The Importance of Manufacturing in Economic Development: Has This Changed?

World Development Vol. xx, pp. xxx xxx, 2017 0305-750X/Ó 2016 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev http://dx.doi.org/10.1016/j.worlddev.2016.12.013 The Importance of Manufacturing in Economic Development: Has This Changed? NOBUYA HARAGUCHI a, CHARLES FANG CHIN CHENG b and EVELINE SMEETS c,* a United Nations Industrial Development Organization, Vienna, Austria b The University of New South Wales, Sydney, Australia c Independent Consultant Summary. Manufacturing has traditionally played a key role in the economic development of developing countries. In recent years, it has been argued that the importance of manufacturing has diminished over the last 20 25 years, resulting in premature deindustrialization or non-industrialization in developing countries. This study explores whether the low levels of industrialization in developing countries are attributable to long-term changes in opportunities available to the sector around the globe. The study s findings show that the manufacturing sector s value added and employment contribution to world GDP and employment, respectively, have not changed significantly since 1970. The declining manufacturing value added and manufacturing employment share in many developing countries has not been caused by changes in the sector s development potential but has instead resulted from a shift of manufacturing activities to a relatively small number of populous countries, thus resulting in a concentration of manufacturing activities in specific developing countries. As was the case in the last millennium, industrialization has continued to play a key role in the growth of developing countries, which have sustained rapid and long-term growth for the last 25 years. Achieving economic development by following the path of industrialization will likely remain important for low-income countries because they are able to take advantage of their backwardness relative to those countries which have already experienced rapid industrialization with a disproportionately large share of manufacturing activities, and could soon enter a mature stage of industrialization. Ó 2016 Elsevier Ltd. All rights reserved. Key words manufacturing, structural change, economic development, concentration 1. INTRODUCTION The objective of this paper is to determine whether the importance of manufacturing in developing countries in terms of the sector s development quality (manufacturing s role as an engine of growth) and quantity (relative share of manufacturing value added and employment in GDP and total employment, respectively) has changed or not. Based on an assessment of these factors, the paper argues that the significance of manufacturing itself in economic development has not changed since 1970. Although the shares of world manufacturing value added and employment have not declined, a greater concentration of manufacturing activities in recent years has led to the decline of these shares in many developing countries. With many developing countries experiencing premature deindustrialization, it has recently been argued that development through manufacturing growth has become a more difficult path for current developing countries to take (Eichengreen & Gupta, 2009; Ghani & O Connell, 2014; Rodrik, 2016). This argument is largely based on the downward shifts of both manufacturing value added and employment share in GDP and total employment, respectively, across all income levels as confirmed in recent studies (Dasgupta & Singh, 2006; Ghani & O Connell, 2014; Palma, 2007; Rodrik, 2016) and confirmed by this study (Figures 1 and 2). 1 It is not entirely clear, however, whether this trend is attributable to sector-specific, structural factors, which could have led to a decline in the manufacturing sector s development quality, quantity or both, relative to other sectors. A decline of manufacturing value added and employment shares in many developing countries might also be caused by countryspecific factors, such as the country s policies and comparative advantages. If the relative importance of manufacturing in developing countries declined due to sector-specific factors or due to a long-term shift in world economic structure, developing countries should rebalance their development strategies and place less emphasis on manufacturing development than those countries that have industrialized in the past. In the case of the latter, i.e., when no change in the relative importance of manufacturing is recorded, the focus should be on policies to tackle opportunities and challenges some of which may be new and specific to the currently developing countries due to a greater degree of globalization in the last 20 years. 2 Given recent discussions on premature deindustrialization and its implications for the sustained growth of developing countries, this paper contributes to a renewed understanding of the role of the manufacturing sector in economic development and in the economic policy of developing countries. To achieve this objective, we evaluate whether the following two conditions hold or not. A. Manufacturing is no longer the driver of economic growth in developing countries based on Kaldor s formulation. 3 B. The share of manufacturing value added (MVA) relative to other sectors and employment has decreased significantly in developing countries. The first condition (A) essentially examines whether the relationship between the share of manufacturing in the economy and economic growth is positive and stronger than the relationship between the share of other sectors and economic growth. The second condition (B) examines the relative size * We would like to thank Adrian Wood, Alejandro Lavopa and Adam Szirmai for feedback on earlier drafts. We are grateful to the editor of the journal and three anonymous referees for their review and useful comments. Final revision accepted: December 18, 2016. 1

2 WORLD DEVELOPMENT Figure 1. MVA share to GDP at constant prices (country average). of MVA and manufacturing employment in the economy. For instance, even though manufacturing might be the main driver of economic growth (hence, the rejection of (A)), we can still consider a scenario in which manufacturing plays a less significant role in the economic development of developing countries than it previously has if its size decreased considerably. In fact, it is widely believed that manufacturing jobs are shrinking globally (Ghani & O Connell, 2014). In turn, despite retaining the same size, manufacturing could be considered to play a less important role if it weakened the ability to boost economic growth. If both (A) and (B) can be rejected, we may conclude that the importance of manufacturing in the growth of developing countries has not in fact changed. In that case, we could claim that premature deindustrialization is not caused by changes in the development characteristics of manufacturing which might have diminished its role in economic development but is attributable to the inabilities of some countries to develop their manufacturing sector relative to others. Quite a number of studies have focused on the first condition (A). Hence, the relevant literature is reviewed to draw conclusions about Condition (A). In contrast, no studies have carried out a detailed comparison of the results between country average and world aggregate share of manufacturing employment. An econometric analysis will therefore be conducted to investigate the trends of country average and world aggregate share of MVA and manufacturing employment in order to draw conclusions about the second condition (B). The paper is organized as follows. A theoretical discussion on manufacturing as the engine of growth is presented in Section 2. Section 3 reviews the literature to draw conclusion about the first condition (A). The data, variables, and econometric models used to examine the second condition (B) are explained in Section 4. Section 5 presents the results. The prospects of manufacturing development in current developing countries are discussed in Section 6, and Section 7 concludes. Figure 2. Manufacturing employment share to total employment (country average). 2. ARGUMENTS FOR ENGINE OF GROWTH Almost no country has achieved and sustained a high standard of living without making significant developments in its manufacturing sector, except for a few oil rich countries and small financial havens (Chang, 2016). Kaldor (1967) empirically tested the manufacturing sector s overriding influence on the rate of economic growth. He put forward various expositions on the special role of manufacturing for growth, including the sector s high productivity (1967: 12), linkage effects (1967: 23) and demand effects (1967: 29). Various other authors have also emphasized these features and have linked them to manufacturing. 4 Manufacturing sector s higher potential for productivity growth (Cornwall, 1977) benefits from the sector s ability to achieve higher levels of capital accumulation, economies of scale and technological progress relative to agriculture and some services (Szirmai, 2013). Lewis (1954) asserts that while there is an unlimited supply of labor at subsistence wage levels, a capitalist sector can expand continuously by investing part of its profits until all the surplus labor has been absorbed. 5 With each round of investment by the capitalist sector, more surplus labor is absorbed, thus increasing the sector s total profits. As the capitalist sector expands with wages in the sector remaining at a subsistence level, the share of profits, savings, and investment in national income rises. Lewis argues that the class of industrialists plays a central role in expanding the capitalist sector because they are the ones who reinvest profits for productive use more than other classes; hence, expanding the capitalist nucleus or industrial development represents a path for capital accumulation and economic growth. 6 In this process, capital accumulation in the capitalist sector may not necessarily increase labor productiv-

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 3 ity if output expansion is proportional to labor expansion. However, the productivity of the economy as a whole increases as labor moves from the subsistence to the capitalist sector, where labor is more productive by using greater amounts of capital, although their wages may remain at the subsistence level in order to increase profits and investment. Once wage levels in the capitalist sector start rising, labor is increasingly substituted with capital to make technology more capital-intensive or the share of capital-intensive industries within the capitalist sector increases (Haraguchi, 2015). Eventually, the labor productivity of the capitalist sector increases as well. The poorer a country is, the faster the productivity in its manufacturing sector will increase, exhibiting a unique tendency of unconditional convergence with the technological frontier (Rodrik, 2013). Thus, the manufacturing sector relative to other sectors has a higher potential for technological progress, as its progress does not depend on country-specific conditions. Manufacturing development tends to accelerate the rate of technological progress of the economy as a whole, partly due to the absorption of surplus labor (Kaldor, 1967) and the creation and dissemination of innovation in certain industries through linkage effects (Marconi, Reis & Araújo, 2016). Economies of scale in manufacturing also drive the sector s productivity growth and the growth of the economy as a whole (Cornwall, 1977; Kaldor, 1967). There are two kinds of economies of scale static and dynamic. The former is related to the reduction of per unit costs due to the reduction of per unit fixed costs, specialization of labor and plant, and the possibility of adopting more efficient production processes as outputs increase, while the latter refers to learning effects (Marconi et al., 2016; Silberston, 1972; Smith, 1904). As the main sector of accumulating physical capital, at least at relatively low income levels, manufacturing benefits from economies of scale more than other sectors. Chenery and Taylor (1968) show that within the manufacturing sector, basic metals, chemicals and petroleum, paper and metal fabricating industries, such as automobiles, in particular benefit from economies of scale. In addition, along the expansion of production scale, dynamic economies of scale through induced technological progress and learning by doing also contribute to the reduction of production costs (Arrow, 1962; Thirlwall, 2002). Linkage effects are also particularly strong in manufacturing, which has much higher backward linkages in general, and forward linkages in resource-based industries, than agriculture and services (Hirschman, 1958). Therefore, an increase in manufacturing output further induces production in the manufacturing sector as well as in other sectors through direct production linkages and indirect multiplier effects, hence driving the growth of the whole economy. Especially at relatively high income level, high tech-industries, such as chemicals and motor vehicles, make significant contributions to manufacturing-related services employment, especially business services, through linkages effects (United Nations Industrial Development Organization, 2013). Business services are increasingly becoming integral part of manufacturing activities. Thus, manufacturing development and structural change play a key role in the development of such modern services. However, as globalization advances, manufacturing production has increasingly been taking place within global value chains. This development, which has in particular accelerated in certain industries since the end of the 1980s, may reduce domestic production linkages and inducement effects to stimulate the domestic economy. Finally, as countries incomes increase, demand effects can provide impetus to manufacturing development, which in turn may stimulate economic development through the manufacturing sector s positive features, its higher productivity, capital accumulation, and linkage effects, resulting in the creation of further demands and leading to a virtuous cycle. Following the logic of Engel s law, an increase in income in developing countries, particularly at a lower income level, tends to raise the share of income spent on manufacturing goods more than or at the expense of the share of income spent on primary products and services. At an early stage of development, finished goods produced from non-metallic minerals, rubber products, wood products and chemical petroleum typically have income elasticities of demand of 1.2 1.5, and durable consumer goods and investment goods have higher elasticities (Chenery & Taylor, 1968). However, manufactured goods that are considered to be essentials, such as foods, leather goods and textiles, usually have elasticities of 1.0 or less. As countries move toward upper middle and higher incomes, consumers tend to increase their share of income spent on services and reduce the share spent on manufactured goods (Cornwall, 1977). The impact of demand effects on manufacturing development is likely to be limited to relatively low-income countries and may possibly be extended if a country succeeds to produce technology-intensive products. 3. ASSESSMENT OF CONDITION (A): MANUFAC- TURING IS NO LONGER THE DRIVER OF GROWTH IN DEVELOPING COUNTRIES Several empirical studies have analyzed Condition (A), i.e., the role of manufacturing as a driver of economic growth in developing countries. Szirmai and Verspagen (2015) tested the relationships between the value-added share of manufacturing and growth of GDP per capita using fixed effects, random effects, Hausman Taylor estimations and between effects models for an unbalanced panel of 92 countries. This relationship was examined for three periods, 1950 70, 1970 90, and 1990 2005, and compared with the results for the service sector. Focusing primarily on the results of conservative Hausman Taylor estimations, the study presents the contribution of manufacturing to GDP per capita growth conditional on the level of education and stage of development. It shows that manufacturing acts as an engine of growth for low- and some middle-income countries, provided that they have a sufficient level of human capital. Such growth engine features are not found in the service sector. Interestingly, their findings for more recent periods indicate that a higher level of human capital (at least seven to eight years of education) is necessary for manufacturing to play a role as an engine of growth in developing countries. Focusing on middle-income economies, Su and Yao (2016) assess, among others, whether the manufacturing sector drives the growth of the services sector. 7 The results from all three methodologies used for the analysis long-run Granger causality tests, cross-sectional regression and panel regression show that manufacturing sector growth drives services sector growth, not the other way around. 8 These findings have led the authors to conclude that manufacturing is indeed the growth engine of economies and, hence, that premature deindustrialization has negative effects on economic growth. Marconi et al. (2016) examine the engine of growth hypothesis by evaluating Kaldor s two laws based on a dynamic panel data for a sample of 63 countries, which includes 32 low and lower middle-income countries (from US$ 1,036 to US$

4 WORLD DEVELOPMENT 4,085 GDP per capita) and 31 upper middle and high-income countries (over US$ 4,085 GDP per capita) for the period 1990 2011. The results confirm the validity of Kaldor s two laws, demonstrating that higher increases in manufacturing output leads to higher economic growth (Kaldor s First Law) and manufacturing productivity (Kaldor s Second Law) for both income groups, with a higher effect on low and lower middle-income countries. Necmi (1999) tested whether Kaldor s conclusions continued to be valid beyond the heydays of rapid industrialization and catch-up of the 1970s, applying an instrumental variable econometric technique for 45 mostly developing countries for the period 1960 94. The results confirm Kaldor s argument that manufacturing is an engine of growth for most of the developing countries studied, with the possible exception of sub-saharan countries. Even for developed countries, McCausland and Theodossiou (2012) find that Kaldor s thesis largely held true for the period 1992 2007. By contrast, the findings of Fagerberg and Verspagen (1999) indicate that manufacturing only acted as an engine of growth in developing, but not in developed countries, in the 1970s and 80s. A cross-sectional regression study by Dasgupta and Singh (2006) involving 48 developing countries from 1990 to 2000 concludes that manufacturing continued to play an engine of growth role, but that services played a similarly important role during that period. Chakravarty and Mitra (2009) and Kathuria and Natarajan (2013) examined the engine of growth hypothesis for India, where the service sector has played a key role in the country s economic development (Aggarwal & Kumar, 2015). In the former (Chakravarty & Mitra, 2009), which covers the period 1973 2004, manufacturing was found to have been one of the drivers of growth, together with construction and services. Kathuria and Natarajan (2013) tested the hypothesis for all 15 states of India in the period 1994 95 to 2005 06, and concluded that manufacturing had indeed acted as an engine of growth in India, despite its declining share in GDP. In a series of recent publications, Rodrik (2013) discusses the nature of manufacturing as a driver of growth, how successful regions have changed their structure to benefit from this driver (Mcmillan, Rodrik, & Verduzco-gallo, 2014), and whether this path of economic development is still available for currently developing countries (Rodrik, 2016). This series reveals that the formal manufacturing sector exhibits a rapid unconditional convergence with labor productivity and that Asian countries grew faster than other regions by moving labor from low to high productive sectors, particularly manufacturing. However, Rodrik (2016) is pessimistic about the continuation of this pattern of economic development for currently developing countries due to premature deindustrialization. In short, evidence from the literature suggests that the engine of growth hypothesis for manufacturing by and large still holds for developing countries particularly those with a higher level of human capital (given their income level). However, availability of the opportunity to use this engine seems to be questionable, which relates to the second condition (B). To shed some light on the second condition (B), it does not suffice to look at the country average share of MVA and manufacturing employment in the respective totals, as is often the case in the literature on pre-mature deindustrialization, and as is also illustrated in Figures 1 and 2. Instead, it is also important to look at manufacturing share at the world aggregate level the share of world MVA and manufacturing employment in world GDP and employment. On the one hand, even though the manufacturing share of a country or region may be decreasing, this drop might be compensated by a rise in manufacturing activities in other parts of the world, if the share of world manufacturing has not changed. On the other hand, if there is a substantial decline in manufacturing share at the world aggregate level, then it is not about a shift among countries but rather about changes in global supply and demand conditions, which have reduced the importance of manufacturing relative to other sectors of the economy. In contrast to studies on the engine of growth hypothesis, there are no studies that have carried out a detailed comparison of the results between the country average and world aggregate share of MVA and manufacturing employment. Thus, the empirical analysis of this paper focuses primarily on the second condition (B). In our analysis, we compare MVA at current prices, constant prices, and manufacturing employment. 9 4. DATA AND VARIABLES This section presents the manufacturing value added and employment database used for the analysis, and describes the definitions of developing and developed country groups for which the analyses were conducted to assess their longterm structural change in terms of MVA and manufacturing employment share. The main data source for sectoral value added is the National Accounts Main Aggregates Database (United Nations Statistics Division, 2015) 10 maintained by the United Nations Statistics Division. The advantage of this database is that it contains national accounts data of essentially all countries over 43 years, i.e., it presents a global picture of changes in sectoral value added based on consistently compiled data of all countries and not based on any estimation by the authors. The database contains data both at current and constant prices 11 which allows us to assess any changes in MVA share attributable to price changes. Unlike manufacturing production data, our manufacturing employment database is based on various sources. Taking the need for intertemporal and international compatibility into consideration, the construction of the database entails four steps and the corresponding approaches and methods, which will be dealt with throughout this section. Table 5 of the Appendix 3 lists the sources included throughout each step. First, the systemic approach combines datasets that include a widely available range of data (across both the international and time spectrums). These databases are merged and their values for comparable data points (i.e., country X in year Y) compared to determine whether the same definition is maintained. To resolve any discrepancies among the values, several procedures are internalized. First, the data sources are ranked to select the majority of values from the most comprehensive databases 12. Simultaneously, their patterns are graphically assessed to establish the gravity of the problem. Finally, severe outliers are excluded from the sample. A large number of observations are still missing within the required sample after the systemic approach is applied. 13 The idiosyncratic approach aims to resolve the majority of these gaps by obtaining data from a multitude of country- or region-specific sources (see Table 5: Sources by step). After the second step, the series (internally) missing observations are linearly interpolated. The available series are converted into their percentage contribution to total employment. The remainder of the values (at the lower or upper end of the time spectrum) is extrapolated in the third phase, the extrapolation approach. This approach

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 5 Table 1. Outline of the estimation procedures Step 1 Step 2 Step 3 Step 4 Systemic approach: core dataset Idiosyncratic approach: additional datasets Extrapolation method: missing observations Aggregate consistency covers three different groups: (i) countries with a maximum of five missing observations at either side of the time spectrum, (ii) countries with between five and 10 years of missing observations, and (iii) countries with gaps in excess of 10 consecutive years. 14 Group (iii) is immediately excluded from the sample, as extrapolation is likely to result in biased estimates. Group (i) undergoes a linear extrapolation process, whereas missing information is resolved in Group (ii) through extrapolation by means of the last observation carried forward. 15 Following these procedures, all percentage values are merged together in a new database, creating the foundation for the fourth and final step. The final step aims to mitigate internal compatibility issues and thereby improves the consistency at the aggregate data level. As already explained in the previous section, consistency is an essential feature of the database. Having derived the shares in the previous steps, they are multiplied in this final step with the aggregate employment values from the Total Economy Database (The Conference Board, 2012). 16 This database provides estimates on total employment levels for a total of 128 countries. The employment series for the remaining nine countries are obtained using idiosyncratic methods. 17 Finally, note that this mechanism still implies that the estimates will better reflect the original data for the series in percentages and covers aggregate patterns as compared to those that are generated on a level-base for individual countries. There are two conditions we have to take into consideration when classifying countries into a group of developing countries for the analysis of the group s long-term structural change in terms of MVA and manufacturing employment. First, the countries classified into the developing country group need to have been developing countries throughout the period of analysis. We are interested in whether manufacturing opportunities in developing countries increased or decreased over the 43-year period to assess any changes in the significance of manufacturing for their economic development. For a consistent and accurate analysis, mature highincome countries that usually experience deindustrialization must be excluded from the group of developing countries for the entire period of analysis. Secondly, ideally, the same number of countries and geographic coverage has to be maintained throughout the period of analysis to ensure that any changes in MVA and manufacturing employment share are attributed to changes in the economic activities of the group analyzed. Due primarily to the break-up of the USSR and Yugoslavia in 1990, and the fact that only some of the former USSR and former Yugoslav countries reached a high income level after 1990, it is not possible to simultaneously maintain the above two conditions. We therefore have two sets of balanced panel data for developing countries before 1990 and after 1990; geographical coverage is maintained within each of the balanced panel data. The difference between the two datasets is largely related to the break-up of the USSR and Yugoslavia. 18 The USSR and Yugoslavia are entered into the dataset of developing countries prior to 1990 as two individual countries. After 1990, the former USSR and Yugoslav states are included in the dataset of developing countries, and with the exception of Estonia, Latvia, Lithuania, Russia, Croatia and Slovenia, reached a high income level in the period 1990 2013. The effects of these changes are minimal as we use shares to assess changes in MVA and manufacturing employment, which does not affect the interpretation of our results, as we focus on the trend (increasing, decreasing or flat) of each period before and after 1990. For the robustness check, we also include the results based on the dataset, which excludes all merged or separated countries, 19 so we can use a single balanced dataset of the same number of developing countries and the same geographic coverage for the entire 43-year period. Developing countries include those that did not reach a high income level in any year until 2013 20 based on the threshold income level for high-income countries defined annually by the World Bank Analytical Classifications (WBAC). 21 This ensures that declines in MVA and manufacturing employment share experienced by developing countries are not caused by the normal pattern of structural change, which usually leads to deindustrialization at high income levels. For countries that have not been classified by WBAC, we use The World Bank (2015) and United Nations Statistics Division (2015) 22 to jointly assess those countries income levels 23 and subsequently define their level of development based on WBAC. For details on the database, see Appendix 2. 5. RESULTS (a) MVA share at current prices Figure 3-A presents the changes in country average share of MVA in GDP of each development group. 24 The share of developed countries has steadily declined since 1970. Developing countries showed a stable trend until 1990, but have experienced a statistically significant declining trend since (see Appendix 4, Table 8, Column 2). The result at current prices confirms the lower share of MVA in developing countries in the post-1990 period. Figure 3-B illustrates the changes in the share of aggregate MVA in aggregate GDP of each development group as a whole (hereafter referred to as aggregate share ). While there is no change in the steadily declining trend of developed countries, the aggregate share of developing countries indicates a different trend. The aggregate share decreased until 1993 and then remained more or less stable until 2013. This sudden change in the aggregate share trend may not reflect the long-term trends of the world manufacturing share due to the economic collapse of the former Soviet Union and subsequent consolidation of manufacturing industries. If we exclude the USSR, Yugoslavia as well as all other merged and separated countries 25 from our dataset, as shown in Figure 3-D, we do not observe a statistically significant increasing or decreasing trend over the 43-year period. In any case, when either including or excluding merged and separated countries, there is no statistically significant declining trend in the aggregate share of developing countries since 1990 (see Appendix 4, Table 8, Columns 4 & 8). 26 (b) MVA share at constant prices At constant prices, the country average MVA share in GDP has also shown a declining trend in developing countries since 1991 (see Figure 4-A and Appendix 5, Table 9, Column 2). In terms of aggregate share (Figure 4-B), developing countries

6 WORLD DEVELOPMENT Figure 3A D. (A and C) MVA share to GDP at current prices (country average). (B and D) MVA share to GDP at current prices (aggregate average). showed a rising trend over 43 years. By contrast, the share in developed countries decreased until 1993 and has remained constant since. The result remains the same if we exclude all merged and separated countries from the dataset to apply a

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 7 Figure 4A D. (A and C) MVA share to GDP at constant prices (Country average). (B and D) MVA share to GDP at constant prices (Aggregate average). single balanced panel data of identical countries for the entire 43-year period (Figure 4-C and D). In short, whether at current or constant prices, the country average share of MVA in GDP in developing countries has

8 WORLD DEVELOPMENT Figure 5A D. (A and C) Manufacturing employment share to total employment (country average). (B and D) Manufacturing employment share to total employment (country aggregate).

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 9 been declining for the last 20 years. This is in line with the premature deindustrialization argument, which contends a declining manufacturing share based on the average picture of developing countries (see Figures 1 and 2, for example). However, using either current or constant prices, the size of MVA in developing countries as a whole has not changed, if not increased, since 1990, as evidenced in the trend of the aggregate share. Moreover, if we exclude merged and separated countries (though the difference in the result is caused by the breakup of the USSR alone), the MVA share in developing countries has not changed since 1970, even at current prices. (c) Employment Developed countries have substantially reduced their share of manufacturing employment in total employment over the last 43 years (see Figure 5-A and B). This is not surprising, as countries at high income levels usually experience deindustrialization, with dwindling numbers of manufacturing jobs as part of the normal pattern of structural change. In the case of developing countries, the country average share has declined since 1991 while the aggregate share has shown a statistically significant increasing trend since 1990, as indicated in the statistical results (see Appendix 6, Table 10, Columns 2 & 4). As illustrated in Figure 5-C and D, developing countries without the USSR and other merged or separated countries show similar trends. However, the declining trend of country average share since 1990 is statistically insignificant. It is also noteworthy that the aggregate share of world manufacturing employment in total world employment (including both developed and developing countries) has hardly changed since 1970. Our analysis of value added and employment trends indicates that no matter how we look at manufacturing aggregate share in developing countries, whether in terms of value added at current prices, constant prices or employment, or whether we include or exclude the USSR and other merged or separated countries the shares have not declined (if not increased) since 1990, and for the entire 43 years since 1970, if the merged and separated countries are excluded. This is quite noteworthy, especially as regards the increasing trend of the share of manufacturing employment in developing countries. An increasing trend is recorded despite the so-called statistical illusion which results from past changes to statistical classifications, and usually indicates a reduction in the total number of manufacturing jobs by reclassifying certain manufacturing jobs to services (Tregenna, 2015). Because premature deindustrialization has been taking place in Africa and Latin America since the mid-1970s (Timmer, de Vries & de Vries, 2015; Tregenna, 2015, p. 104), the fact that the MVA and manufacturing employment shares have at least maintained their levels since 1970 signifies that the decline of manufacturing share as observed in country average share and Figures 1 and 2 is not caused by any long-term, systemic shift in the global economic structure to reduce manufacturing share relative to others. Then why has the manufacturing share in GDP been decreasing in developing countries or why are they experiencing premature de-industrialization? The differences in the results of aggregate share and country average share seem to indicate the possibility of increasing the concentration of manufacturing activities in a small number of (large) developing countries, while the MVA share of developing countries as a whole in their total GDP has not changed. If this is indeed the case, it makes sense that the aggregate share maintained at least a stable trend while a large number of developing countries manufacturing share (country average share) has declined in recent years. In order to test this hypothesis, we examine the level of MVA and employment concentration in developing countries from 1971 to 2013, using the Herfindahl index and Gini. 27 While the discussion here focuses on developing countries, the following figures also include the results of developed countries for reference purposes. Figure 6 shows how the level of MVA concentration in developing and developed countries has changed since 1970 at current prices based on the Herfindahl index (Figure 6-A) and Gini (Figure 6-B). The changes are less drastic in Gini, but the concentration trends are similar in both results. In either case, the concentration level has largely remained stable for developed countries over the 43-year period, but has increased in developing countries since the beginning of the 1990s. This onset of the increase in concentration corresponds with the point in time when the country average share began to decline (Figure 3-A and C). The trends based on constant prices (Figure 7-A and B) illustrate the stability of the level of concentration in developing countries until 1990, and a steady and faster increase in developing countries level of MVA concentration since the beginning of the 1990s. The case of employment requires a more nuanced interpretation. Employment concentration in developing countries rose until the end of the 1980 s according to both the Herfindahl index and Gini (see Figure 8-A and B). This increasing concentration did not result in the decline of country average share (Figure 5-A), probably due to the rise in the share of manufacturing employment in the total employment of developing countries as a whole over the first 20 years (Figure 5-B). From 1991 to 1998, employment concentration increased in developing countries while the aggregate share only changed slightly. This combination is likely to have generated the declining trend of county average share (Figure 5-A). Finally, from 1998 to 2010, country average share generally remained flat due to the decline in both concentration and aggregate employment share during 1998 2002, as well as to the increase in concentration and aggregate employment share from 2002 to 2010. We have also tested the above for developing countries, excluding all merged or separated countries from the database, but there was no change in the result of steady and faster increases in the level of MVA concentration from the beginning of the 1990s, and in the shifting pattern of employment concentration. To further validate the above findings, we exclude China, which is a large country that experienced rapid industrialization, from Figures 6A and B, 7A and B and 8A and B to identify any changes in concentration trends for MVA shares at current and constant prices and manufacturing employment share, respectively. Unlike Figures 6 8, which show an upward trend for concentration of MVA shares since 1991, the figures excluding China give no indication of an increase in concentration since 1990. For employment share, the level of concentration increased until 2002, but has been decreasing since (Figures 9A, B and 10A, B and 11A, B). 6. PERSPECTIVE ON MANUFACTURING DEVELOP- MENT IN CURRENT DEVELOPING COUNTRIES Evidence from the literature and empirical analyses lead us to conclude that Conditions (A) and (B) do not hold, implying that the characteristics of the manufacturing sector itself and its relative importance in the world economy have not changed over the last 40 years. The results indicate that countryspecific factors such as policies and comparative advantage

10 WORLD DEVELOPMENT Figure 6A B. (A) MVA concentration at current prices (HI). (B) MVA concentration at current prices (GINI). Figure 7A B. (A) MVA concentration at constant prices (HI). (B) MVA concentration at constant prices (GINI).

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 11 Figure 8A B. (A) Manufacturing employment concentration (HI). (B) Manufacturing employment concentration (GINI). Figure 9A B. (A) MVA concentration at current prices (HI), excluding China. (B) MVA concentration at current prices (GINI), excluding China.

12 WORLD DEVELOPMENT Figure 10A B. (A) MVA concentration at constant prices (HI) excluding China. (B) MVA concentration at constant prices (GINI) excluding China. Figure 11A B. (A) Manufacturing employment concentration (HI) excluding China. (B) Manufacturing employment concentration (GINI) excluding China.

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 13 Figure 12A B. (A) Developing countries with sustained high growth and the highest growth rate in their manufacturing sector (Pre-1990). (B) Developing countries with sustained high growth (population > 1 million) and the highest growth rate in their manufacturing sector (Pre-1990). seem to have played a greater role in the differences in countries manufacturing performance. However, the results do not mean that the conditions currently developing countries face in terms of manufacturing development are the same prior to and after 1990. As globalization advances, the opportunities and policy space available for manufacturing development might have changed. For example, trade liberalization and the increasing role global value chains play in manufacturing production and trade may have intensified price competition among countries and given distinct advantages to certain countries with economies of scale and agglomeration. While an investigation of these effects on manufacturing development goes beyond the scope of this paper, this last section before the conclusions section provides a general overview of the characteristics of successful countries, namely all countries that have met the two conditions of both fast (more than 7% average annual growth) and sustained growth (more than 25 years), and comparing such countries before and after 1990. 28 As Figure 12-A illustrates, before 1990, out of 143 developing countries there were nine countries that met the long-term, high growth conditions. These included countries with very diverse demographic and geographic characteristics, namely large, small, island and natural resource rich countries. Out of these nine countries, six (66.7%) recorded the highest growth rate in their manufacturing sector during 1971 1990. 29 Small countries tend to have different development patterns from other countries, as their development is more dependent on their given geographic and natural endowment conditions (Armstrong & Read, 1995; Kuznets, 1971; Perkins & Syrquin, 1989). In the case of very small countries, in particular, success in one or a few industries such as financial services, tourism or agri-business, could have a significant impact on their long-term growth rate. Larger countries do not typically follow this development trajectory. Therefore, Figure 12-B excludes countries with a population of less than one million. During 1971 90, the manufacturing sector in six out of eight countries with a population of more than one million (75%) registered the highest growth rate. In post-1990, as illustrated in Figure 13-A, out of 143 developing countries analyzed 10 countries with very diverse demographic and geographic characteristics met the two conditions of at least 7% or higher for 25 years, out of which at least 20 years fell within the period 1990 2013. Six out of 10 countries (60%) registered the highest growth rate in manufacturing during their sustained high growth periods. If we remove the countries with a population of less than one million, six out

14 WORLD DEVELOPMENT Figure 13A B. (A) High sustained growth developing countries with the highest growth rate in manufacturing (post-1990). (B) High sustained growth developing countries (population > 1 million) with the highest growth rate in manufacturing (post-1990). of eight countries (75%) had the highest growth rate in manufacturing during their long-term high growth periods (see Figure 13-B). There is thus not much difference in the strong performance of manufacturing relative to other sectors before and after 1990 in the countries which experienced fast and sustained growth. For countries with a population of more than one million, the number and percentage of countries that recorded their highest growth rate in manufacturing is exactly the same six out of eight countries or 75%. As this review of countries with fast, sustained growth shows, even after 1990, a small country like Cambodia was able to experience rapid, sustained growth through manufacturing development. A small and landlocked country, Lao People s DIR, also belongs to the league of most successful countries, with the fastest growth in manufacturing and doubling the share of manufacturing value added in GDP since 1990, although starting from a small base in manufacturing. While the literature indicates that China s opening to trade did not have a significant effect on the broad sectoral structures of other countries, studies nonetheless reveal that there was a negative impact on labor-intensive industries in East Asian countries (especially small ones) during 1990 2000 (Dimaranan, Ianchovichina, & Martin, 2006; Wood & Mayer, 2011). Despite this external condition, developing East Asian countries including small and landlocked ones have been able to achieve fast, sustained manufacturing growth in recent years (in Figure 13-B). Although the results in this section do not show conclusive evidence of negative or positive effects of current global economic conditions on manufacturing development in developing countries, country level observations lend at least additional empirical support to the premise that manufacturing development has remained important for fast, sustained growth after 1990, and that this path has not been closed even to small or landlocked countries in recent years. China seems to have reached the peak in terms of output and employment in the cloth (combination of the textiles and wearing apparel) industry (Appendix 7). As the country upgrades its industrial structure and moves out of labor-intensive industries, space for industrialization might open further for low-income countries if the right policies are implemented. 7. CONCLUSIONS Despite recent assertions of shrinking opportunities for manufacturing development in developing countries and the decrease in the importance of manufacturing for economic development, this study shows that there is no evidence supporting this argument. Even after 1990, the manufacturing sec-

THE IMPORTANCE OF MANUFACTURING IN ECONOMIC DEVELOPMENT: HAS THIS CHANGED? 15 tor in developing countries continues to meet the conditions to be described as a driver of economic development, especially in achieving high sustained growth while retaining at least the same size in GDP and total employment as in the period 1970 90. Thus, the declining MVA and manufacturing employment share in many developing countries has not been caused by changes in the development quality or quantity of manufacturing activities, but is mostly attributable to the failures of manufacturing development in a large number of developing countries against the backdrop of a rapid development of manufacturing in a small number of countries, resulting in a concentration of manufacturing activities in developing countries. China is an example of an exceptionally successful country. In recent years, China had an MVA share of more than 30% both at current and at constant prices, with the average share of developing countries being around 11% 14%. In the case of manufacturing employment, China has had a share of more than 15% since the end of the 1980s and an 18% to 19% share since 2007, in comparison to an average share of 11% and 12% in developing countries for most of the 43-year period under investigation. In terms of population, China s development is equivalent to that of the 38 average-sized countries that registered rapid simultaneous industrialization. Considering that China s population is greater than the total of all African countries together, its industrialization can also be compared with the rapid industrialization of all African countries together (and more). Thus, the results of this study may not be so counterintuitive. We do not assume that the trends observed in our analysis will continue in the future. However, given the recent claims about the diminishing significance of manufacturing or the increasing difficulty to pursue economic development by following the conventional path of industrialization, the evidence of the significance of manufacturing remained unaltered in the two periods studied, i.e., 1970 90 and 1990 2013 is of relevance. After its success in laborintensive industries, China is likely to upgrade its industrial structure following the path of high-income countries. Once this happens, there may be greater opportunities for current low-income countries to pursue manufacturing activities; manufacturing would then perhaps become more, not less, important for them. Thus, the recommendation for developing countries is to not turn away from manufacturing and abandon the path of economic development through industrialization, but to emulate the experience of rapid industrialization that occurred even in recent years. DISCLAIMER The views expressed here are those of the authors and do not reflect the views of the United Nations Industrial Development Organization. NOTES 1. Following the specifications in Rodrik (2016, p. 6), Figures 1 and 2 illustrate the patterns of manufacturing value added and employment share based on a quadratic estimation using country fixed effects. Please see Appendix 1, Table 2 for the underlying model and time span covered. The estimation results Figures 1 and 2 are presented in Appendix 1, Table 2, column (1) (4), the level and quadratic terms are statistically highly significant for both manufacturing value added and employment share 2. For example, production and trade are increasingly taking place in global value chains, which may promote the participation of developing countries in specific tasks based on comparative advantages, but could confine a country s policy space for upgrading (Chang, 2016). 3. According to one of Kaldor s Laws, the higher the growth of manufacturing output, the higher the growth of GDP (Kaldor, 1967). 4. For a comprehensive review, refer to Szirmai and Verspagen (2015). 5. Lewis explains that in practice, wages paid to employees in the capitalist sector are higher than those in the subsistence sector due to (1) the higher cost of living in the capitalist sector, which is often located in an urban area, (2) the psychological cost of moving from an easy life in the subsistence sector to the better organized urban environment, or the recognition that workers demand a higher wage after having acquired certain tastes and social prestige associated with urban living. 6. Szirmai (2013) demonstrates that manufacturing becomes less capitalintensive relative to other sectors as the country s income increases. 7. The authors use both relative and absolute definitions of middle income economies. The former is defined as those with a GNI per capita lying between 7% and 45% of the US GNI level, while the latter includes those with a GNI per capita between USD 2,250 and USD 14,999 at 2005 international dollar PPP. 8. The dataset for manufacturing value added and service value added mostly covers the 1970s 2013, and includes around 110 countries. 9. In addition to MVA share at constant prices, changes in MVA share at current prices are also included in this study because they indicate the sector s attractiveness (as countries do not want to be specialized in sectors whose prices are declining). 10. Agriculture ISIC A-B, Manufacturing ISIC D, and Services ISIC G-I (United Nations Statistics Division, 2015, updated version as of December 2014). 11. According to the United Nations Statistics Division (2015, updated version as of December 2014), gross domestic product (GDP) at constant prices refers to the volume level of GDP. Constant prices estimates of GDP are obtained by expressing values in terms of a base period. The country specific volume level of GDP is used for quantity revaluation, for instance, by multiplying the current period quantity by the base period (2005) prices. 12. In this case, the ranking is GGDC-ASD and GGDC-10SDB, KILM, ILO, GGDC-WIOD (see Table 5: Sources by step). The databases referred to as being more comprehensive include (1) a larger sample, and (2) less alternative definitions. The ILO recently introduced a database on World Employment Social Outlook (WESO). The WESO database includes the employment data of 174 countries since 1991. The major difference between the WESO and the databases used in this study is the employment data of China. The WESO recorded a lower manufacturing employment for China, especially for recent years, and hence, a lower share of manufacturing employment in total employment. This low figure is in contrast to data from other available employment databases such as those used for this study, namely Asian Productivity Organization s database as well as to China s national census. Nevertheless, the share of industry (including both the manufacturing and non-manufacturing sectors) in total employment between the WESO and our databases are much closer