Intellectual Property Rights, International Migration, and Diaspora Knowledge Networks

Transcription

1 Intellectual Property Rights, International Migration, and Diaspora Knowledge Networks Alireza Naghavi y Chiara Strozzi z Abstract This paper studies the interaction between skilled emigration and intellectual property rights (IPR) in determining innovation in developing countries. We argue that knowledge acquired by emigrants abroad can ow back into their country of origin through diaspora networks. IPR protection in the sending country magni es this e ect by increasing the size of the innovation sector, thereby allowing diaspora gains to fall on a larger range of workers. This transforms brain drain into brain gain. Our empirical analysis uses a panel dataset of emerging and developing countries and con rms a positive correlation between emigration and innovation when IPRs are su ciently strong. J.E.L. Classi cation: O30; F22; J24. Keywords: Intellectual property rights; International migration; Innovation; Knowledge ows; Brain gain; Diaspora. We are indebted to Hillel Rapoport for valuable comments and suggestions that helped us substantially improve the paper. We also thank two anonymous referees, Michel Beine, Maria Elena Bontempi, Maria Comune, Roberto Golinelli, Peter Kuhn, Elisabetta Lodigiani, Giovanni Prarolo, Arsen Palestini, as well as the seminar participants at University of Lille 1, the University of Bologna, and the Italian Trade Study Group 2011 for very helpful remarks. We gratefully acknowledge the European Commission for nancial support through the 7th Framework Programme Project INGINEUS. Financial support from Fondazione Cassa Risparmio di Modena is also gratefully acknowledged. y Corresponding author: University of Bologna. Address: Department of Economics, University of Bologna, Piazza Scaravilli 2, Bologna, Italy. Phone: , Fax: , alireza.naghavi@unibo.it. z University of Modena and Reggio Emilia, IZA. Address: Department of Economics, University of Modena and Reggio Emilia, Viale Berengario 51, Modena, Italy. Phone: , Fax: , chiara.strozzi@unimore.it. 1

2 1 Introduction International trade and foreign direct investment (FDI) have often been identi ed as the main determinants of innovation and growth in developing countries (South) (Saggi, 2002; Keller, 2004). While the relevance of trade and FDI has been con rmed with a signi cant increase in their ratio to world output in the 1990 s, high-skill migration from the South to OECD countries has also witnessed a remarkable growth in the same period (Docquier and Rapoport, 2012). The resulting surge in the outward transfer of the human capital in migrants has created controversial debates about the threats and opportunities that skilled emigration may pose to the South. On the one hand, the traditional literature on migration and brain drain presents mechanisms through which skilled emigration could be detrimental to growth. 1 On the other hand, a growing branch of contributions argues that skilled emigration need not harm the South and may even increase the potential for development. The so-called brain gain e ect derives from an incentive channel that works through the increased expected returns to education brought about by migration prospects (Mountford, 1997; Stark et al. 2007; Beine et al., 2001, 2008). 2 An additional channel is return migration, which can induce innovation through the knowledge possessed by migrants returning from more advanced economies (Domingues Dos Santos and Postel-Vinay, 2003; Mayr and Peri, 2009; Dustmann et al. 2011). Finally, cross-border diaspora networks among skilled emigrants and natives may also promote access to foreign-produced knowledge and foster innovation by encouraging trade, investments, and the recirculation of information back into the sending countries (Agrawal et al., 2011; Kerr, 2008). 3 Referring to Agrawal et al. (2011), The Economist (2009) writes: "[...] a scienti c diaspora gives countries of origin a leg-up in terms of access to the latest research, mitigating some of the problems of a brain drain." There is little doubt in today s role of emigration in creating potential gains to the home country through diaspora networks. Nevertheless, little formal research in the economic literature directly examines the potential link between the knowledge absorbed by emigrants abroad and innovation in their home countries. 4 This research contributes to the literature by exploring how emigration from the South may a ect innovation activities in the home (sending) country. The aim is to investigate the existence of a channel through which the knowledge learned by emigrants after interacting with higher skills 1 Seminal works include those of Berry and Soligo (1969), Bhagwati and Hamada (1974) and Miyagiwa (1991). For a recent complete survey of the literature on brain drain and development, see Docquier and Rapoport (2011). 2 The possibilities of such gains from emigration were rst referred to by Bhagwati and Rodrigues (1975). 3 Studies in other disciplines, such as Meyer (2001), suggest that such informal networks are crucial in turning brain drain into a net brain gain. Student/scholarly networks, local associations of skilled expatriates, short-term consultancies by high-skilled expatriots in their country of origins, and other unestablished intellectual/scienti c diaspora networks are a few examples of such networks (Meyer and Brown, 1999). 4 Williams (2007) and Oettl and Agrawal (2008) focus on the externalities of international migration to emphasize their role in knowledge and technology transfer. More recently, Beine et al. (2011) show the in uence of diasporas on the evolution of migration ows and their composition in terms of skills. 2

3 in developed countries (North) can ow back to the South. 5 We refer to this channel as an "intellectual diaspora", that is, the remote mobilization of intellectuals and professionals abroad and their connection to scienti c, technological and cultural programs at home. This can be thought of as a scientist from the South being more productive in the North due to better facilities and more resources, with some of the bene ts of his innovation owing back to his home country. Diaspora networks make this possible through the cross-border sharing of ideas, for example between Indian computer scientists in Bangalore and their counterparts in Silicon Valley (The Economist, 2009, 2011). In particular, this paper seeks to examine the role of intellectual property rights (IPR) protection in the South by exploring how IPRs interact with emigration in determining innovation performance. The key question we aim to answer is whether an appropriate level of IPR protection could help transform the brain drain caused by skilled emigration into a brain gain. In sum, we argue that although emigration may directly result in a brain drain, it also generates a ow of ideas and inventions back to the sending country, the extent of which depends on the strength of IPR protection. Our theoretical framework is a variant of the Yeaple (2005) and Ohnsorge and Tre er (2007) models of heterogeneous workers, where we introduce an innovation sector, migration, and IPR protection. Emigration reduces e ective innovation activities due to the loss of the most skilled (the extensive margin). Migration, however, also opens a diaspora channel through which the knowledge acquired abroad can ow back into the innovation sector in the home economy and enhance the skills of the remaining workers (the intensive margin). To investigate whether the bene cial e ect of diasporas could outweigh the direct negative e ect of the out ow of skilled workers, we look at the average skills and the size of the innovation sector. While a strong level of IPR protection directly reduces e ective innovation activities by raising returns to skills and hence engaging also the less skilled mass of workers in the innovation sector, the enlarged sector allows diaspora gains to fall on a larger range of workers actively using their skills in the economy. As a consequence, a strong level of IPR protection in the sending country increases the magnitude of potential bene ts from diaspora, making it more likely for the gains to outweigh the negative e ects of brain drain on innovation, and thus facilitating a potential net brain gain. 6 Using a sample of emerging and developing economies, we then perform an empirical analysis to investigate the joint impact of emigration and IPR protection in the sending country on innovation there. The sample we use is a panel of 35 low-income countries ranging from 1995 to We 5 In this framework, the capacity of innovation of Southern innovators, who remain in their origin countries, is related to their access to valuable technological knowledge that is partially accumulated abroad (i.e., brain banks). Fir more on this issue, see Agrawal et al. (2011). 6 These results are in contrast to the theoretical conclusion obtained by McAusland and Kuhn (2011), who claim IPRs to be an obstacle to the international ow of brains. In short, they argue that if brains are emigrating, a country may as well lower its IPRs to free-ride on brains that have moved elsewhere. While their study is to our knowledge the rst contribution that explicitly investigates the link between IPRs and brain circulation, it does not take into account any channels through which the skills acquired abroad can be transferred back into the country of origin. 3

4 measure innovation activities in the South through the number of resident patents, with data taken from WIPO (World Intellectual Property Organization). We use this information together with extensive original data on migration ows and stocks obtained from national statistical o ces and with the index of IPR protection as measured by Park (2008). Our ndings show the impact of emigration on innovation to be positive in the presence of strong IPR protection. Hence, IPRs have a role in promoting the bene cial e ects of the diaspora channel of knowledge, con rming the main conclusions of our theoretical model. To validate that our main empirical ndings are mainly driven by the phenomenon of intellectual diaspora rather than by other compatible explanations, we also apply a variant of the method proposed by Spilimbergo (2009). In particular, we construct a measure of emigration that makes it possible to relate an increase in resident patents granted in a sending country to "where" the emigrants go. The underlying assumption is that emigrants may better promote innovation in their home countries if their host countries have a higher potential for innovation. This approach allows us to con rm that it is the skills learned from abroad by emigrants and transferred back to their home country that increase successful innovation there under strong IPRs, thus corroborating our theoretical conclusions. The role of IPR protection in any study that involves innovation and the developing world is crucial. However, while the trade-o s faced by an emerging economy between imitation and the provision of incentives for domestic innovation through IPRs are clear (Maskus, 2000), the interrelationships between skilled migration and IPR policy in determining innovation remain to be explored. Our work lls this gap and contributes to the above-mentioned strand of research by capturing the diaspora dimension of migration and revealing how IPR protection in the sending country may in uence the e ect of skilled migration on innovation there. 7 On this basis, we shed light on the net impact of emigration on innovation and determine whether a strong IPR regime at home can eventually turn the initial brain drain into a brain gain. By conducting an empirical investigation focused on emerging and developing countries, our work also contributes to the missing empirical analysis on innovation and development in the South. 8 In the remainder of the paper, we introduce the theory in Sections 2 and 3, the empirical exercise in Section 4, and conclude in Section 5. 7 Among the vast literature on intellectual property rights, Chen and Puttinan (2005) and Parello (2008) are perhaps most closely related to our work, as they speci cally focus on domestic skill accumulation and innovation. While the former positively relates IPR protection to innovation, the latter deems it to be ine ective for innovation in less-developed countries. 8 Indeed, while innovation has been deemed central to economic take-o, catch-up, and development in low-income countries, research on innovation tends to neglect developing countries, leaving a large gap in the economic literature (Lorentzen and Mohamed, 2010). 4

5 2 The Model 2.1 The Basic Framework Suppose there are two regions: a developing economy referred to as the South and an alternative North with better economic opportunities and employment possibilities, where skills and wages are assumed to be higher. Because the focus of our study is the Southern market, we concentrate our analysis on goods invented, produced and consumed locally in the South. 9 Consumers have the following utility function: 2 U i = C i = 4 Z N c j di5 ; (1) where individual consumption C i is divided between a continuum of N invented goods subscripted by j 2 (0; N), and 2 (0; 1) represents the inverse measure of product di erentiation. There are two sectors in the economy: a production and an innovation sector. Labor is the only factor of production and innovation and is mobile between sectors. Workers are spread over a continuum of skills z 2 [0; 1), distributed with density g(z) and cumulative distribution G(z). We normalize the mass of workers to one. While production does not require skills, a worker i with skills z i in the innovation sector has productivity h i such that h i (z) = z i + Z; (2) where z i represents own skill endowment and Z (de ned below) is the spillover of knowledge learned by emigrants abroad through what we call the "diaspora" channel. We are interested in observing the initial skill endowment of each individual (their innate ability) used in the innovation sector and how the assumed superior knowledge from the North can ow back to upgrade workers productivity. The timing of the model is as follows. Nature reveals the IPR regime, which is assumed to be exogenous to our model. Emigration takes place in period 0, activating the diaspora channel. Innovation is then carried out in the rst period, and production occurs in the second. Emigration in period 0 is modeled as a movement of labor from the South to the North at a cost F, which allows only the highest skilled to move. Potential diaspora is then realized by means of skilled emigrants transferring their newly acquired knowledge back to the South. We de ne the positive externalities from diaspora networks as Z = b ~ ; (3) 9 Using a reduced form model that abstracts from trade and FDI related issues allows us to single out the impact of South-North migration, but clearly does not provide de nitive answers to how IPRs and diasporas interact. See Iranzo and Peri (2009) for a study of trade and migration and Davis and Naghavi (2010) for the e ects of trade and o shoring on innovation within the same occupational choice setting. These papers however do not deal with the issue of IPRs. 5

6 where the average skills endowment of those who migrate to the North is ~ > 0. Parameter b 0 measures the intensity of diasporas, which is in uenced by factors such as the level of academic and professional interactions and the amount of skills learned in the North, or the successful transmission of knowledge to and the absorptive capacity of the South. Note that b = 0 implies no international knowledge transfer, b = 1 the return of only original (pre-migration) skills of emigrants, and b > 1 the di usion of their improved skills to the South. In period 1, N goods are invented. Each good needs units of skills. The total amount of human capital in the economy can be written as Z 1 H(z) = h i g(z)dz; (4) z 1 where z 1 represents the skills of a threshold worker indi erent between working in the production or the innovation sector. The total number of goods available for consumption are N = N(z) = H(z)=: (5) To work in the innovation sector, each worker must go through training at a cost e, which is paid in the second period. The wage per unit of skill for the high-skilled in the innovation sector is! H and is paid in period 2, giving each individual with skills z i a wage equivalent to h i! H. In period 2, the production sector absorbs all workers who have not worked in the innovation sector in the rst period. The production function is CRS in labor and has a productivity equal to one such that there is a one-to-one relationship between output and labor, n j = l j. Individual wage is identical for all workers in this sector and equals! L. 2.2 Patents and Consumption We use the basic framework presented in Saint-Paul (2003, 2004) as our benchmark, modeling IPR protection as the probability that an innovator can obtain monopoly power over his invention. 10 The probability of being granted a patent is q, which captures the degree of IPR protection. 11 The price of a non-patented good is equal to its marginal cost normalized to one, which also gives us wages in the production sector, p L =! L = 1. Otherwise, if a patent is granted, a rm charges monopoly price p M =, which includes a mark-up over marginal cost: = 1=: (6) 10 Saint Paul (2004) uses this setting to explore the implications of IPRs and redistribution on occupational choice and welfare. 11 Grossman and Lai (2004) also model patent protection in a similar manner. 6

7 Next, consumption is divided between patented and non-patented goods, c P and c N, respectively. Consumers allocate their income y (net of training costs) between the two types of goods by maximizing (1) or equivalently under the budget constraint Max Nqc P + N(1 q)c c N ;c P N; (7) y = Nqc P + N(1 q)c N : (8) The solution to the above maximization problem is: c N = y ; c P = y 1 ( 1) ; (9) where = N(1 q) + Nq ( 1) (10) captures the love of variety e ect =@N > 0 and the disutility caused by monopoly pricing =@q < 0. Using (1), (7), (9) and (10), aggregate consumption index is therefore C = y ( 1) = y P ; where P = ( 1) is the aggregate price index. The value of a patent, which is equal to monopoly pro ts, is equal to = ( 1) Y 1 ( 1) ; (11) where Y is aggregate income (net of training cost). In the above expression, the rst term on the right-hand side (RHS) is the mark-up, while the second is total demand for the patented good. Under a competitive labor market, expected pro t from inventing a new good must equal to its cost in terms of skills such that q =! H : Using (11), this gives! H = q( 1)Y 1 ( 1) = : (12) Recalling that > 1 and < 1, stronger patent protection (higher q) increases wages in the innovation sector! H as! H q = ( 1)Y 1 ( 1) + N(1 ( 1) ) 2 2 > 0; (13) 7

8 making it more attractiveness to work in the innovation sector. 3 Determinants of Innovation We now study the impact of IPR protection, emigration, and the combination of the two by means of diaspora knowledge networks on innovation in the South. 3.1 Intellectual Property Rights Protection We rst de ne innovation as the amount of e ective innovation activities that lead to the creation of new patents, measured by the average level of skills in that sector: Z 1 1 ~z = zdg(z): (14) 1 G(z 1 ) z 1 This allows us to use heterogeneous skills to consider the e ectiveness of human capital in play in the innovation sector rather than simply looking at its absolute size. 12 We rst use the amount of skills workers are endowed with to study the direct e ect of IPRs and migration on innovation (~z) keeping Z xed. We then investigate how the productivity of each worker (h i ) increases in the intensive margin through diasporas in proportion to their initial level of skills. To pin down the threshold skill level z 1 above which individuals choose to train and work in the innovation sector, consider a worker with productivity h i, who can either work in the innovation sector and earn! H h i e, where! H > 1, or become a production worker with wage! L = Choosing the option that generates a higher income, a worker chooses to work in the innovation sector if! H h i e > 1 )! H z i +! H Z e > 1 ) z 1 = 1 + e! HZ! H : (15) Lemma 1 The threshold skill level z 1, which identi es the equilibrium allocation of workers between the production and the innovation sector, decreases with IPR protection q as! H q > 0 and z1! H < 0: higher returns to skills in the South shift workers from the production to the innovation sector. Proof. Follows directly from (13) and (15). 12 Alternatively, one can use total skills in the innovation sector as a proxy for innovation. This takes away an interesting feature of the model made possible by heterogeneity in skills, as adding a worker of any skill level to the innovation sector would monotonically increase innovation in a trivial way. To capture some ambiguity in such case, it would be su cient for average skills to play some role in producing innovation so that adding less clever people to a group would reduce the productivity of the group. This can be done by adding ~z to the productivity of individuals as local spillovers: Equation (2) turns to h i (z) = z i + ~z + b ~, while the essence of the results persists with a slightly more complicated notation. 13 Note that using h i from (2) to nd the threshold implies that workers take into account the potential spillover of knowledge learned by emigrants abroad, Z, when choosing their occupation or deciding whether or not to migrate. In other words, they anticipate potential bene ts from diaspora through migration by others. 8

9 It follows from Lemma 1 that better IPR enforcement q changes average skill level and therefore e ective innovation activities by increasing the returns to working in the innovation sector. IPR protection increases! H as demonstrated in (13) encouraging less skilled workers to move into the innovation sector. 14 E ective innovation activities measured by the average level of skills in (14) therefore falls due to an expansion of that sector towards the less skilled mass of workers. impact of IPR enforcement on innovation in the South is illustrated in Figure Proposition 1 Given z1 q The > 0 from Lemma 1, IPR protection q in the South has a direct negative e ect on its e ective innovation activities ~z as ~z z 1 > 0. Proof. See (14) and Appendix A.3. [FIGURE 1 ABOUT HERE] Di erentiating (14) with respect to z 1 re ects the basic results from the occupational choice model of Roy (1951). Because ~z z 1 > 0, the entry of less-skilled workers in the innovation sector reduces average skills. We interpret this as a direct negative e ect of IPRs on innovation caused by a misallocation of the low-skilled to the innovation sector. Proposition 1 is in line with Glass and Saggi (1998) and Vandenbussche, et al. (2006), who argue that a shift of less skilled workers away from less skill-intensive activities could have adverse e ects for countries far from the technological frontier. 3.2 Migration To account for migration, we introduce a second skill threshold level z 2 above which workers migrate, and rede ne average skills in (14) as Z z 2 1 ~z = zdg(z): (16) G(z 2 ) G(z 1 ) z 1 Throughout the rest of the analysis, we make the assumption that z 1 < z 2 < 1, in order to allow a positive and nite level of migration. 16 A worker migrates to the North if his gains from doing so, net of migration costs, exceed what he would earn in the innovation sector at home:! M z i e F >! H h i e )! M z 2 F =! H z 2 +! H Z ) z 2 = F +! HZ! M! H ; (17) 14 In addition, higher training costs e prevent entry by the low-skilled, whereas higher wages in the innovation sector at home! H and higher potential diaspora gains Z increase returns to working in the innovation sector, thereby shifting workers there. 15 In our analysis below, we show how this reallocation of workers following an improvement of the IPR regime can be bene cial when emigration and the resulting diaspora gains are taken into account. 16 See Appendix A.5 for the feasibility of the assumption. 9

10 where we assume an exogenous wage in the innovation sector of the North higher than that in the South:! M >! H. Lemma 2 The threshold skill level z 2, which distinguishes emigrants from non-emigrants, increases in migration costs F as z2 F Proof. Follows directly from (17). > 0: a fall in the cost of migration encourages emigration to the North. It follows from Lemma 2 that a reduction in migration costs F changes average skills and therefore innovation activities by spurring migration. Emigration decreases the size of the innovation sector by a movement of workers from the upper tail of the distribution of skills out of the country. 17 average level of skills in the South, ~z, therefore falls due to a depletion of skills and knowledge. Proposition 2 Given z2 F on its e ective innovation activities ~z as ~z z 2 > 0. Proof. See (16) and Appendix A.4. The result ~z z 2 The > 0 from Lemma 2, emigration from the South has a direct negative e ect > 0 parallels Roy (1951) in that the loss of the most-skilled workers in the innovation sector reduces average skills. We interpret this as a direct negative e ect of emigration caused by the well-known brain drain syndrome and the ow of skills out of the country. 3.3 Equilibrium The economy is in equilibrium when the allocation of workers across sectors is compatible with the labor and product market clearing conditions. The total number of workers in the production sector in terms of the threshold skill level z 1 is Z z 1 L = L(z 1 ) = g(z)dz = G(z 1 ); (18) and total skills in the innovation sector in terms of z 1 and z 2 is expressed by 0 H(z) = H(z 1 ; z 2 ) = h i g(z)dz: (19) z 1 Market clearing implies that the total output net training cost Y is equal to the total factor income: 18 Y =! H H(z 1 ; z 2 ) + L(z 1 ): (20) Z z 2 17 In addition, higher prospective wages abroad! M encourage the ow of skills away from the country, whereas higher wages in the innovation sector at home! H reduce skilled emigration. Higher potential gains from diaspora Z similarly discourage migration as individuals are aware that they can partially enjoy the knowledge acquired by others who emigrate without bearing the costs of migration themselves. 18 In the following, we assume training costs e to be embedded in Y, which simpli es the notation but does not in uence the results. 10

11 This equilibrium condition can equivalently be written through the labor market clearing condition L(z 1 ) = [N(1 q)] Y + Nq Y 1 ( 1) ; (21) where the rst and the second term on the RHS derive from the total consumer demand for the non-patented and patented goods, respectively. We can close the model by using equations (5), (10), (12), and (20) to solve for the equilibrium wage in terms of z 1 and z 2 :! H =! H (z 1 ; z 2 ) = q( 1) 1 ( 1) L(z1 ) + H(z 1 ; z 2 )[1 q(1 ( 1) )] : (22) Observing (22), an increase in q on the RHS of is always compensated by a fall in z 1 1 )=@z 1 > 0 in the numerator 1 ; z 2 )=@z 1 < 0 in the denominator) and an upward shift in z 2 1 ; z 2 )=@z 2 > 0) to maintain equilibrium. obtain the following two-equation system: Using (22) together with (15), (17), we! H (z 1 ; z 2 )(z 1 + Z) = 1 + e; (23)! H (z 1 ; z 2 )(z 2 + Z) =! M z 2 F: This allows to calculate the dynamics of z 1 and z 2 with respect to changes in the IPR regime, q, and hence analyze how skilled emigration could promote innovation in the South. We then explore the conditions under which the bene cial e ects of cross-border diaspora are likely to outweigh the negative brain drain e ect of emigration on innovation and transform it into brain gain Diaspora This section brings together IPR protection and migration to study their combined e ect on innovation in the sending country through diaspora channels. We are interested in studying the spillover of new knowledge learned by migrants back to their original country. Such potential gains from diaspora are denoted by Z in equation (3) and illustrated in Figure 2. The extent to which diasporas create gains for the sending country depends on the size of the innovation sector, which we aim to pin down in this section. [FIGURE 2 ABOUT HERE] First, note from (17) that the threshold skill level z 2, which distinguishes emigrants from nonemigrants, increases in wages! H as z2! H > 0. That is, higher skilled wages in the South discourage 19 Using (22) to rewrite the equilibrium condition (23), it is easy to see that thresholds z 1 and z 2 must move in opposite directions, 2 =@z 1 < 0. It will be seen in the next sub-section how this general equilibrium e ect reinforces our key results. The formal proof can in turn be found in Appendix A.5. 11

12 emigration to the North. IPR protection therefore works as a force against brain drain by preserving skills in the South. Observing (15) and (17) together reveals that IPR protection increases the size of the innovation sector from both spectrums. Lemma 3 IPR protection increases the size of the innovation sector from both sides of the spectrum as z1! H < 0 and z2! H > 0: better wages in the innovation sector attract less skilled workers into the sector and encourage more skilled workers to remain in the home country. Proof. Follows directly from Lemma 1, (13) and (17). Di erentiating (16) with respect to z 1 and z 2 gives ~z z 1 > 0 and ~z z 2 > 0. A rise in per unit wages of the skilled in the South caused by a tougher IPR regime therefore has an ambiguous e ect on average skills in the innovation sector. We are now in the position to make some conclusions about how IPR protection in uences the magnitude of the diaspora e ects of migration on innovation activities in the sending country. Namely, an increase in the IPR protection level q attracts workers from the production sector and partially prevents migration, increasing the size of the innovation sector in the South. The resulting reduced migration works against the direct negative impact of IPRs on e ective innovation activities. Proposition 3 Given Lemma 3, a stronger level of IPR protection in the South (higher q) fosters potential gains from diaspora by expanding the size of the innovation sector from both ends of the spectrum by reducing z 1 ( dz1 dq Proof. See Appendix A.5. < 0) and raising z 2 ( dz2 dq > 0). Proposition 3 states that for any given F that yields a positive level of migration, more stringent IPR protection allows potential gains from diaspora, Z, to fall on a larger range of workers active in the innovation sector. The channels that determine the extent to which the diaspora mechanism it at work are depicted in Figure 3. [FIGURE 3 ABOUT HERE] 3.5 Brain Drain Versus Brain Gain To measure the net e ect of migration on innovation in the South, we must weigh the magnitude of the negative brain drain e ect against gains brought about by the diaspora channel. Brain drain can be summarized as the direct loss of skills embedded in workers who migrate abroad, i.e., the extensive margin. This is, in other words, the amount of skills initially available prior to migration minus the base skills of the remaining workers post-migration: Z 1 BD = zg(z)dz z 1 Z z 2 Z 1 zg(z)dz = zg(z)dz: (24) z 1 z 2 12

13 Next, we rewrite the aggregate supply of skills as Z z 2 Z z 2 Z z 2 Z z 2 H(z 1 ; z 2 ) = (z + Z)g(z)dz = (z + b )g(z)dz ~ = z 1 z 1 z 1 zg(z)dz + b g(z)dz; ~ (25) z 1 The rst term on the RHS represents the amount of skill workers in the innovation sector are originally endowed with and the second term the aggregate diaspora e ect on the same workers still residing in the South, i.e., the intensive margin. 20 The second term on the RHS of (25) denotes the virtual return of upgraded skills through diaspora and can be rewritten to de ne brain gain as BG = b Z z 2 z 1 Z z 2 g(z)dz ~ = b ~ g(z)dz = b ~ G(z 2 ) G(z 1 ) [G(z 2 ) G(z 1 )] = b zdg(z); (26) 1 G(z 2 ) z 1 z 2 where [G(z 2 ) G(z 1 )] represents the size of the innovation sector, which is then multiplied by the diaspora term b ~ to account for the total e ect of the latter on innovation in the home economy. Recall that an improvement in the IPR regime increases returns to skills (working in the innovation sector) by increasing wages! H. This results in an expansion of the innovation sector by reducing z 1 and increasing z 2. The RHS of Equation (26) reveals that protecting IPRs increases the number of workers in the innovation sector who can bene t from diaspora by enlarging [G(z 2 ) G(z 1 )]. 21 To determine whether the brain gain e ects caused by a diaspora channel could dominate the physical escape of skills caused by brain drain, we must calculate the net e ect of migration on total human capital in the sending country and test whether Z 1 Z 1 G(z 2 ) G(z 1 ) zg(z)dz b zdg(z)? 0 1 G(z 2 ) z 2 z 2 BD BG? 0 (27) Z 1 b G(z 2) G(z 1 ) 1 G(z 2 )? 1: As seen above, the term b G(z2) G(z1) 1 G(z 2) can take a value greater or less than one. Brain gains through diaspora dominate when > 1, which is more likely for high levels of IPR < 0 ) G 0 (z 1 ) > 1 < > 0 ) G 0 (z 2 ) > 2 > 0. As a result, IPRs indirectly promote brain gains by increasing the size of the innovation sector and the quality of diaspora, even if it could directly reduce average skills in the innovation sector (see Lemma 4). Proposition 4 Given Propositions 2 and 3 together with (27), gains from a diaspora could outweigh 20 Note that emigrants are excluded when summing up local skills in the South. 21 Note that in this set up emigration may also increase brain gain by enhancing the level of skills ~ that can be transferred back to the home country. This e ect is however unnecessary for our results and disappears if we take into account the number of emigrants, which implies multiplying the RHS of (26) by 1 G(z 2 ). 13

14 the direct loss of skills caused by migration if the IPR regime in the South is su ciently strong such that b G(z2) G(z1) 1 G(z 2) > 1 holds. This is so because diasporas from the North reach a larger number of workers who use their skills in the innovation > 0. Proof. See (27), Propositions 2-3, and Lemmas Summary and Main Empirical Implications In our theoretical model, we investigate under what circumstances skilled emigration may be bene- cial for development. We show that this occurs in the presence of a strong IPR regime, which may turn a brain drain into a brain gain. IPR protection in uences a country s potential for innovation by changing the average skill level and the size of the innovation sector. This could increase the absorptive capacity of the emigrants country of origin, thus leading to more bene cial e ects from cross-border diaspora networks. The mechanism at work is as follows. Emigration has two e ects. On the one hand, it decreases the average skills of the innovation sector ~z because the implied loss of the most skilled induces a lower z 2 (the extensive margin). On the other hand, it increases the skills of the remaining workers in the innovation sector because of the diaspora channel (the intensive margin). This latter e ect occurs through, ~ which enhances the skills of all remaining individuals in the innovation sector as long as b > 0. The IPR regime in turn in uences innovation by changing the size of the innovation sector. An increase in IPR protection enhances the attractiveness of working in the innovation sector, thus increasing its size from both ends of the spectrum: this causes a ow of low-skilled workers from the production to the innovation sector (i.e., z 1 falls) and reduces emigration (i.e., z 2 rises). Although the presence of the low skilled in the innovation sector may directly reduce e ective innovation activities (~z), the potential for absorption of the newly acquired skills from the North (b ) ~ is higher because the diaspora e ect in uences a larger range of workers (due to a larger H(z)). This line of reasoning leads to the conclusion that the gains in human capital from a diaspora are more likely to outweigh the direct drain of skills caused by emigration under a stronger IPR regime. The main testable implications of our model are the following: 1. IPR protection directly results in less e ective innovation activities in the South by misallocating the unskilled into the innovation sector. 2. Emigration directly harms the origin country s potential for development due to an out ow of skills, leading to brain drain. 3. In the presence of IPR protection, emigration could be bene cial for innovation as IPRs enhance gains from diaspora. When the level of IPR protection in the origin country is su ciently strong, it may help transform brain drain in a net brain gain. 14

15 4 Empirical Analysis 4.1 Data and Speci cation Our empirical analysis focuses on a sample composed of emerging and developing countries (EDC) as classi ed by IMF (2010). We select this sample because our theoretical model speci cally concentrates on the determinants of innovation in the South. The innovation measure we adopt is resident patent applications, i.e., the patent applications submitted by residents of each country to the local national patent o ce. 22 Additional tests will be performed using patent grants. Patent data are from the WIPO database. Our main migration measure is the gross ow of emigrants. An alternative migration measure is gross migrant stock, which we use in our robustness checks. The data on emigration ows and stocks are retrieved by aggregating original bilateral yearly data on immigrant ows and stocks by country of origin into 27 receiving OECD countries. 23 IPRs are measured through the Park (2008) index of IPR protection. The data represent an index of the strength of patent protection for each of the countries of the dataset. The index is the unweighted sum of ve separate scores for the following: coverage, membership in international treaties, duration of protection, enforcement mechanisms, and restrictions. For all details about our data and sources, see Appendix A.2. To investigate whether emigration is more likely to result in brain gains under stronger IPR regimes, we explicitly focus on the interrelationships between migration and IPR protection. To this end, we study the determinants of innovation with the help of an empirical speci cation that introduces the following key variables: migration, IPR protection and the interaction between the two. Our dataset is an unbalanced panel including 35 countries and covering the period from 1995 to The unit of analysis is a country-year. The countries in the sample have been chosen based on data availability and are listed in the Appendix A.1. More speci cally, all EDC countries with 3 or less observations on patent grants for the period from 1995 to 2006 were rst dropped from the sample. 24 This left us with an initial sample of 48 EDC countries, 35 of which have data available for all our key variables of interest: patents, migration ows and IPR protection. All estimations are performed using xed e ects regression methods and include time dummies. The choice of our estimation strategy takes into account both the characteristics of our sample and the speci city of the WIPO patent data at country level. While there are no countries with zero patents in our 22 For the bene ts of using patent statistics to measure innovation, see Griliches (1990). Along with input data such as research and development (R&D) expenditures and the human capital employed in research, patents have become the most common measure of innovation output (Hall et al. 2001) and of knowledge spillovers (Mancusi 2008). 23 Although for our analysis it would have been ideal to use data on skilled emigration, detailed statistics on the skill composition of emigrants by countries of origin are available only for the two most recent census years (1990 and 2000) or, at maximum, every 5 years (from 1975 till 2000), but only with reference to the six major receiving OECD countries. For details about emigration data by skill levels, see Beine et al. (2010) and Defoort (2008). 24 Selection was based on patent grants to follow a more restrictive criterion: the series on patent grants contains more missing values than that of patent applications for most countries. 15

16 sample (see Table 1), it may very well be that for very poor countries a missing data on patents represents a zero. Where the proportion of missing values is relevant, this could result in biased OLS estimates. In our dataset missing observations comprise only 15% of the initial sample in case of patent applications, and 22% for patent grants. 25 The baseline speci cation we adopt is the following: patents it = emigr it IP R it emigr it 5 IP R t 1 + +pop it + gdppc it 1 + i + t + " it ; where i denotes the country and t the year. The dependent variable patents t is our measure of innovation. The variable emigr t 5 represents emigration and is taken with a ve-year lag, to account for the time needed for the emigrants to acquire skills in the destination and interact to transfer the knowledge to their home countries. 26 IP R t 1 is the measure of IPR protection and is taken with a one-year lag to avoid endogeneity issues. The variable emigr t 5 IP R t 1 is the interaction term between emigration and IPR protection. The cumulative e ect of migration on innovation is then captured by 1 and 3 IP R t 1 and varies with the level of IPR protection. pop t and gdppc t 1 are population and GDP per capita, which are included to account for size e ects, with a lag in case of the GDP because of potential endogeneity issues. Finally, the i s are time-invariant country-speci c e ects, the t s are time dummies, and " it is the error term. Following the related literature, we complete the baseline speci cation in our empirical analysis by including a number of relevant controls. First of all, we add patent stock, which can be considered as a proxy for a country s absorptive capacity (Hall et al., 2001). 27 Its potential e ect on the amount of innovation activities is positive. 28 We also add R&D expenditure, which is considered a proxy for a country s potential for innovation and is expected to positively in uence innovation. A further relevant control is education, an additional proxy for the ability to absorb new knowledge. education measure we use is tertiary education, which we believe could best capture this ability. Government spending is next added to measure the degree of economic freedom. The Finally, trade and FDI are included in light of a rich literature on North-South trade and FDI as determinants of 25 More generally, most missing values on WIPO patents data at country level should not represent zeros and are continuously being estimated and updated by WIPO (WIPO, 2008). 26 In the case of patent grants, this also pertains to the time needed to create a patent. 27 To derive the patent stock series, we use the perpetual inventory method (Coe and Helpman, 2005). The patent stock (P S) of country i at time t is P S i;t = P S i;t 1 (1 d) + P i;t 1, where d is the depreciation rate and P is patent ow. The initial value of patent stock (i.e., at time t 0 ) is expressed as follows: P S i;t0 = P i;t0 =(g + d), where g is the average growth rate of patent ow (Griliches, 1980). We assume a depreciation rate of 15% (Hall et al., 2001) and take g as the average growth rate of patents in the rst decade of available and reliable data of the patent series, i.e., starting from year As speci ed in Appendix A.2, the patent series start from However, consistent and complete data are only available from the 1990s. 28 In line with Cohen and Levinthal (1990), absorptive capacity is the capacity to adopt new technologies and to create new inventions. Essential to this concept is the idea that the stock of knowledge accumulated through adoption or invention enhances the capacity to absorb external ideas and to create valuable inventions. In this sense, patent stock, which represents the stock of knowledge accumulated through inventions, can have a positive e ect on innovation. 16

17 innovation in low-income countries. In our empirical analysis, all additional controls except education are taken with a one-year lag to avoid potential endogeneity issues. For details on the sources of the control variables, see Appendix A Table 1 illustrates the summary statistics of the key variables of our analysis with reference to the time interval under consideration. [TABLE 1 ABOUT HERE] 4.2 Results Table 2 presents our results using resident patent applications as the dependent variable. The migration variable is gross emigration ows. We initially consider our baseline speci cation with all the three main variables of interest (migration, IPR protection, and the interaction between emigration and IPRs) and the two key controls for size e ects (population and GDP per capita). 30 [TABLE 2 ABOUT HERE] Column (1) n Table 2 presents our baseline speci cation and shows that our three main variables of interest are highly signi cant. In line with Proposition 2 and the vast literature discussed in the introduction, the negative and signi cant coe cient of emigration suggests that migration by itself could induce brain drain. 31 At the same time, the negative and signi cant e ect of IPRs con rms the theoretical prediction in Proposition 1, that IPRs in the South directly hampers e ective innovation activities by causing a misallocation of the less skilled workers to the innovation sector. This result resembles previous empirical ndings by Qian (2007) that nds IPR protection by itself does not stimulate domestic innovation in developing countries with low educational attainment. It is also in accordance with alternative hypotheses such as Madsen et al. (2010), who show imitation to be a much more important factor in such countries as means of gaining access to essential technological spillovers. Lerner (2009) also nds IPRs to increase foreign rather than domestic patenting in a country, which can result in the capture of national patent monopoly rights mainly by foreign rms (Lanjouw and Cockburn, 2001). 32 The key to our analysis is the interaction term between migration and IPR protection, which tests Proposition 3 and reveals to be highly signi cant and positive. This suggests that IPR protection helps the diaspora channel of knowledge that originates from migration. 29 In our empirical speci cations the following variables are taken in logs: patent grants, patent applications, patent stock (applications and grants), emigration ow, emigration stock, population, and GDP per capita. The rest of the variables (IPR protection, tertiary education, government spending, trade, FDI) are taken using their original values. 30 Since our sample begins in 1995, the rst observation of the lagged migration variable dates back to We are aware of the limitations of our (annual) data, which only allows to capture total migration from developing countries. However, the test in section together with the fact that migration to the OECD area in the 1990s has been increasingly composed of high-skilled immigrants from the South (Docquier and Rapoport, 2011) should reinforce the interpretation of our results and thereby help mitigate related concerns. 32 IPR protection also negatively a ects patenting by delaying spillovers in sequential innovation (Scotchmer and Green, 2000), creating wasteful attempts to invent around the patent (Ja e and Lerner, 2004), and promoting costly disputes and excessive litigation (Bessen and Meurer, 2009). 17

18 It also implies that above a certain threshold level of IPR protection migration induces brain gain, mirroring the conclusion we derive in our theoretical model. 33 In column (1) of Table 2 our two size controls (population and GDP per capita) are positive and signi cant, as expected. Columns (2) to (7) in turn add to the baseline speci cation our additional controls: patent stock, R&D expenditure, education, government spending, trade and FDI. The results show that patent stock and R&D expenditure are positive and signi cant determinants of innovation. The positive sign of patent stock suggests that innovation is stronger in the presence of a higher level of absorptive capacity; this implicitly con rms that the diaspora channel of knowledge is e ective when the ability to absorb new knowledge is high. The positive sign of R&D is intuitive and follows the main predictions of the relevant literature: the more e orts are devoted to R&D, the greater is a country s potential for innovation. Tertiary education appears to be insigni cant in this framework; its negative sign could be due to the fact that highly educated people in developing countries may prefer to apply for patents in more advanced economies. 34 In this speci cation, both trade and FDI are positive as expected but not signi cant. 35 Finally, column (8) simultaneously adds all controls into the speci cation. 36 As the results show, the coe cients of our three main variables of interest remain signi cant and of the same sign as in the baseline speci cation: migration is negative and signi cant, IPR protection is negative and signi cant and the interaction term between migration and IPR protection is positive and signi cant Robustness Tests In Table 3, we perform two di erent robustness checks for our results. The dependent variable is again patent applications. To further investigate the issue of omitted variables, we rst estimate our key speci cations in rst di erences, thus choosing a di erent estimation methodology with respect to xed e ects. While the xed e ects (within) estimator is derived by subtracting the time-average model from the original model, the rst di erence estimator is obtained by subtracting the model lagged by one period from the original model. In other words, the rst di erence model removes the time-invariant individual components by rst-di erencing the data. The relative e ciency of the rst di erence estimator with respect to the xed e ect estimator depends on the properties of the error term. In particular, the rst di erence estimator requires weaker exogeneity assumptions, and 33 Below we will perform a speci c test for this e ect. 34 Although tertiary education here has a non-signi cant coe cient, we also nd that primary education positively and signi cantly a ects the number of patents granted. The results are available upon request. We do not present the results for primary education because we believe tertiary education is more relevant in a study of the determinants of innovation. Moreover, it is worth pointing out that even after controlling for the extent of human capital the e ects of our three main variables of interest remain signi cant. 35 This suggests that international technology transfer is not necessarily due to trade and FDI and at the same time reinforces our view that migration plays an important role in innovation. 36 Here, we exclude R&D expenditure because there is a large number of missing data for this variable, and consequently including R&D reduces our sample to a large extent (i.e., the initial sample loses 200 observations). The results with all controls including R&D are in line with the other results and are available upon request. 37 To determine whether the e ect of emigration on innovation is due to capital independently from knowledge transfers, we also investigated the role of remittances. In our speci cations, remittances are not signi cant. 18