Professional Documents
Culture Documents
665677, 2015
0305-750X/ 2014 Published by Elsevier Ltd.
www.elsevier.com/locate/worlddev
http://dx.doi.org/10.1016/j.worlddev.2014.08.025
1. INTRODUCTION
* Many generous colleagues and two major funding Grants made this
work possible. The research was a part of the studies undertaken in the
Millennium Nucleus for the Study of Stateness and Democracy in Latin
America, Project N NS100014 where David Altman and Anthony
Pezzola provided important inputs. In addition, the work greatly benetted from the contributions of Francisco Gallego, Jeanne Lefortune, Jose
Tessada, Jim Robinson and the nancial support of the Economic History
and Cliometrics lab funding provided by Conicyt/Programa de Investigacion Asociativa Project N SOC1102. At Emory University, we thank
Cliord Carrubba, Tom Clarke, and Jeery Stanton and the members of
Emory Universitys PIM Colloquium. Additional thanks are also due to
Dani Rodrik, Ricardo Hausmann, and Oeindrila Dube. All the usual
caveats apply. Final revision accepted: August 19, 2014.
665
666
WORLD DEVELOPMENT
667
668
WORLD DEVELOPMENT
mechanisms; and restrictions (for example, compulsory licensing in the event that a patented invention is not suciently
exploited). This index was designed to provide an indicator
of the strength of patent protection, not the quality of patent
systems. The ability to generalize across trends in IPR standard opens possibilities for multiple streams of research. Yet
precisely how to measure and compare the strength of intellectual property systems across the globe has been a vexing
issue for researchers interested in analyzing IPR institutions
with a rigorous comparative approach. Initiated in the late
1990s, Ginarte Park Index oered a solution; it simplied
complex systems within a parsimonious analytical framework.
As a result, the Index has prospered. It has become a standard
for evaluating what comprises strong or weak intellectual
property systems, now common currency in innovation literatures. As of this writing, the GP Index (in its original and
extended format) has been cited more than 1,003 times in
diverse and far-reaching studies. Nevertheless, our selection
of the GinartePark Index as a proxy for IPR strength may
cause unease for legal scholars. Those experts deep in the legal
trenches will be privy to the intricate dierences in national
debates, implementation processes, and enforcement of new
IPR systems. They will point to the specic cases, of which
there are many, illustrating that the Index, helpful as it is,
has limitations. Therefore, it is important to review the most
important of these: rst, developing and industrialized countries are prone to sign agreements by which they subsequently
do not abide. While the Index ascribes value to the signing, for
example, of the Paris Convention for the Protection of Industrial Property, this does not indicate a countrys adherence.
Along these lines, the Index does not consider if IPR protocols
are respected, or if, for example, specic issues such as
(a)
(b)
(c)
(d)
Figure 1. (ad) ECI and traditional innovation metrics. Source: Authors calculations. Red line is a linear t. (For interpretation of the references to colour in
this gure legend, the reader is referred to the web version of this article.)
669
670
WORLD DEVELOPMENT
Table 1. Economic complexity (ECI) and intellectual property index (IPP) OLS
Variables
(1)
ECI
(2)
ECI
(3)
ECI
(4)
ECI
(5)
ECI
(6)
ECI
(7)
ECI
(8)
ECI
(9)
ECI
(10)
ECI
(11)
ECIa
0.104***
(0.028)
0.149**
(0.074)
0.090***
(0.028)
0.016
(0.069)
0.116***
(0.029)
0.002**
(0.001)
0.112
(0.072)
0.085***
(0.028)
0.165**
(0.075)
0.082***
(0.027)
0.184**
(0.076)
0.083***
(0.028)
0.149**
(0.075)
0.090***
(0.028)
0.420***
(0.050)
0.125***
(0.026)
0.000
(0.001)
0.018***
(0.007)
0.020
(0.065)
0.005
(0.004)
0.250***
(0.047)
0.148***
(0.038)
Y
N
736
92
0.6809
0.0798
0.120
(0.081)
0.072***
(0.028)
0.001
(0.001)
0.025***
(0.007)
0.190**
(0.091)
0.000
(0.004)
0.141
(0.162)
0.035
(0.103)
Y
Y
736
92
0.9232
0.1134
0.127*
(0.073)
0.077***
(0.027)
lrgdpch
Ipr
0.234***
(0.028)
Openc
0.025***
(0.007)
Cg
0.242***
(0.086)
lyr_sch
polity2
0.000
(0.004)
Lpop
0.006
(0.148)
Lcountryarea
Year FE
Country FE
Observations
Number of id
R-squared
R-squared within
Y
N
775
94
0.4636
0.0231
Y
Y
767
94
0.9167
0.0569
Y
Y
767
94
0.9175
0.0670
Y
Y
767
94
0.9147
0.0612
Y
Y
767
94
0.920
0.0940
Y
Y
758
94
0.918
0.0784
Y
Y
745
94
0.920
0.0703
Y
Y
767
94
0.9175
0.0670
0.024***
(0.007)
0.233***
(0.086)
Y
Y
758
93
0.9200
0.1040
671
Table 2. Economic complexity (ECI) and intellectual property index (IPP) system GMM
Variables
L.ECI
lrgdpch
ipr
(1)
ECIa
(2)
ECIa
(3)
ECIb
(4)
ECIb
(5)
ECIc
(6)
ECIc
0.873***
(0.024)
0.036**
(0.017)
0.080***
(0.026)
0.871***
(0.025)
0.036**
(0.017)
0.081***
(0.026)
0.854***
(0.028)
0.090***
(0.026)
0.073***
(0.024)
0.000
(0.000)
0.009***
(0.003)
0.048*
(0.027)
0.000
(0.003)
0.857***
(0.032)
0.087***
(0.032)
0.077***
(0.029)
0.000
(0.000)
0.009***
(0.003)
0.047
(0.032)
0.000
(0.003)
0.851***
(0.039)
0.066**
(0.030)
0.055*
(0.032)
0.001*
(0.001)
0.020***
(0.005)
0.031
(0.055)
0.006
(0.005)
0.857***
(0.041)
0.060*
(0.031)
0.058*
(0.034)
0.001
(0.001)
0.020***
(0.006)
0.033
(0.055)
0.005
(0.005)
1-step
87
0.171
0.001
0.161
686
94
2-step
87
0.171
0.001
0.161
686
94
1-step
107
0.723
0.001
0.290
663
91
2-step
107
0.723
0.001
0.290
663
91
1step
112
0.768
0.001
0.302
663
91
2-step
112
0.768
0.001
0.302
663
91
openc
Cg
lyr_sch
polity2
# of instruments
Hansen test (p-value)
AR 1 test (p-value)
AR 2 test (p-value)
Observations
Number of id
Notes: System-GMM estimations for dynamic panel-data models. Sample period: 19602005.
a
IPP and L.ECI were treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rst-dierences were
used in the levels equations.
b
IPP, lrgdpch, and L.ECI were treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rstdierences were used in the levels equations.
c
All variables were treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rst-dierences were
used in the levels equations. Two-step results using robust standard errors corrected for nite samples (using Winmeijers, 2005, correction). t-Statistics are
in parenthesis. Signicance level at which the null hypothesis is rejected: ***, 1%; **, 5%, and *, 10%.
(a)
(b)
Figure 5. ECI and IPR. Note: Authors calculations. Panel (a) Separated by mean income; panel (b) Separated by mean ECI.
672
WORLD DEVELOPMENT
Table 3. Economic complexity (ECI) and intellectual property index (IPP) sample divided by income and level of complexity
Method
VARIABLES
(18)
LI
OLS
ECI
(19)
HI
OLS
ECI
L.ECI
lrgdpch
ipr
Openc
Cg
lyr_sch
polity2
# of Instruments
Hansen test (p-value)
AR 1 test (p-value)
AR 2 test (p-value)
Observations
R-squared
Number of id
0.120
(0.119)
0.015
(0.044)
0.002
(0.002)
0.032***
(0.010)
0.431***
(0.139)
0.003
(0.006)
0.259*
(0.132)
0.085**
(0.039)
0.000
(0.001)
0.015
(0.011)
0.136
(0.167)
0.005
(0.006)
341
0.818
50
395
0.939
60
(20)
(21)
(22)
(23)
LI
HI
LI
HI
SGMMa SGMMa SGMMb SGMMb
ECI
ECI
ECI
ECI
(24)
LECI
OLS
ECI
(25)
HECI
OLS
ECI
0.216**
(0.096)
0.087**
(0.037)
0.003**
(0.001)
0.013
(0.010)
0.148
(0.149)
0.001
(0.005)
338
0.927
(26)
(27)
(28)
(29)
LECI
HECI
LECI
HECI
SGMMa SGMMa SGMMb SGMMb
ECI
ECI
ECI
ECI
0.874*** 0.718***
(0.085)
(0.054)
0.013
0.007
(0.051)
(0.034)
0.097* 0.182***
(0.053)
(0.051)
0.000
0.001**
(0.001)
(0.000)
0.011** 0.006
(0.005)
(0.006)
0.103*
0.093
(0.054)
(0.071)
0.001
0.001
(0.003)
(0.003)
56
56
0.199
0.137
0.002
0.000
0.760
0.126
359
304
57
58
0.623***
(0.100)
0.083
(0.057)
0.112
(0.070)
0.002
(0.001)
0.011
(0.011)
0.178
(0.146)
0.013*
(0.007)
62
0.591
0.007
0.964
359
0.862***
(0.083)
0.088
(0.061)
0.095*
(0.051)
0.002***
(0.001)
0.009
(0.010)
0.007
(0.191)
0.012
(0.012)
62
0.807
0.005
0.275
304
57
58
Notes: LI: below mean income, HI: above mean income, LECI: below mean ECI and HECI: above mean ECI. System-GMM estimations for dynamic
panel-data models. Sample period: 19602005.
a
IPP, lrgdpch and L.ECI were treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rstdierences were used in the levels equations.
b
All variables were treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rst-dierences were
used in the levels equations. Principal Components (PCA) were used to keep number of instruments close to the number of panels. Two-step results using
robust standard errors corrected for nite samples (using Winmeijers, 2005, correction). t-Statistics are in parenthesis. Signicance level at which the null
hypothesis is rejected: ***, 1%; **, **, 5%, and *, 10%.
673
(30)
ECI
(31)
ECI
(32)
ECI
(33)
ECI
(34)
ECI
(35)
ECI
(36)
ECI
LHC
(37)
ECI
HHC
0.086**
(0.036)
0.006
(0.025)
0.001
(0.001)
0.108***
(0.034)
0.011**
(0.005)
0.098**
(0.046)
0.001
(0.004)
0.087**
(0.043)
0.432
(0.344)
0.082*
(0.047)
0.004
(0.007)
0.153**
(0.066)
0.110***
(0.033)
0.147**
(0.069)
0.106**
(0.046)
52
0.280
0.015
0.690
266
55
52
0.223
0.000
0.150
420
78
Fdi
0.020
(0.013)
0.006
(0.010)
Cg
0.016**
(0.007)
polity2
0.074**
(0.030)
Lpat
0.065
(0.059)
91
0.211
0.001
0.139
678
93
lyr_sch
# of Instruments
Hansen test (p-value)
AR 1 test (p-value)
AR 2 test (p-value)
Observations
Number of id
93
0.258
0.002
0.138
686
94
86
0.300
0.000
0.381
582
88
93
0.323
0.002
0.156
686
94
85
0.228
0.002
0.326
671
92
70
0.255
0.115
0.236
434
71
Notes: System-GMM estimations for dynamic panel-data models. Sample period: 19602005. IPR, lrgdpch, L.ECI and the interaction variable were
treated as endogenous. Their lagged values used as instruments in the rst dierence equations and their lagged rst-dierences were used in the levels
equations. Principal Components (PCA) were used to keep number of instruments close to the number of panels. Two-step results using robust standard
errors corrected for nite samples (using Winmeijers, 2005, correction). t-Statistics are in parenthesis. Signicance level at which the null hypothesis is
rejected: ***, 1%; 5%, and *, 10%. IPR Interacted is IPR interacted with the other main term. LHC: below mean lyr_sch, HHC: above mean lyr_sch.
Table 5. Economic complexity (ECI) and intellectual property index (IPP) OLS, pre and post TRIPS
(1)
Method
Variables
Lrgdpch
Ipr
Openc
Cg
lyr_sch
polity2
Year FE
Country FE
Observations
Number of id
R-squared
R-squared within
(3)
(4)
All
OLS
ECI
(2)
Pre 1995
LI
OLS
ECI
HI
OLS
ECI
All
OLS
ECI
0.035
(0.118)
0.084***
(0.046)
0.000
(0.001)
0.038***
(0.009)
0.527***
(0.105)
0.005
(0.005)
Y
Y
463
82
0.9272
0.1480
0.084
(0.149)
0.010
(0.068)
0.001
(0.002)
0.038***
(0.010)
0.427**
(0.204)
0.013*
(0.007)
Y
Y
240
51
0.8446
0.3128
0.114
(0.191)
0.104
(0.073)
0.001
(0.002)
0.013
(0.017)
0.351
(0.269)
0.003
(0.006)
Y
Y
223
47
0.9485
0.2354
0.097
(0.130)
0.037
(0.050)
0.001
(0.002)
0.015
(0.017)
0.448
(0.363)
0.013
(0.008)
Y
Y
273
92
0.9740
0.0376
(5)
Post 1995
LI
OLS
ECI
0.410
(0.373)
0.144**
(0.063)
0.003
(0.002)
0.017
(0.029)
0.153
(0.669)
0.021*
(0.011)
Y
Y
101
0.929
0.1323
(6)
HI
OLS
ECI
0.381
(0.245)
0.131**
(0.064)
0.001
(0.002)
0.038
(0.024)
0.015
(0.496)
0.024
(0.021)
Y
Y
172
60
0.9741
0.1059
Notes: All regressions include a trend. Robust standard errors in parentheses; *signicant at 10%; **signicant at 5%; ***signicant at 1%. LI: below mean
income, HI: above mean income.
674
WORLD DEVELOPMENT
675
NOTES
1. For a full review of these literatures, see Allred and Park (2007).
2. Other instruments within IPR institutions include copyright and
industrial design. Firms may rely on these or other alternative strategies
to confront the problem of appropriability. A discussion of why rms select
(or not) formal IPR rights can be found in Hall, Helmers, Rogers, and Sena
(2012), supported by empirical evidence (Anton & Yao, 1994; Bhattacharya
& Guriev, 2006) and in case studies at the industry level (Arora, 1997).
3. For an account of the political process underlying the negotiation of
the agreement see Jawara and Kwa (2003).
4. It is important to note that at the time the US patent standard was 17 years
and that many of the countries adhering to the new rules did not have patents
for products such as pharmaceuticals. For an account of the political process
underlying the negotiation of the agreement see Jawara and Kwa (2003).
5. Implementation of new IP norms has occurred in a staggered time
sequence, commencing in 1996, on the heels of TRIPS approval. Countries
such as Brazil amended their local rules to be consistent with the international
norms within a year. India, by contrast, negotiated a transition time
concluding in 2005. This June (2013), a further extension of treaty exibilities
was granted for a set of least developed countries until 2021. In this sense,
while the shift has been unitary in direction, timing has been heterogeneous.
6. Once these economies have developed, there is evidence that they take into
account the strength of local IPR systems when making overseas investment
decisions (Branstetter et al., 2006; Jeong-Yeon & Manseld, 1996).
7. Patents were the most relevant for pharmaceuticals and chemicals,
where approximately 30% of all products required patents. In three of the
industries surveyed, patents were only marginally relevant (petroleum,
machinery, and fabricated metals) and in another set of seven industries
patents were deemed irrelevant to the development and commercialization
of innovative products. Notably, this group included high-tech products
such as motor equipment and electrical equipment.
8. For a broader discussion of the pros and cons of patent data, see
Griliches (1998).
9. An interesting review of emerging patenting patterns is presented in
Kortum and Lerner (1999).
10. The data on Patents are from Lederman and Saenz (2005): Innovation
and Development around the World, 19602000, World Bank Working
Paper No. 3774. The data are from the U.S. Patent and Trademark Oce
(USPTO) and are the granted patents assigned to countries based on the
country of residence of the rst inventor (Lederman & Saenz, 2005).
11. Although this is a mechanism that has been dicult for a number of
countries to institute, there are a handful of exceptions (Brazil, 2007;
Thailand, 2006, 2007; Ecuador, 2010). For further discussion of the
challenges in the construction of compulsory licenses see Beall and Kuhn
(2012), Sweet and Das (2013)).
12. The United States, consistently the highest scoring country in the
Index for example does not allow for states to be sued by patent holders.
13. As robustness check we re-run all the regressions in Table 1
controlling for autocorrelation of order 1. The results from these
regressions are broadly similar to the ones presented in the paper. The
details are available upon request.
14. The kernel estimations conditional on dierent initial GDP per capita
show that lower income countries have a distribution of ECI that is more dense
and with a lower mean value than countries with higher income per capita.
15. Since dividing the data using average income and average ECI might
be considered somehow arbitrary we also rely on cluster analysis to look
for alternative groupings of the data. Formally, we use k-means cluster
analysis on our three main variables: ECI, IPR and GDP per capita. The
Calinski/Harabasz pseudo-F test indicates the presence of 2 or 3 clusters in
the data. The results from the clustered regressions are broadly similar to
the ones presented in the paper. The details are available upon request.
16. FDI is inward foreign direct investment from the World Bank World
Development Indicators. Lpat is the logarithm of the total number of
patents granted by the USPTO by year for each country. Data from
Lederman and Saenz (2005).
17. Furthermore, we divide the sample in 3 subsequent 15-year periods
and run cross-section instrumental variable estimations of our main
equation. IPR, lrgdpch, and ECI were treated as endogenous and their
initial values used as instruments. The results are supportive of the ones
presented in the paper and are available upon request.
18. We are very thankful to Matheiu Cristelli and co-authors for sharing
their dataset with us.
19. The shorter time series length of the new dataset makes dicult to use the
2-step system GMM technique we used in the longer ECI dataset. Alternatively, we use a less data intensive technique (2-step instrumental variables) to
control for the potential endogeneity of IPR. Results available upon request.
20. We thank an anonymous referee for suggesting this point.
676
WORLD DEVELOPMENT
23. It should be noted that the ruling does allow for the patenting of
cDNA, or so-called synthetic DNA molecules, that require an inventive step.
REFERENCES
Abrams, D. S. (2009). Did TRIPS spur innovation? An empirical analysis of
patent duration and incentives to innovate. Institute for Law and
Economics, University of Pennsylvania Law School, Research Paper
No. 09-24.
Acemoglu, D., & Akcigit, U. (2012). Intellectual property rights policy,
competition and innovation. Journal of the European Economic
Association, 10(1), 142.
Acemoglu, D., Gancia, G., & Zilibotti, F. (2012). Competing engines of
growth: Innovation and standardization. Journal of Economic Theory,
147(2), 570601. http://dx.doi.org/10.1016/j.jet.2010.09.001, e573.
Acha, V., Marsili, O., & Nelson, R. (2004). What do we know about
innovation? Research Policy, 33(9), 12531258. http://dx.doi.org/
10.1016/j.respol.2004.09.001.
Aidt, T., & Eterovic, D. (2011). Political Competition, Participation and
Public Finance in 20th Century Latin America. European Journal of
Political Economy, 27(1), 181200.
Allred, B. B., & Park, W. G. (2007). Patent rights and innovative activity:
Evidence from national and rm-level data. Journal of International
Business Studies, 38(6), 878900.
Anand, R., Mishra, S., & Spatafora, N. (2012). Structural transformation
and the sophistication of production. IMF Working Paper, 12.
Anton, J. J., & Yao, D. A. (1994). Expropriation and inventions:
Appropriable rents in the absence of property rights. The American
Economic Review, 84(1), 190209.
Archibugi, D., & Filippetti, A. (2013). The globalization of intellectual
property rights: Much ado about nothing? SSRN eLibrary, March
(<http://ssrn.com/abstract=2240365>).
Arellano, M., & Bond, S. (1991). Some tests of specication for panel
data: Monte Carlo evidence and an application to employment
equations. Review of Economic Studies, 58, 277297.
Arellano, M., & Bover, O. (1995). Another look at the instrumental
variable estimation of error-components models. Journal of Econometrics, 68, 2951.
Arora, A. (1997). Patents, licensing, and market structure in the chemical
`5), 391403. http://dx.doi.org/
industry. Research Policy, 26(4,A
10.1016/s0048-7333(97)00014-0.
Arora, A., Fosfuri, A., & Gambardella, A. (2001). Markets for technology:
The economics of innovation and corporate strategy. Cambridge, MA:
The MIT Press.
Athreye, S., & Cantwell, J. (2007). Creating competition?: Globalisation
and the emergence of new technology producers. Research Policy,
36(2), 209226, http://dx.doi.org/10.1016/j.respol.2006.11.002.
Barro, R. (1991). Economic growth in a cross section of countries. The
Quarterly Journal of Economics, 106(2), 407443.
Beall, R., & Kuhn, R. (2012). Trends in compulsory licensing of
pharmaceuticals since the Doha declaration: A database analysis.
PLoS Medicine, 9(1), e1001154. http://dx.doi.org/10.1371/journal.pmed.1001154.
Barro, R., & Lee, J. (2010). A New Data Set of Educational Attainment in
the World, 1950-2010. NBER Working Paper 15902.
Bell, M., & Pavitt, K. (1993). Technological accumulation and industrial
growth: Contrasts between developed and developing countries.
Industrial and Corporate Change, 2(2), 157210.
Bhattacharya, S., & Guriev, S. (2006). Patents vs. trade secrets: Knowledge licensing and spillover. Journal of the European Economic
677
ScienceDirect