The Role of Intergovernmental Organizations in Cross-border Knowledge Transfer and Innovation *

Barriers to Cross-border Knowledge Transfers

Innovation occurs through a process of knowledge recombination whereby existing knowledge is reapplied to address a new problem (Schumpeter, 1934). The potential for innovation is enhanced by knowledge transfer that occurs through observing the varied experiences of other entities (Argote and Ingram, 2000; Argote and Miron-Spektor, 2011) and supplementing one’s own path-dependent experience (Ingram and Baum, 1997). But knowledge transfer is a challenging process. Knowledge is broader, richer, and deeper than information and data; it is a fluid mix of experience, important values, contextual information, and expert insight (Bhagat et al., 2002). It reflects the social, cultural, and institutional environment in which it is developed and therefore reflects specific national characteristics, including culture (Hofstede, 1980), technological development (Kogut and Singh, 1988), resource endowments, organization of industry, demand and supply conditions (Porter, 1990), and scientific, technological, and regulatory environments (Phene, Fladmoe-Lindquist, and Marsh, 2006). The national innovation system framework emerging from this perspective indicates that innovation is a country-specific phenomenon with a variety of embedded institutions and organizations responsible for differences across countries (Freeman, 1987; Lundvall, 1992; Nelson, 1993; Bartholomew, 1997; Furman, Porter, and Stern, 2002). In a comparative study of innovation systems, Nelson (1993) documented the consequences for national innovation of differences across countries in the character and effectiveness of education systems, the functioning of public laboratories, government policies toward R&D, and financial institutions. Given the strong effect of national borders, the cross-border transfer of knowledge is particularly challenging.

Building on the work of institutional scholars (Scott, 1995; Kostova, 1997; Busenitz, Gomez, and Spencer, 2000), we propose that differences among countries’ institutional profiles along three dimensions—regulatory, cognitive, and normative—pose specific challenges to international knowledge transfer. The regulatory component of a country’s institutional profile reflects its laws, rules, and standards as generated or adopted by regulatory agencies and other rule-making bodies. Rules that constrain the transfer of certain technologies clearly pose a barrier for international knowledge transfer. Advanced nuclear technology transfer is constrained by law precisely to prevent wide access. Even when formal rules do not restrict knowledge transfer, de facto practices may be a barrier. Many countries have laws governing property rights, but the actual level of protection and enforcement is often poor, making investment and knowledge transfer less likely (Mansfield, 1995; Jandhyala, 2013). Further, differences in regulatory standards may make foreign knowledge less relevant and hence less likely to be transferred. Gruber and Verboven (2001) noted that the lack of a single technological standard lowers the likelihood of technology diffusion in the telecommunications industry.

The cognitive component reflects categories, frames, schemas, and representations that influence what is noticed, how external stimuli are interpreted and categorized, and how meaning is shared with others (Scott, 1995). Although residing in individuals, the cognitive dimension has been viewed as an element of the institutional profile of a country (Kostova, 1997; Busenitz, Gomez, and Spencer, 2000) and can influence thinking, practice, and values. In the context of knowledge, these differences can pose significant barriers. Lam (1997) noted how contrasts in knowledge organization and technical work in the electronic industry influenced knowledge transfer between Japan and Great Britain. Differences in skill formation, education systems, labor market structures, and technological heritage led the two countries to develop distinctive cognitive knowledge. Engineers trained in the British system tend to emphasize theoretical knowledge and specialize in conceptual design and development activities. In contrast, Japanese engineers are more focused on practical knowledge with a broader industry emphasis. Such incompatibility between countries’ knowledge structures leads to poor communication, misinterpretation of specifications, and clashes between approaches to product development. Correspondingly, the ability to learn from constituents in other countries and engage in knowledge transfer is impeded.

The normative component focuses on values and norms that exist within a group or category of people that influence the perception of appropriateness. Normative differences among countries can limit receptivity to new knowledge and pose significant challenges to knowledge transfer. In the electronics example described above, the Japanese may be unwilling to capitalize on the theoretical and specialized knowledge in Britain not only because of cognitive issues but also because the Japanese perceive this knowledge as less appropriate. The need to seek legitimacy within the national innovation system may further increase the reluctance to engage in cross-border transfer. Norms may also deemphasize cross-border knowledge exchanges due to the perceived limited benefits of reciprocity and reputation improvement (Levin and Barnard, 2013).

Organizational scholars have suggested that nonmarket organizations like business associations, cooperatives, or government-supported institutions can help to overcome some of the challenges of domestic knowledge transfer among participating entities and generate reliable channels for collective learning (Lynn, Mohan Reddy, and Aram, 1996; Cohen, Nelson, and Walsh, 2002). But it is not entirely evident that international organizations can overcome deeply entrenched national differences. Promoting international ties or partnerships may be insufficient to overcome the barriers to cross-border knowledge transfer, as is evident from the mixed results of studies of knowledge spillovers from foreign direct investment (Saggi, 2002; Görg and Greenaway, 2004; Smeets, 2008). Even when an organizational form supports international ties or partnerships, cross-border transfer can be challenging. Using the context of a multinational firm as an international organization, prior researchers have identified substantial difficulties in knowledge flows between the various R&D locations as each overseas subsidiary develops distinct roles reflecting the national institutional context (Nobel and Birkinshaw, 1998; Gupta and Govindarajan, 2000; Mors, 2010; Capozzi, Biljon, and Williams, 2013). Researchers have also argued that other international organizational forms are ineffective as they may be epiphenomenal (Mearsheimer, 1994) or lack legitimacy (Zweifel, 2006). Consequently long-term convergence among countries’ structures, skills, and nonmarket organizations, even in the age of globalization, is far from complete, and country-level differences continue to persist (Berry, Guillén, and Hendi, 2014). Thus our baseline expectation is that a country’s connectedness to international nonmarket organizations will have no effect on its innovation.

IGO Connectedness and Innovation

Nonmarket organizations that are themselves international—that develop international ties, forge international partnerships, and draw international participants—may help to overcome the challenges to cross-border knowledge transfer. Some international nonmarket organizations may be well situated to promote cross-border knowledge transfer and enhance innovation, in particular organizations that can simultaneously break down cross-country differences in institutional factors and generate commonalities among participants’ expectations and norms. One of these is the intergovernmental organization (IGO), which, like other international organizations, resolves cross-border issues that cannot otherwise be addressed domestically (Abbott and Snidal, 1998). IGO members are countries, and national governments sign treaties to initiate their participation (Pevehouse, Nordstrom, and Warnke, 2004). Since World War II, the number of active IGOs has increased steadily. Prominent examples include global organizations such as the United Nations, the World Bank, and the World Trade Organization, but more than 300 IGOs operate today with varied membership and mandates (Ingram and Torfason, 2010). Our focus in this paper is on learning-oriented IGOs, those that have a specific mandate for knowledge sharing and transfer by facilitating economic transactions or enhancing information access among members. These IGOs may facilitate cross-border knowledge transfer and innovation by establishing common rules across countries and building international channels for collective learning and knowledge diffusion, so a country’s greater connectedness to these IGOs may promote greater innovation.

Establishing Common Rules

Government participation in IGOs plays a crucial role in lowering the regulatory barriers to cross-border knowledge transfer by establishing common rules, policies, laws, and standards across countries. This would be more difficult to achieve with purely private participation, such as from firms and industry associations. Membership in UNESCO, for example, has influenced domestic science establishments. In an in-depth study, Finnemore (1993) described how member countries established similar national-level science bureaucracies in response to the operating guidelines, budgets, and organizational charts established by UNESCO officials. These bureaucracies shared other common features: entities making science policy decisions were not involved in research, and science policy bodies had access to the highest levels of government, were independent, and coordinated actions across the country.

In addition, IGO membership facilitates the development of similar standards that can enable knowledge sharing and transfer. Countries participating in the North American Plant Protection Organization (NAPPO)—an IGO facilitating the shared development of science-based standards to protect plants against pests—implement approved NAPPO standards through formal domestic regulatory channels. Some IGOs also normalize intellectual property protection across countries (e.g., the World Intellectual Property Organization). The presence of common regulatory institutions that allow the transferring and recipient entities to use similar mechanisms for knowledge protection enhances cross-border knowledge transfer (Bhagat et al., 2002). Because they link various constituencies in the pursuit of technological standards, IGOs can be considered as one type of cooperative technical organization (CTO) identified by Rosenkopf and Tushman (1998). Like CTOs, IGOs help to develop industry-wide procedures and standards. But IGOs also differ in significant ways. Many IGOs have a broader agenda than technical coordination, membership is limited to countries (rather than firms, academia, etc.), and their members (governments) have the sovereign authority to implement coordinated rules or standards through legal, institutional, or bureaucratic means. As Ronit and Schneider (1999) noted, private associations face particular challenges in enabling rule enforcement and the provision of public goods, while IGOs may be better equipped to do so. IGOs also have resources for fact finding and dispute resolution, which aids the enforcement of adopted rules and standards (Rangan and Sengul, 2009).

Building Channels for Learning and Knowledge Diffusion

IGOs also have a bureaucratic form and conduct regular meetings and plenary sessions (Pevehouse, Nordstrom, and Warnke, 2004). This allows them to function as “places where information is exchanged and where people come to appreciate others’ points of view” (Dorussen and Ward, 2008: 192) and enables them to create social knowledge (Barnett and Finnemore, 1999). IGOs develop a bureaucratic form through a permanent secretariat and a staff of international civil servants (Bauer, 2006), who embody not only institutional, scientific, and technical knowledge but also distinctive cultures and norms. These actors also control the flow of information and values among participants through various administrative, scientific, and diplomatic activities. IGOs define technical objectives and acceptable behavior regarding their goals, and their civil servants consider it a part of their mission to spread, inculcate, and enforce these norms globally (Barnett and Finnemore, 1999).

IGOs frequently organize conferences, forums, and meetings that draw participants from member countries (Crane, 1971), enabling interactions among members for social and task-oriented activities. The Department of Social Sciences at UNESCO brings together social science scholars from around the world to increase international understanding (Angell, 1950). Repeated interactions through such events develop embedded ties that lead to trust, fine-grained information exchange, and joint problem-solving efforts (Rosenkopf, Metiu, and George, 2001), which are useful for innovation as they facilitate adapting knowledge to novel problems (Miller, Pentland, and Choi, 2012) and new tasks (Lewis, Lange, and Gillis, 2005). The conferences, forums, and meetings also increase the visibility and availability of information across national boundaries. Technical information and research developments are shared among participating firms, national laboratories, and government representatives from several global locations. The intergovernmental Group on Earth Observations (GEO) is developing a system that integrates environmental data from the many thousands of individual land, sea, air, and space-based observations around the globe, giving participants a wealth of new information that may enhance innovation. When cross-border knowledge is made visible and accessible, innovation through recombination is more likely.

Learning also involves acquiring the ability to act in the IGO’s socially recognized ways (Brown and Duguid, 2001). Through repeated interactions among IGO civil servants and participating members, common knowledge and language are developed. Because IGO-sponsored training programs inculcate values and norms among participants, participating in IGO events shapes the outlook of participants and creates a distinctive common thread among them. A common cognitive and normative frame of reference improves the salience of cross-border knowledge and facilitates greater innovation through cross-border knowledge transfers.

Taken together, these factors suggest that membership in learning-oriented IGOs enables the transfer and use of diverse and unique knowledge across national borders. As innovation is enhanced by this type of knowledge transfer, we expect participation to increase innovation, and the greater the extent of connectedness to learning-oriented IGOs, the greater the opportunities will be to benefit from IGOs. Thus we propose:

Boundary Conditions of IGO Effects: Domestic Context

Though we expect IGO connectedness to increase national innovation, the countries participating in IGOs have varying economic and institutional structures. These structures have a significant impact on the domestic dissemination and diffusion of cross-border knowledge, as Cole (1989) demonstrated in the differing adoption of practices such as quality circles and autonomous work groups in the U.S., Japan, and Sweden. Thus we expect the effects on national innovation of cross-border knowledge obtained through IGO connectedness to be contingent on domestic structures. Prior literature has identified two key characteristics of domestic structures in enabling wider diffusion of cross-border knowledge and the creation of local knowledge: the availability of necessary knowledge to support the understanding and absorption of cross-border knowledge, and interactions and cooperation among local constituents to ensure productive dialogue and use of cross-border knowledge for new local knowledge creation (Xu, 2000; Perez-Aleman, 2011).

Illustrative Caselet

We focused on the CSLF to provide greater depth and nuance to our theoretical arguments. We sketched out how participation in this IGO facilitates innovation among a variety of actors, and we examined the role played by government participation and the organizational form in making the CSLF effective in promoting cross-border knowledge transfer and hence innovation.

The challenge of limiting temperature rise through the reduction of greenhouse gas emissions is a global one. Several technological solutions are in development, but there is disagreement on the suite of technologies that should be pursued or receive government support and attention. For instance, some advocate scaling up renewable energy production, while others propose capturing and storing industrial carbon emissions. Even within carbon capture, countries back different methods: injecting captured emissions into the ground is increasingly accepted in the United States, but India prefers gasification—turning carbon-rich materials into synthetic gas—rather than storage (New York Times, 2009). Different levels of development, technological sophistication, regulations, and expectations around the world shape national technological development and make coordinated action difficult. In this sense, national boundaries pose barriers to the cross-border transfer of knowledge.

In 2003, the CSLF was launched to enable coordinated action on carbon capture and storage (CCS). The CSLF is an intergovernmental organization designed to facilitate the development of cost-effective technologies related to carbon capture, transportation, and long-term storage; promote the implementation of these technologies internationally via collaborative efforts; and determine an appropriate political and regulatory framework needed to promote CCS on a global scale. It is a ministerial-level organization with 23 members. The commitment of participants to cross-border knowledge sharing is reflected in the comments of the Norwegian minister of petroleum and energy (2013), who stated, “We all agree that we need to share knowledge, ideas and experiences in order to find commercially viable solutions. . . . We must learn as much as possible from each other. This requires cooperation and joint efforts from the industry, research community and Governments.”

CSLF initiatives facilitate cross-border knowledge transfer in the area of CCS in several ways. Participating governments have noted that the international forum allows them to share information on important policy initiatives, as well as legal and regulatory developments in member countries (CSLF Ministerial Communique, 2013). For instance, member governments discuss legal, regulatory, and contractual issues regarding liability for the release of stored carbon dioxide (CO₂), ownership of injected CO₂, intellectual property protection for the transfer of technology, and the establishment of a sound regulatory framework (IEA/CSLF Workshop, 2006). Greater coordination of these aspects across countries facilitates the cross-border transfer and application of CCS technologies. The recent policy focus on CCS has been accompanied by a compounded annual growth rate of 46 percent in global CCS patent applications between 2006 and 2011 (IEA, 2013).

The CSLF also provides technological direction and develops norms about appropriate approaches to CCS. For example, the technical group of the CSLF identifies and recognizes projects from member countries that show significant advancement toward commercialization and large-scale deployment. Innovative, leading-edge CCS projects undertaken by Saudi Aramco in Saudi Arabia, E.ON and GDF Suez in the Netherlands, and Chevron in Australia have all been recognized by the CSLF. ² CSLF recognition gives these projects greater global visibility, and information on their technical performance and financial viability is widely shared among members. Best practices and standards are also disseminated. By making this information available, the CSLF shapes beliefs among the scientific and policy communities about possible appropriate technologies for future investment. Further, by creating a common understanding of successful approaches across countries, cross-border knowledge flows and subsequent innovation are enhanced.

The CSLF has also been instrumental in generating technical reports on the state of CCS technologies that provide useful data for firms and other actors to leverage in their own R&D. For instance, U.S. patents filed by Calix Ltd., an Australian minerals-processing and carbon-capture firm, and Corning, Inc., a U.S. glass and ceramics company, in the fields of carbon-cycle processing and CO₂ capture, respectively, cite CSLF technology reports. Further opportunities for knowledge transfer are created by organizing a variety of meetings, conferences, and workshops that involve all sectors of the research community and are hosted in cities worldwide (www.cslfforum.org). At a 2013 technical meeting in Rome, representatives from firms such as ENEL and Shell interacted not only with each other but also with participants from national laboratories and university researchers from around the world to share the latest technology and research developments.

In addition, the CSLF facilitates cross-border partnerships in research, and the number of international collaborations has grown significantly since its founding (CSLF Ministerial Communique, 2011). An international team of researchers from the U.S. and China collaborated to demonstrate carbon storage capacity in China (PNNL, 2009). This cross-border research was funded by the CSLF, and the organization also recognized the study and disseminated the findings broadly. Following the project, the U.S. organization, a government-backed R&D laboratory, continued to file for several patents relating to carbon storage.

Empirical Analysis

For our empirical analysis, we first created a set of all countries identified in the Correlates of War Database and collected data on their innovation by considering their successful patent applications between 1996 and 2006 under the U.S. Patent and Trademark Office (USPTO) using the Patent Network Dataverse (Lai et al., 2011). Patent documents provide information on the inventor, location, and timing of the innovation. Following extant literature (e.g., Thompson and Fox-Kean, 2005), we used the first inventor’s location (country) and the application year to assign patents to different country–year combinations. We collected additional data for our independent variables and controls from the following sources: Correlates of War Database, the Yearbook of International Organizations, the United Nations Commodity Trade Statistics database, the Fraser Institute, and the World Development Indicators. After dropping observations with missing data, our final sample consisted of 83 countries and 688 observations during the time period 1996–2006. Our level of analysis is the country-year, in which the cross-sectional unit is the country and the temporal unit one year. Table 1 lists the countries included in our sample, their patents over the sample period, and the number of learning-oriented IGO memberships by country at the start and end of our sample time period.

OPEN IN VIEWER

Table 1.

Sample Countries with Learning-oriented IGO Membership and Number of Patents

Country	Patents 1996–2006	Learning IGO membership		Country	Patents 1996–2006	Learning IGO membership
		1996	2005			1996	2005
1. Algeria	4	25	49	43. Madagascar	4	31	32
2. Argentina	640	43	47	44. Malaysia	1171	48	50
3. Australia	13922	48	47	45. Mali	5	42	44
4. Austria	6774	55	56	46. Mexico	1170	51	57
5. Bangladesh	6	36	39	47. Morocco	23	49	48
6. Belgium	8771	67	67	48. Netherlands	18663	71	70
7. Brazil	1585	46	54	49. New Zealand	1991	37	40
8. Bulgaria	172	36	42	50. Niger	3	39	40
9. Canada	47323	52	51	51. Nigeria	18	49	45
10. Chile	205	45	48	52. Norway	3386	60	58
11. China	8880	48	51	53. Oman	7	23	29
12. Colombia	107	44	52	54. Pakistan	35	44	43
13. Costa Rica	145	39	46	55. Panama	18	38	43
14. Croatia	53	28	36	56. Paraguay	3	34	40
15. Cyprus	26	36	38	57. Peru	37	41	48
16. Czech Republic	470	41	43	58. Philippines	282	39	43
17. Denmark	6559	63	63	59. Poland	396	47	51
18. Dominican Republic	20	34	41	60. Portugal	189	57	62
19. Ecuador	32	41	47	61. Romania	106	40	42
20. Egypt	78	51	48	62. Russia	2528	44	50
21. El Salvador	10	33	35	63. Singapore	4669	33	37
22. Estonia	54	23	29	64. Slovakia	71	35	42
23. Fiji	7	23	27	65. Slovenia	170	29	39
24. Finland	10673	65	65	66. South Africa	1515	31	40
25. France	48185	76	76	67. South Korea	65714	43	50
26. Germany	132144	69	68	68. Spain	4256	65	68
27. Greece	321	55	58	69. Sri Lanka	42	42	44
28. Guatemala	18	37	39	70. Sweden	18060	67	63
29. Honduras	12	34	36	71. Switzerland	16949	58	57
30. Hungary	681	44	46	72. Syria	13	36	36
31. India	5219	53	57	73. Tanzania	3	35	40
32. Indonesia	166	51	53	74. Thailand	434	44	44
33. Iran	57	34	37	75. Trinidad and Tobago	15	33	36
34. Ireland	2105	47	50	76. Tunisia	11	48	48
35. Israel	13889	34	36	77. Turkey	234	46	47
36. Italy	22530	67	69	78. Ukraine	261	25	33
37. Japan	425684	56	55	79. United Kingdom	48991	65	67
38. Jordan	16	30	34	80. USA	1129878	55	55
39. Kenya	62	35	41	81. Uruguay	30	37	41
40. Kuwait	100	34	34	82. Venezuela	288	51	55
41. Latvia	26	22	30	83. Zimbabwe	9	28	31
42. Lithuania	70	21	30

Independent variables

We built a time lag for the effects of our independent variables to avoid contemporaneous correlations and to account for the delayed innovation resulting from the use of linkages to access, assimilate, and utilize knowledge. Thus we measured our independent (and control) variables in year t−1.

IGO connectedness

Our main independent variable is the extent of a country’s connectedness with learning-oriented IGOs. We first obtained data on all types of IGOs that a country is a member of by drawing on Pevehouse, Nordstrom, and Warnke (2004), the most extensive source on IGO memberships. By construction, these IGOs (1) include three or more members of the Correlates of War–defined state system, (2) hold regular plenary sessions (at least once in ten years), and (3) have a permanent secretariat and corresponding headquarters. Because our focus was on the subset of learning-oriented IGOs, we collected additional data on each IGO’s principal objectives/aims from the Yearbook of International Organizations. ³ We then reviewed these objectives/aims and classified IGOs as learning-oriented if they had a specific mandate for knowledge sharing and transfer, facilitating economic transactions, or enhancing information access among members. Two independent coders read the primary aims of IGOs to classify each IGO under one of the categories, with an interrater reliability of ~0.75. The coders then discussed differences to come to a consensus. Our coding scheme is shown in table 2.

OPEN IN VIEWER

Table 2.

Coding Scheme for IGO Function*

IGO type	Description of organizations	Examples
Learning-oriented	Education, scientific research, and technology organizations; provide standards and harmonization of transactions; protect property rights; technical exchange or cooperation; facilitate information exchange	Council for Technical Cooperation in South and Southeast Asia, Latin American Center for Physics, World Intellectual Property Organization, World Meteorological Organization
	Monitor, enforce, and help process international economic transactions; perform trade-related functions; enhance information access; address issues of structure and operation of specific industries	International Cotton Advisory Committee, World Trade Organization, African Petroleum Producers Association
Others	Umbrella organizations that focus on administration of governments, perform multiple functions, or administer international agreements	UN, Nordic Council, South Asian Association for Regional Cooperation
	Regional political or military alliances; organizations for military/security/defense purposes	Council of Baltic Sea States, Euro-Atlantic Partnership Council, Wassenaar Arrangement
	Address health, disease, disaster, or social welfare; cultural or humanitarian organizations; environmental conservation	International Labor Organization, International Coral Reef Initiative, African Cultural Institute, International Organization for Migration

*

Source: Adapted from Ingram, Robinson, and Busch (2005); authors’ coding.

As an example, a key objective of the Council for Technical Cooperation in South and Southeast Asia is to promote technical cooperation and assist in the sharing and transfer of technology among member countries (www.colombo-plan.org). As a result, we classified this IGO as learning-oriented. Similarly, the Yearbook identified the aims of the International Institute of Refrigeration as to “Further research and promote studies on refrigeration science and technology, in refrigeration and cryogenic systems and heat pumps; [and] develop applications of refrigeration . . .” We classified it as a learning-oriented IGO because of its explicit mandate on research and development. The International Cotton Advisory Committee, which aims to foster a healthy world cotton economy, is also classified as a learning-oriented IGO as it serves as a clearinghouse for technical information about cotton and textiles while providing an objective forum for furthering international collaboration and discussion on cotton. About 58 percent of the active IGOs during the sample time period were classified as learning-oriented IGOs. Following research on IGO membership (Ingram, Robinson, and Busch, 2005; Alcacer and Ingram, 2013), we measured IGO connectedness as the natural log of the total number of active learning-oriented IGOs worldwide that a country is a member of in a given year. Our measure of IGO connectedness varied from 3.13 (= ln(23) for Latvia in 1997) to 4.39 (= ln(81) for France in 1998). On average, the number of learning-oriented IGO memberships of a country increased by 9.8 percent between 1996 and 2005.

Complementary opportunities

We follow Hausmann and Klinger (2006) in suggesting that a country’s revealed comparative advantage (RCA) in exports allows us to determine its current advantages as well as the complementary spaces. Using exports-based RCA as an indicator allows us to consider the complementarity between sectors by extending beyond broad factor endowments, a priori notions of technological sophistication, or assumptions of relatedness in knowledge domains (Makri, Hitt, and Lane, 2010). We constructed the complementary opportunities variable as a combination of a country’s (1) export advantages and (2) product spaces complementary to those in which a country has an export advantage. We built a time-varying country-level measure that incorporates these two facets based on the methodology and intuition developed by Hausmann and Klinger (2006).

We first used product-level exports data to determine the sectors in which a country has an export advantage. Based on the Standard International Trade Classification (SITC Rev.2), we obtained bilateral exports data for each of the 786 four-digit categories for all countries in the world for the time period 1996–2006 from the United Nations Commodity Trade Database. ⁴ For each product-country-year, we calculated whether the country had a revealed comparative advantage (Balassa, 1965) by examining the share of exports of the country for product i in relation to the share of product i in global trade.

R C A ( c , i , t − 1 ) = x ( c , i , t − 1 ) ∑ i x ( c , i , t − 1 ) ∑ c x ( c , i , t − 1 ) ∑ i , c x ( c , i , t − 1 )

where x(c,i,t−1) = exports of country c in the ith product in year t−1. If the RCA in the above calculation for a particular product is ≥ 1, then the share of the product in country c’s exports is equal to or larger than the average share of the product across world exports, revealing a comparative advantage for the country in that product category. In our data, the number of product categories in which a country had a revealed comparative advantage ranges from 12 (Gabon in 1997 and 2002; Algeria in 2006) to 369 (Germany in 2006).

Next, we used the product-level data of exports to determine complementarity between product pairs. ⁵ We followed Hausmann and Klinger (2006) to determine the complementarity between product-sector pairs by examining the likelihood that countries have revealed comparative advantages in both products of the pair. ⁶ For each pair generated from the 786 product categories, we calculated the complementarity between products i and j in time t−1 using the conditional probability as follows:

∅ i , j , t − 1 = min { P ( r i , t − 1 | r j . t − 1 ) , P ( r j , t − 1 | r i , t − 1 ) }

where for any country c,

r c , i , t − 1 = { 1 if RC A i , c , t − 1 ≥ 1 0 otherwise }

and the conditional probability is calculated using all countries in year t−1. This conditional probability is used to determine complementarity between product pairs. For example, the category “Tube and pipe fittings, of iron and steel” has low complementarity with “Cocoa butter and paste” (proximity = .1) but higher complementarity with “Cocks, valves, for pipes, boiler shells, etc.” (proximity = .5). We built on the specification of Hausmann and Klinger (2006) to measure domestic complementary opportunities as:

ComplementaryOpportunitie s i , t − 1 = ∑ i ∑ j [ ∅ i , j ∑ i ∅ i , j ( 1 − r c , j , t − 1 ) r c , i , t − 1 ]

This measure thus captures the complementary weighted value of all new product spaces a country could potentially expand into [ ∅ i , j ∑ i ∅ i , j ( 1 − r c , j , t − 1 ) ] and bases it as conditional on its current export advantages ( r c , i , t − 1 ). We expect the value of this measure to be large if there are several closely related, complementary classes, given the country’s current export advantages, and to be small if a country has few current export advantages or limited sectors complementary to its current advantages. The complementary opportunities measure ranges from a minimum of 12.5 for Kuwait (2001) to a maximum of 163.5 for the U.S. in 2000. In Online Appendix A (http://asq.sagepub.com/supplemental), we provide a detailed example of this calculation. Although this is a measure of export-based advantages, we believe it captures two elements important for innovation in our context. First, it captures the existing related knowledge and absorptive capacity of a country without relying on the dependent variable (patents) to make a priori assumptions of technological sophistication. Second, the measure allows us to estimate relatedness of knowledge domains and hence the potential for the transferred knowledge to be leveraged and applied in multiple sectors.

Domestic coordinating mechanisms

We measured the extent to which the relations between firms and other actors rely on competitive markets by using the Economic Freedom of the World data developed by the Fraser Institute (www.freetheworld.com). In these data, economic freedom is considered to be present in a country when individuals have the freedom to enter and compete in markets, market forces coordinate voluntary transactions, and strong market-supporting institutions like property rights protection and legal enforcement exist. Of this aggregate measure, we used the Regulations subindex, which explicitly measures labor-market regulations, credit-market regulations, and business regulations on their effectiveness in promoting economic freedom. These components mirror the financial and labor-market regulations used by Hall and Soskice (2001) in their classification of countries into liberal market economies (LMEs) and coordinated market economies (CMEs) (see, e.g., their Figure 1.1) and parallel the social institutions identified by Lazonick and O’Sullivan (1996). A higher value on this measure indicates greater reliance on market mechanisms for coordination.

Comparing this measure with Hall and Soskice’s (2001) characterization demonstrates that, as expected, countries classified as CMEs (e.g., Western Europe) rank lower in the regulations index. Further, a simple t-test for the year 2000 shows that for the subsample of Organization for Economic Cooperation and Development (OECD) countries classified as LME and CME by Hall and Soskice (2001), the score on the Regulations index is significantly lower for the CME group than for the LME group (7.3 vs. 7.96)—this pattern is also replicated for the full Economic Freedom index (7.86 vs. 8.38)—indicating that our measure is an appropriate, alternate characterization of coordination mechanisms. ⁷ The data used to construct our measure have the added advantage of being available over a long time period and for more than 100 developed and developing countries. ⁸

Estimation Methods

Our dependent variable is a non-negative integer of the number of ultimately granted patents to inventors of a country that were applied for in a given year. This variable also exhibits overdispersion, with the variance exceeding the mean, as observations range from zero to thousands. The appropriate model for count data characterized by overdispersion is the negative binomial (Hausman, Hall, and Griliches, 1984; Cameron and Trivedi, 1998). We therefore present our results using negative binomial regression. We have panel data involving repeated observations of our set of countries over time, so there may be time-invariant, unobservable country-level factors as well as time trends that influence the innovation process. Thus we present our results with a fixed-effects specification that includes country- and year-fixed effects. ¹² To fit our data to the conditional fixed-effects negative binomial model, we used the xtnbreg command in Stata with the fe option. As Allison and Waterman (2002) indicated, however, this may not control for all stable predictors. Thus we also report our results using alternate fixed-effects specifications in our robustness tests.

Robustness Tests

We conducted a series of tests to check the validity of our results, including endogeneity checks, different modeling specifications, and other robustness tests, none of which change our reported results. Online Appendix B describes the tests we conducted and the results we obtained.