estimators produced different results. For the quantity
component, for example, the OLS estimate of the coefficient on ∆P Rj equals -8.533
and is statistically insignificant. In contrast, the 2SLS estimate equals -12.705 and is
highly statistically significant. To check if the OLS estimator is consistent, I tested the
endogeneity of PRs changes. All test statistics are highly significant, rejecting the null
hypothesis that ∆P Rj is exogenous in (13). Thus, only 2SLS estimator is consistent in
Table 3. The 2SLS estimation results indicate that the coefficient on PRs changes equals
12.947 for the price component and -12.705 for the quantity component. Both coefficients
are highly statistically significant. Thus, for each 1% increase in the developing countries’
index of PRs, the unit prices of U.S. exports in patent-relying industries increased by
12.947% and quantities fell by 12.705% (relative to the comparison industry group).29
Table 4 summarizes the findings. For completeness, I also measured the impact of
PRs changes on overall export growth and its intensive margin. The PRs data indicate
that non-colonies increased their PRs by 7.8% per year on average over the 1990-2000
period. As a result, the annual value of U.S. exports in industries that rely most heavily
on patent protection increased by 4.7% (i.e., 7.8×0.599). This increase in exports was
entirely driven by an expansion in the set of categories exported. New products increased
exports by 7.4% (i.e., 7.8×0.954). Existing U.S. exports, on the other hand, contracted
by 2.8% (i.e., 7.8×0.355) as the quantities of already exported products fell and their
unit prices rose. In terms of the dollar value, the strengthening of non-colonies’ PRs
during the 1990-2000 period added $230 million to patent-relying U.S. exports into 24
non-colonies, a nontrivial effect given that the total of patent-relying U.S. exports into
29
A comparison between Table 2 and Table 3 reveals that the impact of PRs changes on variety is about
13.57 times smaller than on unit prices. It is also true, however, that the sample standard deviation
in (∆EMjk − ∆EMj∗ ) is about 7.89 smaller than in (∆Pjk − ∆Pj∗ ) and hence, one point log change in
the variety component is more important than one point log change in the price component. To make
the comparison more compelling, I computed the standardized coefficients on ∆P Rj . For the variety
component, for example, the standardized coefficient equals (σP R /σEM )β, where σP R and σEM are the
sample standard deviations for ∆P Rj and (∆EMjk − ∆EMj∗ ) respectively. The standardized coefficients
indicate that for each one standard deviation increase in ∆P Rj , the variety of U.S. exports increased by
0.212 standard deviations and the unit prices of U.S. exports increased by 0.364 standard deviations (in
relative terms). Thus the impact on variety is still smaller, but now only 1.72 times smaller.
20
these countries was $493 million in 1990.30
Table 4: Summary of export decomposition
PRs
0.599∗∗∗
0.954∗∗∗
-0.355∗
12.947∗∗∗
-12.705∗∗∗
Overall exports
Extensive margin (variety)
Intensive margin
Unit prices
Quantities
St.Er.
(.190)
(.145)
(.182)
(1.46)
(1.50)
Obs.
1357
1357
1357
1112
1112
Note: See the notes to tables 2 and 3. Overall exports relative growth (∆Xjk − ∆Xj∗ ) is the outcome
variable, computed as the average annual log change in the exports of industry k relative to the average
annual log change in the exports of the comparison group. IM relative growth (∆IMjk − ∆IMj∗ ) is the
outcome variable, computed as the average annual log change in the IM of industry k relative to the
average annual log change in the IM of the comparison group. All averages are over the 1990-2000 period.
6
Sensitivity Analysis
To reinforce the results obtained thus far, I now examine the sensitivity of the results to
my industry grouping, choice of time interval, and the list of countries.
6.1
Industries
The coefficient β = bk − b∗ measures the differential industry impact of stronger PRs on
the outcome variable. In this section, I explore how the impact of PRs changes varies
across industries with different patent effectiveness. Let the individual industry impact
be given by bk = γ ln E k , where E k is the effectiveness of patents in industry k. Similarly,
b∗ = γ ln E ∗ for the comparison industry group, and so β = γ ln(E k /E ∗ ). Now (13)
changes to:
∆Yjk − ∆Yj∗ = β0 + γ ln(E k /E ∗ )∆P Rj + S k + kj ,
(14)
where ln(E k /E ∗ ) > 0 since patent effectiveness is the lowest in the comparison group.
Using the data on patent effectiveness documented in the Appendix B, I estimate γ. If
30
$230=$493×4.7%×10 years. The total value of patent-relying U.S. exports into 24 non-colonies
increased from $493 million in 1990 to $967 million in 2000.
21
γ > 0, the individual industry impact bk is positive and increasing in E k . Hence, the
differential industry impact β is also positive. If γ < 0, then bk is negative and decreasing
in E k ; and so β is negative as well. Table 5 shows the results.
Table 5: Effectiveness and PRs
Relative growth in:
Variety
Unit Prices
Quantities
PRs
St.Er. PRs
St.Er. PRs
St.Er.
∗∗
∗
∗
Effectiveness×PRs changes
1.304
(.597) 19.224 (10.569) -19.123 (10.718)
Constant
-0.209∗∗∗ (.020) -0.340
(0.208 0.291
(0.206)
Industry fixed effects (n=24) Yes
Yes
Yes
Number of observations
1357
1112
1112
R2
0.14
0.04
0.04
Note: See the notes to tables 2 and 3. The score of patent effectiveness averaged across six industries
in the comparison industry group (E ∗ ) equals 22.57.
It is apparent that γ is positive for the relative growth in variety (1.304) and unit
prices (19.224), but negative for the relative growth in quantity (-19.123). The results thus
confirm the previous findings of positive differential impact on variety and unit prices and
negative differential impact on quantity. The differential and individual industry impacts
move in the same direction because industries with the highest patent effectiveness are
impacted the most: they experience the largest expansion in export variety, as well as the
largest increase in unit prices and reduction in quantities of already exported products.
To allow for a difference in the impact of PRs across industries, I interact 24 industry
indicator variables with ∆P Rj and estimate the following specification:
∆Yjk − ∆Yj∗ = β0 + β k I k ∆P Rj + S k + kj ,
(15)
where I k is a vector of industry indicator variables and β k measures the industry-specific
impacts of strengthening PRs. The results are reported in Table 6, where industries are
ranked by patent effectiveness. The final row reports the F statistic from the test that
the industry-specific impacts, β k , are equal across industries. All three F tests reject the
null hypothesis that there is no difference in the impact of PRs across industries.
22
Table 6: Individual industries
Relative growth in:
Pharmaceuticals, Medicinals
Special Purpose Machinery
Autoparts
Office, computing machinery
Miscellaneous Chemicals
Fabricated Metal Products
Car and Truck
Basic Chemicals
General Purpose Machinery
TV & Radio Receivers
Other Chemical Products
Paper and Paper Products
Machine Tools
Electrical Machinery
Petroleum
Plastic Resins
Aerospace
Rubber and Plastics Products
Glass
Precision Instruments
Textiles
Food Products
Publishing and Printing
Constant
Industry fixed effects (n=24)
Number of observations
R2
F -stat (P rob>F )
Variety
PRs
St.Er.
1.247∗
(.722)
∗∗∗
2.727
(.779)
1.836∗∗∗ (.691)
1.426∗∗∗ (.513)
0.383
(.575)
0.039
(.736)
0.867
(.521)
0.448
(1.10)
1.312∗
(.730)
-0.950
(.699)
1.010∗
(.591)
0.959
(.964)
0.099
(.718)
0.790
(.571)
0.330
(.762)
1.949∗∗
(.782)
0.950
(.615)
0.895
(.654)
0.109
(.663)
∗∗
1.758
(.693)
0.984
(.940)
1.683∗
(.891)
1.099∗
(.593)
-0.225∗∗∗ (.034)
Yes
1357
0.16
1.74
(.045)
Unit Prices
PRs
St.Er.
14.628∗
(8.188)
12.642∗
(6.788)
12.559
(8.208)
∗
11.296
(6.299)
12.601∗
(7.125)
13.013∗
(7.069)
∗
14.197
(7.420)
12.997∗
(7.399)
12.162∗
(6.453)
∗
15.513
(8.067)
12.880∗
(7.301)
12.579
(7.788)
13.045
(7.887)
14.329∗∗ (6.964)
11.316
(7.932)
9.475
(7.478)
11.451
(8.096)
11.324
(7.092)
14.713∗
(8.246)
∗∗
15.288
(6.803)
13.482∗
(7.010)
12.215∗
(6.533)
∗
13.692
(7.266)
-0.516
(0.375)
Yes
1112
0.07
1.74
(.048)
Quantities
PRs
St.Er.
-13.518∗
(6.925)
-10.763∗
(6.069)
-11.319
(6.776)
∗
-10.317
(5.864)
-11.539∗
(6.090)
-10.257
(6.181)
∗
-11.110
(6.144)
-10.618∗
(6.266)
-11.121∗
(5.954)
∗
-10.842
(6.484)
-10.143
(6.356)
-11.026
(6.689)
-10.802
(6.667)
-11.407∗
(6.069)
∗
-11.595
(6.612)
-11.180
(6.839)
∗∗
-12.797
(6.106)
-12.591∗
(6.367)
-12.796∗
(6.635)
-9.596
(6.045)
-11.829∗
(5.977)
-9.531
(6.019)
∗
-10.451
(6.193)
0.519∗∗
(0.239)
No
1112
0.05
3.35
(.000)
Note: See the notes to tables 2 and 3. ∗∗∗ , ∗∗ , and ∗ denote 1%, 5%, and 10% significance level.
Two industries (i.e., Concrete, Cement, and Lime and Medical and Surgical Equipment) are omitted
because of lack of observations. When quantity relative growth is the outcome variable industry fixed
effects are not included, because an F-test fails to reject the null hypothesis that industry fixed effects,
S k , are jointly equal to zero at standard confidence levels (i.e., F (24,58)=1.39, P rob>F =0.16). The final
row reports the F statistic from the test that the industry-specific effects, β k , are equal across industries.
The strengthening of PRs affected industries differently, with the strongest impact
observed in industries which rely on patent protection the most. Across industries with
high patent effectiveness (i.e., ranked high in Table 6), the increase in unit prices and
reduction in quantities of products already exported was widespread. The expansion in
23
export variety, on the other hand, was confined to complex product industries, such as
Special Purpose Machinery, Autoparts, and Office and Computing Machinery.
6.2
Time
Up until now, the changes in PRs over the 1990-2000 period were related to the changes in
export components over the same period. In this section, I redefine the outcome variable
as the average annual growth rate of an export component over the 1990-2006 period and
re-estimate (13). The new results, reported in Table (7), confirm the previous findings.
Table 7: Time period
Relative growth in:
1990-2000
1990-2006
Variety
PRs
St.Er.
∗∗∗
0.954
(.145)
0.446∗∗∗ (.092)
Unit Prices
Quantities
PRs
St.Er.
PRs
St.Er.
∗∗∗
∗∗∗
12.947
(1.46) -12.705
(1.50)
6.265∗∗∗ (.897) -5.697∗∗∗ (.883)
Note: See the notes to tables 2 and 3. In the first row, the average annual log change in a given
export component is taken above over the 1990-2000 period. In the second row, the average annual log
change in a given export component is taken over the 1990-2006 period.
The results also suggest that the impact of strengthened PRs is short-lived, since the
coefficients on PRs changes are smaller for the longer period. It could also be, however,
that PRs changes implemented over the 2000-2005 period (which show the lack of noncolonies’ relative progress in strengthening PRs) distorted the results.
6.3
Countries
An important concern is that PRs changes in some countries have resulted from the USTR
pressure.31 Some of the reverse causality of export growth affecting PRs changes could
have influenced the estimated impact of those PRs changes, producing an endogeneity
bias. This concern is relevant for my results provided that (i) the USTR pressure varied
31
Targets for the USTR’s intervention under Section 301 are recommended by four industry associations: Pharmaceutical Research and Manufacturers of America, Biotechnology Industry Organization,
International Intellectual Property Alliance (representing U.S. copyright-dependent industries), and the
Computer and Communications Industry Association.
24
systematically across non-colonies and colonies; and (ii) the USTR selection of countries
for pressure was determined as a function of U.S. export growth in patent-relying industries (distinct from export growth in the comparison industry group). This seems
unlikely, especially since developed countries and the NICs are not considered here. Nevertheless, to reinforce the credibility of the findings, I further modified the list of countries
to exclude six countries that appear to have been influenced by U.S. pressure: Argentina,
Chile, Ecuador, Egypt, Indonesia, and Jamaica.32
Table (8) presents the results. In the first row, the full sample of 64 countries is
considered; while in the second row, the above six countries are excluded. As can be
seen, the results remain much the same. This is in line with the findings in Branstetter
et al. (2006) that the timing of patent reforms is not correlated with U.S. pressure, as
measured by the placement of a country on the 301 Report.
Table 8: Countries
Relative growth in:
64 countries
58 countries
Variety
PRs
St.Er.
∗∗∗
0.954
(.145)
1.027∗∗∗ (.178)
Unit Prices
PRs
St.Er.
∗∗∗
12.947
(1.46)
12.199∗∗∗ (1.27)
Quantities
PRs
St.Er.
∗∗∗
-12.705
(1.50)
-11.658∗∗∗ (1.33)
Note: See the notes to tables 2 and 3. In the first row, 64 countries include 24 non-colonies and 40
colonies. In the second row, 58 countries include 20 non-colonies and 38 colonies. For the restricted
sample of countries, the first-stage F statistic is F (1, 56) = 22.6.
7
Conclusion
Using detailed product level data, this paper estimated the response of U.S. exports
to the strengthening of PRs in developing countries under the TRIPs agreement. The
32
All but Egypt and Jamaica are non-colonies. Indonesia signed a bilateral copyright agreement with
the U.S. in 1989, enacted its first patent law in 1991, and was carrying out “important elements of
the 1992 understanding” with the U.S. when it was implementing regulations for its patent law in 1994
(Source: 1994 Special 301 Report, Office of the USTR). Next, Ecuador and Jamaica signed comprehensive
bilateral intellectual property agreements with the U.S. in 1993 and 1994 respectively. Chile’s progress
in strengthening its IPRs was closely monitored by the U.S. to ensure conformity of the new laws with
NAFTA. Finally, Argentina and Egypt were the only two countries from my sample that were put on
the “Priority Watch List” of the Special 301 report in before 1994.
25
following questions have been considered: (i) Have strengthened PRs changed the variety
of U.S. products exported or the value of products exported already?; and (ii) How have
strengthened PRs affected the prices and quantities of products already exported? While
the practical importance of these issues has been recognized for more than 20 years,
our lack of understanding of these matters continues, as does the international debate.
Several studies have measured the impact of PRs on the value of trade, but it is left
unclear whether that impact is driven by changes in prices, quantities, or new products.
This paper implemented an instrumental variable approach in which colonial origin
was used to explain changes in developing countries’ PRs. To control for unobserved
measures of exports potentially correlated with colonial origin, export growth in a given
industry was measured relative to that in industries with the lowest patent effectiveness. The approach was designed to address potential endogeneity bias due to systematic
measurement errors, confounding factors, and reverse causality.
The findings indicate that the strengthening of PRs in developing countries under the
TRIPs agreement increased the annual value of U.S. exports in industries that rely most
heavily on patent protection by 4.7%. This increase in exports was entirely driven by an
expansion in export variety. As a result of new products exported, the annual value of
U.S. exports increased by 7.4%. Existing exports contracted by 2.8%, as the quantities
of already exported products fell and their unit prices rose. The unit price increase and
quantity reduction of products already exported was observed across a wide range of
industries, while the variety expansion is confined to complex product industries, such as
Special Purpose Machinery, Office and Computing Machinery, and Autoparts.
New products exported in response to the strengthening of PRs under TRIPs added
$366 million to the value of U.S. patent-relying exports into 24 developing countries. This
is an economically large effect, and its positive impact on trade flows is dominating.
The potential increase in product prices and corresponding decrease in unit sales have
been repeatedly recognized as the most troubling aspects of adopting and expanding
26
strong IPRs in developing countries. While the results of this paper do not rule out these
effects, they indicate that developing countries’ access to high-products may rise as the
variety of products imported from the U.S. expands in response to strengthening PRs.
Due to this expansion in export variety, developing countries’ overall imports from the
U.S. rise. The results also confirm the importance of breaking down the trade impact
of PRs into individual components. If the impact of PRs was measured only in terms
of overall export value, the role of PRs in encouraging exports of new products would
have been underestimated and the contraction in the value of products exported already
would have not been observed.
References
Blind, K., J. Edler, R. Frietsch, and U. Schmoch, “Motives to patent: Empirical
evidence from Germany,” Research Policy, 2006, 35, 655–672.
Branstetter, L., R. Fisman, and C. Foley, “Do Stronger Intellectual Property Rights
Increase International Technology Transfer? Empirical Evidence from US Firm-Level
Panel Data,” Quarterly Journal of Economics, 2006, 121, 321349.
Chin, J.C. and G.M. Grossman, “Intellectual property rights and North-South
trade,” in Ronald W. Jones and Anne O. Krueger, eds., The Political Economy of
International Trade: Essays in Honor of Robert E. Baldwin, Oxford: Basil Blackwell,
1990, pp. 90–107.
Co, C.Y., “Do Patent Rights Regimes Matter?,” Review of International Economics,
2004, 12, 359–373.
Cohen, W.M., R.R. Nelson, and J.P. Walsh, “Protecting their Intellectual Assets: Appropriability Conditions and Why U.S. Manufacturing Firms Patent (or Not),”
NBER Working Paper no.7552, 2000.
Deardorff, A.V., “Welfare effects of global patent protection,” Economica, February
1992, 59 (233), 35–51.
Diwan, I. and D. Rodrik, “Patents, Appropriate Technology, and North-South Trade,”
Journal of International Economics, 1991, 30 (1-2), 27–47.
Falvey, R., N. Foster, and D. Greenaway, “Trade, imitative ability and intellectual
property rights,” Review of World Economics (Weltwirtschaftliches Archiv), 2009, 145
(3), 373–404.
27
Feenstra, R.C. and H.L. Kee, “On the Measurement of Product Variety in Trade,”
American Economic Review, 2004, 94 (2), 145–149.
, J. Romalis, and P.K. Schott, “U.S. Imports, Exports, and Tariff Data, 19892001,” NBER Working Paper no.9387, 2002.
Ferrantino, M., “The Effect of Intellectual Property Rights on International Trade and
Investment,” Weltwirtschaftliches Archiv, 1993, 129, 300–333.
Fink, C. and C.A. Primo Braga, “How Stronger Protection of Intellectual Property
Rights Affects International Trade Flows,” Policy Research Working Paper Series No.
2051, The World Bank 1999.
Ginarte, J.C. and W.G. Park, “Determinants of Patent Rights: A Cross-national
Study,” Research Policy, 1997, 26, 283–301.
Glass, A.J. and K. Saggi, “Intellectual Property Rights and Foreign Direct Investment,” Journal of International Economics, 2002, 56 (2), 387–410.
and X. Wu, “Intellectual property rights and quality improvement,” Journal of
Development Economics, 2007, 82, 393415.
Helpman, E., “Innovation, Imitation, and Intellectual Property Rights,” Econometrica,
1993, 61 (6), 1247–1280.
Hummels, D. and P.J. Klenow, “The Variety and Quality of a Nation’s Exports,”
American Economic Review, 2005, 95 (3), 704–723.
Ivus, O., “Do Stronger Patent Rights Raise High-Tech Exports to the Developing
World?,” Journal of International Economics, 2010, 81 (1), 38–47.
, “Trade-Related Intellectual Property Rights: Industry Variation and Technology Diffusion,” Canadian Journal of Economics, 2011, forthcoming.
Lai, E.L.-C., “International Intellectual Property Rights Protection and the Rate of
Product Innovation,” Journal of Development Economics, 1998, 55 (1), 133–153.
Maskus, K.E., Intellectual Property Rights in the Global Economy, Institute for Internatonal Economics, Washington, DC, 2000.
and M. Penubarti, “Patents and International Trade: An Empirical Study,” in et al.
K.E. Maskus, ed., Quiet Pioneering: Robert M. Stern and His International Economic
Legacy, University of Michigan Press, Ann Arbor, MI, 1997, pp. 95–118.
Park, W.G., “International patent protection: 1960-2005,” Research Policy, 2008, (37),
761766.
Rafiquzzaman, M., “Impact of Patent Rights on International Trade: Evidence from
Canada,” Canadian Journal of Economics, 2002, 35, 307–330.
28
Smith, P.J., “Are Weak Patent Rights a Barrier to U.S. Exports?,” Journal of International Economics, 1999, 48, 151–177.
Stegemann, K., “The Integration of Intellectual Property Rights into the WTO System,” The World Economy, 09 2000, 23 (9), 1237–1267.
Stock, J.H., J.H. Wright, and M. Yogo, “A Survey of Weak Instruments and Weak
Identification in Generalized Method of Moments,” Journal of Business & Economic
Statistics, October 2002, 20 (4), 518–29.
Taylor, M.S., “TRIPS, Trade, and Technology Transfer,” Canadian Journal of Economics, 1993, 26, 625–638.
, “TRIPS, Trade, and Growth,” International Economic Review, 1994, 35, 361–382.
Yang, G. and K. Maskus, “Intellectual property rights, licensing, and innovation in
an endogenous product-cycle model,” Journal of International Economics, February
2001, 53 (1), 169–187.
Appendix
A.
Country Groups
Formerly colonized by
Algeria
Bangladesh
Benin
Burkina Faso
Burma
Cameroon
Central African Rep.
Chad
Congo Rep.
Dominican Rep.
Egypt
Fiji
Gabon
Ghana
Britain or France (40)
Guyana
Nigeria
Haiti
Senegal
Ivory Coast Sierra Leone
Jamaica
Somalia
Jordan
Sri Lanka
Kenya
Sudan
Madagascar Syria
Malawi
Tanzania
Mali
Togo
Mauritania Tunisia
Mauritius
Uganda
Morocco
Zambia
Niger
Zimbabwe
Not colonized by Britain or France (24)
Angola
Honduras
Argentina
Indonesia
Bolivia
Mozambique
Burundi
Nepal
Chile
Nicaragua
Colombia
Panama
Congo Dem. Rep. Papua New Guinea
Costa Rica
Paraguay
Ecuador
Peru
El Salvador
Rwanda
Ethiopia
Uruguay
Guatemala
Venezuela
Notes: Data for colonization origin is taken from the CIA World Factbook:
https://www.cia.gov/library/publications/the-world-factbook/index.html.
29
B.
ISIC
Rev.3
3311
2423
2920
3430
3010
2429
2800
3410
2411
2910
3230
2400
2100
2922
3100
2320
2413
3530
2500
2610
2695
3312
3220
3110
2710
3210
2600
1700
2700
1500
2200
Industry Data
Manufacturing Industries
Medical and Surgical Equipment
Pharmaceuticals, Medicinal Chemicals, Botanical Products
Special Purpose Machinery
Autoparts
Office, Accounting and Computing Machinery
Miscellaneous Chemicals
Fabricated Metal Products
Car and Truck
Basic Chemicals
General Purpose Machinery
TV & Radio Receivers, Sound & Video Reprod. Apparatus
Other Chemical Products
Paper and Paper Products
Machine Tools
Electrical Machinery and Apparatus
Petroleum
Plastics in Primary Forms & of Synthetic Rubber
Aerospace
Rubber and Plastics Products
Glass
Concrete, Cement, Plaster
Precision Instruments
Communications Equipment
Electric Motors, Generators, Transformers
Basic Iron and Steel
Electronic Components
Non-metallic Mineral Products
Textiles
Basic Metals
Food Products
Publishing and Printing
Patent
effectiveness
54.70
50.20
48.83
44.35
41.00
39.66
39.43
38.89
38.86
38.78
38.75
37.46
36.94
36.00
34.55
33.33
32.96
32.92
32.71
30.83
30.00
25.86
25.74
25.23
22.00
21.35
21.11
20.00
20.00
18.26
12.08
Number of
categories
70
202
469
44
156
174
340
80
764
540
90
153
318
381
246
89
93
71
240
121
17
153
50
107
383
209
117
878
232
774
60
Notes: Source: Cohen et al. (2000). In bold: industries in the comparison group. Patent effectiveness
is defined as a mean percentage of product innovation for which patenting has been effective in protecting
the “firm’s competitive advantage from those innovations.” Number of categories: the number of distinct
HS codes within a given ISIC code over the 1990-2006 period. The Communications Equipment industry
contains television and radio transmitters and apparatus for line telephony and line telegraphy.
30