Manufacturing Matters...but It's the Jobs That Count

Felipe, Jesus; Mehta, Aashish; Rhee, Changyong Working Paper Manufacturing Matters...but It's the Jobs That Count ADB Economics Working Paper Series, No. 420 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Felipe, Jesus; Mehta, Aashish; Rhee, Changyong (2014) : Manufacturing Matters...but It's the Jobs That Count, ADB Economics Working Paper Series, No. 420, Asian Development Bank (ADB), Manila, https://hdl.handle.net/11540/4216 This Version is available at: https://hdl.handle.net/10419/128534 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK AsiAn Development BAnk 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org Manufacturing Matters… but It’s the Jobs That Count Practically every economy that enjoys a high income today experienced a manufacturing employment share in excess of 18%–20% sometime since the 1970s. Manufacturing output share thresholds are much poorer predictors of rich-country status. We also find that the maximum expected employment share for a typical economy has fallen to around 13%–15%. Industrialization in employment has been more important for eventual prosperity than industrialization in output; and high manufacturing employment shares are becoming more difficult to sustain as incomes rise. Our findings suggest that the path to prosperity through industrialization may have become more difficult. About the Asian Development Bank ADB’s vision is an Asia and Pacific region free of poverty. Its mission is to help its developing member countries reduce poverty and improve the quality of life of their people. Despite the region’s many successes, it remains home to approximately two-thirds of the world’s poor: 1.6 billion people who live on less than $2 a day, with 733 million struggling on less than $1.25 a day. ADB is committed to reducing poverty through inclusive economic growth, environmentally sustainable growth, and regional integration. Based in Manila, ADB is owned by 67 members, including 48 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. MAnufACTurIng MATTers… BuT IT’s The JoBs ThAT CounT Jesus Felipe, Aashish Mehta, and Changyong Rhee adb economics working paper series no. 420 november 2014  ADB Economics Working Paper Series Manufacturing Matters… but It’s the Jobs That Count Jesus Felipe, Aashish Mehta, and Changyong Rhee No. 420 | 2014 Jesus Felipe ( [email protected]) is Advisor in the Economics and Research Department of the Asian Development Bank. Aashish Mehta ([email protected]du) is Associate Professor in the Global Studies Department at the University of California-Santa Barbara. Changyong Rhee ([email protected]) is Director of the Asia and Pacific Department at the International Monetary Fund. This paper develops ideas that we introduced in the Special Chapter on “Asia’s Economic Transformation: Where to; How: and How Fast?” in the Key Indicators for Asia and the Pacific 2013 (Asian Development Bank 2013). We are grateful to Emmanuel Andal, Connie Dacuycuy, Liming Chen, and Rey Galope for their research assistance. ASIAN DEVELOPMENT BANK Asian Development Bank 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org © 2014 by Asian Development Bank November 2014 ISSN 2313-6537 (Print), 2313-6545 (e-ISSN) Publication Stock No. WPS147001-3 The views expressed in this paper are those of the author and do not necessarily reflect the views and policies of the Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” in this document, ADB does not intend to make any judgments as to the legal or other status of any territory or area. Note: In this publication, “$” refers to US dollars. The ADB Economics Working Paper Series is a forum for stimulating discussion and eliciting feedback on ongoing and recently completed research and policy studies undertaken by the Asian Development Bank (ADB) staff, consultants, or resource persons. The series deals with key economic and development problems, particularly those facing the Asia and Pacific region; as well as conceptual, analytical, or methodological issues relating to project/program economic analysis, and statistical data and measurement. The series aims to enhance the knowledge on Asia’s development and policy challenges; strengthen analytical rigor and quality of ADB’s country partnership strategies, and its subregional and country operations; and improve the quality and availability of statistical data and development indicators for monitoring development effectiveness. The ADB Economics Working Paper Series is a quick-disseminating, informal publication whose titles could subsequently be revised for publication as articles in professional journals or chapters in books. The series is maintained by the Economics and Research Department.  CONTENTS TABLES AND FIGURES iv ABSTRACT v  I. INTRODUCTION 1  II. DATA 3 III. CROSS-SECTIONAL EVIDENCE ON THE IMPORTANCE AND FEASIBILITY OF ATTAINING HIGH MANUFACTURING EMPLOYMENT SHARES 5 IV. ARE INDUSTRIALIZED COUNTRIES RICH? 9 V. HAS IT BECOME MORE DIFFICULT TO ACHIEVE HIGH MANUFACTURING SHARES? 17 VI. INTERPRETATION AND CONCLUSIONS 25 APPENDIXES 27 REFERENCES 33  TABLES AND FIGURES TABLES 1 Peak and Current Manufacturing Shares, Selected Economies 3 2 Regressions Corresponding to Figures 1, 2, and 3 6 3 Probabilities of Being Rich, Conditional on Achieving Manufacturing Employment Share Thresholds 11 4 Probabilities of Being Rich, Conditional on Achieving Manufacturing Output Share Thresholds (135 economies) 14 5 Economies Categorized by Industrialization in Output and Employment 16 6 Regressions of (Log) Manufacturing Employment Shares over Time and across Economies 21 7 Country Fixed Effects (Employment Share Regression 5) and Year of Peak Manufacturing Employment 23 8 Regressions of (Log) Manufacturing Output Shares over Time and across Economies 24 FIGURES 1 Peak Manufacturing Employment and Subsequent Prosperity 6 2 Peak Manufacturing Employment and Output Shares in Time 8 3 Per Capita GDP at the Time of Peak Manufacturing Shares in Employment and Output 9 4 Manufacturing Employment Share versus Per Capita GDP, Variables in Natural Logarithms 17 5 Manufacturing Output Share versus Per Capita GDP, Variables in Natural Logarithms 18   ABSTRACT This paper asks, first, whether today’s developing economies can achieve high-income status without first building large manufacturing sectors. We find that practically every economy that enjoys a high income today experienced a manufacturing employment share in excess of 18%–20% sometime since the 1970s. Manufacturing output share thresholds are much poorer predictors of rich-country status than their employment counterparts. This motivates us to ask whether it is becoming more difficult to sustain high levels of manufacturing activity. We find that the maximum expected employment share for a typical developing economy has fallen to around 13%–15%, and that deindustrialization in employment sets in at much lower income per capita levels of $8,000–$9,000, than it once did. Neither manufacturing output shares, nor the level of income at which they decline have fallen as obviously. These results are consistent with the idea that industrialization in employment terms has been more important for eventual prosperity than has industrialization in output terms; and that high manufacturing employment shares are becoming more difficult to sustain as incomes rise. This suggests that the path to prosperity through industrialization may have become more difficult. Keywords: industrialization, inverted U-shape, manufacturing JEL Classification: O14  I. INTRODUCTION This paper asks two complementary questions. First, whether today’s developing economies can achieve high-income status without first building large manufacturing sectors; and second, whether it is becoming more difficult to sustain high levels of manufacturing activity. These are relevant questions because a long tradition in development economics holds that manufacturing is the engine of growth (Kaldor 1966; Chenery, Robinson, and Syrquin 1986). However, as we discuss below, there is a growing perception that economies are finding it more difficult to sustain high levels of manufacturing output and employment while simultaneously increasing wages. The relevance of manufacturing for economic growth derives from the fact that faster growth in manufacturing generates faster growth in productivity in other sectors of the economy. Manufacturing draws resources from traditional sectors of the economy, often without significantly reducing output in these sectors. Channeling these resources into manufacturing is beneficial because manufactured goods have high income elasticities of demand, and because many manufacturing activities are produced under economies of scale. Moreover, manufacturing has a potential for productivity catch-up that is unmatched by most services.1 Along these lines, recent research confirms that manufacturing is key for economic development. Rodrik (2013a) shows that manufacturing exhibits unconditional convergence in labor productivity—industries that start farther away from the labor productivity frontier experience significantly faster productivity growth even without conditioning on variables, such as domestic policies, human capital, geography, or institutional quality. Economies with higher manufacturing employment shares should therefore grow faster. See also Amable (2000), Fagerberg (2000), Peneder (2003), Rodrik (2009), Szirmai (2012), Szirmai and Verspagen (2011), and UNIDO (2013). For these reasons, many national governments have targeted manufacturing in their development plans. For example, India’s 2011 National Manufacturing Policy aims at raising the share of manufacturing in gross domestic product (GDP) to 25% and calls for setting up manufacturing zones to create 100 million manufacturing jobs. The Philippines is also developing a comprehensive manufacturing road map, and Indonesia passed a new Industry Law in early 2014. Developed economies like the United States (US), Australia or the members of the European Union, are also interested in industrializing, or rather, in reindustrializing after decades of deindustrializing (Helper, Krueger, and Wial 2012; Felipe, forthcoming 2015). As noted above, however, there is a growing perception that, in recent times, it has become more difficult for economies to sustain high levels of manufacturing output and employment while simultaneously increasing wages and living standards (Rodrik 2009). This difficulty has been attributed to at least two forces. First, the internationalization of supply chains and increased international competition are making the location of manufacturing activity more sensitive to wage improvements, other things equal (Hasan, Mitra, and Ramaswamy 2007; Rodrik 1997). This phenomenon tends to reduce manufacturing output and employment levels more in high-wage, high-income economies, than in their poorer counterparts. To the extent that manufacturing output and employment shares tend to increase with income per capita when incomes are low, and then to decrease with incomes, following an inverted U-shape, this force should cause the per capita income level at which manufacturing employment and output shares peak to decline. However, to the extent that economies that experience rising wages can substitute capital for labor, this shift should be more apparent in employment than in output shares.  1 See the Symposium on Kaldor’s Growth Laws, published in the Journal of Post Keynesian Economics (1983). 2 | ADB Economics Working Paper Series No. 420  Second, it has long been accepted that technological change has placed downward pressure on the sector’s demand for less-skilled workers (Berman, Bound, and Griliches 1994; Berman, Bound, and Machin 1998; Goldin and Katz 2009), and there is now growing concern that technological change and the efficiencies that derive from globalized mass production are generally labor displacing (Cowen 2013; Brynjolfsson and McAffee 2014). Faster rates of labor-displacing technological change in manufacturing than in other sectors should push manufacturing employment shares down relative to manufacturing output shares. Rodrik (2013b, p. 52) has echoed this sentiment, “Technological changes are rendering manufacturing more capital and skill intensive, reducing the employment elasticity of industrialization and the capacity of manufacturing to absorb large volumes of unskilled labor from the countryside and from the informal sector.” Together, these hypothesized forces make us wonder whether the rapid manufacturing-driven growth experienced by the Republic of Korea, Singapore, or Taipei,China, remains possible elsewhere. If it has become less likely, is there any evidence of an alternative path to prosperity? This paper, therefore, answers two sets of questions. First, are there relevant examples of economies that are rich today but did not attain large manufacturing sectors at some point in their past? Are there economies that did not achieve large manufacturing sectors and became rich anyway? And has manufacturing employment or manufacturing output been more relevant to eventual prosperity? Second, is it indeed the case that high manufacturing employment and output shares are becoming more difficult to sustain? Specifically, if, as hypothesized, economies are experiencing a combination of increasing international competition for manufacturing jobs and rising labor productivity in manufacturing relative to other sectors, we should expect: (H1) peak manufacturing employment shares to decline; and (H2) both employment and output shares to peak at lower and lower income levels over time. Moreover, to the extent that economies can switch to more capitalintensive products and techniques as incomes rise, and to the extent that labor productivity increases drive manufacturing employment down, (H3) both trends (i.e., H1 and H2) should be more apparent in employment than in output shares. The paper tests these three hypotheses. To shed light on these issues, we compile a carefully cleaned data set on manufacturing employment shares for 52 economies that is consistent over time. Reliable data on manufacturing employment shares over time have been available for a relatively small set of economies, most of which have, at some point in their history, supported high levels of manufacturing employment. It is therefore unknown, even as a historical fact, whether economies that failed to generate a large number of manufacturing jobs actually fared worse than those that did generate such jobs. We also have output share data for these same economies and for a further 83 economies. These data sets permit us to test whether there are particular thresholds in manufacturing employment and output that highincome economies have tended to cross. Our main contributions to the literature are two novel findings. First, that manufacturing employment is a better predictor of prosperity than manufacturing output; and second, that economies are finding more difficult to sustain high manufacturing employment shares (but not output shares) as their incomes rise. Table 1 shows the peak and current manufacturing employment and output shares for a group of advanced and developing economies. Employment shares in advanced economies peaked earlier and at significantly higher levels than in developing economies. Current employment shares, on the other hand, are slightly lower in advanced economies. Peak output shares, in contrast, have not fallen nearly as much. Manufacturing Matters… but It’s the Jobs that Count | 9  Figure 3: Per Capita GDP at the Time of Peak Manufacturing Shares in Employment and Output a. Employment (52 countries b. Output (52 countries) c. Output (135 countries) Notes: X-axis shows the year that the 7-year moving average of the manufacturing sector's employment and output shares peaked between 1970 and 2010. Y-axis shows the level at which they peaked. See Appendix Table A.3 for the definition of the codes. Source: Authors’ calculations. In summary, these results confirm that achieving high manufacturing employment shares is a key determinant of subsequent prosperity; and that there are strong headwinds facing manufacturing employment that are likely to constrain economies’ ability to achieve high incomes through industrialization. IV. ARE INDUSTRIALIZED COUNTRIES RICH? Motivated by the findings above, in particular by the positive relationship between income per capita and the peak manufacturing employment shares (Figure 1, and Panel A of Table 2), we investigate here whether all rich economies industrialized, and whether all economies that industrialized are rich. We do this by asking whether there are thresholds for manufacturing output and employment shares that distinguish rich from non-rich economies. We classify an economy as “rich” if its average per capita GDP during 2005–2010 exceeds some cutoff. A cutoff of $12,000 in 2005 prices (not PPP corrected) is a convenient benchmark, corresponding roughly to the World Bank’s definition of a high-income economy. Using this cutoff, exactly half of the 52 economies for which we have employment data are rich. We will also see what happens if we use different cutoffs. We will similarly propose that economies have “industrialized in employment” if their manufacturing employment shares crossed a particular threshold at any point between 1970 and 2010. Industrialization in output is defined analogously. ARG AUS AUT BAN BEL BOL BWA BRA CAN CHL PRC COL CRI DEN SLV FIN FRA GRC GTM HND HKG IND INO IRE ITA JPN KOR LUX MAL MEX NET NOR PAK PAN PER PHI POL POR PTR ROU SMR SIN SPA SUR SWE SWI SYR THA TTO UKG USA VEN 4 6 8 10 12 GDP per capita at time of peak 1970 1980 1990 2000 2010 Year of peak share ARG AUS AUT BAN BEL BOL BWA BRA CAN CHL PRC COL CRI DEN SLV FIN FRA GRC GTM HND HKG IND INO IRE ITA JPN KOR LUX MAL MEX NET NOR PAK PAN PER PHI POL POR PTR ROU SMR SIN SPA SUR SWE SWI SYR THA TTO UKG USA VEN 4 6 8 10 12 1970 1980 1990 2000 2010 Year of peak share AFG ALB DZA AND AGO ARG AUS AUT BAN BEL BLZ BEN BHU BOL BWA BRA BGR BFA BDI CAM CMR CAN CAF TCD CHL PRC COL COD COG CRI CIV CUB DEN DJI DOM ECU EGY SLV GNQ FIN FRA GAB GMB GHA GRC GTM GIN GNB GUY HTI HND HKG HUN IND INO IRN IRQ IRE ITA JAM JPN JOR KEN PRK KOR LAO LBN LSO LBR LBY LIE LUX MAC MDG MWI MAL MLI MRT MUS MEX MON MON MAR MOZ MYA NAM NEP NET NZL NIC NER NGA NOR OMN PAK PAN PNG PRY PER PHI POL POR PTR QAT ROU RWA SMR SAU SEN SLE SIN SOM ZAF SPA SRI SUR SWZ SWE SWI SYR TZA THA TGO TTO TUN TUR UGA ARE UKG USA URY VEN VIE ZMB ZWE 4 6 8 10 12 1970 1980 1990 2000 2010 Year of peak share 10 | ADB Economics Working Paper Series No. 420  We experiment with multiple thresholds for manufacturing shares. For a given income cutoff and threshold manufacturing share, we will conclude that industrialization (I) has been necessary for becoming rich (R) if we observe in the data that all rich economies industrialized (i.e., no industrialization, no high income; Pr(R|~I)=0); and we will say that it is a sufficient condition if we observe that all economies that have industrialized are rich (i.e., industrialization guarantees richcountry status; Pr(R|I)=1). We will also select, for each income cutoff, the threshold manufacturing share that gives us the most separation between rich and non-rich economies (i.e., the level of I— manufacturing share, that maximizes Δ ≡ Pr(R|I) - Pr(R|~I)). The higher this difference is, the more powerful the manufacturing share becomes as a predictor of eventual prosperity. If crossing some manufacturing share threshold is both necessary and sufficient for being rich (i.e., Pr(R|I) - Pr(R|~I)=1), the set of industrialized economies and rich economies would coincide, a situation that would correspond to the traditional usage of the term “industrialized nation.” We will examine these relationships separately for employmentand output-based definitions of industrialization. We emphasize that we use the terms “probability,” “necessary,” and “sufficient” strictly to describe historical data, and not as statements of what is theoretically possible. We will return to the implications of our results for economies’ future prospects in Section VI. Table 3 shows Pr(R|I) in the top panel and Pr(R|~I) in the bottom panel, calculated using employment shares (using data for 52 economies). The first column in each panel gives the percentage of economies that have reached the income per capita shown in each row (e.g., 57.7% of the economies achieved incomes over $6,000). The last row in the top and bottom panels, respectively, provide the percentage of economies that did and did not cross the threshold manufacturing share indicated in each column (e.g., 65.4% of the 52 economies reached a peak manufacturing employment share of 18%, and the other 34.6% did not). We have also marked in boldface the cells within each row corresponding to the employment threshold that maximizes Pr(R|I) - Pr(R|~I) for the income level in that row. For example, for a per capita income of $12,000 this difference is maximized at a threshold of 18% (Pr(R|I) - Pr(R|~I)=0.765-0=0.765, which is the largest difference for that income level. These figures indicate that having crossed an 18% employment threshold between 1970 and 2010 is necessary for achieving $12,000 per capita income today (Pr(R|~I)=0); and having achieved this share fairly strongly predicts being a rich economy today (Pr(R|I)=0.765<1). Table 3 indicates that a 16% threshold is optimal for separating the economies that have and have not crossed $6,000; 18% is optimal for $8,000–$22,000; and 20% is optimal for $24,000–$30,000. In all but one of these cases, achieving the employment share threshold is necessary for crossing the income cutoff; no economy that failed to cross 16% (18%; 20%) achieved a per capita income of $6,000 ($8,000–$22,000; $24,000–$30,000). Achieving these threshold shares is not sufficient for crossing any income cutoff. However, the top panel does reveal that 90.9%–95.5% of economies that attained 24% manufacturing employment shares did reach $6,000–$18,000 per capita income. The takeaway is, therefore, that peak manufacturing employment shares in excess of 18%–20% strongly predict that an economy is rich; while peak shares below this threshold are near perfect predictors that an economy is not rich (i.e., manufacturing employment is necessary for becoming rich). Achieving employment shares of roughly 18%–20% is, therefore, a fairly good definition of industrialization. Manufacturing Matters… but It’s the Jobs that Count | 11 Table 3: Probabilities of Being Rich, Conditional on Achieving Manufacturing Employment Share Thresholds A. Probability that an economy has crossed per capita GDP threshold, given that it crossed the manufacturing employment share threshold Probability that GDPPC > threshold Manufacturing Employment Share Threshold 10 12 14 16 18 20 22 24 26 28 30 Per capita GDP threshold 6K 0.577 0.588 0.638 0.682 0.789 0.824 0.862 0.852 0.955 0.938 0.889 0.875 8K 0.558 0.569 0.617 0.659 0.763 0.824 0.862 0.852 0.955 0.938 0.889 0.875 10K 0.500 0.510 0.553 0.591 0.684 0.765 0.828 0.815 0.909 0.875 0.889 0.875 12K 0.500 0.510 0.553 0.591 0.684 0.765 0.828 0.815 0.909 0.875 0.889 0.875 14K 0.481 0.490 0.532 0.568 0.658 0.735 0.793 0.815 0.909 0.875 0.889 0.875 16K 0.481 0.490 0.532 0.568 0.658 0.735 0.793 0.815 0.909 0.875 0.889 0.875 18K 0.481 0.490 0.532 0.568 0.658 0.735 0.793 0.815 0.909 0.875 0.889 0.875 20K 0.442 0.451 0.489 0.523 0.605 0.676 0.724 0.741 0.818 0.813 0.889 0.875 22K 0.423 0.431 0.468 0.500 0.579 0.647 0.724 0.741 0.818 0.813 0.889 0.875 24K 0.404 0.412 0.447 0.477 0.553 0.618 0.724 0.741 0.818 0.813 0.889 0.875 26K 0.404 0.412 0.447 0.477 0.553 0.618 0.724 0.741 0.818 0.813 0.889 0.875 28K 0.385 0.392 0.426 0.455 0.526 0.588 0.690 0.704 0.773 0.750 0.889 0.875 30K 0.365 0.373 0.404 0.432 0.500 0.559 0.655 0.667 0.727 0.688 0.778 0.750 Probability that manufacturing share > threshold 0.981 0.904 0.846 0.731 0.654 0.558 0.519 0.423 0.308 0.173 0.154 continued on next page 12 | ADB Economics Working Paper Series No. 420  Table 3 continued B. Probability that an economy has crossed per capita GDP threshold, given that it has not crossed the manufacturing employment share threshold Probability that GDPPC > threshold Manufacturing Employment Share Threshold 10 12 14 16 18 20 22 24 26 28 30 Per capita GDP threshold 6K 0.577 0.000 0.000 0.000 0.000 0.111 0.217 0.280 0.300 0.417 0.512 0.523 8K 0.558 0.000 0.000 0.000 0.000 0.056 0.174 0.240 0.267 0.389 0.488 0.500 10K 0.500 0.000 0.000 0.000 0.000 0.000 0.087 0.160 0.200 0.333 0.419 0.432 12K 0.500 0.000 0.000 0.000 0.000 0.000 0.087 0.160 0.200 0.333 0.419 0.432 14K 0.481 0.000 0.000 0.000 0.000 0.000 0.087 0.120 0.167 0.306 0.395 0.409 16K 0.481 0.000 0.000 0.000 0.000 0.000 0.087 0.120 0.167 0.306 0.395 0.409 18K 0.481 0.000 0.000 0.000 0.000 0.000 0.087 0.120 0.167 0.306 0.395 0.409 20K 0.442 0.000 0.000 0.000 0.000 0.000 0.087 0.120 0.167 0.278 0.349 0.364 22K 0.423 0.000 0.000 0.000 0.000 0.000 0.043 0.080 0.133 0.250 0.326 0.341 24K 0.404 0.000 0.000 0.000 0.000 0.000 0.000 0.040 0.100 0.222 0.302 0.318 26K 0.404 0.000 0.000 0.000 0.000 0.000 0.000 0.040 0.100 0.222 0.302 0.318 28K 0.385 0.000 0.000 0.000 0.000 0.000 0.000 0.040 0.100 0.222 0.279 0.295 30K 0.365 0.000 0.000 0.000 0.000 0.000 0.000 0.040 0.100 0.222 0.279 0.295 Probability that manufacturing share < threshold 0.019 0.096 0.154 0.269 0.346 0.442 0.481 0.577 0.692 0.827 0.846 GDP = gross domestic product, GDPPC = gross domestic product per capita. Source: Authors’ calculations. Manufacturing Matters… but It’s the Jobs that Count | 13 Table 4 provides the same information for manufacturing output shares (data for 135 economies). Results indicate that a 22% manufacturing output share achieves maximal separation for all income levels (except for $28,000, where 20% fares slightly better). However, the separation is very poor. The largest observed value for Pr(R|I) - Pr(R|~I) is only 0.296, occurring at an output share threshold of 22% and an income cutoff of $6,000. In contrast, for employment shares, the largest observed value is 0.789, at an employment threshold of 16% and the same $6,000 income cutoff. Unsurprisingly, then, the table also shows that industrialization in output is not sufficient to achieve a high income per capita, as no output share cutoff guarantees that an economy crosses any income threshold; and that it is not necessary either, that is, failure to cross a given output share threshold does not ensure failure to become rich. When we restrict the output sample to the 52 economies for which we have employment data (table available on request), the separation improves, but it is still not as good as it is for employment shares. The values of Pr(R|I) - Pr(R|~I) now range from 0.41 to 0.59; the result that industrialization in output is not sufficient to become rich remains; but now we find that industrialization in output is necessary to cross one income threshold: no economy with a peak manufacturing share of 18% or lower crossed $22,000 per capita income. Compared with the results in Tables 3 and 4, this suggests that output shares are not particularly good predictors of which small economies (many of which lack employment data) will be rich; that output shares provide some signal in the economies in our employment sample; and that employment shares are much better predictors of eventual prosperity than output shares. How do the findings in this analysis classify economies? To see this, we return to the $12,000 cutoff for being rich and separately pick employment and output share thresholds that provide maximal separation between rich and poor economies in our 52-economy sample, that is, 18% for employment, and 22% for output. Table 5 categorizes economies by their industrialization status in employment and output using these thresholds, and highlights the rich economies in bold. Although the threshold for output shares is higher than the threshold for employment shares, these thresholds classify exactly the same number of economies (34) as industrialized in output and employment, respectively. This indicates that the horse-race we are about to run between employment and output shares is fair. We find that all high-income economies have industrialized in employment, but six highincome economies (Canada; Denmark; Greece; Hong Kong, China; Norway; and Portugal) did not industrialize in output. Indeed, if we restrict attention to those that have industrialized in employment, those that did not industrialize in output are more likely to be rich than those that did (6/7 > 20/27). Clearly, if we were to select one target, it would be employment shares, not output shares. Summing up, this analysis indicates that achieving some critical output share has generally been neither necessary nor sufficient for achieving high-income status. On the other hand, achieving a manufacturing employment share of 18%–20% has been almost sufficient and absolutely necessary (in the statistical sense) for achieving high-income status. 14 | ADB Economics Working Paper Series No. 420 Table 4: Probabilities of Being Rich, Conditional on Achieving Manufacturing Output Share Thresholds (135 economies) A. Probability that an economy has crossed per capita GDP threshold, given that it crossed the manufacturing output share threshold  Probability that GDPPC > threshold Manufacturing Output Share Threshold 10 12 14 16 18 20 22 24 26 28 30 Per capita GDP threshold 6K 0.333 0.342 0.362 0.371 0.398 0.436 0.484 0.520 0.500 0.500 0.522 0.438 8K 0.304 0.316 0.333 0.340 0.375 0.410 0.452 0.480 0.450 0.433 0.435 0.438 10K 0.274 0.291 0.305 0.309 0.341 0.372 0.419 0.460 0.425 0.400 0.391 0.375 12K 0.267 0.282 0.295 0.299 0.330 0.359 0.403 0.440 0.400 0.367 0.391 0.375 14K 0.252 0.265 0.286 0.289 0.318 0.346 0.387 0.420 0.400 0.367 0.391 0.375 16K 0.237 0.256 0.286 0.289 0.318 0.346 0.387 0.420 0.400 0.367 0.391 0.375 18K 0.237 0.256 0.286 0.289 0.318 0.346 0.387 0.420 0.400 0.367 0.391 0.375 20K 0.222 0.239 0.267 0.268 0.295 0.321 0.371 0.400 0.375 0.333 0.348 0.375 22K 0.215 0.231 0.257 0.258 0.284 0.308 0.355 0.380 0.350 0.300 0.304 0.313 24K 0.207 0.222 0.248 0.247 0.273 0.308 0.355 0.380 0.350 0.300 0.304 0.313 26K 0.207 0.222 0.248 0.247 0.273 0.308 0.355 0.380 0.350 0.300 0.304 0.313 28K 0.193 0.205 0.229 0.227 0.250 0.282 0.323 0.340 0.300 0.267 0.304 0.313 30K 0.185 0.197 0.219 0.216 0.239 0.269 0.306 0.340 0.300 0.267 0.304 0.313 Probability that manufacturing share > threshold 0.867 0.778 0.719 0.652 0.578 0.459 0.370 0.296 0.222 0.170 0.119 continued on next page Manufacturing Matters… but It’s the Jobs that Count | 15  Table 4 continued B. Probability that an economy has crossed per capita GDP threshold, given that it has not crossed the manufacturing output share threshold Probability that GDPPC > threshold Manufacturing Output Share Threshold 10 12 14 16 18 20 22 24 26 28 30 Per capita GDP threshold 6K 0.333 0.278 0.233 0.237 0.213 0.193 0.205 0.224 0.263 0.286 0.295 0.319 8K 0.304 0.222 0.200 0.211 0.170 0.158 0.178 0.200 0.242 0.267 0.277 0.286 10K 0.274 0.167 0.167 0.184 0.149 0.140 0.151 0.165 0.211 0.238 0.250 0.261 12K 0.267 0.167 0.167 0.184 0.149 0.140 0.151 0.165 0.211 0.238 0.241 0.252 14K 0.252 0.167 0.133 0.158 0.128 0.123 0.137 0.153 0.189 0.219 0.223 0.235 16K 0.237 0.111 0.067 0.105 0.085 0.088 0.110 0.129 0.168 0.200 0.205 0.218 18K 0.237 0.111 0.067 0.105 0.085 0.088 0.110 0.129 0.168 0.200 0.205 0.218 20K 0.222 0.111 0.067 0.105 0.085 0.088 0.096 0.118 0.158 0.190 0.196 0.202 22K 0.215 0.111 0.067 0.105 0.085 0.088 0.096 0.118 0.158 0.190 0.196 0.202 24K 0.207 0.111 0.067 0.105 0.085 0.070 0.082 0.106 0.147 0.181 0.188 0.193 26K 0.207 0.111 0.067 0.105 0.085 0.070 0.082 0.106 0.147 0.181 0.188 0.193 28K 0.193 0.111 0.067 0.105 0.085 0.070 0.082 0.106 0.147 0.171 0.170 0.176 30K 0.185 0.111 0.067 0.105 0.085 0.070 0.082 0.094 0.137 0.162 0.161 0.168 Probability that manufacturing share < threshold 0.133 0.222 0.281 0.348 0.422 0.541 0.630 0.704 0.778 0.830 0.881 GDP = gross domestic product, GDPPC = gross domestic product per capita. Source: Authors’ calculations. 16 | ADB Economics Working Paper Series No. 420 Table 5: Economies Categorized by Industrialization in Output and Employment Employment Share Relative to 18% Not Industrialized Industrialized (0/18 economies are rich) (26/34 economies are rich) Output Share Relative to 22% Not Industrialized (6/18 economies are rich) Bangladesh, Bolivia, Botswana, Chile, Colombia, Honduras, India, Pakistan, Panama, Peru, Syria (0/11 economies are rich) Canada, Denmark, Greece, Hong Kong, China, Mexico, Norway, Portugal (6/7 economies are rich) Industrialized (20/34 economies are rich) Brazil, People’s Republic of China, Indonesia, Philippines, Suriname, Thailand, Venezuela (0/7 economies are rich) Argentina, Australia, Austria, Belgium, Costa Rica, El Salvador, Finland, France, Guatemala, Ireland, Italy, Japan, Republic of Korea, Luxembourg, Malaysia, Netherlands, Poland, Puerto Rico, Romania, San Marino, Singapore, Spain, Sweden, Switzerland, Trinidad and Tobago, United Kingdom, United States (20/27 economies are rich) GDPPC = gross domestic product per capita. Note: In bold if GDPPC during 2005–2010 exceeds $12,000. Source: Authors. Manufacturing Matters… but It’s the Jobs that Count | 17 V. HAS IT BECOME MORE DIFFICULT TO ACHIEVE HIGH MANUFACTURING SHARES? This section turns to panel data to deepen the analysis in Figures 2 and 3, and panels B and C of Table 2. Specifically, the section asks whether it has become more difficult to achieve high manufacturing shares over time. Figures 4 and 5 show scatterplots of logged manufacturing employment and output shares against logged per capita GDP, with quadratic best fit lines produced separately by decade. Each graph involves multiple observations from each economy. As can be seen, the employment and output shares expected at each income level have generally fallen. Also, as expected, the fall in predicted employment shares is deeper than the fall in predicted output shares. Figure 4: Manufacturing Employment Share versus Per Capita GDP, Variables in Natural Logarithms Source: Authors’ calculations. 1 2 3 4 4 6 8 10 12 ln(per capita GDP, constant 2005 $) 1970–1980 data 1981–1990 data 1991–2000 data 2001–2010 data 1970–1980 fitted 1981–1990 fitted 1991–2000 fitted 2001–2010 fitted ln(Manufacturing employment share) 18 | ADB Economics Working Paper Series No. 420  Figure 5: Manufacturing Output Share versus Per Capita GDP, Variables in Natural Logarithms Source: Authors’ calculations. These two graphs also allow us to examine the hypothesis that manufacturing shares follow an inverted U-shape vis-à-vis income per capita. The inverted U-shape is an old idea that goes back to the pioneering work of Chenery (1960) and Kuznets (1966), among others. While Figure 5 confirms this result, Figure 4 is more interesting. Pooling economies and years within decades, we see no general trend for manufacturing employment shares to decline with income, during the 1970s, 1980s or 1990s, even at high incomes. However, the concavity of this profile increases with each decade, to the point that there is actually a peak expected employment share by the 2000s at a per capita income of around $13,000. The regression lines in Figures 4 and 5 are partly identified of differences between economies. They are not representative of economies’ trajectories over time. For example, the apparently constant upward trajectory of manufacturing employment shares with income in the first three decades of our sample might simply reflect the fact that the original OECD economies were richer and more industrialized than the developing economies during this period. In this case, the results offer little insight into whether developing economies would have been able to continually increase their manufacturing shares as they became richer. Therefore, they would not provide the right comparisons for analyzing the structural possibilities that economies have faced, or for testing the three hypotheses generated by the discussion in the introduction to the paper. That task is better accomplished by comparisons within economies over time. This requires regressions with country fixed effects. We therefore estimate logarithmic regressions of output (Y) and employment (L) manufacturing shares (, ) on income per capita (LGDPPC) and income per capita squared as well as a time trend (T), and the interaction between the time trend and income per capita. This specification embeds the three hypotheses we motivated in the introduction. Specifically, the time trend allows us to test whether manufacturing shares, conditional on income levels, have declined over time (H1); and, where the inverted U-shape is confirmed, whether expected peak shares have fallen. 1 2 3 4 4 6 8 10 12 ln(per capita GDP, constant 2005 $) 1970–1980 data 1981–1990 data 1991–2000 data 2001–2010 data 1970–1980 fitted 1981–1990 fitted 1991–2000 fitted 2001–2010 fitted ln(Manufacturing output share) Manufacturing Matters… but It’s the Jobs that Count | 25 We find that, compared with the employment shares, the log output shares expected at any income level have drifted down much more slowly. The resulting decline of employment shares relative to output shares is exactly what we would expect to see if rates of labor-saving technological change have been higher in manufacturing than in the rest of the economy. With respect to H2, we likewise find that the per capita income at which output shares peak has not decreased significantly. This, in combination with the fact that the income level at which manufacturing employment peaks has fallen, is consistent with the possibility that economies substituted capital-intensive for laborintensive manufacturing activities as incomes rose. Together, these results confirm that sustaining manufacturing output is becoming somewhat more difficult, and that rising incomes do not weigh heavily on the output levels that economies can sustain. The real problem economies face is with sustaining manufacturing employment. Thus, they confirm H3. VI. INTERPRETATION AND CONCLUSIONS We have explored thoroughly the relationships between economic prosperity and manufacturing output and employment shares. We have shown that all economies that are rich today have, at some point in the last 40 years, enjoyed high manufacturing employment shares; while only a few economies that attained high manufacturing employment shares are not rich economies. Manufacturing employment has therefore been, as a matter of historical record, necessary but not sufficient for eventual prosperity. This is quite consistent with two points demonstrated by Rodrik (2013a). First, that economies that create many manufacturing jobs grow faster because manufacturing has an “escalator” quality—labor productivity in manufacturing industries rises rapidly towards the global frontier. Second, this is insufficient to ensure that poorer economies will grow faster than richer economies because manufacturing constitutes a relatively small share of total employment. Our second contribution is to show that manufacturing employment shares have fallen, and now go into decline at lower levels of per capita income than they once did. Therefore, the fraction of national workforces for which manufacturing serves as an escalator has declined, even in low-income economies. Once again, manufacturing output shares do not display these trends as strongly. Taken literally, these results would appear to highlight the maintenance of manufacturing employment as one of the most important and difficult challenges that developing economies face today. Of course, these results cannot be taken literally without a clear understanding of the causal mechanisms that drive them. We have shown that the data are consistent with explanations involving two forces: rising own-price elasticities of manufacturing labor demand, and labor productivity that rises faster in manufacturing than in non-manufacturing activities. However, two other alternative explanations need to be considered. One is that, the result is purely mechanical. Suppose, in contrast to our story, that global labor productivity in manufacturing and in other sectors grew at the same pace, and that the growth rates of manufacturing outputs and of other products were identical. It follows that global manufacturing employment shares would have remained constant.5 In this case, recent small increases in national manufacturing employment shares in some populous developing economies (e.g., the PRC, India)  5 Aggregating across developing economies, Haraguchi (2014) finds that: (i) manufacturing’s share in total output (added up across economies) has not changed since the 1970s, hovering around 20%–23%; and (ii) the aggregate manufacturing employment share increased since 1970. 26 | ADB Economics Working Paper Series No. 420  would have to be accompanied by large declines in the corresponding shares in economies with smaller populations. While we are confident that the large populations of recent industrializers are part of the story, the relatively small movements in manufacturing output shares are not consistent with this purely mechanical explanation. If our results for employment shares were entirely driven by the Southward migration of manufacturing activity, output shares should have moved along with employment shares. Another possibility is that, we are simply capturing the increased outsourcing of manufacturing-related services activity to dedicated service companies.6 Again, it seems likely that this explains part of the declines in measured manufacturing shares, but not all of them. After all, this classification problem should afflict manufacturing output shares as well, and yet, these have not shifted downwards very fast, or leftwards at all. Finally, it is worth noting that if we add UNIDO’s (2013) estimate of outsourced or manufacturing-related jobs to our figures, manufacturing employment shares would increase by around 20%. If we apply this increase to developing and developed economies alike, many low and middle-income economies would still fail to reach the 18%– 20% manufacturing employment share threshold (e.g., Bangladesh, Indonesia, the Philippines, and most economies in the Middle East and North Africa, Central Asia, and Central America). Two final comments, offered in closing. First, why is it that manufacturing employment shares have much clearer relationships with concurrent and subsequent income levels than output shares? One possibility is that output shares are influenced by the price of manufactured relative to nonmanufactured outputs, which vary with incomes and across economies. We have already argued that these relative price differences may explain why output shares peak at lower incomes than employment shares. Variations in this relative price may also confound efforts to identify relationships with output shares. Second, with the scope for manufacturing growth limited by the structural forces we have identified as well as by the increasing awareness of the high carbon footprint of many industries, and the possibility of restrictions on carbon emissions to avoid the negative effects of climate change (Gutowski 2007, Stern 2007), we need to consider whether economies can get rich by shifting to services without achieving high manufacturing employment shares. While it is impossible to rule out this possibility (given the growing array of new services and service-delivery modes, some of which appear to have rather high economies of scale (e.g., Maroto-Sánchez and Cuadrado-Roura 2009), Section IV demonstrates there are not yet any examples of economies that have done so successfully.  6 UNIDO (2013) reckons that manufacturing employment is underestimated because informal manufacturing jobs and jobs in manufacturing-related services are not properly counted. UNIDO has estimated the latter worldwide for 1970– 2009. In 2009, the number of these jobs was 95 million, or almost half of the direct formal jobs globally in manufacturing, and that 32 million of these jobs were in developed economies. We have spoken with UNIDO staff about how these jobs were estimated and their reliability. UNIDO has advised that their estimates are a first approximation that can certainly be used, but with great caution.  APPENDIX A: EMPLOYMENT DATA The cleaning of the ILO’s LABORSTA data proceeded as follows. We began with the full LABORSTA database. In any given economy and year, these data can include estimates from more than one source, and the sources may use different sectoral classifications. From this, we kept observations collected according to the International Standard Industrial Classification, versions 2, 3, or 4, and dropped the others. We then dropped sources that exclude major sections of the workforce (e.g., rural residents, agricultural workers). In those instances where employment levels in some sectors were missing, but could be inferred from total employment and employment in other sectors, we filled in the blanks and checked to see whether this yielded discontinuities in the series. Where discontinuities were observed, the series corresponding to that economy and source were dropped. After these adjustments, some economies still had multiple sources in some years. For these economy-year pairs with overlapping series, we opted to use the longest continuous series. When the series had the same length, we chose the one that used ISIC revision 2. The final series were checked graphically for anomalies. Table A.1: List of Economies by Source LABORSTA only Bangladesh, Botswana, El Salvador, Guatemala, Honduras, Pakistan, Panama, Poland, Portugal, Puerto Rico, Romania, San Marino, Suriname, Switzerland, Syria, Trinidad and Tobago, United Kingdom OECD only Belgium, Denmark, France, Ireland, GGDC only Argentina, Bolivia, India, Peru LABORSTA + OECD growth rate Australia, Austria, Canada, Finland, Greece, Italy, Luxembourg, Netherlands, Norway, Spain, Sweden, United States LABORSTA + GGDC growth rate Chile; Costa Rica; Hong Kong, China; Japan; Republic of Korea; Malaysia GGDC + LABORSTA growth rate Brazil, Colombia, Indonesia, Mexico, Philippines, Singapore, Thailand, Venezuela National Census People’s Republic of China Source: Authors.  28 | Appendix A: Employment Data  Table A.2: Coverage of Manufacturing Employment Shares Economy # of Obs. Earliest Obs. Latest Obs. Economy # of Obs. Earliest Obs. Latest Obs. Argentina 36 1970 2005 Korea, Republic of 39 1970 2008 Australia 38 1971 2008 Luxembourg 40 1970 2009 Austria 34 1976 2009 Malaysia 34 1975 2008 Bangladesh 9 1984 2005 Mexico 39 1970 2008 Belgium 40 1970 2009 Netherlands 40 1970 2009 Bolivia 38 1970 2007 Norway 40 1970 2009 Botswana 7 1985 2006 Pakistan 36 1973 2008 Brazil 38 1970 2007 Panama 35 1970 2008 Canada 39 1970 2008 Peru 36 1970 2005 Chile 39 1970 2008 Philippines 38 1971 2008 China, People’s Republic of 6 1982 2010 Poland 28 1981 2008 Colombia 39 1970 2008 Portugal 35 1974 2008 Costa Rica 39 1970 2008 Puerto Rico 39 1970 2008 Denmark 40 1970 2009 Romania 37 1970 2008 El Salvador 20 1975 2007 San Marino 29 1978 2008 Finland 40 1970 2009 Singapore 39 1970 2008 France 39 1970 2008 Spain 40 1970 2009 Greece 29 1981 2009 Suriname 25 1973 2004 Guatemala 7 1981 2006 Sweden 40 1970 2009 Honduras 29 1970 2007 Switzerland 32 1970 2008 Hong Kong, China 32 1974 2005 Syrian Arab Republic 15 1970 2007 India 35 1971 2005 Thailand 39 1970 2008 Indonesia 39 1970 2008 Trinidad and Tobago 26 1970 2008 Ireland 40 1970 2009 United Kingdom 39 1970 2008 Italy 40 1970 2009 United States 40 1970 2009 Japan 39 1970 2008 Venezuela, Bolivarian Republic of 36 1970 2005 Source: Authors. Appendix A: Employment Data | 29  Table A.3: List of Codes for Each Economy Economy Code Economy Code Afghanistan AFG Finland FIN Albania ALB France FRA Algeria DZA Gabon GAB Andorra AND Gambia, The GMB Angola AGO Ghana GHA Argentina ARG Greece GRC Australia AUS Guatemala GTM Austria AUT Guinea GIN Bangladesh BAN Guinea-Bissau GNB Belgium BEL Guyana GUY Belize BLZ Haiti HTI Benin BEN Honduras HND Bhutan BHU Hong Kong, China HKG Bolivia BOL Hungary HUN Botswana BWA India IND Brazil BRA Indonesia INO Bulgaria BGR Iran, Islamic Republic of IRN Burkina Faso BFA Iraq IRQ Burundi BDI Ireland IRE Cambodia CAM Italy ITA Cameroon CMR Jamaica JAM Canada CAN Japan JPN Central African Republic CAF Jordan JOR Chad TCD Kenya KEN Chile CHL Korea, Democratic People's Republic of PRK China, People's Republic of PRC Korea, Republic of KOR Colombia COL Lao People's Democratic Republic LAO Congo, Democratic Republic of the COD Lebanon LBN Congo, Republic of the COG Lesotho LSO Costa Rica CRI Liberia LBR Cote d'Ivoire CIV Libya LBY Cuba CUB Liechtenstein LIE Denmark DEN Luxembourg LUX Djibouti DJI Macau, China MAC Dominican Republic DOM Madagascar MDG Ecuador ECU Malawi MWI Egypt, Arab Republic of EGY Malaysia MAL El Salvador SLV Mali MLI Equatorial Guinea GNQ Mauritania MRT continued on next page 30 | Appendix A: Employment Data  Table A.3 continued Economy Code Economy Code Mauritius MUS Saudi Arabia SAU Mexico MEX Senegal SEN Monaco MON Sierra Leone SLE Mongolia MON Singapore SIN Morocco MAR Somalia SOM Mozambique MOZ South Africa ZAF Myanmar MYA Spain SPA Namibia NAM Sri Lanka SRI Nepal NEP Suriname SUR Netherlands NET Swaziland SWZ New Zealand NZL Sweden SWE Nicaragua NIC Switzerland SWI Niger NER Syrian Arab Republic SYR Nigeria NGA Tanzania TZA Norway NOR Thailand THA Oman OMN Togo TGO Pakistan PAK Trinidad and Tobago TTO Panama PAN Tunisia TUN Papua New Guinea PNG Turkey TUR Paraguay PRY Uganda UGA Peru PER United Arab Emirates ARE Philippines PHI United Kingdom UKG Poland POL United States USA Portugal POR Uruguay URY Puerto Rico PTR Venezuela, Bolivarian Republic of VEN Qatar QAT Viet Nam VIE Romania ROU Zambia ZMB Rwanda RWA Zimbabwe ZWE San Marino SMR Sources: World Bank and Asian Development Bank. Table A.4: Economies with Employment Share Data Are Different With Employment Data (52 economies) With or Without Employment Data (135 economies) Mean per capita GDP (2005–2010) $21,200 $7,466 Mean per capita GDP at time of peak manufacturing output share $11,607 $5,277 Mean population over the sample period. 76.8 million 11.6 million Mean manufacturing output share in year of peak 25.5% 16.2% Median year of manufacturing output share peak 1977 1988 Source: Authors’ calculations.  APPENDIX B: DIFFERENCES BETWEEN SMALL ISLAND NATIONS AND OTHER NATIONS We began by examining the distribution of output shares with respect to log per capita income. The data in Figure B1 are raw. They have not been corrected for exogenous determinants. While an inverted U-shape is observed for non-small island nations in both time periods, no evidence of an inverted U-shape appears for small island nations in the latter 20 years of our sample. Small island nations are also clearly different, in that they have lower manufacturing output shares than other economies. Figure B1: All Economies, Split by Small Island, No Corrections 1970–1990 1991-2010 Notes: Green dots are island nations. Blue dots are non-island nations. Source: Authors’ calculations. Next, to examine these structural issues further, we regressed the output shares on the year, year-squared, and several structural features. These are: natural resource exports as a share of total exports in 1990, log population, log per capita land endowment, the share of land that is usable for agriculture, and the ratios of retirees and young people to the working-age population. Results are available on request. The key differences between small island nations and other nations are: (i) small island nations have a lower intercept (smaller manufacturing shares, other things equal), suggesting that manufacturing activity is hard to sustain in more remote economies; and (ii) while natural resource intensity reduces manufacturing output in other nations, it increases it in islands, again suggesting that island nations without natural resources of their own find it difficult to sustain manufacturing. Finally, figure B2 graphs the residuals from these regressions against log per capita GDP. With these corrections for underlying structural features firmly in place, we now see much clearer evidence of an inverted U with respect to per capita income for island nations. All these results suggest that per capita GDP has been a relatively less important determinant of manufacturing output shares in island nations, but that structural factors have been relatively important. To avoid complications, we therefore do not include them in our analysis. -2 0 2 4 4 6 8 10 12 lgdppc -2 0 2 4 4 6 8 10 12 lgdppc 32 | Appendix B: Differences Between Small Island Nations and Other Nations  Figure B2: All Economies, Split by Small Island, Structural Corrections Notes: Green dots are island nations. Blue dots are non-island nations. Source: Authors’ calculations.  –3 –2 –1 0 1 2 4 6 8 10 12 lgdppc 1970–1990 -4 -2 0 2 4 6 8 10 12 lgdppc 1991–2010  REFERENCES Amable, B. 2000. International Specialisation and Growth. Structural Change and Economic Dynamics. 11. pp. 413–431. Asian Development Bank. 2013. Special Chapter on Asia’s Economic Transformation: Where to; How: and How Fast? in the Key Indicators for Asia and the Pacific 2013. Manila. Bah, E. M. 2011. Structural Transformation Paths Across Countries. Emerging Markets Finance & Trade. Vol. 47, Supplement 2 (June). pp. 5–19. Barro, R. J. and J.-W. Lee. 2010. A New Data Set of Educational Attainment in the World, 1950-2010. National Bureau of Economic Research Working Paper. No. 15902. Baumol, W. J. 2012. The Cost Disease: Why Computers Get Cheaper and Health Care Doesn't. Yale University Press. Berman, E., J. Bound, and Z. Griliches. 1994. Changes in the demand for skilled labor within UnitedStates Manufacturing: Evidence from the Annual Survey of Manufactures. Quarterly Journal of Economics. 109. pp. 367–397. Berman, E., J. Bound, and S. Machin. 1998. Implications of skill-biased technological change: International evidence. Quarterly Journal of Economics. 113. pp. 1245–1279. Brynjolfsson, E. and A. McAffee. 2013. The Second Machine Age. London and New York: W.W. Norton & Co. Chenery, H. 1960. Patterns of Industrial Growth. The American Economic Review. 50 (4). pp. 624–654. Chenery, H., S. Robinson, and M. Syrquin. 1986. Structural Transformation. In H. Chenery, S. Robinson, and M. Syrquin, eds. Industrialization and Growth: A Comparative Study. Washington: Oxford University Press. Cowen, T. 2014. Average is Over. New York: Dutton. Dabla-Norris, E., A. Thomas, R. Garcia-Verdu, and Y. Chen. 2013. Benchmarking Structural Transformation Across the World. IMF Working Paper. No. 13/176. Fagerberg, J. 2000. Technological Progress, Structural Change and Productivity Growth: A Comparative Study. Structural Change and Economic Dynamics. 11. pp. 393–411. Felipe, J., ed. Forthcoming 2015. Development and Modern Industrial Policy in Practice. Issues and Country Experiences. Edward Elgar. Goldin, C. D. and L. F. Katz. 2009. The Race Between Education and Technology. Harvard University Press.   ADB recognizes China as the People’s Republic of China. 34 | References  Gutowski, T. G. 2007. The Carbon and Energy Intensity of Manufacturing. 40th Seminar of CIRP, Keynote Address, Liverpool University, Liverpool, UK. Haraguchi, N. 2014. Patterns of Structural Change and Manufacturing Development. Mimeo. United Nations Development Organization. October draft. Hasan, R., D. Mitra, and K. V. Ramaswamy. 2007. Trade Reforms, Labor Regulations, and LaborDemand Elasticities: Empirical Evidence from India. Review of Economics and Statistics. 89. pp. 466–481. Hausmann, R., J. Hwang, and D. Rodrik. 2007. What you export matters. Journal of Economic Growth. 12. pp. 1–25. Helper, S., T. Krueger, and H. Wial. 2012. Why Does Manufacturing Matter? A Policy Framework. Washington, DC: Brookings Policy Framework Policy Report. Journal of Post Keynesian Economics. 1983. Symposium on Kaldor’s Laws. Spring. 5 (3). Kaldor, N. 1966. Causes of the Slow Rate of Growth of the United Kingdom. An Inaugural Lecture. Cambridge: Cambridge University Press. Krugman, P. R. 1991. Geography and Trade. Cambridge, MA: MIT Press. Kuznets, S. 1966. Modern Economic Growth: Rate, Structure and Spread. New Haven and London: Yale University Press. Li, C., and J. Gibson. 2013. Rising Regional Inequality in China: Fact or Artifact? World Development. 47. pp. 16–29. Maroto-Sánchez, A. and J. Cuadrado-Roura. 2009. Is Growth of Services an Obstacle to Productivity Growth? A Comparative Analysis. Structural Change and Economic Dynamics. 20. pp. 254–265. Nickell, S., S. Redding, and J. Swaffield. 2008. The Uneven Pace of Deindustrialization in the OECD. The World Economy. 31 (9). pp. 1154–1184. Peneder, M. 2003. Industrial Structure and Aggregate Growth. Structural Change and Economic Dynamics. 14. pp. 427–448. Rodrik, D. 1997. Has Globalization Gone Too Far? Washington, DC: Institute for International Economics. ———. 2009. Growth After the Crisis. Mimeo. (May 12 version). Cambridge, MA: Harvard Kennedy School. ———. 2013a. Unconditional Convergence in Manufacturing. Quarterly Journal of Economics. 128. pp. 165–204. ———. 2013b. The Past, Present, and Future of Economic Growth. Global Citizen Foundation Working Paper 1. June.