Intellectual Property Rights, Product Complexity, and the Organization of Multinational Firms

Naghavi, Alireza; Spies, Julia; Toubal, Farid Working Paper Intellectual Property Rights, Product Complexity, and the Organization of Multinational Firms Quaderni - Working Paper DSE, No. 773 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Naghavi, Alireza; Spies, Julia; Toubal, Farid (2011) : Intellectual Property Rights, Product Complexity, and the Organization of Multinational Firms, Quaderni - Working Paper DSE, No. 773, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/4466 This Version is available at: https://hdl.handle.net/10419/159612 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ Intellectual Property Rights, Product Complexity, and the Organization of Multinational Firms Alireza Naghavi Julia Spies Farid Toubal Quaderni - Working Paper DSE N°773 Intellectual Property Rights, Product Complexity, and the Organization of Multinational Firms∗ Alireza Naghavi†1, Julia Spies2, and Farid Toubal3 1University of Bologna 2International Trade Centre, Geneva 3Ecole Normale Supérieure de Cachan, Paris School of Economics and CEPII March 2, 2013 Abstract This paper studies how the Intellectual Property Right (IPR) regime in destination countries influences the way multinationals structure the international organization of their production. In particular, we explore how multinationals divide tasks of different complexities across countries with different levels of IPR protection. The analysis studies the decision of firms between procurement from related parties and outsourcing to independents suppliers at the product level. It also breaks down outsourcing into two types by distinguishing whether or not they involve technology sharing between the two parties. We combine data from a French firm-level survey on the mode choice for each transaction with a newly developed complexity measure at the product level. Our results confirm that firms are generally reluctant to source highly complex goods from outside firm boundaries. By studying the interaction between product complexity and the IPR protection, we obtain that (i) for technology-sharing-outsourcing IPRs promote outsourcing of more complex goods to a destination country by guaranteeing the protection of their technology, (ii) for non-technology-related-outsourcing IPRs attract the outsourcing of less complex products that are more prone to reverse engineering and simpler to decodify and imitate. Keywords: Outsourcing, product complexity, intellectual property rights, technology sharing JEL: F12, F23, O34 ∗Research for this paper was partially funded by the European Community’s Seventh Framework Programme (Project INGINEUS, Grant Agreement No.225368, www.ingineus.eu). The authors alone are responsible for its contents which do not necessarily reflect the views or opinions of the European Commission, nor is the EC responsible for any use that might be made of the information appearing herein. The same disclaimer holds for the International Trade Centre (ITC). We thank the participants of the ETSG 2010 for valuable comments. †Corresponding author. University of Bologna. Dipartimento di Scienze Economiche, Piazza Scaravilli 2, I-40126 Bologna, Italy. Email: [email protected], Phone: +39 051 209 8873, Fax: +39 051 209 8040. 1 Introduction In a global survey of senior executives conducted by the Economist Intelligence Unit, 84 percent of all respondents state that they perceive the lack of Intellectual Property Rights (IPRs) protection in emerging markets as a challenge when outsourcing their R&D (The Economist Intelligence Unit Report, 2004). This qualifies IPR protection as the biggest challenge of globalized R&D among all factors listed. Yet, we continue to see actions and decisions by top managers of successful enterprises that contradict these figures. An example is Apple, the company known for boasting that its machines are “made in the USA”. In 2007, it decided to outsource the production of its unscratchable glass for i-phone and i-pad, to a plant in China. Today, “almost all of the 70 million iPhones, 30 million iPads and 59 million other products Apple sold last year were manufactured overseas” (New York Times, 21 January 2012). Apple’s shift, or better revolution, in its organizational approach has drawn a great deal of curiosity in the business world. The company is said to be breaking a taboo by outsourcing and transferring its intellectual property to competitors and countries that have a history of stealing good ideas. Are IPRs truly essential to outsourcing decisions, and if so how do they influence the organizational mode through which multinational firms source inputs? The aim of this research is to take a step in understanding the equivocal connection between IPRs and outsourcing by breaking down products into different levels of complexity and outsourcing into two types by distinguishing whether or not they involve technology sharing between the two parties. A series of influential papers on firm organization highlight the choice faced by firms between purchasing from an affiliate or from an independent supplier, where the latter gives rise to a hold-up problem when contracts are incomplete (see Antràs and Helpman, 2004, 2008 among others). The availability of firm-level data has pushed the literature forward to study the concept of heterogeneity not only among firms, but also in terms of products or tasks embedded in their produc- 1 tion. Grover (2007) interacts the intensity of the sourced input with technology transfer costs and shows that the results from Antràs and Helpman (2004, 2008) only hold for a certain range of technological complexity of the input. More in line with our approach, Costinot et al. (2011) reinterpret the source of contractual frictions as arising from the non-routineness of tasks. Since these cannot be fully specified ex-ante, ex-post adaptation becomes necessary. Due to better communication and less opportunistic behavior among affiliated parties, outsourcing only takes place for tasks below a certain complexity threshold. Focusing on the relation between technology and the outsourcing decision, the message is clear: higher technology complicates the relation with the supplier and makes it optimal to vertically integrate. Despite the clear findings on the relation between outsourcing and technological complexity, the role of IPR protection as a determinant of outsourcing upper parts of the value chain has attracted little attention in the economic literature. Indeed, Antràs and Rossi-Hansberg (2009) suggest that past literature has focused too much on hold-up inefficiencies as the main drivers of the internalization decision and underline the importance of the effects of the non-appropriable nature of knowledge on the internalization decision of firms. Does the IPR regime play the same institutional role as contractual enforcement when outsourcing goods embodied with more complex tasks? A limited number of earlier studies have investigated the impact of IPR protection on outsourcing (see e.g. Glass, 2004). This branch of literature, however, treats imitation in the destination country as an equal threat for all goods and is therefore only testable at the aggregate level. This makes it impossible to study the interaction between IPR protection and task complexity as a determinant of the organization of multinationals. This paper tries to fill in this gap by studying how the IPR regime influences the way multinationals structure the international organization of their production. In particular, it sheds light on how multinationals divide tasks of different complexities across countries with different levels of IPR protection. A multinational active in a given country can 2 choose to source an intermediate good from a related party or from independent suppliers. The outsourcing contract could be an alliance that involves technology sharing, or pure procurement of parts from a supplier. In contrast to the existing studies, we depart from the hold-up problem and emphasize the role of the prevailing IPR regime in the host country in multinationals’ international portfolio of tasks. We test existing theoretical explanations to see whether multinationals are induced to keep technologically complex tasks in the circle of related parties within firm boundaries. Moreover, we check whether a stronger IPR regime in a given country encourages a multinational to procure its intermediates from a foreign independent supplier. But does a stronger IPR regime in a host country encourage multinationals to outsource more complex tasks to foreign independent suppliers? The response is not obvious. Intermediates requiring more complex tasks are more difficult to copy and reproduce and therefore create less infringement concerns. We therefore expect IPR protection to be more crucial for outsourcing (easily imitable) less complex goods. A firm can however still lose information and profits when it intentionally shares technology with its partner and IPR protection is low or absent. In this case intermediates containing a more complex technology are especially sensitive and can be associated with a larger loss in the case of imitation. IPR protection here mitigates imitation risk and promotes outsourcing of more complex goods. To test the validity of the above channels, we study the interaction between the IPR regime in a host country and the complexity of tasks required in the production of intermediates in determining the decision of multinationals on the organization of their production. Evidence on the two opposing effects of complexity on the need for IPR protection was first introduced in Mansfield (1994) who surveyed 100 major US firms with international operations in 1991. He suggested that although more technologically sophisticated industries place a larger emphasis on IPRs, those that require complex inputs are protected from imitation and may therefore be unaffected by IPRs. 3 We empirically examine the choice between intra-firm trade and outsourcing by combining firm-level data on the mode choice for each transaction with a newly developed complexity measure at the product-level. The complexity of a product group is derived by merging three different data sets, (i) ratings of occupations by their intensities in ‘problem solving’ from the U.S. Department of Labor’s Occupational Information Network, (ii) employment shares of occupations by sectors from the Bureau of Labor Statistics Occupational Employment Statistics and (iii) French make tables from Eurostat. The firm-level data comes from a French survey which provides information on import transactions at the product-level of multinational firms and their sourcing mode by country. This information is merged with balance sheet and income statement data from which we compute firm-level productivity. The French data have the advantage of dealing with firms that are part of a multinational network. This allows us to explore the decision of firms with a related party in a given country whether or not to acquire their different inputs from foreign outside suppliers. We argue that this choice is influenced by the complexity level of the tasks involved in the production of the sourced products and the IPR regime in the host country. Conceptually, such data allows us to study how IPRs affect each firm differently rather than the widely studied hypothesis that there are fewer arms-length transactions in countries with weak IPRs. 1 In line with the previous literature, our baseline results suggest that multinationals tend to outsource less complex goods from foreign independent suppliers. These findings confirm that firms are generally reluctant to source highly complex goods from outside firm boundaries. The level of IPRs however does not affect the outsourcing share of inputs with an average level of complexity. We then analyze the interaction between IPR protection and product complexity and find that more complex tasks mitigate the need for IPRs when outsourcing to independent suppliers. Firms tend to outsource more 1In other words, we study the decision of multinationals on the intensive margin (what proportion of their products to source from independent suppliers in a given country) rather than the extensive margin (whether to use integration versus outsourcing in different countries). 4 complex products in countries with relatively less stringent IPR regimes as complexity itself provides a shelter against reverse engineering of goods that are simpler to decodify and imitate. This provides an explanation for the recent behavior by Apple with respect to the outsourcing of some of its most complex tasks to countries with poor records of IPR enforcement. Using the information on the sample of transactions from external suppliers for a more in-depth study of the interaction between product complexity and the IPR protection in the sourcing country, we next look at the type of outsourcing to see whether sharing information on technology has an impact on the outsourcing behavior of firms. Confirming the baseline results, we find a negative and significant interaction term in the sample of outsourced imports from external suppliers without technology sharing. The share of outsourcing of more complex inputs is therefore larger in countries with lower levels of IPRs because of the lower imitation possibilities. We however find a positive interaction term in the sample of transactions that require technology sharing. The share of outsourcing of more complex inputs is therefore larger from countries with a stronger IPR regime when the transactions require technology sharing. IPRs protect firms from the risk of losing their technology (or having it used against their interest), the value of which increases with the level of complexity. In other words, in a weak IPR environment, costly intangible assets embodied in complex goods can only be effectively protected against imitation within firm boundaries. We can therefore conclude that (i) for technology-sharing-outsourcing IPR protection promotes outsourcing of more complex goods to a destination country by guaranteeing the protection of their technology, (ii) for non-technology-related outsourcing more IPR protection attracts the outsourcing of less complex products that are more prone to imitation. Our work is most closely related to Ivus (2010, 2012), who takes the opposite perspective and investigates the impact of improved IPRs in the destination country on exports. She finds that patent protection increases the value of exports from developed to devel- 5 oping countries in patent-sensitive industries, especially for industries that rely heavily on patent protection (Ivus, 2010). This work is extended to differentiate between exports at the intensive and the extensive margin (existing and new products respectively) to find a shift from the former to the latter caused by stronger IPRs (Ivus, 2012). Another closely related paper is Canals and Sener (2012), who study the impact of IPR reform on US offshoring decisions using a measure based on intermediate goods trade. They find that intra-industry offshoring in high-tech industries is most influenced by IPR reforms. Data availability restricts their analysis to industry-level observations and to 16 receiving countries. We take the analysis to a more disaggregate level by looking at the outsourcing decision at the firm-level for products of different complexities to 99 countries. In addition, our data allows us to distinguish between different types of outsourcing by observing whether or not they involve technology sharing. Only such distinction allows us to detect the dual character of product complexity on imitation risk. Considered from a different perspective, our results are also consistent with Branstetter et al. (2006) in that stronger IPRs in a host country stimulate (direct) technology transfer by multinational firms. Finally, related empirical works on complexity include Berkowitz et al. (2006), who show that higher quality legal institutions located in the exporter’s country enhance international trade in complex products. They argue this to be due to a production cost effect, assuming that the production of complex products contains some degree of outsourcing, and hence depends on contracts. Better institutions enable the exporting country to cheaply and quickly enforce contracts and resolve business disputes by reducing the likelihood of hold-up on the production chain. This in turn lowers the production costs of complex products. Since these issues are less important for simple goods, better legal institutions enhance a country’s comparative advantage in complex goods. While Berkowitz et al. (2006) deals with contractual complications of business transactions involving complex products, we explore the relevance of IPRs and the appropriability of knowledge for the type of trade (intraversus extra-group) 6 Classification (SOC).5Each occupation, o, embodies a complexity of zo=i0.25 o+l0.75 o(3) where the weights give the contributions of the two complexity components, importance i∈[1,5] and level l∈[0,7].6Higher values correspond to a higher “importance” or “level” of skills. The difference between “importance” and “level” can best be understood with an example. The subtask “problem identification” has an importance of 3 for both purchasing managers and vessel inspectors. Since purchasing managers are more often confronted with the task of identifying problems, they are assigned a level score of 4.2 whereas vessel inspectors are assigned a 2.8 only. The resulting single score is therefore higher for purchasing managers. After combining the two components into a single complexity score for each occupation, we derive the complexity embodied in each industry. To this end, we use information on the employment of the different occupations by industry. In line with Costinot et al. (2011), we assume that every country in the sample uses the same technology and rely therefore on employment information from the U.S. Bureau of Labor Statistics’ Occupational Employment Statistics (OES). The 1999 data contains the number of employees by occupation in every 3-digit industry k(according to the Standard Industrial Classifi- 5Since the skill-content of occupations may change over time, we use version 3.0 of the O*NET database, released in 2000, to match the data on French multinationals. Note that this version still distinguishes eight categories, namely “problem identification”, “information gathering”, “information organization”, “synthesis/reorganization”, “idea generation”, “idea evaluation”, “implementation planning”, “solution appraisal”, which have been from version 4.0 on summarized into a single category “complex problem solving skills” (Boese and Lewis, 2001). 6We normalized the different scales of the complexity components to a [0,1] scale using the min-max method, I=io−min(i) max(i)−min(i)L=lo−min(l) max(l)−min(l). The weighting scheme is the same as in Jensen and Kletzer (2010). We have tried different weights that have been used in the literature (see also Blinder, 2009) without any effect on the results. 13 cation, SIC).7The occupational intensity, bk o, of each industry is then given by bk o=Lk o PoLk,(4) where Lk ois the employment level of occupation oin industry k. Although the SIC provides information on products or services which are generated under the same industry heading, it does not relate atypical (or secondary) products. In order to compute the skill content embodied in primary and secondary outputs of each industry, we employ a make table for France from 1999 (provided by Eurostat). This allows us to derive a precise complexity measure at the product-level,8 zp(ω) = X k∈p xk(ω) Pkx(ω)zobk o,(5) where xk(ω) Pkx(ω)gives the share of industry kin the production of each product. Table 4 in Appendix summarizes the 32 product categories in our sample ranked according to their complexity. Intellectual Property Rights and Other Controls We use a measure of IPR protection from Park (2008) for the countries of origin available in our sample. This measure of IPR protection is the updated version of the worldwide used Ginarte and Park index (Ginarte and Park, 1997). It incorporates the effects of the TRIPS agreements of 1995 and it takes into account the revisions in national patent laws required to conform to international and regional agreements (such as the North American free trade agreement (NAFTA), European patent convention (EPC), African Regional industrial property organization (ARIPO), Cartagena agreement, among others). All the technical details related to the construction of the index are in Park (2008). We measure the strength of 7Crop production, animal production and private households are not surveyed. After matching the O*NET data to the OES data, 695 occupations remain in the sample. 8Since direct concordance tables of the NACE Rev. 1.1 classification and the SIC 1987 classification are not available, correspondence is achieved via the NAICS 2002 classification. We use simple average to go from the 3-digit level of the SIC to the 2-digit of the NACE. 14 IPR protection in 1995 which is available for the 99 countries of the full sample. The variable is transformed into logs. Table 5 in the Appendix summarizes the information for the countries of our sample. The cross-country rankings of IPRs tend to be correlated with other developmentrelated variables such as rule of law or the level of corruption.9In this case, we attribute some effects to the IPR level while they might be driven by other country characteristics. We therefore introduce other development related variables in the estimation. We control first for the rule of law. The variable measures the independence of the judiciary and the extent to which rule of law prevails in civil and criminal matters. It is taken from the Freedom in the World database. We also control for the level of corruption by including a measure that is taken from the Heritage Foundation database.10 Because the strength of IPR enforcement varies with the overall level of economic development, we include per-capita GDPs in the estimation. The data are taken from the Penn World Tables. We include other explanatory variables that might influence the sourcing decision such as market size, as the number of potential suppliers is larger/thicker in foreign markets (McLaren (2000); Grossman and Helpman (2002)). We also include the distance between the location of the French firm and the countries of origin of the imports to control for the overall monitoring and trade costs (Mugele and Schnitzer, 2006; Ottaviano and Turrini (2007)). Table 6 in the Appendix summarizes information related to the means and standard deviations of the variables used in the estimations. Estimation Methodology Our analysis draws inferences on the impact of IPR and input complexity on the organization of international production. More crucially, our theoretical framework suggests that the share of outsourcing is jointly determined by the 9The correlation between the IPR level and the development-related variables is above 75% as seen in Table 7 in the Appendix. 10This indicator is based on the Corruption Perception Index from Transparency International and assessments of the US Department of Commerce, the Economist Intelligence Unit, the Office of US Trade Representative and official government publications of each country. 15 level of IPR and the product complexity. We introduce thus an interaction term between the product complexity and the IPR level in our empirical analysis. Several characteristics of the data aid in the empirical identification strategy. First, a single French multinational firm often imports different inputs from different countries.11 We use the variation in the countries of origin and imported inputs within each firm to identify which product is outsourced, controlling for firm fixed effects. The introduction of firm-level fixed effects helps moreover controlling for all unobserved firm-level characteristics that might influence the sourcing decision. We estimate the baseline equation using a linear probability model. yijp =λ1IP Rj+λ2Compp+λ3(IP Rj×Compp)(6) +Controlsj+νi+ijp All estimations include a set of firm specific effects, νi. The standard errors are allowed to be adjusted for clustering at the country-product-level to account for heteroskedasticity and non-independence across the repeated observations within firms.12 4 Results Baseline Results Table 2 presents our basic results for the outsourcing share in the intermediate input sample. The empirical model includes firm-specific effects. In Appendix A.2, we present and briefly discuss the results of a model that substitutes the firm-specific effects with the firm-level variables commonly used in the literature. Both the IPR and the complexity variables have been centered around their mean. From column (1) to (3), we include the level of IPR protection, product complexity and the 11Firms in the intermediate inputs sample import on average five product categories from about eight countries. 12This level of clustering is chosen as a benchmark. For robustness we have nonetheless also considered a two-way clustering of the errors at the product- and country-level. The results remain similar. 16 interaction term. In column (4) we introduce other country-level characteristics that may independently impact the outsourcing share, namely the rule of law, corruption, per capita GDP, market size and distance between the home and the host country. Finally, column (5) presents the estimates for the full sample of products. The coefficient of the IPR variable is statistically not significant, suggesting that the IPRs do not affect the oustourcing share of inputs with an average level of complexity. Looking at the complexity measure, we find a negative and significant effect on the outsourcing decision of multinationals. The negative coefficient of the product complexity variable implies a lower outsourcing share for more complex inputs imported from a country with an average level of IPR protection. A one-standard-deviation increase in the complexity coefficient reduces the share of outsourcing by roughly 2.3 percentage points. Corcos et al. (2012) complement these results by providing evidence that complex goods and inputs are more likely to be produced within firm boundaries. Note however that the statistical significance of this effect disappears once we include the full sample of products. Our results suggest that the importance of product complexity is limited to the sourcing of inputs. Turning to the interaction term, looking across each specification in Table 2 reveals strong evidence that the share of outsourcing of complex inputs is smaller in countries with a stronger IPR regime. The negative interaction term is robust and remains significant throughout the different specifications, even after controlling for other countrylevel characteristics. These findings can be interpreted as evidence for a larger share of outsourcing of complex inputs because of the lower imitation possibilities. Complexity works as a shield against reverse engineering mitigating the importance of IPR protection. Firms therefore outsource less complex products (and inputs) that are easier to decodify and imitate to independent suppliers in countries with relatively strong IPR regimes. 17 Table 2: Baseline results (S1) (S2) (S3) (S4) (S5) IP Rj-0.019 -0.019 -0.019 -0.035 -0.040 (0.014) (0.014) (0.014) (0.025) (0.025) Compp-0.184*** -0.177*** -0.169*** -0.010 (0.047) (0.047) (0.044) (0.065) IP Rj×Compp-0.404** -0.374* -0.540*** (0.201) (0.200) (0.196) Rule of Lawj0.001 -0.001 (0.002) (0.002) Corruptionj0.004* 0.000 (0.003) (0.003) Per Capita GDPj0.005 0.002 (0.011) (0.011) Market Sizej0.002 0.003 (0.002) (0.003) Distanceij -0.012*** -0.010*** (0.004) (0.004) Firm-level FE yes yes yes yes yes Sample Inputs Inputs Inputs Inputs Full Observations 25,077 25,077 25,077 25,077 39,711 Adjusted R20.583 0.584 0.584 0.584 0.517 Robust standard errors clustered at the country-product identity. Standard errors in parentheses. ***, **, * significantly different from 0 at 1%, 5% and 10% level, respectively. Main Results In the evidence presented so far, we have explored how the interaction between product complexity and the IPR regime affects the share of arms length trade without distinguishing between the type of outsourcing. The French firm-level data allows us take a step further and distinguish between the type of relationship that links French firms with their foreign suppliers. To do this, we use the information provided on whether or not an outsourcing relation involves technology sharing to analyze whether this has an impact on the outsourcing behavior of firms in our sample. The results are reported in Table 3. Columns (1-2) report the results corresponding to the sample of transactions from external suppliers classified as pure procurement of parts from a supplier without technology sharing, while columns (3-4) present the results using the sample of transactions that involve sharing information on the technology of multinational firms. 18 Table 3: Outsourcing with and without technology sharing External Supplier Without Technology Sharing With Technology Sharing (S1) (S2) (S3) (S4) IP Rj-0.015 -0.047* -0.003 0.012 (0.014) (0.026) (0.005) (0.008) Compp-0.161*** -0.152*** -0.016 -0.017 (0.050) (0.046) (0.013) (0.013) IP Rj×Compp-0.563*** -0.531*** 0.159** 0.156** (0.202) (0.201) (0.071) (0.071) Rule of Lawj0.000 0.001 (0.002) (0.001) Corruptionj0.004* -0.000 (0.003) (0.001) Per Capita GDPj0.013 -0.008* (0.011) (0.005) Market Sizej0.002 -0.000 (0.002) (0.001) Distanceij -0.014*** 0.002** (0.004) (0.001) Firm-level FE yes yes yes yes Sample Inputs Inputs Inputs Inputs 25,077 25,077 25,077 25,077 0.592 0.593 0.543 0.543 Robust standard errors clustered at the country-product identity. Standard errors in parentheses. ***, **, * significantly different from 0 at 1%, 5% and 10% level, respectively. Confirming the baseline results, we find a negative and significant interaction term in the sample of outsourced imports from external suppliers without technology sharing.13 The share of outsourcing of more complex inputs is therefore larger in countries with lower levels of IPR because of the lower imitation possibilities and therefore less infringement concerns. We however find a positive interaction term in the sample of transactions that require technology sharing. The share of outsourcing of more complex inputs is therefore larger from countries with a stronger IPR regime when the transactions require technol- 13The regression models estimated in Table 3 generate similar goodness-of-fit results regardless of the sample; note that the Adjusted R2is nearly the same across the different specifications presented in Tables 2 and 3. 19 ogy sharing. IPRs protect firms from the risk of losing their technology, the value of which rises with complexity. Complex inputs involve costly R&D efforts that are likely to be more effectively protected against imitation within firm boundaries. Due to the greater amount of intangible assets embodied in complex goods, outsourcing to foreign independent suppliers is only a profitable option if the IPR regime is sufficiently strong to protect firms against the dissipation of their technology. This result is consistent with the alternative property-rights explanation in the spirit of Grossman and Hart (1986) and Hart and Moore (1990) provided by Carluccio and Fally (2012), who find that multinationals are more likely to integrate suppliers of more complex inputs that are located in countries with poor financial institutions. They argue this to be an attempt by multinationals to reduce exposure to opportunism in countries where financial development is low. The results after splitting the sample according to the two outsourcing types confirm our conceptual framework and provide novel results. First, for technology-sharing- outsourcing IPR protection promotes outsourcing of more complex goods by mitigating the risk of imitation, thereby guaranteeing the protection of their technology. Second, for outsourcing contracts that do not require technology sharing, more IPR protection attracts the outsourcing of less complex products that are more prone to imitation. This is the first work to our knowledge that encompasses the dual role of complexity and the role of IPRs in protecting intangible assets introduced in Mansfield (1994). Although more technologically sophisticated industries place a larger emphasis on IPRs, those that require complex inputs are automatically protected from imitation making IPRs instead more important for easily imitable goods. Finally, it is worth noting that the direct effect of product complexity is not significant in the technology sharing sample. The level of product complexity hence does not matter for the inputs that are imported from an average IPR country. We find moreover a positive effect of distance on the sourcing share of inputs from external suppliers with technology sharing. In line with the 20 transaction-costs literature (e.g. Williamson, 1975, 1985), this finding suggests higher monitoring costs when the relationship requires technology sharing making arms length trade a more desirable option. Finally, we find a negative coefficient for the per capita GDP variable suggesting a lower share of outsourcing to less developed countries. 5 Conclusion This paper has investigated the decision of multinational firms in the sourcing of products of different complexity levels. The analysis has proposed and tested the hypothesis that technological complexity of a product and the level of IPR protection are the co-determinants of the sourcing mode of multinational firms. As measures of the technological intensity at the product-level are not available, we have built a new measure reflecting the complex problem-solving skills involved in the production of a good. The estimations confirm the predictions made about firms’ decision between intra-firm trade and procurement from independent suppliers. The results first shed light on the significance of IPRs given the nature of products or inputs. In contrast to conventional wisdom, IPRs tend to play a more important role in the outsourcing of simple goods. Our explanation for this result relies on the hypothesis offered by Mansfield (1994) that complexity may work as a self-defense mechanism against the infringement of goods. In addition, the availability of a rich French firm- and productlevel data has made it possible to conduct the first study on multinationals and IPRs that distinguishes between outsourcing transactions by whether or not they involve technology sharing. Doing so reveals that firms recognize IPRs as an important factor in their decision regarding the location of outsourcing, when the contract involves technology transfer. Only in this case, firms outsource highly complex products to countries where their intellectual property is recognized. Here IPRs work to protect a multinational against the dissipation of their knowledge-based intangible assets. As much as the paper contributes to understanding the internationalization strategy 21 of firms and the differences between intra-firm trade and outsourcing, it bears an important policy conclusion: The results suggest that attracting the upper part of the value chain requires building trust into the protection of IPRs. 22 Table 7: Correlation Matrix (Intermediate Inputs Sample: 25077 Observations) yijp IP RjComppIP Rj×ComppRule of LawjCorruptionjPer Capita GDPjMarket SizejDistanceij yijp 1.00 IP Rj-0.01 1.00 Compp-0.09 0.05 1.00 IP Rj×Compp0.01 -0.10 -0.09 1.00 Rule of Lawj0.00 0.75 0.01 -0.06 1.00 Corruptionj-0.01 0.77 0.04 -0.07 0.77 1.00 Per Capita GDPj-0.01 0.87 0.05 -0.09 0.81 0.81 1.00 Market Sizej-0.01 0.20 0.07 0.01 -0.10 0.11 0.11 1.00 Distanceij -0.04 -0.44 0.09 0.07 -0.51 -0.46 -0.46 0.23 1.00 29 A.2 The results including firm-specific variables Table 8 presents the results based on an empirical model that includes specific effects at the 3-digit sector levels. We substitute the firm-fixed effects with a set of firm-specific factors for which we have information. The firm-level variable are identified in the literature to be important determinants of the outsourcing share (See the remarkably detailed paper by Nunn and Trefler (2013)). We include the size of the firm, measured as the total number of employees, its age, and its capital and skill intensity. The skill intensity is measured as the number of skilled production workers in total employment.14 With the exception of age, we take the logarithm one year lagged value of each variable. Our results remain mostly robust. In particular, we find strong evidence that the share of outsourcing of complex inputs is larger in countries with lower levels of IPR. The direct effect of IPR is not statistically significant while the product complexity is statistically negative. Turning to the firm-level variables, we find that older firms have a statistically significant larger share of outsourcing. As argued by Bartel et al. (2008), older firms might have had time to learn about the the quality and reliability of potential foreign input suppliers. The size variable is also positive and significant meaning that larger firms have a larger share of outsourcing. In line with the recent literature on the choice between outsourcing and intra-firm trade, we find a negative impact of the capital and skill intensities on the outsourcing share (Nunn and Trefler (2013)). 14We follow Biscourp and Kramarz (2007) and define the skill groups according to the French Classification of Occupation and Social Categories. We divide these categories into six groups: executives, technicians, intermediate administrative occupation, clerks, skilled production workers and unskilled production workers. The share of skilled production workers is the ratio of the number of skilled production workers to the total number of production workers. 30 Table 8: Results including firm-specific variables (S1) (S2) (S3) (S4) (S5) (S6) IP Rj-0.010 -0.010 -0.010 -0.046 -0.029 -0.035 (0.015) (0.015) (0.015) (0.032) (0.034) (0.029) Compp-0.209*** -0.201*** -0.187*** -0.195*** -0.010 (0.055) (0.055) (0.052) (0.055) (0.067) IP Rj×Compp-0.453** -0.415* -0.619** -0.581*** (0.212) (0.213) (0.248) (0.201) Rule of Lawj0.001 0.001 -0.001 (0.002) (0.003) (0.002) Corruptionj0.004 0.003 0.000 (0.003) (0.004) (0.003) Per Capita GDPj0.004 0.006 0.006 (0.015) (0.016) (0.013) Market Sizej0.005 0.007** 0.008** (0.003) (0.003) (0.003) Distanceij -0.016*** -0.013*** -0.008** (0.004) (0.004) (0.004) Sizei0.010*** 0.011*** (0.003) (0.003) Agei0.001*** 0.001*** (0.000) (0.000) Capital Intensityi-0.006* -0.006** (0.003) (0.003) Skill Intensityi-0.054*** -0.064*** (0.007) (0.006) 3-digit Sector FE yes yes yes yes yes yes Sample Inputs Inputs Inputs Inputs Inputs Full Observations 25,077 25,077 25,077 25,077 19,916 31,570 Adjusted R20.142 0.143 0.143 0.144 0.162 0.133 Robust standard errors clustered at the country-product identity. Standard errors in parentheses. ***, **, * significantly different from 0 at 1%, 5% and 10% level, respectively. 31 