Innovation, intellectual property rights and
technological diffusion: A cross-sectional study of
developed and developing countries
University of Groningen Faculty of Economics and Business
Master Thesis International Economics and Business
Name Student: Wout Bobbink Student ID number: 1903128
Student email: w.bobbink@student.rug.nl Date: 9th January 2015
ABSTRACT
This paper examines the role of IPRs and international technological diffusion in determining a country’s rate of domestic innovation, including innovation new to the world and innovation of imitative nature new to countries. The empirical analysis employs a cross-sectional pooled dataset including 33 developing and 40 developed countries for the 2007 – 2012 time period in which strong global uniform IPRs are effective in law due to the TRIPs agreement. Results suggest a negative relationship exists between IPR protection and domestic innovation. The reductive effect is larger for technologically advanced countries. Technological diffusion does not materialize in terms of more innovation.
1. INTRODUCTION
At the end of the 1980s, developed nations started to integrate the issue of intellectual property rights (henceforth: IPRs) with discussions on free trade. During the Uruguay Round negotiations that took place between 1986 and 1994, negotiations on free trade were complemented with negotiations on IPRs. As a consequence, the Uruguay Round not just led to the creation of the World Trade Organization (WTO), but also to the signing of the Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPs). Global trade liberalization and minimum levels of intellectual property protection became top issues on the world economic agenda. With the TRIPs agreement becoming fully effective in 2006 (except for a few specific sectors), relatively strong worldwide harmonized IPRs seem to be a de facto situation, even for least developed countries (Verspagen, 1999).
From these developments it appears that strengthening IPR protection on a worldwide scale is assumed to be a logical step towards economic development. The rationale behind this is that IPR protection has an impact on technological change, which is deemed a main driver of economic growth in endogenous growth models (Rushin & Thompson, 1996). Technological change can be achieved by innovation (technology new to the world) or by imitation (technology new to a country) (Wu, 2009). Previous literature, both theoretical and empirical in nature, points out that increasing IPR protection levels can enhance domestic innovation in three ways (e.g. Chen & Puttitanun, 2005; Schneider, 2005). Firstly, stronger IPR protection creates a direct incentive for innovation as it reduces the profit loss of the IPR holder by granting him/her a temporary monopoly. Secondly, as stronger IPR protection enhances innovation, domestic spillover effects might occur (Verspagen, 1999): domestic firms can imitate domestic innovators leading to more innovation of imitative nature. Thirdly, stronger IPR protection can increase imitative innovation by technological diffusion from developed to developing countries resulting from international trade: it encourages international transactions such as imports or foreign direct investment (FDI), channels that can embody foreign technological knowledge. This is one of the main motivations behind the TRIPs agreement (Verspagen, 1999). As firms from developed countries run less risk of their technological knowledge being copied by firms from developing countries, technological development levels in less developed countries can converge towards technological development levels in developed countries (Verspagen, 1999; Schneider, 2005).
innovation only new to the country. Especially imitation resulting from trade is an important driver of technological change and thereby growth in less developed countries as they can hardly contribute to innovation that enhances global technological levels (Wu, 2009). In other words: different needs for technological progress (innovation or imitation) exist at different stages of economic development (Chen & Puttitanun, 2005; Wu, 2009).
This is only one side of the story though. Convincing literature exists that shows IPR protection can also negatively affect innovation. Firstly, too strong IPR protection can reduce domestic technological diffusion and thereby domestic innovation as it leaves no legal space for other domestic economic agents to explore and exploit protected technological knowledge of domestic innovators (Verspagen, 1999). Secondly, as developing countries’ innovative activities might be of more imitative nature, enhancing IPR protection in developing countries might suppress international technological diffusion resulting from trade. Due to legal restrictions, developing countries are not able to benefit from foreign technology embodied in imports or FDI and thereby possibilities for innovation that are new to the developing country reduce (Schneider, 2005; Wu, 2009). Finally, in the presence of international trade strong international IPR protection can negatively affect innovation in developing countries, but also in developed countries through a feedback effect: if technological spillovers from a developed country to a developing country cannot take place, this will strike back economically to the developed country – global strong uniform IPRs decrease imitation and thereby competitiveness of intermediate goods in the developing countries, meaning developed countries have to buy more costly intermediates from the developed world, leaving less resources available to invest in domestic innovation. (Azevedo, Afonso, & Silva, 2014).
These latter two observations imply the importance of distinguishing between developed and developing countries in examining the relationship between IPR protection and innovation. Not doing so might lead to erroneous results and policy implications (Schneider, 2005). Stronger IPR protection in less developed countries might increase innovation by technological transfers from technologically advanced countries through international trade, but whether or not technological diffusion materializes into more domestic innovation depends on a country’s absorption capacity for technological knowledge (Verspagen, 1999).
this study addresses the questions whether IPR protection is indeed enhancing innovation and if the relationship between IPR protection and innovation is conditional on a country’s technological development level. Whilst answering these questions, the role that technological diffusion plays through international trade is thoroughly examined as well. Not only does this paper contribute to previous literature by empirically investigating the relationship between IPR protection and innovation in a time period in which a minimum level of globally uniform IPRs have become effective due to the TRIPs agreement, it also adds value by introducing a new measure of IPR protection to the academic field that attempts to take into account actual enforcement of IPRs next to the presence of effective IPRs, as the mere existence of IPRs does not mean anything if they are not enforced properly.
The empirical analysis is performed using a personally composed cross-section dataset of 40 developed and 33 developing countries pooled into one sample, in which the country observations take a value representing an average year in the time period from 2007 to 2012. Patent data is used as a proxy for domestic innovation. This measure includes both regular and utility patents. To obtain a utility patent a technology is not required to have an inventive step. Moreover, only slight technological novelty is necessary for the territory in which the patent will become effective (WIPO, 2014). Regular patents increase innovation in developed countries and utility patents increase innovation in developing countries (Choo, Kim, Lee, & Park, 2012). Therefore, this measure includes both innovation which is new to the world and innovation of imitative nature arising from technological diffusion through trade, new to the country.
conferences, which might be more important spillover channels for developed countries than developing countries.
As this study reveals a negative relationship between IPR protection and innovation in a ‘post-TRIPs’ time period, economic policy makers might want to rethink whether strong global uniform IPR protection imposed upon the world is the appropriate way to stimulate technological progress and thereby economic growth.
This paper is organized as follows. Section 2 briefly describes both theoretical and empirical literature on private property rights and extensively reviews theoretical and empirical literature on the interrelated connections amongst IPR protection, innovation and technological diffusion. Section 3 discusses the data collection and methodology implemented in the study. Section 4 describes and discusses the empirical results. Finally, section 5 concludes the paper and opens perspectives of future work.
2. LITERATURE REVIEW 2.1 Private property rights
To get a complete overview of the relationship between IPRs and innovation, this section starts with an analysis of the economic importance of private property rights in general. Subsequently, IPRs are discussed.
Theoretical background
Besley and Ghatak (2010) explain how four aspects of private property can influence economic activity by means of an illustrative analytical framework capturing previous empirical and theoretical work. Based on Barzel (1997) they define property rights as an individual’s ability to consume a good (service or asset) directly, to consume it indirectly through exchange, or to use it as a means of income generation (Besley & Ghatak, 2010).
The first factor Besley and Ghatak (2005) identify is expropriation risk. When insecure property rights exist an increase in the risk of expropriation leads to a decrease in investment in terms of labor supply, and reduces outputs and profits: weak property rights reduce the opportunity for individuals to gain full benefit from their production. Without secure property rights human capital is not allocated to the most productive activities, which hinders innovation, hence economic growth (Torstensson, 1994).
over producing and guard labor. From their analysis they conclude that on the one hand increasing the use of guard labor can lead to an increase in productive labor in terms of single worker output, when the insecurely protected asset is involved in the production process (e.g. as in agricultural land). On the other hand, when the insecurely protected asset is not involved in the production process (e.g. as in residential property), an increase in guard labor can lead to either a decrease or increase in productive labor.
A third aspect of how property rights influence economic activity is by their effect on trade. Property rights facilitate the mobility of an asset, which is desirable in the sense that the total economy becomes more productive when assets are available to be used by the ones able of using these assets most productively. However, if property rights’ strength does not compensate for the risk of expropriation, the owner of an asset will not divert resources to someone who can use it more productively. Torstensson (1994) adds to this a distributional effect of private property rights. Given the existence of private property, investment is allocated to sectors with the highest private returns. When property is state-owned and property rights are ineffectively enforced, the absence of competitive markets makes that the sectors invested in are not per definition the sectors with the highest rates of return. This means investments are allocated inefficiently, which distorts economic growth.
The fourth factor identified by Besley and Ghatak (2005) is the influence of the relationship between property rights and collateralizability of assets on economic activity. A strong property rights system facilitates the ability of borrowers to use assets as collateral, which again enhances a more efficient distribution of assets.
Empirical studies
There are several empirical studies that stress the importance of private property rights by examining its relationship with economic growth using cross-section data. Johnson, McMillan and Woodruff (2002) investigate the relationship between property rights security and investment by means of a micro approach using firm data. They find that firms in five different Eastern-European transition countries in the year of 1997 are less willing to reinvest profits when property rights are perceived as not secure.
Besley and Ghatak’s (2005) results for 172 countries in the year of 2000 show that countries with a higher risk of expropriation (first proxy for property rights quality) have lower levels of GDP per capita; and that in countries where it is more difficult to register property (second proxy for property rights quality) also lower levels of GDP per capita exist.
In a cross-sectional study by Torstensson (1994) amongst 68 developed and developing countries in which the variables he uses capture growth ratios between the years of 1976 and 1985, he concludes that a significant negative relationship exists between arbitrary seizure of property and economic growth due to inefficient allocation of investment funds and inefficient use of available human capital in case of weakly secured private property. However, he is not able to empirically relate state ownership of property and economic growth.
In another empirical analysis by Acemoglu and Johnson (2003) the findings encompass that property rights institutions (institutions protecting against expropriation) affect long-run economic growth, investment and financial development. Contracting institutions (institutions enabling private contract formation) seem to affect financial intermediation, but only have limited effects on growth and investment. In their study, they use a sample of countries representing the whole world in the year of 1990. They employ settler mortality and population density before colonization as instruments for property rights institutions; colonial power identity is used as instrument for contracting instruments, as it represents a colony’s legal origin (common or civil law) and exogenously determines colonial legal systems, according to Acemoglu’s and Johnson’s (2003) reasoning.
2.2 Intellectual property rights
IPRs are designed to tackle the issue arising from goods that embody technological knowledge. A normal economic good (e.g. an orange) is excludable and rival. A supplier can exclude persons from consuming the orange (those who are not willing to pay for it) and if the good is supplied, only one person can use or consume the orange (rivalry). Technological knowledge is a good that lacks these two characteristics. If the specific knowledge is not protected by IPRs the private agent can neither exclude others from using the knowledge he or she has developed, nor can he/she avoid that, even though the good embodying the knowledge is consumed, the knowledge can still be used by others at the same time (non-rivalry) (Verspagen, 1999). In this situation private economic agents have no incentive to invest in technological knowledge as competitive agents can free-ride on the efforts of the inventor and, assuming imitation is less costly than innovation, offer the product on the market for a lower price than the original inventor. Without IPRs market failure occurs: the market will not produce knowledge goods or only in amounts too low to meet actual demand. With IPRs this market failure can be solved (Verspagen, 1999).
Different types of IPRs exist. Trademarks protect brands and therefore marketing activities. Copyrights provide exclusive rights of usage and distribution for the creator of an original creative work, such as books or films. Patents provide an inventor or a firm a temporary monopoly in using specific technological knowledge, whether or not embodied in a specific good. For this study, the former two are of less interest as they induce less spillovers and thus have less consequences for the growth potential of an economy (Verspagen, 1999). Intellectual property law needs to find a balance between private gain and public welfare (Gould & Gruben, 1996). Length (duration of monopoly right) and breadth (scope of protection) of patents are important determinants of this balance. A patent that is in force for a longer period of time increases profits for the inventing/innovating company, but also increases welfare losses for the consumer due the firm’s monopolistic position making it able to supply the good at prices higher than as in a situation of perfect competition. The same applies to patents with a broad scope (covering the complete scope of the technology). Moreover, patent breadth defines the space for spillovers. If certain aspects of the protected technological knowledge are not protected, others might be able of adopting and exploiting these aspects (Verspagen, 1999). Therefore, both length and breadth of a patent impact economic development and determining an optimum is far from straightforward.
Notwithstanding the type of relationship that is found, two topics concerning IPRs are recurrently discussed in previous literature: 1) the role of IPRs in explaining innovative and imitative activities and 2) the relationship between country development levels and optimal IPR protection levels.
Theoretical background
The neoclassical growth model originally proposed by Solow (1956) describes economic growth as the function of the inputs of capital, labor and total factor productivity: the latter being the residual comprising anything but accumulation of capital and labor. Solow assumes constant returns to scale and diminishing returns to capital. The sole way to escape diminishing returns is by improving disembodied technology, which shifts the production function. The main policy implication is that economic growth is achieved by investing heavily in capital and labor and hoping that for some reason technology improves to avoid diminishing returns.
To explain disparities in economic growth between countries it is therefore crucial to understand technological change. A main shortcoming of neoclassical growth models is that technological change is captured by the residual. Therefore, endogenous growth models were developed, recognizing that technological change results in profit opportunities which in turn can lead to the prevention of stagnation of economic growth (Rushin & Thompson, 1996). In an explanation of how the Asian NICS experienced such rapid growth compared to other developing economies, Nelson and Pack (2001) consider technological change as crucial and define adoption of new technology as a process of learning the tacit knowledge to employ a technology that requires new skills, novel ways of organizing economic activity and creating familiarity with new markets. Entrepreneurs need to take the risk of committing resources herein, but eventually this learning process rises productivity levels of the modern sector at the expense of less productive traditional sectors, increasing national productivity.
technology, whereas innovation tries to improve productivity based on the local technological level.” (p. 169). For a country with a large distance from the world technology frontier imitation is the most important source for productivity growth and for convergence towards the world technology frontier. For a country close to the world technology frontier innovation is the most important source for productivity growth. Both activities are costly, but are also affected by how employed labor is distributed over both alternatives. If employed labor is used for imitation (innovation), this lowers the costs of imitation (innovation) relative to innovation (imitation).
Wu (2009) assumes that a higher (lower) degree of IPR protection implies more (less) difficulty in copying the world technology frontier making imitation more (less) costly. A higher (lower) degree of IPR protection implies a reduction (increase) in innovator’s loss of profit due to less (more) competition from imitators. Strengthening IPR protection creates a direct incentive for innovation enhancing the world technology frontier and prevents imitation leading to innovation that is new to the country. The key assumption Wu (2009) makes is that the more a country approaches the world technology frontier, the more important this former type of innovation gets as a way of productivity enhancement (and profit increases) and the more likely entrepreneurs will switch from allocating labor from imitative to innovative activities. Based on this assumption two growth paths exist for an economy. In the first growth path an economy starts with high effort on imitation and when a specific technological level is reached it will switch to high effort on innovation and converge to world technology. In the second growth path an economy starts with low effort on imitation and when a specific technological level is reached it will switch to high effort on innovation and converge to world technology.
It can be concluded from Wu’s (2009) study that the level of development of an economy matters in avoiding traps where economic growth stagnates and consequentially, stops converging towards the world technology frontier. To avoid these traps countries need to switch to a strategy of high effort on innovation. IPRs play a very important role, because they affect the moment where a country will switch from imitative to innovative activities. If IPR protection strengthens, a country will switch to innovative activities earlier as it becomes more profitable to innovate than to imitate. IPR protection is positively affecting innovation, which can prevent stagnation in economic growth (Wu, 2009).
Countries will choose an optimal level of IPRs, depending on their level of economic development (Chen & Puttitanun, 2005). Just as Wu (2009) Chen and Puttitanun (2005) hace created a theoretical model, which recognizes that different optimal levels of IPR protection exist for different levels of development. They measure output of the model in welfare (consumer and social surplus) and apply the model to developing countries only. Their model consists of a local sector and an import sector. The former consists of two domestic firms of which one has the potential to develop new patented technology, while the other one can merely imitate technology. The latter includes a foreign firm with patented technology and a domestic firm with technology of lower quality than the foreign firm. However, this domestic firm can increase quality by imitating the foreign firm’s technology The model has two main results, which are both in line with Wu’s (2009) findings.
Firstly, increasing IPR protection has a positive impact on innovation. Increased IPR protection makes imitation more difficult in both sectors. In the local sector the firm that has the potential to develop technology new to the world will focus more on innovation if imitation is more difficult, which increases the economy’s surplus. In the import sector the domestic firm is less able of imitating the foreign firm’s patented technology. This implies a lower quality of domestic technology, less competition for the foreign firm and higher prices for the foreign firm’s technology. This decreases the economy’s surplus. Optimal levels of IPR protection enhances innovation and balances the trade-off between the increasing and decreasing surplus associated with more innovation (Chen & Puttitanun, 2005).
level exceeds a certain technological threshold, domestic innovation becomes more efficient, which makes it desirable to increase IPR protection with the level of development. This technological threshold is the minimum value in the U-shaped relationship – the moment for a developing country to strengthen IPR protection and switch from imitative to innovative strategies to prevent from ending up in the middle-income or poverty trap (Wu, 2009).
As in Wu (2009), Chen and Puttitanun (2005) report a positive relationship between IPR protection and innovation of the type that introduces innovation new to the world, but a negative relationship between IPR protection and innovation of imitative nature new to the country. However, as explained by Verspagen (1999), two mechanisms of technological diffusion exist through which enhanced IPR protection can also increase innovation of imitative nature, meaning that innovation new for a country can also be encouraged by stronger IPR protection.
The first mechanism is domestic technological diffusion: domestic firms can imitate technological knowledge developed by a domestic innovator. The protected knowledge can provide new ideas for innovations as it increases the knowledge stock of a country. In other words: domestic innovations embodying protected knowledge create positive externalities that increase domestic innovation even more. With more innovations come more externalities. So as innovations increase with IPR protection, the stronger the level of IPR protection, the more spillovers might occur. On the other hand, stronger IPR protection, for instance in the form of broader patents, can also counteract as it leaves less space for other agents to exploit parts of the patented technology in which they see other/new economic purposes (Verspagen, 1999).
The second mechanism is international technological diffusion. Imports of capital goods or FDI can embody technological knowledge, making domestic firms able of imitating foreign innovators. Weak domestic IPR protection might discourage foreign companies of exporting capital goods that embody technological knowledge they developed in the foreign country, due to fear of imitation by domestic competitors. Stronger protection might thus be required to attract imports embodying technological knowledge developed abroad. The same applies to international transactions involving FDI. International technological diffusion is one of the main motivations behind the signing of the TRIPs agreement, as it can enhance technological levels in less developed countries (Verspagen, 1999).
technological change (as in Chen and Puttitanun (2005) and Wu (2009)), but also foreign country technological change. In accordance with Chen and Puttitanun (2005) and Wu (2009), Azevedo et al. (2014) consider that different levels of country development require different optimal levels of IPR protection.
Azevedo et al. (2014) examine optimal international IPR enforcement in a North-South open-economy growth model in which a final good sector, an intermediate good sector and an R&D sector exist in both economies. Intermediate goods embody technological knowledge and are used as inputs for the final goods sector. Innovation and imitation (technological diffusion) in the R&D sector increase quality of the intermediate goods. These are used as inputs for the final goods sector. Output of the final goods sector determines economic growth, so economic development is set by innovation and technological diffusion within the R&D sector.
The model of Azevedo et al. (2014) differs from the models by Chen and Puttitanun (2005) and Wu (2009). It measures the effect of IPR protection on technological development and thereby economic growth, solely through its effect on imitation (technological diffusion), instead of through its effect on both imitation and innovation: Azevedo et al. (2014) assume IPR protection is aimed at preventing imitation. They conclude that when trade between North and South occurs, increases in IPR protection make technological diffusion in the South more costly, which decreases spillover effects in the South, which in turn decreases innovation in the North through a feedback effect: in the absence of technological diffusion in the South caused by too strong IPR protection, the Southern intermediate goods sector lacks competitiveness to compete against the Northern intermediate goods sector, meaning that the North will obtain its inputs for its final goods sector from the Northern intermediate goods sector, which is more costly than obtaining inputs from the Southern intermediate goods sector. Therefore, the North has less resources available to invest in R&D, decreasing the probability of innovation in the North. In other words: increasing international IPR protection levels decreases innovation new to the South (imitation) and through the feedback effect also reduces innovation in the North (innovation new to the world). This seems to undermine the rationale behind the TRIPs agreement.
So far, previous theory points out that IPR protection can enhance domestic innovation in three ways. Firstly, it creates a direct incentive to innovate as the innovator can collect profits that he/she might not have been able of collecting if the innovation would have been unprotected due to competition by imitators (Wu, 2009). This concerns the more radical type of innovation that is new to the world. Secondly, IPR protection can increase domestic innovation as it can enhance domestic technological diffusion: technological knowledge embodied in innovations increases the general knowledge stock of a country and leads to positive externalities that can increase innovation even more (Verspagen, 1999). Finally, IPR protection can enhance domestic innovation as it encourages international transactions that can embody technological knowledge, for instance through imports and FDI (Verspagen, 1999). These latter two types of innovation are of imitative nature and mainly lead to innovation new to the country. As IPR protection can enhance domestic innovation both new to the country and new to the world, hypothesis 1 can be formulated as follows:
H1: IPR protection is positively correlated with domestic innovation.
However, IPR protection can also reduce domestic innovation by suppressing innovation of imitative nature arising from technological diffusion if IPR protection is too strong to produce externalities. It can even negatively affect foreign innovation due to a feedback effect (Azevedo et al., 2014). Moreover, the relationship between IPR protection and domestic innovation is dependent on a country’s technological abilities. Despite the presence of channels for technological diffusion, stronger IPR protection might not materialize into enhanced innovation as less developed countries lack the required resources and human capital to contribute to the advancement of the world technology frontier (Verspagen, 1999): different needs for technological progress exist at different stages of economic development (Chen & Puttitanun, 2005; Wu, 2009). Therefore, hypothesis 2 of this study is defined as:
H.2: The effect of IPR protection on domestic innovation is conditional on a country’s technological development level.
The following section examines the empirical findings of previous studies that analyze these theories.
Empirical studies
intellectual property protection (measured as a self-designed IPR protection index (Ginarte & Park, 1997b)) affects economic growth indirectly by making more investment activities (R&D) possible, since long-term growth is stimulated by investments in tangible and intangible capital. Ginarte and Park (1997a) conclude that IPR protection does not matter for R&D activities in less developed countries, in which R&D activities are replaced by imitation.
Rushin and Thompson (1996) draw a similar conclusion. By means of a panel dataset including 112 countries during the period of 1970-1985 they provide evidence of a positive relationship between IPR protection measured as patent protection and economic growth, but only for countries with an initial GDP per capita of at least 3,400 US 1980 dollars. This again confirms that less developed countries benefit less from enhanced IPR protection, due to their focus on imitation. Employing a panel data set of 47 developed and developing countries from 1970 – 1990, Schneider (2005) emphasizes this finding. She uses Ginarte and Park’s (1997b) IPR protection index as explanatory variable and finds that IPRs ‘have a stronger impact on domestic innovation for developed countries and might even negatively impact innovation in developing countries’ (p. 453). Chen and Puttitanun (2005) complement their theoretical model with an empirical analysis and conclude that even within developing countries IPR protection can have different effects on domestic innovation. Their results show that IPR protection, measured as the Ginarte and Park (1997b) index, has a positive impact on domestic innovation (which confirms hypothesis 1, but is at odds with Schneider’s (2005) result), measured as R&D expenditures and as number of patent applications in developing countries. More importantly, they empirically underpin the U-shaped relationship between IPR protection and level of technological development. After a threshold of a GDP per capita of US$854 (in 1995 prices) IPR protection starts increasing, meaning that innovative activities dominate for more developed countries. Chen and Puttitanun’s (2005) and Schneider’s (2005) empirical findings are in line with hypothesis 2. Less developed countries might want to implement weaker IPR protection to encourage imitation and achieve economic growth.
and a higher degree of domestic IPR protection actually enhances imports to developing countries (Verspagen, 1999), this is at odds with theory by Chen and Puttitanun (2005) and Wu (2009), which explains that developing countries with a low level of technological development first need low IPR protection levels to encourage imitative activities.
That the strength of IPR protection is related to international trade is an important observation for this study, because Wang and Xu (1999), using panel data for 21 OECD countries for the period 1983 – 1990 and Schneider (2005), employing a panel data set of 47 developed and developing countries from 1970 – 1990, show that capital goods imports and high-technology imports, respectively, can enhance domestic innovation through technological diffusion in both developed and developing countries. On the other hand, for a sample of 14 OECD countries for the time period 1975 – 1995, Fagerberg and Verspagen (1998) find that technological diffusion embodied in trade is only a minor contributor to productivity growth in the importing country. Disembodied spillovers (spillovers through channels such as labor mobility, international contacts at conferences, scientific literature, patent specifications) have a much stronger impact on productivity growth.
International trade is not the only channel through which embodied technological diffusion might occur. An alternative could be FDI inflows. Just as imports of capital goods, FDI could lead to technological transfers from foreign to domestic firms (Verspagen, 1999). Empirical findings on the effect of FDI inflows on domestic innovation are inconclusive (Schneider, 2005). Either way, to attract FDI inflows developing countries might want to consider to strengthen IPR protection (Verspagen, 1999).
All in all, empirical research seems to emphasize the importance of IPR protection as it has an influence on economic growth. Moreover, it seems to confirm that the relationship between IPR protection and domestic innovation can take various forms as the effect of IPR protection on domestic innovation appears to be depending on the technological development level of the domestic country. Also, it highlights that international transactions can indeed contribute to domestic innovation due to technological diffusion.
protection, domestic innovation and technological diffusion by means of an empirical analysis using hypotheses 1 and 2 as starting point.
3. METHODOLOGY
Schneider’s (2005) innovation equation includes all relevant aspects of domestic innovation, IPR protection and technological diffusion as described in the previous section. By means of this equation she explains the level of domestic innovation using IPR protection and the channels for technological diffusion high-tech imports and FDI as explanatory variables. She controls for technological development levels by including human capital and infrastructure level. Hypothesis 1 and 2 are tested using her equation as an example, albeit with a few adaptations: as explained further on she also controls for R&D expenditures and market size, controls that are not included in this specific study.
The empirical analysis of this study employs a cross section data set including both developed and developing countries for the years of 2007-2012 with six-year averages of annual observations where data availability allows for this. This time frame allows to examine IPR protection and innovation in a world with fairly strong globally harmonized IPRs. Moreover, it allows to use a measure for IPR protection that has the benefit that it attempts to include IPR enforceability as well. The innovation equation is estimated using ordinary least squares (henceforth: OLS).
3.1 Innovation regression
To explain domestic innovation Schneider’s (2005) innovation equation is used as an exemplary equation. In this equation the variable to be explained is the level of domestic innovation. The main explanatory variable is the strength of the IPR regime. First of all, as formulated by hypothesis 1, stronger IPR protection increases domestic innovation as it provides a direct incentive to innovate. Thus, IPR protection is included in the equation explaining domestic innovation. The relationship between IPR protection and domestic innovation is expected to be positive.
stronger positive correlation between human capital stock measured as mean years of secondary education and domestic innovation for developed countries than for developing countries. If a country does not have enough human capital available, innovation might not occur, even though the property right system is strong enough to encourage it (Verspagen, 1999). Therefore, human capital is used as measure for a country’s technological development level and included in the innovation regression as part of an interaction term with the IPR variable. Marginal effects of IPR protection on domestic innovation for all countries’ human capital scores (technological development levels) are calculated to test hypothesis 2. To allow for the estimation of the marginal effects of IPR protection on domestic innovation given specific country development levels both the constitutive terms (the elements constituting the interaction term) need to be included autonomously in the regression as well (Brambor, Clark, & Golder, 2005). Before adding the interaction term to the regression, the human capital variable can also be interpreted as in Schneider (2005), to check if there is indeed a positive correlation between human capital and domestic innovation in both developed and developing countries. As soon as the interaction term is added to the regression, the coefficients of the IPR protection and human capital variables cannot be interpreted as autonomous estimators of domestic innovation effects anymore, but are used to calculate marginal effects.
Next to IPR protection and human capital, Schneider (2005) defines two economic transactions that could enhance local technological diffusion and increase domestic innovation. Firstly, she finds that high-tech imports from developed countries are strongly predicting domestic innovation in both developed and developing countries. Keller (1999) stresses that countries importing primarily from technological leaders receive more technology embodied in intermediate goods than countries that import from non-technological leaders. Therefore, it can be expected that high-tech imports from developed countries are positively correlated with domestic innovation.
Foreign technologies embodied in capital goods imports or FDI are not all relevant for developing countries (Verspagen, 2005). If the developing country lacks the infrastructure required for the imported technology, technological diffusion is irrelevant or restricted. To control for this effect, infrastructure is added to the regression, just as in Schneider (2005). A positive correlation between the level of infrastructure on domestic innovation can be expected. Schneider (2005) uses two variables in her regression that are not part of the regression in this study. Firstly, she includes R&D as a predictor variable. R&D is supposed to have a positive influence on domestic innovation. However, Maggio and Sweet (2015) reject R&D expenditures as an effective indicator of innovation, for reasons that also undermine its usability as an explanatory variable for innovation. The most important reason is that data on R&D expenditures is scattered and inconsistent for developing countries. The collection of R&D data in developing countries is far from reliable due to institutional challenges in finding standardized data collection systems. During the data collection process for this study, it turned out data on R&D in developing countries is hardly available. A possible explanation for this could simply be that these countries lack sufficient resources to invest in research (Verspagen, 1999). Moreover, as assimilation of technology does not materialize in terms of innovation anyway if there is a lack of sufficient quality in local human capital (Verspagen, 1999), human capital might be more relevant than R&D expenditures in explaining domestic innovation, especially for developing countries. Therefore, R&D expenditures are not included in this model as an explanatory variable.
In addition, Schneider (2005) recognizes that the size of a country’s market determines profitability of innovation. For this reason, she includes market size measured as GDP per capita in the innovation regression. Her explanation for this, based on Furman, Porter, and Stern (2000) is that market size can be used as a proxy for a country’s knowledge stock. As a country’s knowledge stock is already explained by the human capital variable, adding GDP per capita to the regression is unnecessary and might cause issues of multicollinearity. Hence, GDP per capita not included in this study.
In short, the innovation equation can be defined as:
Ii = β0 + β1IPRi + β2HCi + β3 (IPR*HC)i +β4IMPi + β5FDIi + β6INFi + μi, (1)
in which dependent variable Ii is the (natural log of the) domestic rate of innovation in country
i; independent variable IPRiis the strength and enforcement level of IPR protection in country
i; independent variable HCi is the level of human capital in country i; independent variable
and human capital (technological development level) in country i; independent variable IMPi is
the natural log of the import level of high-technology goods from developed countries into country i; independent variable FDIi is the (natural log of the) inflow of foreign direct
investment into country i;; independent variable INFi is (the natural log of) country’s i
infrastructure level; and β0 and μiare the intercept and error term, respectively.
Hypothesis 1 can be tested by observing the coefficient for the IPR protection variable and its related significance level after running the regression. If the coefficient is positively and significantly correlated with domestic innovation, it can be rejected that there is no relationship or a negative relationship between IPR protection and domestic innovation and there is sufficient evidence to conclude that a statistically significant positive relationship exists between IPR protection and domestic innovation.
To test hypothesis 2 some additional calculations are required. As recognized by Brambor et al. (2005), after running the regression, merely checking whether or not the coefficient of the interaction term between IPR and HC (β3) is positive and significant does not
provide sufficient evidence that the effect of IPR protection on domestic innovation is conditional on country development level, which should hold for hypothesis 2 to be accepted. Marginal effects of IPR protection on domestic innovation for the country values of technological development (human capital) can be insignificant even if β3 is significant, or they
can be significant for specific values of country development even if β3 is insignificant. Of
primary interest are the marginal effects of IPR protection on domestic innovation for values of the conditioning variable human capital. These are calculated after running the multivariate regression. If a marginal effect of IPR protection for a specific value of human capital is significantly correlated with domestic innovation it can be rejected that there is no marginal effect of IPR protection on domestic innovation for that specific value of human capital (technological development level) and there is sufficient evidence to conclude that a statistically significant marginal effect of IPR protection on domestic innovation exists for that specific value of human capital. Following theory by Verspagen (1999) and empirical results by Schneider (2005), the marginal effects of IPR protection on domestic innovation might be lower for developing countries than for developed countries, or might even be negative for developing countries.
3.2 Data
the World Bank (2014a)) and is presented in Table 1, including the averages for all variables included in equation (1). The time frame for this study are the years from 2007 till 2012 and the OLS regression is performed on a cross section of country observations based on six-year averages of annual observations during these six years.
Cross-sectional observations are used because of the nature of IPR protection. The goal of this study is not to detect effects of developments of the independent variables on developments in innovation, but to compare the magnitude of the relationship between several possible factors (including IPR protection) and domestic innovation amongst different countries. Six-years averages between 2007 and 2012 are used as differences amongst countries might exist in years in which a possible impact of the TRIPs agreement on innovation is actually measured by the variable representing IPR protection. IPR protection will likely not be enforced as strongly for all countries in the sample at one specific moment in time, notwithstanding the existence of the TRIPs agreement.
Data collection
First of all, the dependent variable rate of domestic innovation (I) in equation (1) is measured as the natural log of the total count of national and international patents applied for registered by the patent registration office in a country, scaled by the country’s population. Patent applications are preferred over patents granted as they are a more accurate measure in terms of timing of innovative activity than is the granting of patents (Schneider, 2005). In fact, according to Johnson, Juhl and Popp (2003) the average time lag between patent application and grant for US patents is almost 3.5 years. To speed up the patent review process and increase the encouragement of incremental innovation utility models were introduced in a significant number of countries. They are intended to provide protection of adjustment to existing technology that is not major enough to be qualified for full patent protection (IP Australia, 2011). To obtain a utility patent a technology is not required to have an inventive step and only slight technological novelty for the territory in which the patent will become effective is necessary (WIPO, 2014). Choo et al. (2012) find that in developing countries utility patent protection positively affects innovation using panel data for 70 countries during the time frame 1975 – 2003, while there is no effect for developed countries. By means of the same dataset they also find that regular patent protection positively affects innovation in developed countries, but do not have an impact on innovation in developing countries. Utility patents seem to be a measure of innovation of imitative nature, which is new for the (developing) country, whereas regular patents seem to be a measure of innovation in developed countries, that enhances the world technology frontier. As regular patent and utility patent protection both seem to have an influence on innovation, the variable on domestic innovation includes both types of patents.
institutions governing intellectual property. However, according to Furman et al. (2000) patents ‘are the most concrete and comparable measure of innovative outputs across countries and time’ (p. 18). They make comparisons across countries more reliable as they are more universal than other measures of innovation such as high-tech industry success (Furman et al. 2000).
Human capital (HC) is measured as the index of human capital per person provided by the Penn World Tables coordinated by the Groningen Growth and Development Centre (Feenstra, Inklaar, & Timmer, 2013). The index is composed of years of schooling and returns to education. Observations lack for the year of 2012. However, as changes in the index over time are only minor, an average of the annual observations between 2007 – 2011 is assumed to be representative for a 2007 – 2012 average. Human capital is expressed in its real value.
The strength and enforcement level of IPR protection is measured as the IPR dimension in the International Property Rights Index developed by the Property Rights Alliance (2014). The time varying IPR index measures annual country scores for an average of 124 different countries between the years of 2007 and 2012. The indexation also exists for the year of 2006, but this year is dropped from the analysis in this study since the 2006 index includes a much smaller amount of countries in the indexation. The complete International Property Rights Index consists of three core components: legal and political environment, physical property rights and intellectual property rights. Of these three only the latter, the intellectual property rights dimension, is used as a measure of IPR protection in this study. The IPR index measures country-specific de jure and de facto outcomes for three sub-categories: protection of intellectual property rights, patent protection and copyright piracy. The indexation is based on data made available by established international organizations and academic researchers and includes both observational data (e.g. number of legal procedures and days necessary to register property) and subjective data such as expert opinion surveys comprising the perception of strength and enforceability of IPR protection. It is measured on a continuous scale ranging from 0 to 10. This means that the higher the score (with 10 as a maximum) the higher the strength and enforcement level of IPR protection is for that specific country. Again, for each country the average IPR protection index over the 2007 – 2012 time frame is calculated. IPR protection is expressed in its real value.
data is included based on expert opinion surveys that assesses the perception of actual enforceability levels by the respondents. Finally, the included measure of copyright piracy is based on de facto quantitative data on country piracy levels. The latter measure might not seem extremely relevant. However, it is valuable as it gives a good overview of a country’s overall effectiveness in protecting intellectual property (Di Lorenzo, 2013).
The fact that the enforcement level is measured is very valuable since a country might have strong IPR protection, but if corresponding law is weakly enforced, strong IPR protection de jure is not representative for IPR protection in reality. Schneider (2005), for instance, solely employs Ginarte and Park’s (1997b) widely used index of intellectual property rights protection as a measure of the degree of IPR protection. However, she explicitly mentions that one of the major disadvantages of Ginarte and Park’s (1997b) index is that, although it does include enforcement mechanisms of patent laws, it does not take into account enforcement itself. For this reason, the IPR index of the Property Rights Alliance (2014) includes expert opinions on strength of enforceability. A drawback of this index is that it is only available from 2006 – 2012, meaning that the analysis cannot be extended to a larger time frame to perform, for instance, a regression using panel data. Regarding the TRIPs agreement this time frame is perfect though. The 2008 – 2012 reports describing the IPR index are publicly available (Americans for Tax Reform Foundation, 2010), as well as the 2013 report (Di Lorenzo, 2013), and provide the index based on data measured in the previous year.
One of the two channels for technological diffusion is the annual import level of high-technology goods (IMP) from five developed countries in US dollars, scaled by country population and expressed in natural log. As in Schneider (2005), data on this variable comprises the value of combined imports of capital goods from Standard International Trade Classes (Rev. 1) 7 (machinery and transport equipment), 86 (optical, medical and photographic instruments) and 89 (miscellaneous manufactured good). Collecting and composing the data for the IMP variable was an iterative process. Downloading relevant data was limited to five countries and five years at once, meaning that each download included only five countries and was lacking one year that had to be collected separately. For each country the years and trade classes had to be matched, sorted and added together resulting in the final imports per country per year. Yearly imports were then divided by the matching annual country population count. Finally, the six-years averages of annual high-tech imports per capita were calculated per country.
from technological leaders receive more technology embodied in intermediate goods than countries that import from non-technological leaders, the import levels of high-technology goods considered in this study are the ones coming from the five biggest developed exporters worldwide in 2013: China, the United States, Germany, the United Kingdom and Japan (Central Intelligence Agency, 2014). Whether or not China is a developed country is debatable, but if China is excluded the next country meeting the criteria of being a developed exporter would be France. This would place large emphasis on European countries as Germany and the United Kingdom are also included as exporters of high-tech trade goods. This could potentially lead to a geographic bias causing larger imports of high-tech goods for European countries. Since China, the US, Germany, the UK and Japan are all included in the sample and imports from one country of these five countries to one and the same country are non-existent (e.g. Germany cannot import from Germany), the average of imports from the remaining four countries is used as replacement.
The second possible channel for technological diffusion is FDI. Data on the FDI variable is collected from the World Development Indicators (World Bank, 2014b). FDI is measured as ‘the net inflows of investment to acquire a lasting management interest (10 percent or more of voting stock) in an enterprise operating in an economy other than that of the investor. It is the sum of equity capital, reinvestment of earnings, other long-term capital, and short-term capital as shown in the balance of payments’ (World Bank, 2014b). FDI is measured as a percentage of GDP and expressed in its natural log. Observations of this variable for this study represent the average of annual FDI as percentage of GDP over the time period 2007 – 2012.
a replacement for electricity production and as a proxy for the infrastructure variable (INF) in this study. The variable is expressed as natural log.
4. RESULTS 4.1 Descriptive statistics
Table 1 provides a summary of descriptive statistics of the variables. Not surprisingly, it can be concluded from the table that during the years of 2007 - 2012 developed countries have a higher average rate of domestic innovation, a higher average IPR protection index and a higher human capital index than developing countries. In addition, developed countries enjoy larger high-tech imports, larger FDI inflows and a better infrastructural level than developing countries.
As can be observed in table 1, for several variables differences between means and medians are quite large, which implies that the samples for most variables are not normally distributed but take a skewed form. To check whether or not the samples of the variables come from normally distributed populations, the Shapiro-Wilk test is performed on all variables expressed in their real values and expressed as their natural logs. Table 2 displays the test scores for all non-transformed variables and for the variables transformed expressed as natural logs.
V-values presented in table 2 should be close to 1 for samples to be normally distributed. P-values lower than 0.01 reject the hypothesis that the sample is normally distributed at a 1% significance level. All non-transformed variables are not normally distributed. IPR protection and human capital move further away from being normally distributed after transformation. High-tech imports, FDI and infrastructure all enjoy normal distributions after they have been log-transformed as their p-values are larger than 0.01. Domestic innovation (I) is neither normally distributed before, nor after transformation. However, its V-value moves substantially closer to a desired value of 1. Hence, high-tech imports, FDI and infrastructure are expressed in natural logs when equation (1) is estimated, as well as domestic innovation, since it approaches a normal distribution more than when it is expressed in its real values and it makes interpretation (in terms of percentage changes) of the coefficients easier.
Table 1: Descriptive sample statistics
Variables* Full sample Developing
countries (33)
Developed countries (40)
Mean Median Min Max Mean Mean
Rate of domestic innovation (patent applications per capita) 0.0003671 0.0001342 0.00000003797 0.0035291 0,0000895 0,0005962 IPR protection (index score) 5.95 5.72 2.30 8.55 4.60 7.04 Human capital (index score) 2.83 2.92 1.35 3.61 2.53 3.08 High-tech imports per capita 1,989 843.22 20.02 27,442.01 450.38 3,258.84 FDI (% of GDP) 5.39 3.64 0.25 31.44 4.69 5.97 Infrastructure (index score) 3.11 2.96 1.92 4.26 2.63 3.5
*Not expressed in natural logs
Note: country list:
(1) Developing countries: Albania, Argentina, Armenia, Brazil, Bulgaria, China, Colombia, Egypt,
Guatemala, Hungary, India, Indonesia, Jordan, Kazakhstan, Kenya, Malaysia, Mali, Mexico, Moldova, Morocco, Pakistan, Panama, Philippines, Romania, Serbia, South Africa, Sri Lanka, Thailand, Tunisia, Turkey, Uganda, Ukraine, Zambia.
(2) Developed countries: Australia, Austria, Belgium, Canada, Chile, Croatia, Cyprus, Czech Republic,
Denmark, Estonia, Finland, France, Germany, Greece, Hong Kong, Iceland, Ireland, Israel, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Netherlands, New-Zealand, Norway, Poland, Portugal, Russia, Singapore, Slovak Republic, Slovenia, South-Korea, Spain, Sweden, Switzerland, United Kingdom, Uruguay, US.
Table 2: Shapiro-Wilk test results
Observations W* W** V* V** P* P** I 73 0.57 0.89 27.6 7.1 0.000 0.000 IPR 73 0.95 0.92 3.2 4.8 0.006 0.001 HC 73 0.93 0.87 4.3 8.3 0.001 0.000 IMP 73 0.47 0.98 33.5 1.5 0.000 0.176 FDI 73 0.72 0.99 17.9 0.8 0.000 0.661 Infrastructure 73 0.94 0.95 3.6 2.9 0.003 0.011
* Test results for real (non-transformed) variables ** Test results log-transformed variables
Table 3 shows the simple correlations amongst all variables included in equation (1). High correlations between the dependent innovation variable and all independent variables are desirable. All independent variables seem to be positively and moderately correlated with the rate of domestic innovation, except for the weak correlation between FDI and innovation. This weak correlation might already suggest that FDI is not a channel for technological diffusion, as found by Schneider (2005). The correlation matrix measures linear correlations between non-transformed variables. Log transformations of variables can improve linear correlation between them. Therefore, as several variables, including the dependent variable domestic innovation, are log transformed before linear estimation using OLS, correlations between the independent variables and dependent variable are most likely stronger when they are expressed in their natural logs.
Table 3 suggests that the strongest correlates of domestic innovation are infrastructure, the interaction term between IPR protection and human capital, and IPR protection independent from human capital. Moreover, IPR protection and infrastructure are strongly correlated, which may indicate that countries with high levels of infrastructure are also countries that have stronger IPR protection, possibly because they are more developed. High-tech imports and FDI are rather strongly correlated with each other, which could suggest countries receiving high-tech imports from developed countries might also receive more FDI, possibly because they are more open to trade.
Table 3: Correlation matrix
I IPR HC IPR*HC FDI IMP INF
I 1.000 IPR 0.4470 1.000 HC 0.4032 0.4758 1.000 IPR*HC 0.5129 0.9425 0.7324 1.000 FDI 0.0738 0.1225 0.1250 0.1241 1.000 IMP 0.4242 0.4177 0.2352 0.3982 0.6983 1.000 INF 0.5189 0.9066 0.4822 0.8769 0.1586 0.5027 1.000
Multicollinearity inflates standard errors of the model parameters. However, as stressed by Brambor et al. (2005), significance of the parameters of the interaction variable and its constitutive terms is not of interest. Of interest are the marginal effects and these may be significant for certain values of human capital, even if all model parameters are insignificant. For this reason, the interaction term is not omitted from the model, but simply added as final variable to make sure interpretation of the other coefficients is not distorted by multicollinearity.
4.2 Empirical results
The relationship between IPR protection and domestic innovation is of main interest for this study. Therefore, figure 1 graphically presents the simple correlation between both variables. As can be observed from the figure, no clear positive linear relationship can be detected, which undermines hypothesis 1. The most notable observations are several Asian countries with high levels of IPR protection and very high rates of domestic innovation, such as South-Korea, Japan, Hong Kong, Singapore. For South-Korea and Japan this can be explained due to the fact both countries also offer utility patents. Hong Kong and Singapore have very large patents counts relative to their populations.
It seems that hypothesis 1 is confirmed for several countries such as the US, New-Zealand and Canada. However, there are also many countries with high IPR protection levels and relatively low rates of domestic innovation. They are slightly difficult to distinguish in the graph, but examples include developed countries such as Belgium, Finland, Austria and The Netherlands. Moreover, there are several developing countries with fairly low IPR protection levels, but with rates of domestic innovation that are quite large relative to developed countries with strong IPR protection. Examples of such countries are China, Russia, Ukraine and Uruguay. Apparently, strong IPR protection is not necessarily required for relatively high domestic innovation rates.
Figure 1: Scatterplot IPR protection level and domestic innovation rate
observations of economic units of varying sizes. Heteroskedasticity causes incorrect calculations of standard errors in a least square estimator meaning that confidence intervals and hypothesis tests might be misleading (Hill, Griffiths, & Lim, 2011). To check for the existence of heteroskedasticity a White test is performed on the residuals of the model. The low p-value of the White test (0.0003) indicates that the null hypothesis of homogeneous variance has to be rejected, meaning that heteroskedasticity indeed exists in the variance of the residuals.
To correct for heteroskedasticity in the least squares regression and ensure reliability of the confidence intervals and hypothesis tests the model defined by equation (1) is estimated using White’s robust standard errors. Table 4 reports the estimates of equation (1) using OLS including robust standard errors.
ML
GTM MORUGNIND
PAN IDNKNY ZMB THLEGY TRKTNA BRZ CHL PRT CHN SA URG PHLMXC CRT SNG JDN SRB AGN UK ITL AUT PNA MLD BLG PLN SWT CYP FNL DNM LXM MLS CHL KZ SPN HK MLT RMN ABN LTV FR ARM GRC ICL BLG LTH NL SRL UKR SVK ISR RSA SWD JPN IRL HNG SLVEST GRM SK AUS CDA CZR NRW NZLUS
0
.0
0
1
.0
0
2
.0
0
3
.0
0
4
D
o
me
st
ic
in
n
o
va
ti
o
n
r
a
te
2
4
6
8
10
Table 4: OLS estimation innovation regression DV: Log I OLS (1) OLS (2) OLS (3) OLS (4) OLS (5) OLS (6) Log INF (β6) 6.169 (6.56)* 10.610 (4.28)* 8.578 (4.92)* 9.056 (7.90)* 8.663 (7.47)* 8.69 (7.53)* IPR (β1) -0.636 (-2.27)** -0.774 (-2.98)* -0.736 (-3.77)* -0.767 (-3.79)* -0.407 (-0.71) Log IMP (β4) 0.532 (3.01)* 0.030 (0.17) 0.155 (0.69) 0.137 (0.58) HC (β2) 2.053 (6.07)* 1.912 (5.82)* 2.57 (2.58)** Log FDI (β5) -0.240 (-1.44) -0.251 (-1.50) IPR*HC (β3) -0.120 (-0.73) Constant (β0) -15.928 (-14.05)* -17.075 (-13.02)* -17.500 (-14.95)* -20.763 (-18.32)* -20.257 (-21.27)* -22.117 (-8.62)* Observations 73 73 73 73 73 73 Adj. R-squared 0.502 0.554 0.624 0.740 0.747 0.745 F 43* 32.83* 26.05* 41.23* 36.30* 31.32*
t-statistics are in parentheses.
* Significant at 0.01 level
** Significant at 0.05 level
*** Significant at 0.10 level
Before discussing the results presented by table 4, normality of the residuals need to be examined to make sure valid hypothesis tests can be performed. As can be seen in table 2, data for several independent variables and the dependent variable rate of domestic innovation are not normally distributed, which makes it highly likely the residuals are not normally distributed either. Once again the Shapiro-Wilk test is performed, but this time on the residuals of the model including all variables defined by equation (1). As the resulting p-value of the test is lower than 0.01, the hypothesis that the sample is normally distributed at a 1% significance level is rejected, meaning that the residuals are not normally distributed. The fact that the residuals are not normally distributed does not necessarily mean the coefficients estimated using OLS are not the best linear unbiased estimators anymore. However, it does mean confidence intervals and tests concerning significance might not be reliable. A solution could be to increase the sample size and collect more observations as for sufficiently large sample sizes least squares estimators move towards a distribution approximating a normal distribution (Hill et al., 2011). For now, the results are discussed based on the sample presented in table 1, meaning that one needs to be cautious about the conclusions drawn based on hypotheses tests resulting from the OLS regression displayed in table 4.
regression the highest adjusted R-squared is found, meaning that without including the interaction term between IPR and human capital the explanatory power of the model is at its best. 74.7% in the variation of the rate of domestic innovation is explained significantly, as indicated by the F-test result. The adjusted R-squared is highly valuable in building a model predictor by predictor as it generates comparability of each model that is estimated. The regular squared increases with every predictor that is added to the model, whereas the adjusted R-squared only increases if the explanatory power of the model increases whilst controlling for the number of predictors in the model. As soon as the adjusted R-squared reduces, it is clear the previously estimated model creates the most reliable estimates. That adding the interaction term to regression 6 reduces the adjusted R-squared and the explanatory power of the model to 74.5% can probably be attributed to the fact that the interaction term introduces multicollinearity to the model. As the interaction term poses the issue of multicollinearity the coefficients of the variables estimated by regression 6 are firstly left out of the discussion. Once the marginal effects of IPR protection on domestic innovation need to be calculated regression 6 will come into play.
The IPR variable is of main interest for testing hypothesis 1. As can be observed in Table 4, the IPR variable enters regressions 2 till 5 with a negative coefficient that is highly significant. Assuming H0 represents IPR protection sharing a negative correlation with domestic innovation
and H1 the alternative hypothesis 1, the negative significant coefficient for the IPR variable
indicates there is no sufficient evidence to reject H0. To the contrary of hypothesis 1, it has to
be concluded that IPR protection is negatively correlated with domestic innovation: countries with stronger IPRs that are enforced more effectively appear to have lower rates of domestic innovation. Regression 5 shows that if the IPR index score of a country increases with 1, this is associated with a reduction of 76.7% in the domestic innovation rate, controlling for the effects of infrastructure, high-tech imports, human capital and FDI. This might seem to be a very large effect, but it has to be taken into account that the IPR index ranges from 0 – 10, hence meaning that an increase of 1 on the index is relatively substantial.
