WPS5739 Policy Research Working Paper 5739 Entrepreneurship Capital and Technical Efficiency The Role of New Business/Firms as a Conduit of Knowledge Spillovers Leopoldo Laborda Jose Luis Guasch Daniel Sotelsek The World Bank Latin America and the Caribbean Region Finance and Private Sector July 2011 Policy Research Working Paper 5739 Abstract Increasingly, entrepreneurship is being discussed and (or divergence) in the efficiency estimated for individual considered as a source of high economic growth and countries. competitiveness. A conceptual process of creative The empirical evidence and results here tend to support construction that characterizes the dynamics between the hypothesis. Specifically, the empirical analysis shows entrants and incumbents can prove quite useful to that the rate of expenditure on research and development analyze the impact of countries’ entrepreneurship in relation to new businesses registered has a positive capital on economic performance and can be a guide for and significant effect in increasing technical efficiency. economic policy. These factors facilitate the dissemination of existing This paper applies a Stochastic Frontier Analysis knowledge, develop entrepreneurship capital, and thus approach to test the hypothesis that entrepreneurship provide the missing link to economic performance— capital promotes economic performance by serving as entrepreneurship capital. The authors also show the a conduit of knowledge spillovers. In addition, kernel trends and dynamics of changes in countries’ technical density functions are employed to analyze convergence efficiency. This paper is a product of the Finance and Private Sector, Latin America and the Caribbean Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at jguasch@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Entrepreneurship Capital and Technical Efficiency: The role of new business/firms as a conduit of knowledge spillovers1 Leopoldo Laborda* World Bank and Institute of Latin American Studies University of Alcalá (Spain) Jose Luis Guasch** World Bank and University of California, San Diego (USA) Daniel Sotelsek*** Institute of Latin American Studies University of Alcalá (Spain) Keywords: Technical Efficiency, Entrepreneurship Capital, Stochastic Frontier Analysis (SFA). * Research Associate at the Institute of Latin American Studies, University of Alcalá c/ Trinidad nº1, Colegio de Trinitarios, Alcalá de Henares, 28801, Madrid. (Spain). Phone: +34 91 8820399. e-mail: llabordacastillo@gmail.com ** Senior Adviser at the World Bank and Professor of Economics at University of California, San Diego (USA) e-mail: jguasch@worldbank.org *** Professor of Economics at the University of Alcalá (Spain). e-mail: dsotelsek@uah.es 1 This research is part of the program in LAC Private Sector and Finance Unit in supporting private sector development and entrepreneurship in Latin America and Caribbe. This paper is forthcoming in the Entrepreneurship Research Journal. For additional information contact Jose Luis Guasch. 1.- Introduction The knowledge-based view of the firm argues that wealth creation in a firm is a function of its ability to create new knowledge and to exploit it in the market (Teece, Pisano and Schuen, 1997). However, the investment that a firm makes in knowledge-related activities has important implications beyond its boundaries because of its lack of ability to appropriate and exploitat all the benefits. As a result, existing organizations may be characterized as having an abundance of underexploited knowledge (Agarwal et al., 2004). For Agarwal, Audretsch and Sarkar (2010), individuals that perceive unexploited opportunities created by knowledge investments by incumbent organizations, may choose to venture out using the human capital/knowledge they acquired during their tenure at the knowledge-generating organization. Entrepreneurs starting a new venture not only create new firms, but also provide a conduit for the spillover of knowledge that might otherwise not have been commercialized and would have remained dormant in the incumbent firm. The literature that links knowledge spillovers to entrepreneurship capital emphasizes that existing (incumbent) organizations are an important source of new entrants, as they embody knowledge that can be appropriated, and thus facilitate new entry into the sector (Audretsch, 1995; Agarwal, et al., 2004; Colombo, 2005). The knowledge spillover theory of entrepreneurship suggests that knowledge spillovers serve as the source and create the entrepreneurial opportunities for new firms. This theory suggests that entrepreneurship is the missing link in the process of economic growth because it facilitates the spillover of knowledge from universities and private firms, resulting in commercialization of ideas that otherwise might remain unexploited or uncommercialized.2 According to Audretsch and Keilbach (2003), by starting up a business, an entrepreneur literally “betâ€? on the product he offers and thus is willing to take the risk that this venture bears. Acs and Audretsch (2003) observe that, “by commercializing ideas that otherwise would not be pursued and commercialized, entrepreneurship serve as one mechanism facilitating the spillover of knowledge.â€? The empirical evidence supporting the knowledge-spillover theory of entrepreneurship is based on the analysis of variations in startup rates across different industries reflecting different underlying knowledge contexts. As pointed out by Caves (1998), those industries with a greater investment in new knowledge also exhibited higher startup rates, while those industries with less 2 “Although characteristics of entrepreneurial activity differ across countries, the importance of entrepreneurship for economic development is widely acknowledgedâ€? (Global Entrepreneurship Monitor (GEM) 2007 Executive report (2008: 12 )) 2 investment in new knowledge exhibited a lower startup rate, where startups are interpreted as a conduit, transmitting knowledge spillovers.3 Agarwal, Audretsch and Sarkar (2010) deal with these questions by developing the creative construction approach, which identifies knowledge spillovers as a key mechanism that underlies new firms’ formation and development at the micro level, and economic performance at the macro level. Yet little analytical, and particularly empirical, work has been undertaken to support that general hypothesis. Here we are advancing the knowledge on this subject by analyzing the impact of countries’ entrepreneurship capital on economic performance. The main objective of this work is to analyze whether the entrepreneurship capital promotes economic performance (in terms of technical efficiency) by serving as conduit of knowledge spillovers. With this objective in mind, the work is organized as follows: in the next section we present the conceptual framework proposed to clarify the relation between new business (as a conduit of knowledge spillovers) and economic performance (in terms of technical efficiency). In Section 3 we develop the methodology of the analysis. In Section 4 we present the main empirical results. Section 5 ends with a summary of the main conclusions and policy implications. 2. Background: Creative Construction and Economic Performance 2.1. The Process of Creative Construction Agarwal, Audretsch and Sarkar (2010) argue that creative construction is similar to creative destruction in highlighting the creation of value through entrepreneurial entry. However, it differs from creative destruction in two aspects: (i) it identifies the “constructionâ€? of these new entrants due to incumbent investments in knowledge, and (ii) it questions whether incumbents are necessarily destroyed in the process, given the potential for simultaneous (co- existent) growth of both incumbents and entrants alike, and for strategic management by incumbents of the knowledge spillovers that may result in “spill- insâ€?. In addition, Agarwal and Bayus (2002) show that sales and growth in the industry are linked to a critical mass of entry in the industry. Other authors like Saxenian (1994) have explicitly linked the growth of regions and industries to spinout/spinoff activity. These authors document the positive synergies and agglomeration economies caused due to geographical clusters enabled by knowledge spillover, strategic entrepreneurship, and they also provide several reasons for a win-win rather than a win-lose outcome. 3 In relation with this issue, and in order to evaluate a potential reverse causality concern in our analysis, we perform a Granger causality test (see more details in Section 4). 3 The first reason stems from agglomeration and legitimacy effects, which can lead to increase in demand that permits simultaneous growth of both the parent and the progeny. Klepper (2007) argues that growing industries and regions attract not only additional human capital, but also supporting infrastructure related to the supply-chain operations needs as well as venture financing. Not only does this serve to reinforce the supply-side effects for the incumbent organization, but it can also lead to enhanced demand of the product it sells. Thus, particularly in the growth stages of the industries, both parent and progeny organizations may grow, and the growth of one is not at the expense of the other. The second reason stems from “spill-inâ€? or capability enhancement effects which arise when spinouts occupy complementary rather than competitive positions, and their growth in capabilities provides a potential for learning (and even subsequent acquisition of the spinout firm) by the parent organization. According to Somaya, Williamson and Lorinkova (2007), an incumbent firm may be able to leverage off the capabilities of the spinout it has spawned, and use it as a complementary asset. While much has been documented about spinouts occupying competing positions in the supply chain, the authors have systematically documented that employee mobility to firms that are vertically linked, or produce complementary products, can have beneficial effects on the incumbent firms. Finally, Somaya, Williamson and Lorinkova (2007) argue that an incumbent can access new knowledge, competencies and capabilities created in the new venture, by relying on social capital links to the new venture. Such linkages, either formally through contractual agreements, or informally through interactions of personnel from both the incumbent and new venture, can facilitate the access of valuable know-how and competencies generated by the new venture, thereby enabling the “spill-inâ€? of knowledge from the new venture generated by the spillover back to the incumbent. 2.2. Economic Performance The dynamics at the firm level also have implications at macro levels on the performance of regions, industries and economies. As endogenous growth theory (Romer, 1990) suggests, a greater degree of knowledge spillovers will spur higher rates of growth, employment and international competitiveness. Entrepreneurial new ventures are an important mechanism for knowledge spillovers, as their use of knowledge and ideas serves as the crucial resource driving the competitive advantage of the industries, regions and economies that they are associated with. Regions and industries with a high degree of 4 entrepreneurial activity will also facilitate more knowledge spillovers, which, in turn, will increase economic growth, employment creation, economic performance and international competitiveness. Thus the virtuous cycle. In others words, endogenous-growth models improve on the earlier traditional models of growth by providing insights regarding the underlying growth-transmission mechanisms, and, focusing on economic performance as being driven by explicit firm action, either due to investments in knowledge by existing organizations, or due to research activity undertaken by new entrants. These models advance our understanding of the underlying mechanisms by relating growth to exogenous spillovers of endogenous investments in knowledge. However, this approach assumes that spillovers are randomly generated. As we show in the next section, our conceptualization highlights the active role of entrepreneurial action in the spillover process; thus, in addition to endogenous investments in knowledge by incumbent organizations, spillovers occur due to subsequent endogenous pursuit of innovation by individuals immersed in these institutional contexts. As a result, economic performance is due to deliberate investment and activity both by incumbent organizations, and by entrepreneurial individuals within these organizations who then carry it over to new entities through the creation of new ventures. Entrepreneurship is an important conduit of knowledge spillovers, In its absence, that existing knowledge might not have been commercialized, so that there would have been no growth emanating from the investments in knowledge made by incumbent organizations. Importantly, such a conceptualization draws attention to the fact that economic performance occurs due to path-dependent action that is local or non-random in nature. 3. Methodology: Empirical Model, Dynamic Convergence, Data and Variables According to economic theory (Leibenstein, 1968), an enterprise can be categorized as technically efficient if it is able to produce maximum output given available resources. According to the literature, a gap normally exists between a firm’s actual and potential (feasible) levels of economic performance. Efficiency will be defined herein as the activity which produces maximum production given a certain set of resources, alternatively, the action which consumes the least possible volume of resources in order to achieve a certain volume of production. There are three different efficiency categories to consider: scale, assigned or technical, but this paper focuses on technical efficiency which measures total production volume produced with allocated (given) resources. In this context, Farrell (1957) proposed that the efficiency of a firm consists of two components: technical efficiency, which reflects the ability of a 5 firm to obtain maximal output from a given set of inputs, and allocative efficiency, which reflects the ability of a firm to use the inputs in optimal proportions, given their respective prices and the production technology. Although there is no consensus among researchers regarding the way to establish the process to evaluate the influence of capital entrepreneurship variables on technical efficiency levels, in this paper we have attempted to detect/link the repercussion of certain intermediate factors – like R&D activity - by using deterministic frontier production functions. In this context, a Stochastic Frontier Analysis (SFA) approach is applied to estimate technical efficiency rates for individual countries. In addition, kernel density functions are employed to analyze convergence (or divergence) in the efficiency estimated. SFA estimates an efficient frontier incorporating the possibility of measurement error or chance factors in its estimation. To separate inefficiency and noise, strong assumptions are needed on the distribution of noise among each observed firm. A production frontier reveals technical relationships between inputs and outputs of firms and represents an alternative when cost frontiers cannot be calculated due to lack of data. The estimated output is the maximum possible output for given inputs of an individual firm. The output difference obtained in the estimation is interpreted as technical inefficiency of each individual country. 3.1. Empirical Model Following to Coelli et al. (2005), a production function expresses one output as function of inputs. Mathematically, all these different functions can be written in the form: y  f x1 , x 2 ,...,x N  , where y is the dependent variable; the xn n  1,..., N  are the explanatory variables; and f  is a mathematical function. In this context, the first step in estimating the relationship between the dependent and explanatory variable is to specify the algebraic form of f  . In this study we use specifications such as the Cobb-Douglas (CD) function with constant returns to scale and the TransLog (TL) with variable elasticity of factor input substitution. Also we account for technological change by including a time trend as suggested by Coelli et al. (2005). The next expressions 1 (for CD specification) and 2 (for TL specification) account for technological change: N ln y  A0  ï?± t  ï?¢ n 1 n ln x n 1 6 N N N   ï?¢ 2 1 ln y  ï?¢ 0  ï?±1 t  ï?± 2 t 2  ï?¢ n ln x n  nm ln x n ln x m n 1 2 n 1 m 1 In 1 and 2 , t is a time trend; and ï?± , ï?±1 and ï?± 2 are unknown parameters to be estimated. Including time trends in the previous models makes implicit assumptions about the nature of technological change. Following to Coelli et al. (2005), the CD specification implicitly assumes that technological change is constant related to y; the TL model allows the technological change effect to increase or decrease with time (depending on whether ï?± 2 is positive or negative). The percentage change in y in each period due to technological change is given by the derivate of ln y with respect to t in 1 and 2 . Continuing with Coelli et al. (2005), one method for estimating a production frontier using data is to envelop the data points using an arbitrary- chosen function. In basic stochastic production frontier models, the output is specified as a function of a non-negative random error which represents technical inefficiency, and a symmetric random error which accounts for noise. Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977) proposed the stochastic production frontier model of the form: ln qi  x i' ï?¢  vi  ui 3 In 3 , q i represents the output of the i-th firm; x i is a K ï‚´ 1 vector containing the logarithms of inputs; ï?¢ is a vector unknown parameters; vi is a symmetric random error to account for statistical noise; and u i is a non-negative random variable associated with technical inefficiency. The resulting function is called stochastic production frontier because the output values are bounded from   above by the stochastic variable exp x i ï?¢  vi . The random error vi can be ' positive or negative and so the stochastic frontier outputs vary about the   deterministic part of the model, exp x i ï?¢ . In the case of firms that produce the ' output q i using one input, xi , the CD stochastic frontier model take the form: qi  expï?¢ 0  ï?¢1 ln xi  vi  ui  4 7 The most common output-oriented measure of technical efficiency is the ratio of observed output to the corresponding stochastic frontier output: TEi  qi   exp x i' ï?¢  vi  u i   exp u  5 exp  x i' ï?¢  vi  exp  x i' ï?¢  vi  i It measures the output of the i-th firm relative to the output that could be produced by a fully-efficient firm, using the same input vector. According to Coelli et al.(2005), panel data often allows us: i) to relax some of the strong distributional assumptions that were necessary to disentangle the separate effects of inefficiency and noise, ii) to obtain consistent predictions of technical efficiencies, and iii) to investigate changes in technical efficiencies over time. Panel data versions (Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977)) can be written in the general form: ln qit  x it ï?¢  vit  uit ' 6 The expression 6 is identical to the model 4 except we have added a subscript “tâ€? to represent time. The expectation is that inefficient firms improve their efficiency levels over time, with managers learning from experience, and for their technical efficiency levels to change systematically over time. One model (Battese and Coelli 1993) allows for time-varying technical inefficiency and takes the form: uit  f t   ut , where f  is a function that . determines how technical inefficiency varies time: f t   expï?›ï?¨ t - Tï?? 7 In 7  , ï?¨ is unknown parameter to be estimated. The Battese and Coelli (1993) function has the properties f t   0 and f T   1 , and is either non- increasing or non-decreasing depending on the sign of ï?¨ . However, it is convex for all values of ï?¨ . The Battese and Coelli (1993) model can be estimated under  the assumption that u i has a truncated normal distribution: u i ~ iddN ï?­ , ï?³ u . 2   3.2 . The Efficiency Distribution Dynamic 8 To understand the dynamic of the whole efficiency distribution, the intention is to use stochastic kernel estimators in much the same way as Birchenal and Murcia (1997) employed them to analyse convergence. Figure 1 illustrates this approach, showing a possible distribution of efficiency in two time periods, t and t+s. The distribution in period t indicates that there is an average efficiency level shared by most of the economies considered, and that there are few with extremely high or low efficiency. By contrast, t+s groups the most and least efficient economies to create two clearly differentiated groups, while the medium-efficiency groups have disappeared. Figure 1. Change in the efficiency distribution Source: Prepared on the basis of Birchenal and Murcia (1997). The arrows in Figure 1 show the internal dynamic of the distribution. For example, arrows 2 and 3 indicate the “mobilityâ€? of the economies within the distribution, and arrows 1 and 4 indicate the “persistenceâ€? of the economies that keep their original position between periods t and t+s. To analyse this dynamic without distorting it, the idea is to divide the efficiency space into an infinite number of regions or a continuum. In this case, the corresponding transition probability matrix will tend towards a continuum of rows and columns, becoming a stochastic kernel. 3.2. Data and Variables 9 Table 1 presents a summary of the key variables used to empirically validate the combined stochastic-inefficiency model for 61 countries from 2002 to 2005. Table 1. Function production variables Variables Definition GDP at purchaser's prices is the sum of gross value added by all resident producers in YNTx1 : GDP 1 (constant the economy plus any product taxes and minus any subsidies not included in the value 2000 US$) of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant 2000 U.S. dollars. Dollar figures for GDP are converted from domestic currencies using 2000 official exchange rates. For a few countries where the official exchange rate does not reflect the rate effectively applied to actual foreign exchange transactions, an alternative conversion factor is used. Gross fixed-capital formation (formerly gross domestic-fixed investment) includes land K NTx1 : Gross fixed improvements (fences, ditches, drains, and so on); plant, machinery, and equipment capital formation2 (constant purchases; and the construction of roads, railways, and the like, including schools, 2000 US$) offices, hospitals, private residential dwellings, and commercial and industrial buildings. According to the 1993 SNA, net acquisitions of valuables are also considered capital formation. Data are in constant 2000 U.S. dollars. Total labor force is comprised of people who meet the International Labor Organization LNTx1 : Labor force,3 total definition of the economically active population: all people who supply labor for the production of goods and services during a specified period. It includes both the employed and the unemployed. While national practices vary in the treatment of such groups as the armed forces and seasonal or part-time workers, in general the labor force includes the armed forces, the unemployed and first-time job-seekers, but excludes homemakers and other unpaid caregivers and workers in the informal sector. Cyclical and Hicks neutral technological progress. TNTx1 : Time Expenditures for research and development are current and capital expenditures (both R & DNTx1 : Research public and private) on creative work undertaken systematically to increase knowledge, and development including knowledge of humanity, culture, and society, and the use of knowledge for expenditure4 new applications. R&D covers basic research, applied research, and experimental development. New businesses registered are the number of new firms, defined as firms registered in NBR NT : New the current year of reporting. Businesses Registered5 Time-varying inefficiency effect. TNTx1 : Year Notes: 1 International Finance Corporation's micro, small, and medium-size enterprises database (http://www.ifc.org/ifcext/sme.nsf/Content/Resources). 2 World Bank national accounts data, and OECD National Accounts data files. 3 International Labour Organization, using World Bank population estimates. 4 United Nations Educational, Scientific, and Cultural Organization (UNESCO) Institute for Statistics. 5 International Finance Corporation's micro, small, and medium-size enterprises database (http://www.ifc.org/ifcext/sme.nsf/Content/Resources). Source: World Bank’s World Development Indicators (2009), and authors’ calculations. The data source used for this analysis is the World Bank’s World Development Indicators (WDI). This database provides more than 800 development indicators, with a time series for 209 countries and 18 country groups from 1960 to 2007. From the World Bank’s World Development Indicators (WDI), we have time series observations (T=4) for 2002, 2003, 2004 and 2005. We are able to form a balanced panel data, as seen in the descriptive statistics of variables in Table 2. 10 Table 2. Descriptive statistics of variables of the production frontier model 2002- 2005 Variable Year Mean Std. Dev. Min Max GDP (in millions) 2002 456684 1419363 9997600 2371 2003 466695 1451176 10249800 2703 2004 483326 1503767 10651700 2987 2005 497255 1547630 10995800 3402 Gross Fixed Capital 2002 91219 274765 1835000 448 Formation (GFKF) (in 2003 93539 280281 1890700 600 millions 2004 98363 293680 2004600 637 2005 103604 309418 2132800 664 Labor Force (LF) 2002 20738 55685 408342 168 (in miles) 2003 21005 56640 416285 169 2004 21353 57546 422759 173 2005 21656 58537 430607 177 Research and 2002 10336237 38943279 265884666 215 Development Expenditure 2003 10553995 39853975 272239510 354 (R&D) 2004 10714079 40311619 275048835 988 (in miles) 2005 11210290 42355334 287770196 1168 New Businesses 2002 49496 101292 650843 31 Registered (NBR) 2003 52404 104156 618503 57 2004 57433 111332 657195 24 2005 59459 114565 676830 23 Notes: Number of observations = 61 countries. Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). 4.- Empirical Results There is a wide difference and variation of expenditure on Research and Development – R&D (as a percentage of GDP) across countries (see Appendix tables 6 and 7). There are also differences on New Business Entry Rate – NBER (New Businesses Registered - NBR as % of total business) across countries when grouped in terms of High-Medium and Medium-Low-Tech (see Table 3). Some examples are Sweden4 (high expenditure on R&D and low NBER), Turkey5 (high NBER and low expenditure on R&D), India6 (low expenditure on 4 The industrial sector plays an important role in Sweden’s economy. Recently the Agricultural sector has also contributed a lot to the country's Gross Domestic Product. The major industries in the country are iron, steel, wood pulp, paper products, and motor vehicles. The important agricultural products in the country are wheat, barley, sugar and milk. 5 In the case of Turkey the strong and rapidly growing private sector is a landmark of its economic success. 6 Agriculture is a major component of India’s economy. The industrial sector includes manufacturing industries, textiles and handicrafts, etc. However, the service sector is greatly expanding and has started to assume an increasingly important role (e.g., India has become a hub 11 R&D and low NBER) and Germany7 (high expenditure on R&D and high NBER). Table 3. Average of variables of production frontier model 2002-2005 by technological level Sample GDP GFKF LF R&D Entrepreneurship capital 2002-2005 Total Total Total Total % GDP NBR NBER All 475990 96681 21188 10703.65 1.49 54698 8.83 High-Medium Tech 996684 199474 19562 24907.70 2.82 91304 9.91 countries1 Medium-Low-Tech 114397 25298 22317 839.73 0.565 29277 8.07 countries2 Notes: 1 High-Medium Tech countries: R&D (% GDP) > 1 (year 2002) 2 Medium-Low-Tech countries : R&D (% GDP) < 1 (year 2002) Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). With a medium level on R&D and NBER, Italy has an robust small and medium enterprise sector (77.8% of total firms), but it has not been as successful in establishing multinational corporations8 . 4.1. Efficiency and Productivity Analysis When comparing efficiency levels, we can see large differences between countries (see Appendix Tables I and II). The above differences are greater when comparing High-Medium and Medium-Low-Tech countries (see Table 4). Using a TransLog production function to estimate Technical Efficiency, we observe important differences across countries (see Appendix Tables I and II). We can see that the U.S. has a high level of technical efficiency while Bolivia has a very low level of technical efficiency, and can/should significantly improve its performance, making better use of its resources. In between we have the cases of Japan,9 United Kingdom10 and Canada11 (high efficiency), Georgia,12,Mexico13 of outsourcing activities in the areas of technical support and customer services for some of the major economies of the world). 7 In Germany’s economy, the average annual growth rate of GDP has been on a decline since the 1980s. Germany’s economy is currently recovering, ending a phase in stagnation on the back of its traditionally strong, competitive and innovative export-oriented manufacturing sector. 8 Many of these companies do not have a high level of technology sophistication. Italian services today make up 69% of the economy, industry 29%, and agriculture 2%. 9 Japan's economy is the second largest economy in the world and the largest in Asia, based on real GDP, market exchange rates, and nominal GDP. Japan uses planned development of science and technology, and has a strong work culture. However, in the 1990s Japan experienced a “Lost Decadeâ€?, a period when the Japanese economy was stagnant. 10 The main economic activity of United Kingdom is the service sector (76.2% of GDP in 2008). Industry and manufacturing (22.8%) and agriculture (0.9%) are other important industrial sectors. 12 and New Zealand14 (medium efficiency) and Iceland15 (low efficiency). There are determinants of the ranking and there are some plausible explanations in the corresponding footnotes. Table 4. Estimates of Technical Efficiency, Marginal effects, and inefficiency Function TransLog (TL) Cobb-Douglass (CB) Mean 2002- Efficiency Marginal Inefficiency Efficiency Marginal Inefficiency 2005 Effects E(u/e) Effects E(u/e) All sample 0.937 0.068 0.068 0.991 0.011 0.011 High-Medium 0.953 -0.824 0.049 1.000 -0.872 0.000 Tech countries1 Medium-Low- 0.926 0.688 0.082 0.985 0.624 0.018 Tech countries2 Notes: 1 High-Medium Tech countries: R&D (% GDP) > 1 (year 2002) 2 Medium-Low-Tech countries : R&D (% GDP) < 1 (year 2002) Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). United Kingdom’s economy has a large trade deficit in manufacturing and has become a net importer of energy. 11 Canada has moved from agriculture straight to services (this industry is very diverse and employs 75% of the total million working population). Manufacturing has never been a dominant sector of the Canadian economy, but it has been an important secondary industry. 12 The main economic activity of Georgia is agriculture. Mining, construction, financial services and communication are other sector making significant contribution towards Georgia’s GDP. Georgia has a good supply of hydropower, however it imports a major part of its energy resources. 13 Mexico has one of the largest economies in the world. The industrial sector in Mexico is very heterogeneous (the industrial sector combines technologically advanced businesses and antiquated industries). The agricultural sector is also an important part of the Mexican economy. The private sector has started assuming an increasingly important role in both the agricultural and the industrial sectors. 14 Manufacturing and creative media largely constitute the New Zealand economy. Some of the major industries of New Zealand include iron and steel, natural gas processing, printing, publishing and recorded media, wood processing, cement, and fishing. Other minor industries in New Zealand include paper, tanning, transport equipments, wine making, tourism, and timber trade. Manufacturing industries in New Zealand contribute over 15% of GDP and over 44% of export receipts. Agriculture also contributes significantly to the economic growth of New Zealand. 15 Iceland has a Scandinavian-type economy. This means that the main economic activity of Iceland is the fishing industry. Iceland also exports animal products and aluminum. New businesses in Iceland are tourism, software production, financial services and biotechnology. 13 Table 5. Maximun Likelihood Estimates (MLE): Cobb-Douglass (CD) and TransLog (TL) Stochastic Production Frontier Model Function TransLog (TL) Cobb-Douglass (CD) Mo- Variable1/Tech- All Medium- High- All Medium- High- del nological level Low-Tech Medium Low-Tech Medium Tech Tech Dep. GDP Const../ Const./ Const./ Const./ Const./ Const./ Var. Std. Err. Std. Err. Std. Err. Std. Err. Std. Err. Std. Err. Prod. Constant 0.079* 0.055* 0.154*** 0.011 0.141 3.97E-02 Fron- [0.048] [0.033] [0.041] [0.013] [0.104] [0.039] tier 1.007*** 0.975*** 0.928*** 0.991*** 0.936*** 9.85E-01*** Kt [0.013] [0.022] [0.023] [0.010] [0.032] [0.025] 0.038*** 0.133*** 0.078** 0.043*** 0.104*** 0.025 Lt [0.014] [0.027] [0.032] [0.012] [0.020] [0.034] -0.059*** -0.043** -0.019 -0.026** -0.054 -0.029 Tt [0.0203] [0.0203] [0.021] [0.011] [0.035] [0.023] -0.043*** 0.005 -0.120*** - - - K t2 [0.016] [0.025] [0.037] K t  Lt 0.043*** 0.026 0.108*** - - - [0.016] [0.026] [0.039] K t  Tt -0.009 -0.021 0.009 - - - [0.009] [0.015] [0.013] Lt  Tt 0.000 0.003 -0.017 - - - [0.011] [0.015] [0.014] -0.028 0.033 -0.117*** - - - L2 t [0.018] [0.031] [0.035] 0.021 0.013 -0.002 - - - Tt2 [0.029] [0.033] [0.025] constant -5.665*** -9.836*** -8.930*** -63.278* -4.689 -1.01E+01** u it [1.803] [2.733] [2.210] [32.711] [3.487] [4.893] 0.248 3.420** -4.304*** 11.164 0.415 -3.61E+00* R & Dt [0.398] [1.473] [1.327] [7.598] [0.952] [2.001] 0.265 1.477 1.113 3.236 -0.385 1.52E+00 NBRt [0.400] [1.009] [1.264] [9.564] [0.878] [1.567] -1.081 -1.088 -0.091 -0.338 -0.609 7.55E-03 Tt [0.675] [0.901] [0.991] [1.287] [1.774] [1.665] 0.061 -0.468 -1.540*** 0.299 -0.118 -1.02E+00 R & Dt2 [0.138] [0.362] [0.564] [0.907] [0.293] [0.681] 0.420 0.416 0.007 13.744 -0.109 0.926 NBRt2 [0.489] [0.424] [0.701] [11.211] [0.414] [1.249] R & Dt  NBRt -0.078 -0.830 0.345 -1.534* 0.027 9.57E-02 [0.074] [0.524] [0.602] [0.909] [0.387] [0.880] - 0.161 0.038 - -0.030 1.05E-01 R & Dt .Tt [0.272] [0.363] [0.509] [0.496] - -0.055 -0.085 - -0.183 -1.01E-01 NBRt .Tt [0.282] [0.276] [0.649] [0.394] constant -3.426*** -3.524*** -4.627*** -3.319*** -3.154*** -4.11E+00*** vit [0.125] [0.158] [0.252] [0.092] [0.200] [0.169] Sigma-squared 0.180 0.172 0.099 0.190 0.207 1.28E-01 [0.011] [0.013] [0.012] [0.009] [0.0207] [0.011] Wald chi2 15408.11 4980.84 21248.55 19794.51 1728.23 7729.43 Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 Log likelihood 57.275 27.675 71.755 56.894 15.317 59.712 Number of obs 244 144 100 244 144 100 1 Description in natural logs. Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). 14 When we consider all countries, Table 5 shows that the combined effect of R&D expenditure and New Businesses Registered (NBR) has a positive and significant effect to reduce the Technical inefficiency (in the Cobb-Douglass specification). We observe a very different behavior between countries with Medium- Low-Tech level and countries with High-Medium Tech level (using a TransLog production function). In countries with High-Medium Tech level, the R&D expenses have a positive and significant effect on technical efficiency (in the short and long term). In contrast, in countries with Medium-Low Tech level, the R&D expenses have a negative and significant effect on technical efficiency (in the short term). When we use a Cobb-Douglass production function, in countries with High-Medium Tech level, the R&D expenses have a positive and significant effect on technical efficiency (only in the short term). Finally, since earlier GEM reports demonstrated a systematic, U-shaped relationship between a country’s level of economic development and its level and type of entrepreneurial activity, there might be a potential reverse-causality issue. To address this issue, in applying a second step SFA procedure, we performed the Granger causality test which showed a statistically significant positive value of the Wald test on all coefficients of distributed lags16 but only in the TransLog specification. 4.2. Efficiency and Dynamic Convergence Analysis The issue here is to analyze the trend and dynamics of the distribution of technical efficiency, along the lines suggested by Quah (1993, 1997), that is, changes in the form of the distribution and distributional dynamic within that distribution,17 based on the estimation of kernel density functions as proposed by Lucy, Aykroyd and Pollard (2002). The results are shown in Figure 2 and illustrate the trends in convergence (divergence) and persistence (mobility) in the level of technical efficiency attained by the countries during the period 2002-2005. 16 The ratio between R&D expenses and the annual number of new firms - in the first OLS regression - and a non- statistically significant value of the Wald test on all coefficients of distributed lags of country’s technical efficiency - in the second OLS regression. 17 Quah (1997) argues that convergence coalitions, or clubs, can form endogenously across all countries, and the different convergence dynamics will depend on the initial distribution of country characteristics. 15 Figure 2. Bivariate Kernel Density (a) Total sample (b) High-Medium Tech countries (c) Medium-Low-Tech countries Source: Authors’ calculation. 16 The interpretation of the graphs is as follows. If the whole distribution maintains its characteristics between periods t and t+s, we say the distribution of technical efficiency is persistent, that is, efficient countries remain efficient and inefficient ones remain inefficient. For the distribution of technical efficiency to show mobility, it would have to show a complete (at the extreme) reversal of the countries’ starting conditions, so that those deemed inefficient in period t would become efficient in period t+s, while those deemed efficient would become inefficient.18 Lastly, if the distribution clusters around a plane parallel to the t axis over time, whereas efficiency was distributed normally in the whole of the cross- section to begin with (i.e., with grouping around the value t+s=1), the distribution is said to be converging on equality in the countries’ efficiency levels. Focusing on TL specification for the entire sample (the variance in efficiency index is bigger than CD specification), what seems to come out from the data is a pattern of non-mobility regarding the efficiency level attained over the years, with a degree of convergence upon higher efficiency levels (this is reflected in the decreasing proximity of the dots marking out the different level curves to the axis drawn across the graphs). For the High-Medium Tech countries technical efficiency over the whole period, we detect a pattern of divergence and mobility. And for the Medium-Low- Tech countries technical efficiency index over the whole period, we find a pattern of convergence and non-mobility with polarization towards higher technical efficiency values. 4. Conclusions and Policy Implications We have shown the impact of R&D investments and entrepreneurship and of its interplay on economic performance. The results point to the positive effects of entrepreneurship on economic performance, and particularly when linked to R&D. Specifically, we have shown that the rate of R&D expenditures in relation to New Businesses Registered (NBR) has a positive and significant effect to reduce the technical inefficiency. Moreover, the dynamic analysis hints that, with the proper policies in those two areas, countries could, over time, significantly improve their economic performances as measured by technical efficiency. The empirical results of this work indicate that not only the traditional factors associated with economic performance and growth are important, but in 18 According to Birchenal and Murcia (1997), a simple way of appreciating these things is to observe whether the outlines of the distribution are concentrated on the 45 degree line marked on the t–t+s plane (in this case, the distribution persists during the periods). If the outlines of the distribution are concentrated on a line perpendicular to the 45 degree line, there is total mobility within the distribution. 17 addition, entrepreneurial activity also plays an important role in generating economic efficiency and in fostering a conducive environment for productivity, competitiveness and growth. In this context, the level of Expenses on R&D in relation to NBR plays a decisive role as a determinant of levels and change in technical efficiency. Thus the policy implications are clear. In the context of the endogenous growth theory, the focus of public policy ought to shift towards policies and instruments that would increase investments in knowledge and in human capital, as well as research and development and facilitation for the formation of new firms and start ups. Some of the specific policies that could be promoted would be: ï‚· Developing and nurturing entrepreneurship and innovation through placing entrepreneurship modules in the curriculum of engineering and business schools; ï‚· Celebrating innovation and entrepreneurship by establishing media programs and highly visible awards; ï‚· Facilitating the creation of technology transfer offices at leading universities or through a consortia of universities; ï‚· Implementing programs to facilitate and finance start-ups, particularly technology based; ï‚· Fostering networks of incubators; ï‚· Providing a coherent fiscal and financial incentives for R&D and for spin-offs; ï‚· Inciting the development of supporting R&D infrastructure and networks of knowledge; ï‚· Providing technological based training and knowledge transfer programs. These actions should also be complemented by revising and simplifying procedures and costs for the registering of new firms, now still quite cumbersome and costly in many developing countries. Authors like Suyanto, Salim and Bloch (2009) suggest that policies for strengthening the absorptive capacity of domestic entrepreneurship through investing in knowledge and human capital formation are critical and perhaps superior to those oriented at the development of entrepreneurship. As far as issues for future research in these themes, rethinking and codifying what exactly constitutes entrepreneurship capital and how public policy can more effectively and directly contribute to its formation would be most 18 useful. Another critical area would be to explore the linkages between the creation of entrepreneurial opportunities, their implementation through launching a new venture, and the subsequent impact on regional economic growth and development. Such analysis would help to quantify the overall performance consequences and social welfare gains of knowledge spillover through strategic entrepreneurship Finally, on the econometric methodology for these type of analyses, it would be useful to use others’ non-parametric approaches in order to evaluate how robust these findings are, since while the SFA accounts for data noise, such as data errors and omitted variables, the separation of noise and inefficiency relies on strong assumptions on the distribution of the error term. References Acs Z. and Audretsch D. (2003). The International handbook of entrepreneurship. Kluwer Academic Publishers, Dordrecht. Agarwal, R., Audretsch, D. B. and Sarkar, M. B. (2007). “The Process of Creative Construction: Knowledge Spillovers,â€? Strategic Entrepreneurship Journal, Forthcoming. __________ 2004). “Knowledge Transfer through Inheritance: Spin-out Generation, Development and Performance,â€? Academy of Management Journal, 47: 501-522. Agarwal, R. and Bayus, B. L. (2002). “The Market Evolution and Sales Takeoff of Product Innovations,â€? Management Science, 48(8): 1024-1041. Agarwal, R., Audretsch, D. and Sarkar, M. (2010). “Knowledge spillovers and strategic entrepreneurship,â€? Strategic Entrepreneurship Journal, 4 (4): 271–283. Aigner, D., Lovell, C.A.K. and Schmidt, P. (1977). “Formulation and Estimation of Stochastic Frontier Production Function Models,â€? Journal of Econometrics, 6: 21-37. Audretsch, D.B. (1995). Innovation and Industry Evolution. MIT Press, Cambridge MA. Audretsch, D. and Keilbach, M. (2003). “Entrepreneurship capital and Economic Performance,â€? Regional Studies, 38 (8): 949 – 959. Battese, G.E. and Coelli, T.J. (1993). “A Stochastic Frontier Production Function Incorporating a Model for Technical Inefficiency Effects,â€? Working Papers in Econometrics and Applied Statistics, No. 69, Department of Econometrics, University of New England, Armidale. Birchenal, J.A. and Murcia, G.E. (1997). “Convergencia regional: una revisión del caso colombiano,â€? Archivos de macroeconomía, 69, Bogotá, D.C., National Planning Department. 19 Caves, R. (1998). “Industrial Organization and New Findings on the Turnover and Mobility of Firms,â€? Journal of Economic Literature, 36: 1947-1982. Coelli, T., Rao, P., O’Donnell, C. and Battese, E. (2005). An introduction to Efficiency and Productivity Analysis. Second Edition. Springer. USA. Colombo, M.G. (2005). “Founders’ human capital and the growth of new technology-based firms: A competence-based view,â€? Research Policy, 34(6): 795-816. Farrell, M.J. (1957). “The Measurement of Productive Efficiency,â€? Journal of the Royal Statistical Society, 120(3): 253-290. Leibenstein, H. (1968). “Entrepreneurship and development,â€? American Economic Review, 38 (2): 72-83. Lucy, D., Aykroyd, R.G. and Pollard, A.M. (2002). “Nonparametric calibration for age estimation,â€? Applied Statistics, 52 (2). 185-196. Meeusen, W. and van den Broeck, J. (1977). “Technical Efficiency and Dimension of the Firm: Some Results on the Use of Frontier Production Functions,â€? Empirical economics, 2 (2): 109-122. Quah, D. (1997). “Empirics for growth and distribution: stratification, polarization, and convergence clubs,â€? Journal of Economic Growth, 2 (1): 27-59. __________ (1993). “Galton's fallacy and tests of the convergence hypothesis,â€? Scandinavian Journal of Economics, 95 (4): 427-443. Romer, P. (1990). “Endogenous Technological Change,â€? The Journal of Political Economy, 98 (5): 71-102. Saxenian, A. (1994). Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Harvard University Press, Cambridge, MA. Somaya, D., Williamson, I. and Lorinkova, N. (2007). “The effects of Employee Mobility Between Competitors and Cooperators on Firm Performance,â€? Intellectual Property Research Institute of Australia Occasional paper number 1/07. Suyanto, R., Salim, R.A. and Bloch, H. (2009). “Does Foreign Direct Investment Lead to Productivity Spillovers? Firm Level Evidence from Indonesia,â€? World Development, Article in Press, doi:10.1016/j.worlddev.2009.05.009. Teece, D. J., Pisano, G. and Shuen, A. (1997). “Dynamic Capabilities and Strategic Management,â€? Strategic Management Journal, 18(7): 509-533. 20 Appendix Table 6 Basic statistic High-Medium Tech countries: Mean 2002-2005 Mean 2002-2005 GDP GFKF LF RDTotal RD NBR NBER Efficiency mfx TL Inefficiency Efficiency mfx Inefficiency % GDP TL TL CD CD CD Australia 450718 117059 10106 7827.95 1.74 78448 9.00 0.969 -1.932 0.032 1 -1.94 0 Austria 202210 42790 3957 4550.24 2.25 13303 8.06 0.954 -0.884 0.048 1 -0.907 0 Belgium 243266 49457 4428 4592.78 1.89 23351 7.18 0.956 -1.036 0.046 1 -1.055 0 Canada 788585 163060 17316 15852.70 2.01 84052 6.07 0.967 -2.27 0.034 1 -2.293 0 Croatia 21827 6039 1959 241.04 1.11 7039 6.66 0.947 1.171 0.055 1 1.07 0 Czech Republic 63260 18320 5166 810.58 1.28 30945 11.31 0.961 -0.001 0.04 1 -0.078 0 Denmark 165430 33530 2860 4145.09 2.51 21918 10.24 0.955 -0.628 0.047 1 -0.65 0 France 1400735 272224 26919 30384.72 2.17 130423 11.20 0.973 -2.799 0.027 1 -2.823 0 Georgia 3865 1542 2316 8.19 0.21 3650 7.14 0.959 2.517 0.042 0.988 2.438 0.012 Germany 1937468 366735 40625 48390.03 2.50 73416 16.26 0.969 -3.098 0.032 1 -3.136 0 Iceland 9723 2224 172 277.61 2.86 2741 12.94 0.905 2.33 0.103 1 2.19 0 Ireland 115643 26987 1940 1388.19 1.20 15247 9.81 0.956 -0.398 0.046 1 -0.416 0 Italy 1128176 236345 24254 12483.23 1.11 102575 6.26 0.966 -2.654 0.036 1 -2.679 0 Japan 4826556 1117403 66943 155185.37 3.21 110958 4.34 0.984 -4.261 0.016 1 -4.261 0 Luxembourg 22418 4906 197 368.04 1.64 2199 11.27 0.886 1.434 0.127 1 1.367 0 Macedonia, FYR 3637 656 856 889.21 24.45 9713 6.70 0.918 3.424 0.09 1 3.299 0 Netherlands 399163 80179 8461 6989.90 1.75 98500 10.04 0.958 -1.538 0.043 1 -1.562 0 New Zealand 57919 13911 2105 685.08 1.18 55750 16.91 0.953 0.307 0.049 1 0.238 0 Norway 180168 32819 2482 2920.18 1.62 42073 14.66 0.951 -0.605 0.051 1 -0.621 0 Russian Federat. 317857 59584 73366 3753.84 1.19 372577 8.64 0.94 -1.266 0.064 1 -1.353 0 Slovenia 22101 5798 997 319.43 1.45 2698 7.49 0.927 1.236 0.079 1 1.134 0 Sweden 259794 43463 4674 9995.17 3.85 18568 6.32 0.948 -0.897 0.055 1 -0.928 0 Switzerland 253141 55655 4159 7413.53 2.93 12781 9.66 0.957 -1.165 0.045 1 -1.17 0 United Kingdom 1569712 270380 30323 27984.48 1.78 318825 16.62 0.979 -2.787 0.022 1 -2.821 0 United States 10473725 1965775 152467 275235.80 2.63 650843 12.87 0.999 -4.804 0.001 1 -4.855 0 Note: GDP, GFKF and Total R&D in thousands, and LF in millions. Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). 21 Table 7 Basic statistic Medium-Low-Tech countries: Mean 2002-2005 Mean 2002-2005 GDP GFKF LF RDTotal RD NBR NBER Efficiency mfx TL Inefficiency Efficiency mfx Inefficiency % GDP TL TL CD CD CD Algeria 64382 13806 12711 123.24 0.20 12138 14.41 0.933 0.24 0.072 1 0.166 0 Argentina 276607 40192 17813 1180.97 0.42 43500 10.96 0.938 -0.788 0.065 1 -0.864 0 Armenia 2866 710 1298 6.53 0.23 8914 7.65 0.917 3.32 0.09 0.989 3.231 0.011 Bolivia 9221 1290 3966 25.51 0.28 1634 7.37 0.798 2.593 0.245 0.75 2.563 0.309 Botswana 7496 1316 665 28.87 0.39 7549 11.23 0.837 2.771 0.194 1 2.619 0 Chile 86135 19282 6407 580.11 0.67 29044 17.24 0.953 -0.052 0.049 1 -0.129 0 Costa Rica 18032 3309 1854 65.84 0.37 40193 10.82 0.906 1.773 0.104 1 1.662 0 Estonia 7377 2315 666 61.95 0.83 7858 11.86 0.944 2.206 0.059 1 2.066 0 Finland 132214 25261 2634 4535.38 3.43 7343 6.50 0.941 -0.327 0.062 1 -0.367 0 Greece 167682 41299 5020 828.99 0.49 2289 7.20 0.934 -0.842 0.07 1 -0.88 0 Guatemala 21410 3400 3954 6.44 0.03 3924 6.21 0.826 1.683 0.205 0.977 1.602 0.023 Hong Kong, China 187965 44928 3512 1268.25 0.67 59706 11.32 0.955 -0.939 0.047 1 -0.95 0 Hungary 55602 13657 4221 519.82 0.94 21584 9.88 0.953 0.306 0.049 1 0.223 0 India 570124 151914 419498 4058.18 0.71 31435 4.77 0.969 -2.236 0.032 1 -2.348 0 Jordan 10329 2148 1691 34.79 0.34 6028 6.47 0.903 2.214 0.106 1 2.104 0 Kazakhstan 26244 5565 7828 68.46 0.26 2896 9.66 0.916 1.166 0.09 1 1.106 0 Latvia 10194 3128 1120 45.75 0.44 8656 4.80 0.946 1.881 0.056 1 1.76 0 Lithuania 14858 3317 1637 106.41 0.71 3811 5.85 0.913 1.787 0.095 1 1.674 0 Madagascar 4005 756 8243 9.06 0.23 1048 5.92 0.932 2.995 0.073 0.846 3.115 0.182 Mexico 607966 123545 41601 2811.43 0.46 306400 7.14 0.942 -1.971 0.061 1 -2.057 0 Morocco 44488 11747 10682 269.04 0.60 11342 8.24 0.954 0.409 0.048 1 0.335 0 Pakistan 85312 12600 53907 284.58 0.33 2478 5.89 0.87 0.196 0.145 0.965 0.193 0.039 Peru 60307 11074 12573 71.60 0.12 27621 5.39 0.925 0.455 0.081 1 0.387 0 Philippines 87462 14535 35427 123.49 0.14 13328 - 0.912 0.108 0.096 1 0.069 0 Poland 187777 36071 17235 1046.32 0.56 23683 4.79 0.947 -0.72 0.056 1 -0.8 0 Portugal 116059 28018 5498 891.49 0.77 15923 6.17 0.953 -0.437 0.05 1 -0.501 0 Romania 45067 9370 10309 176.37 0.39 76152 10.27 0.937 0.629 0.068 1 0.56 0 Senegal 5420 1312 4507 4.86 0.09 34 3.30 0.996 2.562 0.004 1 2.545 0 Slovak Republic 23562 6248 2666 127.58 0.54 5027 7.09 0.937 1.118 0.067 1 1.019 0 South Africa 150311 24743 19540 1293.88 0.86 33484 6.35 0.942 -0.351 0.062 1 -0.429 0 Spain 648385 177132 19839 6863.20 1.06 128168 6.37 0.969 -2.354 0.032 1 -2.382 0 Sri Lanka 18234 3946 8054 34.21 0.19 3990 7.65 0.911 1.47 0.098 1 1.424 0 Tunisia 22495 5281 3680 192.60 0.85 5757 10.43 0.928 1.261 0.078 1 1.169 0 Turkey 297732 58288 25612 2049.28 0.68 70560 11.66 0.954 -1.191 0.048 1 -1.272 0 Ukraine 41178 8558 22812 435.11 1.06 27670 6.28 0.952 0.649 0.051 1 0.623 0 Zambia 3787 651 4743 0.68 0.02 2815 5.20 0.887 3.18 0.128 0.928 3.233 0.082 Note: GDP, GFKF and Total R&D in thousands, and LF in millions. Source: Authors’ calculation from the World Bank’s World Development Indicators (WDI). 22