WPS5931 Policy Research Working Paper 5931 Innovative and Absorptive Capacity of International Knowledge An Empirical Analysis of Productivity Sources in Latin American Countries Leopoldo Laborda Castillo Daniel Sotelsek Salem Jose Luis Guasch The World Bank Latin America and Caribbean Region Finance & Private Sector January 2012 Policy Research Working Paper 5931 Abstract This paper examines two sources of global knowledge factor productivity) is decomposed using a generalized spillovers: foreign direct investments and trade. Empirical Malmquist output oriented index, in order to evaluate evidence demonstrates that foreign direct investment the specific effect in technical change, technical efficiency and trade can contribute to overall domestic productivity change, and scale efficiency change. Using country-level growth only when the technology gap between data for 16 Latin American countries for 1996–2006, the domestic and foreign firms is not too large and when empirical analysis shows positive productivity spillovers a sufficient absorptive capacity is available in domestic from foreign direct investment and trade only when firms. The paper proposes the terms research and the country has absorptive capacity in terms of research development and labor quality to capture the innovative and development. Foreign direct investment and trade and absorptive capacity of the country. The spillover spillovers are found to be positive and significant for scale effects in productivity are analyzed using a stochastic efficiency change and total productivity factor change. frontier approach. This productivity (in terms of total This paper is a product of the Finance & Private Sector, Latin America and 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 INNOVATIVE AND ABSORPTIVE CAPACITY OF INTERNATIONAL KNOWLEDGE An empirical analysis of productivity sources in Latin American countries Leopoldo Laborda Castillo* World Bank and Institute of Latin American Studies University of Alcalá (Spain) Daniel Sotelsek Salem*** Institute of Latin American Studies University of Alcalá (Spain) Jose Luis Guasch**1 World Bank and University of California, San Diego (USA) Keywords Technical Efficiency, Stochastic Frontier Analysis (SFA), Malmquist index. * 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 ** Professor of Economics at the University of Alcalá (Spain). e-mail: daniel.sotelsek@uah.es *** Senior Economist at the World Bank and Professor of Economics at University of California, San Diego (USA) e-mail: jguasch@worldbank.org 1 The authors are thankful to Esperanza Lasagabaster and two anonymous referees for very useful comments. 1 1. Introduction Damijan et al. (2003) examine different channels of global technology transfer to transition countries. These authors study the impact of direct technology transfer through foreign direct investment (FDI), intra-industry knowledge spillovers from FDI, firms‘ own research and development (R&D) accumulation and R&D spillovers through trade for total factor productivity (TFP) growth of local firms. Using firm-level data for eight transition countries for the period 1994–1998, this research found that technology is being primarily transferred to local firms through direct foreign linkages. In the specific context of Latin American countries, Ramirez (2010) estimates whether FDI flows and other relevant variables have had a positive and significant effect on private investment spending over the 1980–2002 period. On the other hand, Schiff and Wang (2010) examine the impact on TFP in Latin America and the Caribbean (LAC) and in other developing countries (DEV) of trade-related technology diffusion from the North, education and governance. The main findings are: i) education and governance have a much larger direct effect on TFP in LAC than in DEV, while the opposite holds for the North‘s R&D; and ii) education and governance have an additional impact on TFP in R&D-intensive industries through their interaction with trade-related technology diffusion from the North in LAC but not in DEV. According to Suyanto et al. (2009), the mixed evidence of productivity spillovers leads to the celebrated argument that firm-specific characteristics or absorptive capacity may influence the ability of domestic firms in gaining productivity spillovers from FDI and trade. The most commonly used measure of absorptive capacity in the literature about this topic is the extent of R&D expenditure (Findlay, 1978; Glass and Saggi, 1998; Wang and Blomstrom, 1992). Kathuria (2000) shows – in the context of the Indian manufacturing sector - that local firms that invest in learning or R&D activities receive high productivity spillovers, whereas the non-R&D local firms do not gain much from the presence of foreign firms. This result indicates that the productivity spillovers are not automatic consequences of the presence of foreign firms; rather they depend on the efforts of local firms‘ investment in R&D activities. Kinoshita (2001) finds similar evidence in a study on Czech manufacturing firms during 1995–98. Griffith et al. (2004) also confirm that R&D plays an important role in knowledge transfer, besides its role as a medium of innovation. In this context, empirical evidence demonstrates that FDI and trade can contribute to overall domestic productivity growth when the technology gap between domestic and foreign firms is not too large and when a sufficient absorptive capacity is available in domestic firms, in terms of R&D and Labor Quality (Borensztein et al., 1998). There is evidence that positive developmental impacts of FDI flows are conditional on high levels of human capital and thus on the existence of ‗good‘ infrastructure in recipient countries (Yamin and Sinkovics, 2009). For Suyanto et al. (2009, p. 4), ―the empirical studies usually assume that productivity advantage from FDI is exclusively contributed by technology transfers as is consistent with the use of conventional approach of production function. Technical and scale efficiencies are hardly studied in relation to productivity gains from FDI.‖ In this context Smeets (2008) argues that the productivity spillovers from FDI and trade should be defined broadly, as they arise from new knowledge rather than from new technology only. Smeets defines knowledge as including technology; managerial, 2 and production skills, which may contribute to technical efficiency and the ability to exploit scale efficiency. This line of research finds its basis in the pioneer works of Farrel (1957) and Aigner et al. (1977), which searched for a measurement of efficiency through the decomposition of the growth of productivity. We make use of the stochastic frontier analysis (SFA) approach to estimate productivity spillovers in Latin American countries for the period 1996-2006. In a second step, we compute the Malmquist index to decompose total factor productivity (TFP) growth into technical efficiency change (TEC), technological progress (TP), and scale efficiency change (SEC). In this context Orea (2002) provides a parametric decomposition of a generalized Malmquist productivity index that takes scale economies into account. (See Section 3 for more details about this approach.) This paper is organized as follows: in the next section we present a critical review of the theoretical and empirical studies on productivity spillovers. In Section 3 we develop the methodology of the analysis. Section 4 presents an empirical analysis using country level evidence from Latin American countries (we present the data sources, construction of variables, and the main empirical results obtained). Section 5 ends with a summary of the main conclusions and policy implications. 2. Background: Theory and evidence According to Yao and Wei (2009), although FDI and trade are widely believed to have a positive effect on economic growth, the exact mechanism of how FDI and trade impact upon the development process of the newly industrializing economies is far from being well understood. Three approaches provide theoretical explanations regarding this issue: (1) industrial organization theories, (2) international trade theories, and (3) endogenous growth theories. The industrial organization approach investigates explicitly the role of FDI and trade in technology transfer and the diffusion of knowledge, as well as the impact of FDI and trade on market structure and competition in host countries (Findlay, 1978; Das, 1987; Dunning, 1993; Aitken and Harrison, 1999; Cheung and Lin, 2004). In particular, the effect of trade competition may result in either positive or negative productivity spillovers for domestic firms. Aitken and Harrison (1999) argue that, in the short-run, the presence of foreign firms in an imperfectly competitive domestic market may raise the average cost of production of domestic firms through the ‗‗market stealing‖ phenomenon. Foreign firms with a lower marginal cost have an incentive to increase production relative to their domestic competitors. The productivity of domestic firms will fall as they have to spread fixed costs over a smaller amount of output. However, in the long-run, when all costs can be treated as variable costs, there is a possibility for domestic firms to reduce their costs by allocating their resources more efficiently and imitating foreign firms‘ knowledge (Wang and Blomstrom, 1992). If the efficiency effect from foreign presence is larger than the competition effect, there can be positive productivity spillovers. On the other hand, Caves (1971, 1996) argues that firms must possess specific advantages in order to overcome the difficulties of doing business abroad. Specifically, Caves suggests that when multinational corporations establish subsidiaries overseas, they experience disadvantages in the form of access to resources and domestic demand, when compared to their local counterparts. In order to compete with the domestic firms, multinational corporations need to possess superior knowledge. With this superior knowledge, multinational corporations are often assumed to have 3 higher performance levels than domestic firms, being more efficient and productive, in particular. The firms investing in foreign countries therefore have distinctive characteristics that may differ from firms in host countries. FDI is not merely a source of capital, it is also a conduit for technology transfer and human skills augmentation in host countries. As a result, the effect of competition, demonstration and learning-by-doing on local industry may lead to an increase in productivity (Blomstrom and Kokko, 1996; Blomström and Sjöholm, 1999). Carstensen and Toubal (2006) show that the traditional determinants, such as market potential, low relative unit labor costs, a skilled workforce and relative endowments, have significant and plausible effects. In addition, transition-specific factors, such as the level and method of privatization and country risk, play important roles in determining the flows of FDI into the ―transition economies‖ and help to explain the differing attractiveness of individual countries to foreign investors. In international trade theories, the main focus is to examine why FDI occurs and how firms choose between exporting, FDI and licensing as an entry mode (Brainard, 1993). Empirical evidence underscores the importance of international trade as a vehicle of international knowledge spillovers to developing countries (Co et al., 1997; Gorg and Strobl, 2005). International trade works as a channel of technology transfer, either through imports of intermediate products and capital equipment or through learning-by-exporting into industrial countries (Jacquemin and Sapir, 1991; Kinoshita, 2001). Kohpaiboon (2006) examines technology spillover from foreign direct investment (FDI) based on a cross-industry analysis of Thai manufacturing. The analysis is built around the hypothesis of Bhagwati that technology spillover is conditioned by the nature of the trade policy regime. The result, based on a two-equation model that allows for the two-way link between the foreign presence and productivity of locally owned industries, provides support for the hypothesis. Finally Markusen and Venables (1999) have formally shown how it is possible for FDI to act as a catalyst, leading to the development of local industry through linkage effects. The endogenous growth model considers FDI and trade as an important source of human capital augmentation, technology change and spillovers of ideas across countries and therefore FDI and trade is expected to have a positive effect on growth (Glass, and Saggi, 1998; Griffith, et al. 2004). The magnitude of spillovers depends on the extent to which local firms respond positively to the technology gap and invest in ‗learning activities‘ (Grossman and Helpman, 1995). Within the endogenous growth framework, Liu (2008) offers an explanation on how foreign direct investment (FDI) generates externalities in the form of technology transfer. A new insight gained from the theory is that the level and rate effects of spillovers can go in opposite directions. The negative level effect underscores the fact that technology transfer is a costly process—scarce resources must be devoted to learning. The positive rate effect indicates that technology spillovers enhance domestic firms' future productive capacity. In the model of Wang and Blomstrom (1992), technology transfer channeled through FDI is considered as an endogenized equilibrium phenomenon which results from strategic interaction between foreign firms and local firms. 4 In an important effort to establish a framework that synthesizes the previous three theoretical approaches, Gachino (2007) incorporated four spillover channels: competition, linkage, labor mobility and demonstration effects. Technological spillovers occurring through each of these channels are further conceptualized in the same way – technological changes, learning and capability building. This author argues that firms respond to external stimuli, skills, knowledge or technology transferred by implementing dynamic technological changes. These technological changes include modifications, improvements, and extensions meant to improve efficiency and increase firm productivity. Based on the previous studies, the present research focuses on whether there is evidence that FDI and trade facilitate technological progress in Latin American countries. 3. Methodology: Model specification and estimation techniques This section proposes an assessment methodology for productivity spillovers in order to examine when spillovers from foreign direct investment (FDI) contribute to productivity growth. The spillovers effects from FDI will be analyzed using a stochastic frontier (SFA) approach (Kumbhakar and Lovell, 2003). This approach uses the stochastic frontier production function, following Battese and Coelli (1988, 1993, 1995), and a generalized Malmquist output-oriented index to decompose productivity growth (Orea, 2002). 3.1. Deterministic frontier production functions: The stochastic frontier-inefficiency model Following Battese and Coelli (1995), the stochastic frontier approach (SFA) is used to estimate a production function and an inefficiency function simultaneously. The Battese–Coelli model can be expressed as follows: yit  f xit , t; ï?¢   exp vit  uit  ï?›1ï?? where yit implies the production of the ith firm i  1,2,..., N  in the tth time period t  1,2,..., T  , xit denotes a 1ï‚´ k  vector of explanatory variables, and ï?¢ represents the k ï‚´ 1 vector of parameters to be estimated. The error term consists of two components: vit and uit , which are independent of each other. In addition, vit denotes the time-specific and stochastic part, with idd N 0, ï?³ v2 , and uit represents technical inefficiency, which is a normal distribution, but truncated at zero with mean zitï?¤ and variance ï?³ u2 . The technical inefficiency effects, uit , are assumed as a function of a 1ï‚´ j  vector of observable non-stochastic explanatory variables, zit , and a  j ï‚´ 1 vector of unknown parameters to be estimated, ï?¤ . In a linear equation, the technical inefficiency effects can be specified as follows: uit  zitï?¤  wit ï?›2ï?? where wit is an unobservable random variable, which is defined by the truncation of the normal distribution with zero mean and variance, ï?³ u2 , such that the point of truncation is  zitï?¤ . 5 Equation ï?›1ï?? shows the stochastic production function in terms of the original production value, and equation ï?›2ï?? represents the technical inefficiency effects. The parameters of both equations can be estimated simultaneously by the maximum-likelihood method. The likelihood function is expressed in terms of variance parameters ï?³ s2  ï?³ v2  ï?³ u2 and ï?§  ï?³ u2 / ï?³ s2 e. If ï?§ equals zero, then the model reduces to a traditional mean response function in which z it can be directly included into the production function. Based on the theoretical model in Equations ï?›1ï?? and ï?›2ï??, we start with a flexible functional form, namely, a translog production function. By adopting a flexible functional form, the risk of errors in the model specification can be reduced. Moreover, the translog form is useful for decomposing the total factor productivity growth. The functional form of the translog production function is as follows: N 1 N N N ln yit  ï?¢ 0   ï?¢ n ln xnit  1  ï?¢ nk ln xnit ln xkit  ï?¢ t t  2 ï?¢ u t 2   ï?¢ nt ln xnitt  vit  uit ï?›3ï?? n 1 2 n1 k 1 n 1 where y implies output, x represents variables that explain output, t is time, i is firm. And u it is defined as: J uit  ï?¤ 0   ï?¤ j zit  wit ï?›4ï?? j 1 where z is the set of explanatory variables that explain technical inefficiency. Given the specifications in equations ï?›3ï?? and ï?›4ï??, the technical efficiency of production for the ith firm at the tth year is defined as the ratio of the actual output of firm i , ln yit , to its potential output, ln yitp :  Eï?›ï€­ uit vit  uit ï??  Eï?›ï€¨ï€­ zit ï?¤  wit vit  uit ï?? ï?›5ï?? ln yit TE  ln yitp where N 1 N N N ln yitp  ï?¢ 0   ï?¢ n ln xnit  1  ï?¢ nk ln xnit ln xkit  ï?¢ t t  2 ï?¢ u t 2   ï?¢ nt ln xnitt  vit ï?›6ï?? n 1 2 n1 k 1 n 1 3.2. Decomposing productivity growth: A generalized Malmquist index According to Orea (2002), if firm i ‘s technology in time t can be represented by a translog output-oriented distance function D0  yit , xit , t  where yit , xit , and t are defined as above, then the logarithm of a generalized output-oriented Malmquist productivity growth index, G0,it 1 , can bet decomposed into TEC, TP, and SEC between time periods t and t  1 : G0,it 1  TECti ,t 1  TPit ,t 1  SEC ti ,t 1 t ï?›7ï?? 6 where TECti ,t 1  lnD0  yi ,t 1 , xi ,t 1 , t  1  lnD0  yi ,t 1 , xi ,t 1 , t  ï?›8ï?? 1  lnD0  yi ,t 1 , xi ,t 1 , t  1 lnD0  yi ,t 1 , xi ,t 1 , t  TPit ,t 1     ï?›9ï?? 2 t  1 t  1 N  ï?¥ i ,t 1  1 ï?¥ 1   x  SEC ti ,t 1    ï?¥ ï?¥ i,t 1,n  itï?¥ ï?¥ itn   ln  ix,t 1,n  ï?›10ï?? 2 n1  i ,t 1  it   itn   N lnD0  yit , xit , t  where ï?¥ it   ï?¥ itn is the scale elasticity such that ï?¥ itn  n 1  ln xitn If the output is only one, then a translog output-oriented distance function can be defined as lnD0  yit , xit , t   ln yit  ln yitp  vit ï?›11ï?? Given the technical efficiency measure in Equation ï?›5ï?? , the technical efficiency change (TEC) between periods t  1 and t can be estimated by following Coelli et al. (2005): TECti ,t 1  lnTEi ,t 1  lnTEit ï?›8ï?? The technical progress (TP) index can be obtained from equations ï?›6ï?? , ï?›9ï?? , and ï?›11ï?? as follows: 1 N N  TPi ,t 1,t   ï?¢ tn ln xi ,t 1,n   ï?¢ tn ln xitn  2ï?¢ t  2ï?¢ tt ï?›ï€¨t  1ï??  t  ï?›13ï?? 2  n1 n 1  From equation ï?›3ï?? , the scale elasticity can be written as 1 K ï?¥ nit  ï?¢ n   ï?¢ nk xnit  ï?¢ ntt ï?›14ï?? 2 k 1 The index of scale efficiency change then can be calculated by using equations ï?›10ï?? and ï?›14ï?? . 4. Empirical analysis: Database, variables and results This section examines the productivity spillovers from FDI in the Latin American countries by using a unique and extensive country-level panel data covering the period 1996–2006. The intra-country productivity spillovers are examined through the FDI and trade variables, and the roles of labor skills and R&D effort in extending spillovers from FDI and trade are evaluated to test the absorptive capacity of productivity spillovers. 7 4.1 Statistical source and variables The statistical source used for this analysis is the World Bank‘s World Development Indicators (WDI). This database provides more than 800 development indicators, with time series for 209 countries and 18 country groups from 1960 to 2007. From the World Bank‘s World Development Indicators (WDI), we have temporal observations (T=10) for 16 Latin American countries for the period 1996–2006. We are able to form a balanced panel of data. (See descriptive statistics of the variables in Table 1.) [INSERT TABLE 1] Table 2 presents a summary of the key variables used to empirically validate the combined stochastic-inefficiency model (and the control variables used in the second step analysis). [INSERT TABLE 2] 4.2 Empirical results The indices of TEC, TP, SEC and G 0 are calculated using equations ï?›7ï?? a ï?›14ï?? and the average of these indices for the selected period (2001-2003) is presented in Table 3. [INSERT TABLE 3] Table 3 shows that the major contribution to productivity growth in the Latin American countries is from technological progress. In contrast, the Technical Efficiency Change Indices are relatively low, suggesting that this component does not contribute much to productivity growth. As to the negative contribution of SEC to productivity growth, we find several explanations with regard to this issue in Ventura-Dias, Cabezas and Contado (1999). These authors find that, with the exception of Mexico, the majority of Latin American economies are fully exploiting comparative advantage rooted in abundant natural resource endowments. Mexico firstly and Central American countries more recently have developer manufacturing activities oriented to the United States market based on a second source of comparative advantage, low-paid unskilled labor. Only Argentina, Brazil and Chile have developed competitive industries that can be classified as raw material processing (pulp and paper, nonmetallic minerals) and scale-intensive industries (steel, basic chemicals). Finally, Figure 1 shows the indices of TEC, TP, SEC and G 0 by country using the Hodrick and Prescott Filter. 8 [INSERT FIGURE 1] 1) FDI, absorptive capacity and productivity spillovers The estimation results of a translog stochastic production frontier (see table 4) show that the coefficients of labor and capital have the expected positive signs (in models 4). The positive and highly significant coefficients confirm the expected positive and significant output effects of labor and capital. In contrast, the squared variable of labor ï?›ln( Lt )ï?? in models 1, 2 and 3 is negative and 2 statistically significant at a 1% level, which indicates a decreasing return to labor. The same is not true of the squared capital. The squared variable of capital ï?›ln( K t )ï?? in models 1, 2 and 3 is positive 2 and statistically significant at a 1% level, which indicates an increasing return to capital. Furthermore, the estimated coefficient of the interacting variable between labor and energy ln( Lt ) * ln( Et ) in models 1, 2 and 3 is positive and significant at a 1% level, suggesting a substitution effect between labor and energy. [INSERT TABLE 4] A particular interest of this study is in regard to the estimated coefficients of the inefficiency function in the second part of Models in Table 4. The coefficient of the FDI is positive and significant at the 1% level, suggesting that countries with high FDI, on average, have lower efficiencies compared to those with low FDI. In order to explain why the FDI could have a positive correlation to inefficiency, Nourzad (2008) suggests ―the Bhagwati hypothesis,‖ which suggests that the efficiency-enhancing effect of FDI could depend on the degree of development of the host country. According to Nourzad (2008), the results suggest that increased FDI increases potential output in both developed and developing countries, with the effect being more profound in the former. Nourzad also found that increased FDI reduces technical inefficiencies the more open the economy, but that this effect holds only for developed economies. The negative and significant coefficient of the interacting variable between FDI and R&D in Models 1, 2, 3 and 4 in Table 4 implies a positive and significant efficiency spillover in Latin American countries. This result suggests that Latin American countries with high R&D effort gain more spillovers from FDI. Given this result, it is possible to infer that countries with high R&D effort can reap benefits from foreign firms‘ presence by upgrading their knowledge and fostering innovation. This finding confirms that firms‘ absorptive capacity (or firm-specific characteristics) determine productivity spillovers from FDI, as argued in several previous studies, for example, by Kathuria (2000, 2001). 2) Trade, absorptive capacity and productivity spillovers In table 5 the estimated parameters of the production functions have a similar sign and significance as in the baseline models shown in Table 4. The coefficient of trade is positive and significant at the 9 1% level, suggesting that Latin American countries with high trade, on average, have lower efficiencies compared to those with low trade. The positive correlation between trade and inefficiency can be explained using the justification given by McCalman, Stähler and Willmann (2011). These authors develop an efficiency theory of contingent trade policies by modeling the competition for a domestic market between one domestic and one foreign firm as a pricing game under incomplete information about production costs2. Using this theoretical framework McCalman, Stähler and Willmann (2011) show that the foreign firm must price more aggressively to overcome its cost disadvantage. For these authors the resulting possibility of an inefficient allocation justifies the use of contingent trade policy on efficiency grounds3. However, the negative coefficient of the interacting variable between trade and R&D suggests that countries with high R&D effort gain more spillovers from trade firms. From these findings, it may be inferred that domestic firms operating in an open economy with high R&D effort in Latin American countries will gain spillover benefits in an open economy. According to Suyanto et al. (2009) higher trade level is an inverse measure of the static competition that can protect inefficient firms. However, higher trade level can also be the result of the dynamic competition among firms of differential efficiency that removes inefficient firms from the industry according to Demsetz (1973) and Peltzman (1977). The first argument suggests that trade is associated with greater inefficiency, while the latter argument suggests that trade is associated with lower inefficiency. [INSERT TABLE 5] 3) Sources of productivity growth and FDI and trade spillovers .After obtaining the indices of Malmquist productivity growth G0  , TEC, TP, SEC, the next step is to estimate the contribution of FDI spillovers on total factor productivity growth and its sources. Using the indexes of TEC, TP, SEC, and G 0 obtained from the decomposition, we then estimate the impact of FDI spillovers on total factor productivity growth and this sources (see table 6). [INSERT TABLE 6] 2 According to McCalman, Stähler and Willmann (2011) the cost the cost distributions are asymmetric because the foreign firm has to pay a trade cost. 3 For McCalman, Stähler and Willmann (2011), contingent trade policy that seeks to maximize global welfare can avoid the potential inefficiency. These authors show how National governments make excessive use of contingent trade policy due to income-shifting considerations. As a result, the expected inefficiency of national policy is larger (smaller) for low (high) trade costs compared to the laissez-faire case. Finally, these authors conclude that in general, there is no clear ranking between the laissez-faire outcome and a contingent national trade policy. 10 Table 6 reveals that FDI contributes to SEC, TP and G0  (as shown by a statistically significant estimate of FDI variable on SEC, TP and G0 ). Moreover, a negative and statistically significant estimate of Latin American countries on technical efficiency suggests that higher FDI may decrease TEC. The same relation is found for trade, which indicates that countries with a high trade level have higher SEC and G0 than those with a low trade level. The negative correlation between trade level (and also with FDI) and TEC may be due to several factors. For example Ventura-Dias, Cabezas and Contado (1999) argue that operating under very unstable macroeconomic and political conditions, Latin American enterprises, in general, have not had the incentives for long-term investments in human and capital resources. As a result, those activities are not likely to generate endogenous sources of innovation and accumulation in the long term, primarily through innovative inter-sectoral linkages. A positive and significant estimate is found for Researchers in R&D, which indicates that countries with high numbers of researchers in R&D level have higher TC than those with low numbers of researchers (the opposite is found for Research & development expenditure). Based on the existent empirical evidence, it is difficult to explain the negative contribution of R&D expenditure to TP that we have found. Nevertheless, in the specific context of Latin American economies, Cimolli and Katz (2003) suggest that the present pattern of production specialization — strongly biased in favor of industries featuring low domestic knowledge generation and value added content — and the inhibition of local R&D and engineering activities resulting from the rapid expansion of internationally integrated production systems could be pushing Latin American economies into a ‗low development trap.‘4 5. Conclusions and policy implications The empirical analysis shows positive productivity spillovers from FDI and trade only when the country has absorptive capacity in terms of R&D; higher competition (in terms of trade) is associated with larger spillovers; and countries with high R&D effort gain more spillover benefits compared to those with less R&D effort. FDI and trade spillovers are found to be positive and significant for scale efficiency change and total productivity factor change. The empirical results show that intra-country productivity spillovers are present in the Latin American countries. Countries with R&D expenditure receive more productivity spillovers than those without R&D expenditure. Furthermore, technological progress is the major driver of productivity growth in the Latin American countries. The number of researchers in R&D has been found to be positive and significant for TP. Despite the presence of positive spillovers from FDI and trade in countries with absorptive capacity, the policy implications of these findings are not straightforward. These results may support the 4 In conclusion, Cimolli and Katz (2003) argue that new institutions and new forms of public-to-private interaction in the field of technology generation and dissemination seem to be a sine qua non condition for faster productivity growth and for the improvement of international competitiveness. 11 continuing fiscal and investment incentives provided by public and private institutions on R&D and human capital. According to Bjorvata and Eckel (2006), with many countries competing for FDI and trade, particularly in the presence of an asymmetric competition among countries, there are undesirable welfare effects for developing countries. Authors like Suyanto et al. (2009) suggest that policies for strengthening the absorptive capacity of domestic firms through investments in knowledge and human capital formation may be superior to policies that provide concessions for FDI and trade. In this context more general policies should be pursued, that not only attract FDI but also benefit domestic firms, for example, building modern infrastructure, increasing and strengthening the institutions for accelerating and sustaining economic growth. References Aigner, D., Lovell, C.A.K. and Schmidt, P. (1977): ―Formulation and Estimation of Stochastic Frontier Production Function Models‖, Journal of Econometrics, (6). 21-37. Aitken, B. J., and Harrison, A. E. (1999). Do domestic firms benefit from direct foreign investment? Evidence from Venezuela, The American Economic Review, 89(3), 605–618. Battese, G. E., and Coelli, T. J. (1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics, 20(2), 325–332. Battese, G.E. and Coelli, T.J. (1988), Prediction of Firm-Level Technical Efficiencies with a Generalised Frontier Production Function and Panel Data. Journal of Econometrics, (38). 387-399. Battese, G.E. and Coelli, T.J. (1993): ―A Stochastic Frontier Production Function Incorporating a Model for Technical Inefficiency Effects‖, Working Papers in Econometrics and Applied Statistics No. 69. Bjorvata, K., & Eckel, C. (2006): ―Policy competition for foreign direct investment between asymmetric countries‖, European Economic Review, 50 (7), 1891–1907. Blomström, M., Sjöholm, F. (1999). Technology transfer and spillovers: does local participation with multinationals matter?, Eur. Econ. Rev. 43, 915–923. Blomstrom, M., Kokko, A., (1996): ―Multinational Corporations and Spillovers‖. CEPR Discussion Paper, No. 1365. Borensztein , E, J. De Gregorio, J., Lee, J. W., (1998): ―How does foreign direct investment affect economic growth?‖, Journal of International Economics, (45) 115–135. Brainard, L.S., (1993): ―A Simple Theory of Multinational Corporations and Trade with a Trade- Off Between Proximity and Concentration‖, NBER Working Paper, No. 4269. 12 Blomstrom, M., Kokko, A., (1996): Multinational Corporations and Spillovers. CEPR Discussion Paper, No. 1365. Carstensen, K, and Toubal, F. (2004): ―Foreign direct investment in Central and Eastern European countries: a dynamic panel analysis‖, Journal of Comparative Economics, (32) 3–22 Caves, R. E. (1971): ―International corporations: The industrial economics of foreign investment‖, Economica, 38(149), 1–27. Caves, R.E., (1996): Multinational Enterprise and Economic, Analysis, 2nd ed. Cambridge University Press, Cambridge. Cimoli, M., and Katz, J. (2003): ―Structural reforms, technological gaps and economic development: a Latin American perspective‖, Industrial and Corporate Change, (12) 2, 387-411. Co, C. Y., (2001): ―Trade, foreign direct investment and industry Performance‖, International Journal of Industrial Organization, (19)163–183. Coelli, T. J., Rao, P., O‘Donnell, C. J., and Battese, G. E. (2005). An introduction to efficiency and productivity analysis (2nd ed.). Springer. Cheung, K.-Y., and Lin, P. (2004): ―Spillover effects of FDI on innovation in China: Evidence from the provincial data‖, China Economic Review, 15(1), 25–44. Coelli, T. J., Rao, P., O‘Donnell, C. J., and Battese, G. E. (2005). An introduction to efficiency and productivity analysis (2nd ed.). Springer. Damijan, J. P, Knell , M, Majcen, B., and Rojec, M. (2003): The role of FDI, R&D accumulation and trade in transferring technology to transition countries: evidence from firm panel data for eight transition countries, Economic Systems, (27) 189–204. Das, S. (1987): ―Externalities and technology transfer through multinational corporations‖, Journal of International Economics, 22(1–2), 171–182. Demsetz, H. (1973): ―Industry structure, market rivalry, and public policy‖, Journal of Law and Economics, 16(1), 1–9. Dunning, J.H., (1993): Multinational Enterprises and the Global Economy. Addison-Wesley, Reading. Farrel, M.J. (1957), ―The Measurement of Productivity Efficiency‖, Journal of the Royal Statistical Society, Series A, CXX, Part 3. 253-290. Findlay, R. (1978): ―Relative backwardness, direct foreign investment, and the transfer of technology: A simple dynamic model‖, Quarterly Journal of Economics, 92(1), 1–16. 13 Gachino, G. (2007): ―Technological spillovers from multinational presence towards a conceptual framework, UNU-MERIT Working Papers, No. 17. Glass, A., and Saggi, K. (1998): ―International technology transfer and the technology gap‖, Journal of Development Economics, 55(2), 369–398. Gorg, H., and Strobl, E. (2005): ―Spillovers from foreign firms through worker mobility: An empirical investigation‖, Scandinavian Journal of Economics, 107(4), 693–739. Griffith, R., Redding, S., and van Reenen, J. (2004): ―Mapping the two faces of R&D: Productivity growth in a panel of OECD industries‖, Review of Economics and Statistics, 14(5), 922–940. Grossman, G., and Helpman, E., (1995): ―Innovation and Growth in the Global Economy‖. MIT Press, Cambridge, MA. Jacquemin, A., Sapir, A., (1991): Competition and imports in the European market. In: Winter, A., Venables, A. (Eds.), European Integration: Trade and Industry. Cambridge University Press, Cambridge. Kathuria, V. (2000): ―Productivity spillovers from technology transfer to Indian manufacturing firms‖, Journal of International Development, 12(2), 343–369. Kinoshita, Y. (2001): ―R&D and technology spillovers through FDI: Innovation and absorptive capacity‖. CEPR discussion paper 2775. Kumbhakar, S. C., and Lovell, C. A. K. (2003): ―Stochastic Frontier Analysis‖, Cambridge University Press: New York. Liu, Z. (2008): ―Foreign direct investment and technology spillovers: Theory and evidence‖, Journal of Development Economics, (85) 176–193. Nourzad, F. (2008): ―Openness and the Efficiency of FDI: A Panel Stochastic Production Frontier Study‖, International Advances in Economic Research, Volume 14, No. 1, DOI: 10.1007/s11294- 007-9128-5. Markusen, J.R., and Venables, A.J., (1999): ―Foreign direct investment as a catalyst for industrial development‖, European Economic Review, 43, 335–356. McCalman, P., Stähler, F. and Willmann, G. (2011): ―Contingent Trade Policy and Economic Efficiency‖, CESIFO Working Paper No. 3424 Orea, L. (2002): Parametric decomposition of a generalized Malmquist productivity index. Journal of Productivity Analysis, 18(1), 5–22. Peltzman, S. (1977): ―The gains and losses from industrial concentration‖, Journal of Law and Economics, 20(2), 229–263. 14 Ramirez, M. R. (2010): ―Is Foreign Direct Investment Productive in the Latin America Case? A Panel Co-integration Analysis, 1980–2002‖, The International Trade Journal, 25:1, 35-73. Schiff and Wang (2010): ―North-South Trade-Related Technology Diffusion: Virtuous Growth Cycles in Latin America‖, IZA DP No. 4943 Smeets, R. A. (2008): ―Collecting the pieces of the FDI knowledge spillovers puzzle‖, The World Bank Research Observer, 23(2), 107–138. Suyanto, R., Salim, R.A., and Bloch, H. (2009). Does Foreign Direct Investment Lead to Productivity Spillovers? Firm Level Evidence from Indonesia, World Development, doi:10.1016/j.worlddev.2009.05.009 Ventura-Dias, V., Cabezas, M., and Contado, J. (1999): ―Trade reforms and trade patterns in Latin America‖, ECLAC, International Trade Serie, 5. Wang, J. W., and Blomstrom, M. (1992): ―Foreign investment and technology transfer: A simple model‖, European Economic Review, 36(1), 137–155. Yamin, M., and Sinkovics, R., (2009): ―Infrastructure or foreign direct investment? An examination of the implications of MNE strategy for economic development‖, Journal of World Business, (44) 144–157. Yao, S, and Wei, K. (2007): ―Economic growth in the presence of FDI: The perspective of newly industrialising economies‖, Journal of Comparative Economics, (35) 211–234 15 TABLES AND FIGURES Table 1 Descriptive statistics of variables by country: Mean 1996-2006. country/mean Y K L E AVA EVI FDI IVA MT R&D SVA RRD SET LFT Argentina 85,000,000,000 46,940,000,000 16,400,000 60,904.36 7.42 113.75 2.96 31.07 26.28 0.43 61.51 733.17 58.53 27.76 Bolivia 8,775,000,000 1,410,000,000 3,683,423 4,819.09 15.17 138.04 6.19 30.46 41.38 0.30 54.37 92.70 36.61 16.27 Brazil 665,400,000,000 108,300,000,000 84,800,000 195,304.73 5.96 131.97 2.97 27.40 18.86 0.87 66.63 409.85 19.15 7.16 Chile 79,300,000,000 17,990,000,000 6,280,403 26,002.64 5.42 131.20 6.47 39.76 53.06 0.57 54.83 495.63 41.33 25.20 Costa Rica 16,500,000,000 3,003,000,000 1,703,393 3,393.27 10.10 99.56 3.88 30.02 77.70 0.33 59.88 120.31 19.66 16.49 Ecuador 17,650,000,000 4,346,000,000 5,112,170 9,023.91 7.05 130.22 3.54 35.82 48.23 0.07 57.14 63.98 21.06 25.45 El Salvador 13,380,000,000 2,215,000,000 2,324,244 4,148.27 11.11 96.87 2.31 30.74 56.80 0.08 58.15 28.31 18.41 23.80 Guatemala 19,840,000,000 3,226,000,000 3,627,581 7,065.18 17.71 100.14 1.37 23.87 46.92 0.03 58.42 30.53 9.25 5.45 Honduras 7,570,000,000 1,879,000,000 2,332,121 3,416.09 16.14 105.08 3.79 30.28 110.45 0.05 53.57 70.03 15.72 5.45 Mexico 572,900,000,000 115,400,000,000 40,100,000 155,317.73 4.50 96.43 2.95 30.25 54.99 0.41 65.25 299.77 22.32 19.75 Nicaragua 4,003,000,000 926,500,000 1,871,618 2,899.00 20.63 101.84 5.56 28.59 62.18 0.06 50.78 70.03 17.77 5.45 Panama 12,080,000,000 2,265,000,000 1,307,084 2,551.45 7.37 99.68 7.19 17.79 34.61 0.32 74.84 107.21 43.28 18.10 Paraguay 7,499,000,000 1,327,000,000 2,536,386 4,052.64 18.93 137.76 1.52 22.25 56.01 0.09 58.82 81.36 21.06 25.45 Peru 56,480,000,000 11,610,000,000 12,200,000 12,417.64 8.12 141.56 3.22 31.38 29.31 0.11 60.51 227.77 33.12 30.46 Uruguay 22,490,000,000 3,050,000,000 1,567,053 2,879.18 8.85 114.26 2.21 26.00 31.69 0.27 65.15 290.09 34.43 16.98 Venezuela, RB 19,600,000,000 24,710,000,000 10,400,000 57,239.00 4.50 97.75 2.90 50.03 44.50 0.37 45.46 157.55 39.74 30.46 Total 119,279,187,500 21,787,343,750 12,265,342 34,465 11 115 4 30 50 0 59 205 28 19 Source: Authors‘ calculation from the World Bank‘s World Development Indicators (WDI). 16 Table 2 Key variables. Variables Definition Frontier model GDP at purchaser's prices is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the YNTx1 : GDP 1 (constant 2000 US$) value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant 2000 U.S. dollars. Dollar figures for GDP are converted from domestic currencies using 2000 official exchange rates. For a few countries where the official exchange rate does not reflect the rate effectively applied to actual foreign exchange transactions, an alternative conversion factor is used. Gross fixed capital formation (formerly gross domestic fixed investment) includes land improvements (fences, ditches, drains, and so on); plant, machinery, and equipment K NTx1 : Gross fixed capital formation 2 purchases; and the construction of roads, railways, and the like, including schools, offices, hospitals, private residential dwellings, and commercial and industrial buildings. (constant 2000 US$) 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 comprises people who meet the International Labour Organization definition of the economically active population: all people who supply labor for the LNTx1 : Labor force , total 3 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. Energy use refers to use of primary energy before transformation to other end-use fuels, which is equal to indigenous production plus imports and stock changes, minus E NTx1 : Energy use 4 (kt of oil equivalent) exports and fuels supplied to ships and aircraft engaged in international transport. Cyclical and Hicks neutral technological progress. TNTx1 : Time Inefficiency Foreign direct investment are the net inflows of investment to acquire a lasting management interest (10 percent or more of voting stock) in an enterprise operating in an model FDI NTx1 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 : Foreign direct investment5, net payments. This series shows net inflows in the reporting economy and is divided by GDP. inflows (% of GDP) Merchandise trade as a share of GDP is the sum of merchandise exports and imports divided by the value of GDP, all in current U.S. dollars. MTNTx1 : Merchandise trade 6 (% of GDP) Expenditures for research and development are current and capital expenditures (both public and private) on creative work undertaken systematically to increase R & DNTx1 : Research and development knowledge, including knowledge of humanity, culture, and society, and the use of knowledge for new applications. R&D covers basic research, applied research, and experimental development. expenditure7 (% of GDP) Labor force with tertiary education is the proportion of labor force that has a tertiary education, as a percentage of the total labor force. LTENTx1 : Labor force with tertiary 8 education (% of total) Time-varying inefficiency effect. TNTx1 : Year Control variables Agriculture corresponds to ISIC divisions 1-5 and includes forestry, hunting, and fishing, as well as cultivation of crops and livestock production. Value added is the net (second stage) AVANTx1 : Agriculture , value added (% of 2 output of a sector after adding up all outputs and subtracting intermediate inputs. It is calculated without making deductions for depreciation of fabricated assets or GDP) depletion and degradation of natural resources. The origin of value added is determined by the International Standard Industrial Classification (ISIC), revision 3. Note: For VAB countries, gross value added at factor cost is used as the denominator. Industry corresponds to ISIC divisions 10-45 and includes manufacturing (ISIC divisions 15-37). It comprises value added in mining, manufacturing (also reported as a IVANTx1 : Industry , value added (% of GDP) 2 separate subgroup), construction, electricity, water, and gas. Value added is the net output of a sector after adding up all outputs and subtracting intermediate inputs. It is calculated without making deductions for depreciation of fabricated assets or depletion and degradation of natural resources. The origin of value added is determined by the International Standard Industrial Classification (ISIC), revision 3. Note: For VAB countries, gross value added at factor cost is used as the denominator. Services correspond to ISIC divisions 50-99 and they include value added in wholesale and retail trade (including hotels and restaurants), transport, and government, SVANTx1 : Services , etc., value added (% of 2 financial, professional, and personal services such as education, health care, and real estate services. Also included are imputed bank service charges, import duties, and any GDP) statistical discrepancies noted by national compilers as well as discrepancies arising from rescaling. Value added is the net output of a sector after adding up all outputs and subtracting intermediate inputs. It is calculated without making deductions for depreciation of fabricated assets or depletion and degradation of natural resources. The industrial origin of value added is determined by the International Standard Industrial Classification (ISIC), revision 3. Note: For VAB countries, gross value added at factor cost is used as the denominator. Export values are from UNCTAD's value indexes or from current values of merchandise exports. EVI NTx1 : Export value index 12 (2000 = 100) Researchers in R&D are professionals engaged in the conception or creation of new knowledge, products, processes, methods, or systems and in the management of the RRD NTx1 : Researchers in R&D 13 (per projects concerned. Postgraduate PhD students (ISCED97 level 6) engaged in R&D are included. million people) Gross enrollment ratio is the ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the level of education shown. SETNTx1 : School enrollment 13 , tertiary (% Tertiary education, whether or not to an advanced research qualification, normally requires, as a minimum condition of admission, the successful completion of education gross) at the secondary level14 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. 17 4 International Energy Agency. 5 International Monetary Fund, International Financial Statistics and Balance of Payments databases, World Bank, Global Development Finance, and World Bank and OECD GDP estimates. 6 World Trade Organization, and World Bank GDP estimates. 7 United Nations Educational, Scientific, and Cultural Organization (UNESCO) Institute for Statistics. 8 International Labour Organization. 12 United Nations Conference on Trade and Development, Handbook of Statistics, and International Monetary, International Financial Statistics. 13 United Nations Educational, Scientific, and Cultural Organization (UNESCO) Institute for Statistics. 14 Note: Break in series between 1997 and 1998 due to due to change from International Standard Classification of Education (ISCED76) to ISCED97. Recent data are provisional. Source: World Bank‘s World Development Indicators (2009). Table 3 Sources of productivity growth by sector for 1996-2006. Sector Mean Technical Technical Scale Total Efficiency Change Efficiency productivity Change (TC) Change Change (TEC) (SEC) G0  Argentina -0.018 -0.598 8.943 8.328 Bolivia 0.003 1.131 -31.481 -30.347 Brazil -0.001 0.072 -20.542 -20.471 Chile 0.004 -0.978 7.605 6.631 Costa Rica -0.049 2.305 2.986 5.242 Ecuador -0.044 1.126 -0.980 0.103 El Salvador -0.039 1.447 5.427 6.834 Guatemala -0.049 0.687 -5.285 -4.647 Honduras -0.066 2.202 10.610 12.746 Mexico -0.006 -0.873 24.829 23.950 Nicaragua -0.022 0.873 -6.197 -5.347 Panama -0.050 2.296 2.930 5.176 Paraguay 0.036 0.669 -48.384 -47.679 Peru -0.053 4.175 -7.457 -3.336 Uruguay 0.006 2.843 -16.183 -13.334 Venezuela, RB -0.053 -3.023 53.662 50.586 Total -0.025 0.897 -1.220 -0.348 Source: Author‘s calculations. 18 Table 4 Maximum Likelihood Estimates of stochastic production frontier with inefficiency coefficient as function of FDI and Spillovers (by R&D and Labor skills). Variable Parameter Model 1 Model 2 Model 3 Model 4 Model 5 ï?¢0 Production Constant -27.22564 13.06251 -9.744519 2.224334*** 1.845238*** frontier1 [31.62189] [29.14972] [30.45277] [0.662596] [0.7073365] ï?¢1 -3.878797 -6.060685** -3.978034 0.9214329*** 0.965304*** ln( K t ) [2.91302] [2.913091] [2.989339] [0.0380093] [0.039755] ï?¢2 11.31664*** 6.373234** 7.922446*** 0.1023971** ln( Lt ) [2.928154] [2.640194] [2.756052] [0.0489112] 0.035645 [0.048911] ï?¢3 -0.8336474 4.137326 1.371627 -0.0286705 ln( Et ) [3.902221] [3.632856] [3.759156] [0.0581869] 0.006911 [0.059408] ï?›ln( K t )ï??2 ï?¢11 0.4626481*** [0.1686924] 0.5463026*** [0.1729435] 0.451646** [0.1803403] ï?›ln( Lt )ï??2 ï?¢ 22 -1.433836*** [0.2831406] -1.014223*** [0.3055332] -1.034772*** [0.3207719] ï?›ln( Et )ï??2 ï?¢ 33 0.1840744 [0.2445038] 0.4857334** [0.2367923] 0.3501664 [0.2404679] ln( Lt ) * ln( K t ) ï?¢12 0.0769518 [0.1456199] 0.1513515 [0.1486073] 0.0737517 [0.148731] ln( K t ) * ln( Et ) ï?¢13 -0.7367614*** [0.2031064] -0.8313027*** [0.2150417] -0.6972762*** [0.2177063] ln( Lt ) * ln( Et ) ï?¢ 23 1.011279*** [0.2117392] 0.6528089*** [0.1981608] 0.7150243*** [0.2048636] Tt ï?¢t -0.5227036*** [0.1866401] 0.0111609*** [0.00423] ln( K t ) * Tt ï?¢1t 0.0234228** [0.0110004] ln( Lt ) * Tt ï?¢ 2t 0.0376901*** [0.0119172] ln( Et ) * Tt ï?¢ 3t -0.0618408*** [0.0160849] Tt 2 ï?¢ tt -0.0051394* [0.0027086] -0.0053111* [0.0029266] ï?¤ u0 constant -4.328997*** 4.950119*** -4.546313*** -5.758383*** u it [0.4797635] 0.7979365] [0.6643834] [1.00458] Equation FDI NTx1 ï?¤ FDI 0.6134197*** [0.1219802] 0.7214421*** 0.161151] 0.6777422*** [0.1430362] 0.8398162*** [0.2021684] FDI NTx1 * R & DNTx1 ï?¤ FDIxR & D -1.737438*** [0.3882507] -1.919297*** 0.5163247] -1.699004*** [0.4369535] -2.461248** [1.239592] FDI NTx1 * LTENTx1 ï?¤ FDIxLTE 0.0015189 [0.0050572] -0.0021701 0.0066998] -0.0031846 [0.0061448] -0.0001214 [0.0079937] ï?¤ v0 constant -4.231141*** -3.829503*** -3.855063*** -3.366986 vit [0.2353143] 0.1963722] [0.2259758] [0.14119] Equation 19 ï?³v Sigma 0.12056450.014 0.1473784 0.1455069 0.1857242 0.2173873 1853 0.0144705 0.0164405 0.0131112 0.0116145 Wald chi2 14120.00 10799.55 10344.89 6736.84 8765.61 Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000 Log likelihood 71.093044 60.943097 55.942379 34.67139 18.853981 Number of obs 176 176 176 176 176 Notes: Model 1 is a translog production function. Models 2 and Model 3 represent a Hicks-neutral and a no-technological progress production functions, respectively. Model 4 is a Cobb–Douglas production function. Model 5 represents a no-inefficiency production function: lnsig2v: coefic. -3.05215 and std. err. 0.106855; lnsig2u: coefic. -12.28412 and std. err. 206.2603; sigma_u: coefic.0.0021505 and std. err. 0.2217808; sigma2: coefic. 0.0472619 and std. err. 0.0050745; lambda: coefic.0.0098925 and std. err. 0.2228764; Likelihood-ratio test of sigma_u=0: chibar2(01) = 0.00 Prob>=chibar2 = 1.000 Standard errors are in parentheses and presented until two significant digits, and the corresponding coefficients are presented up to the same number of digits behind the decimal points as the standard errors: * Denotes significance at 10%;** Denotes significance at 5%;*** Denotes significance at 1%; * p  0.1 : * * p  0.05; * * * p  0.01 Source: Authors‘ calculation. Table 5 Maximum Likelihood Estimates of stochastic production frontier with inefficiency coefficient as function of trade and Spillovers (by R&D and Labor skills). Variable Parameter Model 1 Model 2 Model 3 Model 4 ï?¢0 Production Constant -63.75117** -46.10637 -53.34125* 2.140026*** frontier1 [27.09068] [29.30114] [30.4911] [0.6358066] ï?¢1 -0.1441388 -0.8605591 -0.0976849 0.9157772*** ln( K t ) [2.644289] [2.948618] [3.026257] [0.03715] ln( Lt ) ï?¢2 13.27632*** [2.485212] 10.43628*** [2.544593] 10.91683*** [2.693568] 0.1225255** [0.0498045] ln( Et ) ï?¢3 -5.312538 [3.384722] -2.565611 [3.627755] -3.685716 [3.771481] -0.0405479 [0.0561805] ï?›ln( K t )ï?? 2 ï?¢11 0.1718158 [0.1616403] 0.2344044 [0.1777567] 0.2190375 [0.1823832] ï?›ln( Lt )ï??2 ï?¢ 22 -1.730239*** [0.2589777] -1.356151*** [0.2878417] -1.293024*** [0.3031869] ï?›ln( Et )ï??2 ï?¢ 33 -0.1547828 [0.2251013] 0.1074325 [0.2372708] 0.0199565 [0.2443523] ln( Lt ) * ln( K t ) ï?¢12 0.1250482 [0.1313789] 0.1034678 [0.1361103] 0.0328078 [0.1420171] ln( K t ) * ln( Et ) ï?¢13 -0.5156052*** [0.1898468] -0.5572378*** [0.2134383] -0.4847293** [0.2184586] ln( Lt ) * ln( Et ) ï?¢ 23 1.182345*** [0.1918807] 0.9118755*** [0.1966486] 0.9298185*** [0.2069266] Tt ï?¢t -0.4637614*** [0.1702002] 0.0144658*** [0.0044314] ln( K t ) * Tt ï?¢1t 0.0138253 [0.0098161] 20 ln( Lt ) * Tt ï?¢ 2t 0.0471882*** [0.0116229] ln( Et ) * Tt ï?¢ 3t -0.0601819*** [0.0146426] Tt 2 ï?¢ tt -0.0019538 [0.0027455] -0.0022101 [0.0029859] ï?¤ u0 constant -5.223112*** -5.340834*** -4.696655*** -3.885253*** u it [0.6244387] [0.7625763] [0.7706064] [1.230057] Equation MTNTx1 ï?¤ MT 0.0400736*** [0.0078391] 0.0379042*** [0.0081021] 0.0333523*** [0.0078793] 0.0470576*** [0.0130393] MTNTx1 * R & DNTx1 ï?¤ MTxR&D -0.1398774*** [0.0478615] -0.1312944** [0.0630352] -0.1414705** [0.0665409] -0.6226683*** [0.226363] MTNTx1 * LTENTx1 ï?¤ MTxLTE 0.0011217** [0.0004378] 0.0010888** [0.0005149] 0.0007393 [0.0005251] 0.0004228 [0.0007061] ï?¤ v0 constant -4.269993*** -3.948264*** -3.857357*** -3.367329 vit [0.2090885] [0.1905863] [0.1974477] [0.1319885] Equation ï?³v Sigma 0.118245 0.1388818 0.1453401 0.1856923 0.0123618 0.0132345 0.0143485 0.0122546 Wald chi2 16458.78 12741.27 11842.23 8003.14 Prob > chi2 0.0000 0.0000 0.0000 0.0000 Log likelihood 71.852987 58.757756 52.907567 33.430661 Number of obs 176 176 176 176 Notes: Model 1 is a translog production function. Models 2 and Model 3 represent a Hicks-neutral and a no-technological progress production functions, respectively. Model 4 is a Cobb–Douglas production function. Standard errors are in parentheses and presented until two significant digits, and the corresponding coefficients are presented up to the same number of digits behind the decimal points as the standard errors: * Denotes significance at 10%;** Denotes significance at 5%;*** Denotes significance at 1%; * p  0.1 : * * p  0.05; * * * p  0.01 Source: Authors‘ calculation. 21 Table 6 Sources of productivity growth and spillovers: 1997-2005. Dependent Variable SEC TP TEC G0 Sources product. & FDI spillovers Model FE Model RE Model FE Model RE Model FE Model RE Model FE Model RE Independent variables Coef/E. St. Coef/E. St. Coef/E. St. Coef/E. St. Coef/E. St Coef/E. St Coef/E. St Coef/E. St 5.877959 7.699713* 0.1022834** 0.1800723*** -0.0118866 -0.0154555** 5.968356 7.86433* FDI NTx1 : Foreign direct investment [4.478186] [4.335772] [0.0476449] [0.0602199] [0.0074611] [0.0073282] [4.471485] [4.327804] 0.8951474 1.030092* 0.0037336 -0.0131566 -0.0008391 -0.0008214 0.8980418 1.016114* MTNTx1 : Merchandise trade [0.6119442] [0.6015853] [0.0065107] [0.0083555] [0.0010196] [0.0010168] [0.6110285] [0.6004797] 22.3687 -2.857874 -6.025811*** -6.041163*** 0.0181406 0.04199 16.36103 -8.857051 R & DNTx1 : Research & development expenditure [85.03727] [85.73711] [0.9047393] [1.190811] [0.1416801] [0.1449097] [84.91002] [85.57955] 0.7145777 0.2487539 -0.0407188** -0.0090431 -0.0005725 -0.0007888 0.6732863 0.238922 LTENTx1 : Labor force with tertiary education [1.786074] [1.778848] [0.0190026] [0.0247066] [0.0029758] [0.0030065] [1.783401] [1.775579] Control variables (dropped) (dropped) (dropped) (dropped) (dropped) (dropped) (dropped) (dropped) AVANTx1 : Agriculture, value added 1.250211 2.006595 0.0174974 -0.0673268 -0.0018926 -0.0006553 1.265816 1.938613 IVANTx1 : Industry, value added [3.278244] [3.200899] [0.0348783] [0.0444576] [0.0054619] [0.00541] [3.273339] [3.195017] 0.7648675 1.303565 0.1488787*** 0.066263 -0.0019491 -0.0005158 0.911797 1.369312 SVANTx1 : Services, etc., value added [3.173707] [3.103752] [0.0337661] [0.0431083] [0.0052877] [0.0052458] [3.168958] [3.098049] EVI NTx1 : Export value index 0.4479707 1.106331*** 0.0231497*** -0.0125423*** -0.0008023 -0.0009962* 0.470318 1.092792*** [0.5297272] [0.3433629] [0.0056359] [0.004769] [0.0008826] [0.0005803] [0.5289346] [0.3427319] RRD NTx1 : Researchers in R&D (in ln) -5.127533 [20.66532] -3.014281 [20.57651] 0.4586041** [0.2198651] 0.9032746*** [0.2857891] 0.0114921 [0.0344304] 0.0014447 [0.0347777] -4.657436 [20.6344] -2.10956 [20.5387] SETNTx1 : School enrollment, tertiary 0.1818811 0.322796 -0.0106004 -0.0499516*** 0.0005303 0.000842 0.1718109 0.2736863 [1.267745] [1.229769] [0.0134879] [0.0170804] [0.0021122] [0.0020785] [1.265848] [1.227509] Constant -217.2321 -356.9921 -11.14535*** -0.8447519 0.2826265 0.2318622 -228.0948 -357.605 [283.756] [264.4889] [3.018973] [3.673511] [0.4727644] [0.4470295] [283.3314] [264.0029] Number of obs 128 128 128 128 128 128 128 128 R-squared 0.0497 0.0436 0.5688 0.3935 0.0423 0.0404 0.0493 0.0439 Hausman test Prob>chi2 = 0.9563: RE Prob>chi2 = 0.0000: FE Prob>chi2 = 0.9885: RE Prob>chi2 = 0.9628: RE * p  0.1 : * * p  0.05; * * * p  0.01 Source: Author‘s calculations. 22 Figure 1 Hodrick Prescott Filter of TPFC. Argentina Bolivia Brazil Chile 300 250 80 200 200 200 60 150 150 40 100 100 100 20 50 0 50 0 0 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -100 0 -20 -50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -40 -200 -100 -100 -60 -150 -300 -150 -80 -200 -400 -200 -100 -250 -500 -250 -120 -300 TPFC HP Trent HP Cicle T PFC HP T rent HP Cicle TPFC HP Trent HP Cicle T PFC HP T rent HP Cicle Costa Rica Ecuador El Salvador Guatemala 200 200 100 200 150 80 150 150 100 60 100 100 50 0 40 50 50 -50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 20 0 0 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -100 0 -50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -50 -150 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -20 -100 -200 -100 -40 -150 -250 -150 -300 -60 -200 T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle Honduras Mexico Nicaragua Panama 200 120 250 300 100 200 150 200 80 150 100 60 100 40 100 50 0 20 50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 0 -100 0 0 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -20 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -50 -200 -50 -40 -100 -100 -300 -60 -150 -80 -150 -400 T PFC HP T rent HP Cicle TPFC HP Trent HP Cicle T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle Paraguay Peru Uruguay Venezuela RB 100 200 300 600 150 200 50 400 100 100 0 200 50 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 0 -50 0 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 0 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -100 96-97 97-98 98-99 99-00 00-01 01-02 02-03 00-04 00-05 00-06 -50 -100 -200 -200 -100 -150 -300 -400 -150 -200 -200 -400 -600 T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle T PFC HP T rent HP Cicle Source: author‘s calculations. 23