Economics Letters 220 (2022) 110878 Contents lists available at ScienceDirect Economics Letters journal homepage: www.elsevier.com/locate/ecolet Drivers of productivity change in global value chains: Reallocation vs. innovation✩ ∗ Ivan Savin a,b , , Philipp Mundt c a Institute of Environmental Science and Technology, Autonomous University of Barcelona, Bellaterra, Spain b Graduate School of Economics and Management, Ural Federal University, Yekaterinburg, Russia c Department of Economics, Otto-Friedrich-Universität, Bamberg, Germany article info a b s t r a c t Article history: We investigate the role of technological improvement and market share reallocation in determining Received 26 July 2022 global changes in sectoral labor productivity. Contrary to previous work that neglects dependencies Received in revised form 22 September 2022 between suppliers in global value chains, we account for input linkages that impact both channels of Accepted 26 September 2022 productivity improvement. Using sector-level data from the World Input–Output Database, we show Available online 30 September 2022 that reallocation between countries has a larger effect on productivity change than innovation within JEL classification: countries. C67 © 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license E24 (http://creativecommons.org/licenses/by/4.0/). L14 L16 O47 Keywords: Input–output analysis Market share reallocation Productivity decomposition Production network Technological improvement 1. Introduction productivity gains accumulate along the production chain (Mc- Nerney et al., 2022). On the other hand, Cantner et al. (2019) Changes in productivity are a major source of growth and argue that competitive selection leads to higher market shares development. Prior studies argued that the weighted average of producers with the most productive suppliers in their value productivity of an industry can change through two main chan- chain, but not necessarily with the highest individual produc- nels (Metcalfe, 1994). While the first mechanism operates at the tivity. Employing multinational input–output tables, this paper intensive margin and stresses the role of technological improve- asks whether reallocation between countries has a larger effect on ment within producers (Mohnen and Hall, 2013), the second productivity change than innovation within countries in a given operates at the extensive margin and captures the reallocation of industry. market shares between them (Alfaro and Chen, 2018). Under the hypothesis that competition increases the shares of more produc- 2. Productivity Measure and Decomposition tive actors, the second channel explains why average productivity can improve even in the absence of innovation. The main idea of the present investigation is to measure Recent literature suggests that input linkages in global value productivity on the level of value chains instead of individual chains (GVCs) affect both aforementioned channels of productiv- countries in a given sector. Then we aggregate this measure ity change. On the one hand, technological improvement does not only increase the productivity of the innovator but also the pro- across countries and assess the contribution of innovation and ductive performance of its downstream customers, implying that reallocation on global productivity improvement. Our analysis focuses on global labor productivity in a given industry instead ✩ I.S. received support from the Russian Science Foundation [RSF grant of total factor productivity (TFP) across industries in a country for several reasons. First, the choice of labor productivity eases number 19-18-00262]. We thank Jeroen van den Bergh and Ilfan Oh for feedback. ∗ Corresponding author at: Institute of Environmental Science and Technol- comparability to prior work because several important studies on ogy, Autonomous University of Barcelona, Bellaterra, Spain. productivity growth decomposition considered labor productivity E-mail address: ivan.savin@uab.cat (I. Savin). instead of TFP (e.g., Bernard and Jones, 1996; Dietzenbacher et al., https://doi.org/10.1016/j.econlet.2022.110878 0165-1765/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). I. Savin and P. Mundt Economics Letters 220 (2022) 110878 2000; McMillan et al., 2014). Second, the computation of value and reporting their percentage shares in global sectoral labor pro- chain productivity and country weights simplifies considerably ductivity improvement. A positive and dominant absolute within for labor productivity. Third, TFP might be biased in the presence effect would imply that ∆Π (s) improved mainly because indi- of technologically heterogeneous producers and input comple- vidual layers of value chains have increased their productivity mentarities (Dosi and Grazzi, 2006). Since the consideration of through technological change. In contrast, a positive and domi- disaggregated data is crucial for the measurement of competition, nant absolute between effect would suggest that the majority of we conduct the decomposition separately for each industry. improvement originates in growing market shares of countries Building on the work by Timmer and Ye (2017), we employ with a higher value chain productivity in that sector. input–output accounting to compute labor productivity as the ratio of the sum of value added across all layers of the production 3. Evidence from World Input–Output Tables chain to the sum of both direct and indirect labor required for producing final output. Hence, our measure reflects that country- Our empirical application rests on the 2013 release of the sectors source a substantial amount of indirect labor through the World Input–Output Database (WIOD; Timmer et al., 2015), which use of their suppliers’ inputs.1 Moreover, it excludes the portion provides data on employment, value added, and production link- of labor producing intermediate output for other country-sectors. ages for 35 ISIC Rev. 3 industries in 40 countries for the period Direct and indirect labor demand is obtained from 1995–2009. All data have been converted into US Dollars and L = l̂(I − A)−1 f, (1) adjusted for inflation using sectoral price indexes with base year 1995. where l̂ is a diagonal matrix with labor coefficients, measuring Considering the median across all industries and taking value the direct labor demand of a particular country-sector per unit of chain linkages into account, we find that 60% of productivity im- gross output. (I − A)−1 is the Leontief inverse with the identity provement are due to reallocation, and only 40% originate in in- matrix I. A = (ai,j (s, r)) is a matrix of technical coefficients, novation (Fig. 1).3 The decomposition results by economic activity defined as output shipped from sector s in country i to sector in Table 1 further show that the between effect consistently dom- r in country j, divided by the output of the receiving country- inates the within effect in agriculture, mining, manufacturing, sector. f represents a diagonal matrix with final demand. We and services, whereas the opposite holds for construction.4 One then compute the sum of direct and indirect labor necessary to way to interpret this result is that construction is a non-tradable produce final output from the column sum of L = (li,j (s, r)). Using sector, whereas output in agriculture, mining, manufacturing, and the accounting identity, which implies that the sum of value a part of the service sector is tradable (see OECD, 2018, p. 61). added along all layers of the value chain coincides with the output Therefore, we would expect that sectors exposed to international sold to end consumers, we measure the former as final demand. competition through trade are subject to a larger reallocation Thus, our value-chain labor productivity measure for sector s in effect. country i is:2 fi (s) πi (s) = ∑ ∑ . (2) j r lj,i (r , s) To compare the role of technological improvement and market share reallocation for productivity improvement in sector s, we consider the global labor productivity Π of sector s ∑ Π (s) = wi (s)πi (s), (3) i which weights productivities πi (s) by the labor shares wi (s) of the respective countries in that sector: li,j (s, r) ∑∑ j r wi (s) = ∑ ∑ ∑ . (4) i j r li,j (s , r) In line with Griliches and Regev (1995), we then decompose the change in Π (s) according to ∑ ∑ ∆Π (s) = w̄i (s)∆πi (s) + ∆wi (s)π̄i (s) (5) i i into technological improvement (within effect) in the first term, and the reallocation of market shares (between effect) in the second term on the right-hand side of Eq. (5), where ∆ is the Fig. 1. Decomposition of aggregate productivity change. Box plots for the within and between effects across all industries are shown. White markers indicate difference operator, and a bar over a variable stands for its aver- median values. age over two consecutive periods. Our quantitative comparison of the two effects is eased by summing the effects over years 3 Note that the absolute within and between effects in Eq. (5) are not always positive, implying that the total productivity change may be smaller than one 1 An implicit assumption underlying our approach and data is that all output, of its constituent elements. This explains why percentage shares of within and within a given country-industry, is produced with the same input mix, while in between effects can exceed unity in absolute value. reality the latter might depend on the downstream use of output (de Gortari, 4 Despite moderate technological improvement in mining and agriculture, 2019). Since the present study focuses on the global reallocation of sectoral mar- productivity in these industries decreased due to a negative and dominant ket shares through competition, we believe that our results are not materially between effect, which can be explained by the decline of these industries in affected by this assumption. several high-income countries in Europe and North America and their growing 2 To simplify the notation, we omit time indices here and in the following importance in low-middle and high-middle income economies such as India, exposition. Indonesia and Brazil over the sample period. 2 I. Savin and P. Mundt Economics Letters 220 (2022) 110878 Table 1 in productive performance at the level of individual producers. Productivity decomposition by economic activity. Furthermore, our analysis relates to the literature on the factor Economic activity Within Between content of trade (Davis and Weinstein, 2001) where it shows Panel A: Value-chain productivity how country-specific labor endowments are implicitly being re- Agriculture −0.05 1.05 allocated through intermediate input linkages across the world. Manufacturing 0.43 0.57 Construction 1.00 0.00 A direction for future work would be to extend the analysis to Mining and quarrying, utilities 0.01 0.99 total factor productivity. We will address this question in future Services 0.38 0.62 research. Panel B: Idiosyncratic productivity Agriculture −0.09 1.09 Data availability Manufacturing 0.69 0.31 Construction 1.65 −0.65 Mining and quarrying, utilities 4.00 −3.00 Data will be made available on request. Services 0.47 0.53 References Note: Reported values represent the median within and between effect across the industries in a given category. Alfaro, L., Chen, M., 2018. Selection and market reallocation: productivity gains from multinational production. Am. Econ. J.: Econ. Policy 10 (2), 1–38. Bernard, A., Jones, C., 1996. Productivity and convergence across U.S. states and To assess the contribution of global input–output linkages to industries. Empir. Econ. 21, 113–135. Bernard, A., Moxnes, A., 2018. Networks and trade. Annu. Rev. Econ. 10, 65–85. the sectoral labor productivity decomposition, we compare our Cantner, U., Savin, I., Vannuccini, S., 2019. Replicator dynamics in value chains: value-chain productivity measure to the idiosyncratic produc- explaining some puzzles of market selection. Ind. Corp. Change 28 (3), tivity of countries in each sector, which neglects the existence 589–611. of input linkages. The results in Fig. 1 and Table 1 testify to Davis, D., Weinstein, D., 2001. An account of global factor trade. Amer. Econ. a considerably weaker role of reallocation for the idiosyncratic Rev. 91 (5), 1423–1453. Dietzenbacher, E., Hoen, A., Los, B., 2000. Labor productivity in Western Europe productivity measure, with a median between effect across all 1975–1985: an intercountry, interindustry analysis. J. Reg. Sci. 40 (3), industries of merely 21%. This stresses the relevance of input 425–452. linkages in studies of productivity change and implies that the Dosi, G., Grazzi, M., 2006. Technologies as problem-solving procedures and role of reallocation is downward biased when these linkages are technologies as input-output relations: some perspectives on the theory of production. Ind. Corp. Change 15 (1), 173–202. ignored. de Gortari, A., 2019. Disentangling Global Value Chains. Technical Report 25868. NBER Working Paper Series. 4. Conclusion Griliches, Z., Regev, H., 1995. Firm productivity in Israeli industry: 1979–1988. J. Econometrics 65 (1), 175–203. McMillan, M., Rodrik, D., Verduzco-Gallo, I., 2014. Globalization, structural Although production networks have been a vital research change, and productivity growth, with an update on Africa. World Dev. 63, topic in macroeconomics and international trade in recent years 11–32. (Bernard and Moxnes, 2018), a quantitative assessment of the McNerney, J., Savoie, C., Caravelli, F., Carvalho, V., Farmer, D., 2022. How different channels of productivity change in the setting of a production networks amplify economic growth. Proc. Natl. Acad. Sci. 119 (1), 1–11. production network has been missing in previous studies. Against Metcalfe, J.S., 1994. Competition, Fisher’s principle and increasing returns in the this background, we measure labor productivity in GVCs and selection process. J. Evol. Econ. 4 (4), 327–346. compare the importance of technological improvement and mar- Mohnen, P., Hall, B., 2013. Innovation and productivity: an update. Eurasian Bus. ket share reallocation in productivity improvement. We find Rev. 3 (1), 47–65. that reallocation effects are quantitatively more important than OECD, 2018. Productivity and Jobs in a Globalised World: How Can All Regions Benefit?. OECD Publishing, pp. 57–93, chapter. Thinking global, developing innovation once input linkages are accounted for, especially in local: tradable sectors, cities and their role for catching up. tradable industries. Our results have implications for the lit- Timmer, M., Dietzenbacher, E., Los, B., Stehrer, R., De Vries, G.J., 2015. An erature on growth and development, but also for competition illustrated user guide to the World Input-Output Database: the case of global analysis since network effects may counteract disadvantages automotive production. Rev. Int. Econ. 23 (3), 575–605. Timmer, M., Ye, X., 2017. The Oxford Handbook of Productivity Analysis. Oxford University Press, pp. 699–724. 3