Investigating Externalities of Knowledge Spillover among Selected European Countries

International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 631 Investigating Externalities of Knowledge Spillover among Selected European Countries Fatemeh Taleghani 1 PhD student, College of Management and Economics, Shahid Bahonar University, Kerman, Iran Seyed Abdolmajid Jalaee Esfandabadi Faculty Member, College of Management and Economics Shahid Bahonar University, Kerman, Iran Fatemeh Irani Kermani Professor, College of Management and Economics Shahid Bahonar University, Kerman, Iran Abstract Generally, knowledge spillovers result in the creation of new knowledge, increased competitive advantages, and economic cooperation. Since the investigation of spillover flows among countries is considered to be highly important, in this study, knowledge spillovers and its resulting externalities were considered among a number of selected European countries during 1995 to 2011 using spatial econometric analysis. The results indicated an indirect effect and positive feedback caused by changing human development index, research and development expenditure, and knowledge-bearing imports, which confirmed the existence of spillovers and adsorption capacity in this region. Keywords: Externality, Knowledge Spillovers, Spatial Econometrics Cite this article: Taleghani, F., Jalaee Esfandabadi, S. A., & Kermani, F. I. (2015). Investigating Externalities of Knowledge Spillover among Selected European Countries. International Journal of Management, Accounting and Economics, 2(7), 631-645. 1 Corresponding author’s email: [email protected] International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 632 Introduction Knowledge and technical development are among the principle resources of dynamism in economic growth models. Therefore, patent activities include use, application, and transfer of scientific and technical knowledge in problem solving and this knowledge is different from the information which is conventionally applied. On the other hand, knowledge transfer occurs via movement of skilled labor, simulation and reverse engineering by local firms, and increasing competition and commercial communication between domestic and international firms; however, such transfer is limited by the adsorption capacity of the host country via various channels. This capacity provides an opportunity for learning and applying knowledge. Furthermore, geographical distance can facilitate or limit knowledge transfer; this distance is specifically important in knowledge transfer, because source countries are looking for the destination countries that, beside structural similarity, have minimum geographical distance from them. Thus, by the formation of a knowledge life cycle between the source and destination countries, both of them could benefit from this knowledge transfer with minimum cost. This cycle indicates a situation in which a foreign country transfers a less sophisticated knowledge to a domestic country. In this case, foreign and domestic countries are called source and destination countries, respectively. At the next stage, knowledge of the foreign country is combined with the domestic technology to generate new knowledge; then, it is returned to the foreign country. In this case, positions of source and destination countries are changed. In other words, domestic and foreign countries are called source and destination countries, respectively. On the other hand, not only knowledge spillovers are not directly observable, but also their transfer is accompanied by some externalities which might have positive or negative effects on the economic growth and development. This article was aimed to investigate the nature of these externalities using intraand interregional patent responses to the changes in the inputs of knowledge production function. Therefore, it is necessary to study the patent response of each region and other regions to changes in inputs of every region and the neighboring areas. This response indicates Marshall-Arrow-Romer (MAR) externalities. Based on MAR perspective, spillovers occur among similar units with sharing common knowledge; yet, on the opposite side, there are Jacob's externalities among complementary units. Therefore, the regions with various products must create more patents. Furthermore, Jacob's externalities are decreased with more increased distance compared with MAR externalities (Autant-Bernard and LeSage, 2011). Thus, in this study, direct and indirect effects on patent were investigated, which resulted from changing the inputs of knowledge production. Therefore, in order to investigate the externalities resulting from knowledge production inputs, research literature and model description are presented in Sections 2 and 3, respectively. In addition, model estimation is presented in Section 4 and conclusions are delivered in the final section. Literature Considering the importance of technology spillover in international economy, numerous studies have been conducted on the presence and efficacy of technology spillovers in different time periods and using different methods. In this section, a quick review of the available research on technology spillover is presented. In this regard, International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 633 Anselin et al. (1997) investigated spatial spillovers among academic research and patent and used data from 29 states in the USA in 1982. They concluded that there was weak evidence on spillovers among universities and research institutes inside the states; but, the resulted externalities went beyond the borders. Also, they found a correlation between academic studies and patent activities in the private sector along with externalities. Paci and Usai (2000) studied knowledge spillovers and geographical distribution of patent using data from 85 industrial sectors in Italy between 1990 and 1991. Results demonstrated spatial correlation in the distribution of patent activities and positive influence of industrial sectors from similar industrial sectors in the neighboring regions. Their analysis showed two externalities in this study. Fischer and Varga (2003) analyzed knowledge spillovers from research activities in scientific centers to high-technology industries in 72 regions inside Australia and used data related to 1991. The results indicated the presence of intra-regional spillovers; however, some externalities were produced which were smaller than MAR ones. Moreno et al. (2003) conducted a study entitled "Spatial spillover, patent activity, and role of knowledge production process" using data from 138 regions in 17 European countries between 1978 and 1997. They found a significantly positive spatial auto-correlation in patent; i.e. knowledge production in the studied region was influenced by spatial spillovers which resulted in increased patent activity in other regions. The most important effective factors for patent generation were international research and development expenditure. Driffield and Love (2003) studied direct foreign investment, technology source, and reverse spillovers in UK industries; by focusing on 1984-1992, they found that the technology made by spillovers in domestic firms spilt over to foreign firms; but, effect of these spillovers was limited to developed sectors. Nonetheless, both sides benefited from these spillovers. Also, they found that technology spillovers were influenced by spatial concentration of industries. Bernard and LeSage (2011) investigated knowledge spillovers using spatial econometric analysis models and data from 1992 to 2000. They stated that estimating spillovers, regardless of spatial dimension, can be biased and inconsistent; accordingly, they applied spatial TOBIT method and concluded that the biggest direct and indirect effects of technology spillovers in 94 Asian regions were related to research and development activities of private sector; these externalities decreased with distance from the source and the results referred to optimal regional strategies. Model description In this study, endogenous growth model was used instead of exogenous growth model, since theory of exogenous growth model states that capital flows from countries with low efficiency to those with high efficiency; however, studies have shown such a flow and confirm capital transfer from poor to developed countries. In fact, the reason can be found in endogenous growth models. In endogenous growth models, the unanimous view is that accumulation of physical capital does not make countries richer, but human capital is placed beside physical capital and a ground is developed for the technology formation and its absorption capacity via research and development department. In this regard, Mingyong et al. (2006) applied Romer's (1990) model to study technology spillovers and stated that production was made using a large number of incomplete alternative inputs, since technical process is originated from the invention of new inputs via research and development activities. Two foreign and domestic countries are thus considered. In the International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 634 domestic country, economy is composed of three parts of research and development, intermediate goods, and final goods. For final goods, good y is produced under perfect competition and the production function is as follows: 𝑌=𝐴 𝐻𝑦 𝛼[∫𝑥𝑖𝛽 𝑁 0𝑑𝑖+∫𝑥𝑖∗ ∗𝛽 𝑁∗ 0𝑑𝑖∗] 𝛼,𝛽>0 ,𝛼+ 𝛽=1 (1) where A, Hy, xi, and 𝑥𝑖∗ are total productivity level, human capital used for final goods, and numbers of N intermediate domestic and foreign inputs denoted by i, respectively. N and 𝑁∗ indicate the number of domestic and foreign intermediate inputs, respectively. Intermediate goods are invented or completed in research and development department and are then purchased from two domestic and foreign producers. Production in research and development department depends on international spillovers of research and development through commerce, human capital investment in this department, and technical knowledge of domestic country. Since technical knowledge is shown by different variables of capital goods, developing plans for answering new needs of domestic country can be defined as: 𝑁•= 𝛿 𝐻N [N+G(D,H)𝑁∗] (2) where 𝛿, 𝐻N , and H are a productivity constant, amount of human capital used in research and development department, and total amount of human capital, respectively, which indicates the constant amount of knowledge and skill in economy (H=HN+Hy). Also, G(D,H) shows adsorption capacity, which is determined by total domestic human capital and extent of economy openness so that 𝐷𝜖(0,∞). For intermediate goods, after the development or invention of a project, an intermediate firm buys the project and produces the inputs under perfect competition. For the sake of simplicity, it is assumed that intermediate inputs are spent on the production of a unit of Y. Market balance Price of product Y is considered one. 𝑊Hy and W𝐻N are the payments of human capitals in research and development along with final goods departments, respectively. Also, 𝑃xiand 𝑃xi are the price of domestic and foreign intermediate inputs. Since intermediate goods are converted into capital, capital price is also considered one unit and interest rate (r) is determined in a perfect financial market. It is also assumed that markets Y and H are competitive for Y production firms. Two criteria exist for intermediate goods: first, it is assumed that inventors are free to enter the business and, second, each intermediate good is produced by a monopoly on the sale. Therefore, the problem can be presented as follows for a final producer: 𝑀𝑎𝑥 𝜋=𝑌{𝐻𝑦 ,𝑥𝑖,𝑥𝑖∗ ∗} − 𝑊Hy𝐻𝑦−∫𝑃𝑥𝑖𝑥𝑖 𝑁 0𝑑𝑖−∫𝑃𝑥𝑖∗ ∗𝑥𝑖∗ 𝑁∗ 0𝑑𝑖 ∗ (3) By adopting the required first-order conditions for the above maximization problem, we have: International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 635 𝑊Hy=αY 𝐻𝑦 ⁄ (4) 𝑥i=𝐻𝑦 [A𝛽 𝑃xi] ⁄1 α , i.e. 𝑃xi=A𝛽𝐻𝑦𝛼x−𝛼 (5) 𝑥𝑖∗ ∗=𝐻𝑦 [A𝛽 𝑃x𝑖∗ ∗]⁄ 1 α , i.e. 𝑃𝑥∗=A𝛽𝐻𝑦𝛼𝑥∗ −𝛼 (6) According to the above equations, it is clear that all the intermediate goods are employed for the production of final goods; therefore, the same demand function is shared. Producers in intermediate input department use price 𝑃𝑥 to maximize current profit at any given time. 𝑉(𝑡)=∫ ( ∞ 𝑡𝑃 x .x − 1 .x) e−r (s ,t)(s−t)𝑑𝑠 X is total intermediate inputs which are produced by demand function at any time and r(s ,t)=[1 𝑠−𝑡 ⁄ ]∫𝑟(𝑣)𝑑𝑣 ∞ 𝑡 indicates average interest rate between times t and s. Therefore, by assuming a constant value for interest rate, the problem of intermediate firms for selecting a price which could maximize profit is given as follows: max 𝑃 x 𝜋𝑚=𝑃 x .(x − 1) .x (7) The solution for price of monopoly on the sale is presented below: 𝑃 xi =𝑃 x =1 𝛽 ⁄ (8) A similar method for obtaining the price of foreign intermediate inputs is: 𝑃x𝑖∗ ∗= 𝑃x∗ Since domestic economy could be completely combined with global economy, degree of economy openness is used so that, to obtain every X unit from a foreign intermediary, x 𝑒D units should be sent. Therefore, optimum price for a foreign monopoly can be obtained. 𝑃x𝑖∗ ∗= 𝑃x∗=𝑒𝐷𝛽 ⁄ (9) By inserting Equations 8 and 9 in Equations 5 and 6, balances value of xi and xi∗ are obtained. 𝑥i=x=A1α ⁄𝛽2α ⁄𝐻𝑦 (10) 𝑥𝑖∗=𝑥∗=A1α ⁄𝛽2α ⁄𝐻𝑦𝑒−𝐷 α ⁄ (11) Using Equations 1, 10, and 11, balance level of the product is determined as: 𝑌=𝐴𝐻𝑦𝛼(𝑁𝑥𝛽+𝑁∗𝑥∗𝛽)=A1α ⁄𝐻𝑦 𝛽2𝛽 α ⁄(N+𝑁∗𝑒−𝐷𝛽 α ⁄)= International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 636 A1α ⁄𝐻𝑦 𝛽2𝛽 α ⁄[𝑁+𝐹(𝐷)𝑁∗] (12) In this case, 𝐹(𝐷)=𝑒−𝐷𝛽 α ⁄ , 𝐹(0)=1, 𝐹(∞)=0, and 𝜕𝐹 𝜕𝐷 ⁄<0 and PN is the patent price of the intermediate product. By guaranteeing free entry into the intermediate department, the reduced value of profit would be equal to the patent price. 𝑃𝑁=𝑉(𝑡)=∫𝜋𝑚(𝑠) ∞ 𝑡 e−r (s ,t)(s−t)𝑑𝑠 (13) Assuming a constant value for interest rate, a specific solution can be obtained owing to the constant value for PN. In this case, the following equation is obtained: 𝑃𝑁=𝑉(𝑡)=1 𝑟𝜋𝑚(𝑡)=1 𝑟(𝑃𝑥−1)x=1 𝑟(1 𝛽−1)x=1 𝑟(𝛼 𝛽)x (14) In the research and development department, total income of research and development activities is as follows: 𝑇𝑅=𝑃𝑁𝑁•=𝑃𝑁𝛿𝐻𝑁[𝑁+𝐺(𝐷 ,𝐻)𝑁∗] And total costs are as follows: 𝑇𝐶=W𝐻N.𝐻𝑁 Therefore, free entry into research and development department is guaranteed and the payment for human capitals in the research and development department would be: W𝐻N=𝛿𝑃𝑁[𝑁+𝐺(𝐷 ,𝐻)𝑁∗] (15) Below, Romer assumes that human capital moves between departments. Furthermore, by determining the balance condition and allocating human capital for departments of final production and research and development, Romer states that payments for human capital must be equal in all the departments: W𝐻N=W𝐻y (16) Considering Equations 4, 12, 14, and 15, Equation 16 can be rewritten as follows: α 𝐴1α ⁄ β2β α ⁄[𝑁+𝐹(𝐷 )𝑁∗]=𝛿(α 𝑟βx) [𝑁+𝐺(𝐷 ,𝐻)𝑁∗] By inserting Equation 10 in 𝑥 in the above equation, the following equation is obtained: 𝐻y𝛿 𝑟[𝑁+𝐺(𝐷 ,𝐻)𝑁∗]=1 β[𝑁+𝐹(𝐷 )𝑁∗] Consequently, the above equation can be simplified as Equation 17. 𝐻y=𝑟[𝑁+𝐹(𝐷 )𝑁∗] 𝛿β[𝑁+𝐺(𝐷 ,𝐻)𝑁∗] (17) International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 637 Considering Hy, for calculation simplicity, it is assumed that: 𝑁𝑤𝑜𝑟𝑙𝑑 =𝑁+𝑁∗ , 𝑁∗ 𝑁=𝑢 (18) If 𝑢≥0, then there is technology gap between foreign and domestic countries. 𝑁= 1 1+𝑢𝑁𝑤𝑜𝑟𝑙𝑑,𝑁∗=𝑢 1+𝑢𝑁𝑤𝑜𝑟𝑙𝑑 (19) By inserting Equation 19 in Equation 17, the following relation is obtained: 𝐻y=𝑟[1+𝑢 𝐹(𝐷 )] 𝛿β[1+𝑢 𝐺(𝐷 ,𝐻)] (20) Since 𝐻N=H−𝐻y, Equations 2 and 19 can be used to calculate technology growth rate: 𝑔𝑁=𝑁• 𝑁=𝛿𝐻N[1 + 𝑢 𝐺(𝐷 ,𝐻)]=𝛿(𝐻−𝐻y)[1 + 𝑢 𝐺(𝐷 ,𝐻)] By inserting Equation 18 in Equation 12, the following relation is obtained: 𝑌= 𝐴1α ⁄ 𝐻yβ2β α ⁄[1 + 𝑢 𝐹(𝐷 )]𝑁 (21) If r is constant, Equation 20 indicates that Hy is also constant; also, 𝑥 is constant according to Equation 10. In this economy, using capital and product, total production grows at an equal rate N. Growth rate 𝑔 for achieving a static balance growth path for all the variables can be written as: 𝑔=𝑔𝑦=𝑔𝐶=𝑔𝑁=𝛿𝐻N[1 + 𝑢 𝐺(𝐷 ,𝐻)]=𝛿(𝐻−𝐻y)[1 + 𝑢 𝐺(𝐷 ,𝐻)] (22) Equation 22 shows a positive correlation between growth rate with economic stability g, human capital in research and development HN, and adsorption capacity 𝐺(𝐷,𝐻) so that, with increasing human capital in research and development department and improving domestic adsorption capability, growth rate with economic stability is increased. If 𝐻N=0, there is no long-term growth; if 𝐻N is positive and less than H, then g would be positive. Therefore, stable growth rate would be equal to: 𝑔=𝜎𝐻[1+𝑢𝐺(𝐷,𝐻)−(𝜌 β)[1+u F(D)] ⁄ 1+(𝜎 β)[1+u F(D)] ⁄ (23) Below, in order to study externalities in the knowledge production function, spatial econometric is used so as to obtain intraand inter-regional externalities. According to the studies by LeSage and Pace (2009), when data samples have a spatial component, two cases would occur: 1) spatial dependence between observations, and 2) spatial heterogeneity (spatial structure). International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 638 Conventional econometric analysis ignores these two issues to a great extent, which could be due to the violation of Gauss-Markov's assumptions used in regression models. Therefore, in order to apply this method, its concepts should be understood. Below, a brief description of spatial heterogeneity and dependence and the way of determining location and spatial lags is presented. Spatial dependence Spatial dependence in a set of sample data means that observations in location i depend on other observations in location j. In other words: 𝑌𝑖=𝑓(𝑌𝑗), 𝑖=1,2,…,𝑛 𝑖 ≠𝑗 (24) This dependence may exist among different observations and disturbing components and must correspond with fundamental theorems of regional science; i.e. closer observations must reflect a higher degree of spatial dependence than the ones which are distant from each other. In other words, spatial dependence and its effects on observations must decrease with increasing distance between observations. Spatial heterogeneity The term spatial heterogeneity refers to deviation in relations between observations at the level of geographical locations. In most cases, different relations are expected for each point in the space. In other words, linear relation is expressed as follows: 𝑌𝑖= 𝑋𝑖𝛽𝑖+ 𝜀𝑖 (25) where i, Xi, Yi, and 𝜀𝑖 indicate the obtained observations at points i=1,2,3,…,n in the space, 𝑋𝑖 shows n×k vector of descriptive variables along with its related 𝛽𝑖 parameter set, 𝑌𝑖 is dependent variable in observation or location i, and 𝜀𝑖 is random error in the above relation. A more complex presentation of this concept is as follows: 𝑌𝑖=𝑓(𝑋𝑖𝛽𝑖+ 𝜀𝑖) (26) Considering Equation 15, it is not expected to estimate a set of n parameters from vector 𝛽𝑖 considering one sample of observations and unique estimation for each point in the space. Generally, spatial heterogeneity also violates the Gauss-Markov's assumption which suggests only one definite linear relationship with constant variance between sample observations. Therefore, after rejecting the null hypothesis, which states lack of any spatial auto-correlation among disturbing components, spatial error model (SEM), simultaneous auto-regression-regression (SAR) model, general spatial model, or spatial Durbin model can be utilized. Simultaneous auto-regression-regression model This model explains y variations as a linear combination of neighboring countries like auto-regressive time series and emphasizes on what occurs in these countries, because knowledge production in every country can be affected by variations in knowledge production and spillovers of neighboring countries. In this regard, method of maximum International Journal of Management, Accounting and Economics Vol. 2, No. 7, July, 2015 ISSN 2383-2126 (Online) © IJMAE, All Rights Reserved www.ijmae.com 639 likelihood can be used to estimate parameters of this model. The above mentioned model is as follows: 𝑦i=ρ∑Wijyj+∑𝛽𝑘𝑥𝑘𝑖 𝑘 𝑘=1 +εi= n j=1 ρW y + Xβ+εi (27) εi~N(0,σ2In) Spatial error model In spatial econometric analysis, a model is spatial error model, in which knowledge production is influenced by shock creation in neighboring countries. This model can be presented as: 𝑦i=∑𝛽𝑘𝑥𝑘𝑖 𝑘 𝑘=1 +εi=Xβ+ui (28) ui=λWui+εi , εi~N(0,σ2In) Spatial Durbin model This model which has the spatial coefficient of dependent variable and descriptive is written as follows: 𝑦i=ρWy+Xβ+𝑊Xθ+εi (29) εi~N(0,σ2In) It must be mentioned that spatial Durbin model is preferred to simultaneous spatial regression-auto-regression and spatial error models, because when 𝜃=0, SDM is converted into simultaneous spatial regression-auto-regression model, and when 𝜃= −𝛽𝜌, this model is changed to spatial Durbin model. In addition to simultaneous spatial regression-auto-regression model, in spatial Durbin model, direct effects can be also distinguished from indirect ones. It is worth mentioning that spatial Durbin model is less biased than simultaneous spatial regression-auto-regression model; also, spatial error model results in the elimination of spatial spillovers. Technology patents resulted from spatial knowledge spillovers from neighboring regions are examples of positive externalities or positive spillovers. A large part of knowledge is implicit and its transfer requires cooperation and common activities. On the other hand, there is explicit knowledge, since conducting ideas to technological patents ensures the existence of people who have relationship with inventors' experience. This knowledge existence increases and is often transferred as a result of discovering new ideas in a region. Knowledge of a particular region and its neighboring regions is a good general commodity, which provokes spatial explanation for knowledge. Spatial regression models can be used to determine spatial extent of spillovers through investigating indirect effects using expanded series 𝐼𝑛+𝜌𝑊+𝜌2𝑊2+⋯. Generally, in order to obtain direct effects, first, effect of increasing descriptive variable in country i on dependent variable in country i is calculated (i.e. own-partial derivative is equal to 𝜕𝑦𝑖 𝜕𝑥𝑖) and since i=1,2,3,…,n, all the effects in the entire region is averaged. To calculate indirect accumulative effect,