www.merit.unu.edu

number 3, 2020

Overview Many middle-income developing Industrialisation and countries are experiencing stalled industrialisation or ‘premature de-industrialisation’. Premature de- De-Industrialisation in industrialisation refers to an empirical observation that some the Developing World developing countries have reached ‘peak manufacturing’ in employment and GDP shares at lower levels than ince the 1990s, many middle-income developing countries have experi- earlier industrialisers. In other words, Senced processes of stalled industrialisation or ‘premature de-industriali- the opportunities manufacturing sation’ (a term first coined by UNCTAD) accompanied by a tertiarisation of presents for economic development employment and /or value-added. Put simply, this means that the social and and employment creation are being economic benefits that manufacturing presents are being exhausted earlier than exhausted earlier and with fewer of has historically been the case. These contemporary patterns of economic devel- the benefits that have typically accrued opment are now well documented in a general sense in many middle-income to countries through manufacturing. developing countries. Such trends are likely to exacerbate global income This brief discusses the phenomenon inequality between countries and within countries by weakening economic and the policy implications for growth (and impacting on inequality between countries) as well as weakening developing countries looking ahead. low and medium skill employment generation (and thus impacting on within- country inequality) in developing countries. Written by Andy Sumner, King’s College London (1). Edited by Howard The ‘New Normal’ Path of Economic Development Hudson, UNU-MERIT It is now well established how the structure of an economy can change © United Nations University 2020 through economic development (e.g. Chenery, 1960; Chenery and Syrquin, ISBN 978-92-808-5014-7 1975; Kuznets, 1965; Syrquin, 1988). Specifically, agriculture dominates employment and value-added at low levels of GDP per capita. The share of agri- Licensed under the Creative Commons culture diminishes as GDP per capita rises (to very low shares at high income). Deed ‘Attribution-NonCommercial- In contrast, manufacturing follows an inverted-U pattern as manufacturing NoDerivs 2.5’ shares are small at low GDP per capita levels and rise over time to a peak before falling again. Finally, the service sector share is low at low levels of GDP per The views expressed in this publication capita and rises over time to very high levels at high GDP per capita. Duarte and are those of the authors and do not Restuccia (2010, p. 135) empirically concur with this historic research and note: necessarily reflect the views of United Nations University. ...all countries follow a common process of structural transformation. First, all coun- tries exhibit declining shares of hours in agriculture, even the most advanced countries in this process, such as the United Kingdom and the United States. Second, countries at an early stage of the process of structural transformation exhibit a hump-shaped share of hours in industry, whereas this share is decreasing for countries at a more advanced stage. Finally, all countries exhibit an increasing share of hours in services.

www.unu.edu

This ‘traditional’ pattern of structural Studies of Industrialisation and transformation is now in question. In De-industrialisation

fact, it is now likely to be harder than Many seminal studies exist has historically been the case to which detail ‘good’ and ‘bad’ de- achieve the ‘traditional’ pathway of industrialisation in OECD coun- structural transformation – industri- tries. The former, ‘good’ alisation. Furthermore, new pathways de-industrialisation, generally entails – tertiarisation – may become a ‘new new employment to absorb laid-off normal’ at lower levels of per capita manufacturing workers in tertiary income than earlier developers expe- sectors (see Alderson, 1999; Bacon rienced, which has implications for and Eltis, 1976; Bazen and employment growth, real wage Thirlwall, 1986, 1989, 1992; growth, and value-added. Blackaby, 1978; Bluestone and

Harrison, 1982; Cairncross, 1979; Kim and Sumner (2019) Fontagné and Harrison, 2017; De present a typology of varieties of Groot, 2000; Kucera and Milberg, structural transformation. They 2003; Linkon, 2018; Rowthorn and identify patterns of industrialisation Ramaswamy, 1997; Saeger, 1997; – primary and upgrading – and pat- Thirlwall, 1982; Wren, 2013). There About the Author terns of ‘new’ economic development, is also a set of influential papers namely, stalled industrialisation and Andy Sumner is a Professor of which discuss the inverted-U curve International Development at de-industrialisation. Figure 1 shows of manufacturing shares of employ- King’s College London. He is also this typology. The typology is based ment in OECD countries as the Director of the UK Economic and on the changes in manufacturing Social Research Council (ESRC) defining pattern of de-industrialisa- value-added and employment shares Global Challenges Strategic tion (e.g. Rowthorn and Wells, 1987; in the post-1990 period. The cate- Research Network on Global Rowthorn and Coutts, 2004; gorisation has been constructed Poverty and Inequality Dynamics. Rowthorn and Ramaswamy, 1997; He has 20 years’ international based on the recent direction of Singh, 1977, 1987). research experience using both changes in the manufacturing shares qualitative and quantitative and not on the absolute levels of methods and has published Caution is needed when those shares. It is important to note extensively, including ten books using these studies as a template for that not all developing countries and fifty journal papers and middle-income developing countries would fit neatly into this typology. book chapters. His most recent because the effects of de-industriali- books are ‘Global Poverty’ (2016, More precisely, these varieties are not sation may be different in the con- Oxford University Press) and suited for very poor and aid-depen- temporary developing world. This is ‘Development and Distribution’ dent developing countries that (2018, Oxford University Press). because the process is happening at remain largely agrarian or for those He was appointed at King’s lower levels of both per capita which are heavily dependent on the in 2012 and was a founding income and ‘peak manufacturing’ (in mining sector. A separate analysis is Co-Director of the King’s terms of value-added and employ- International Development needed for these countries. It is also ment shares) than was historically Institute, which later became important to note that countries are the case. Furthermore, contemporary the Department of International not fixed within the variety of struc- Development. He is also a de-industrialisation occurs at a dif- tural transformation to which this member of the ESRC Peer Review ferent point in history wherein analysis has assigned them. Shifting College. Correspondence to: global production is now highly frag- between different varieties is possible [email protected] mented into global value chains and, in some cases, desirable. (GVCs). Moreover, in middle-

income developing countries, the

2 Policy Brief www.merit.unu.edu

Figure 1. Varieties of Industrialisation and De-industrialisation. Source: Elaborated from Kim and Sumner (2019), by Arief Anshory Yusuf

welfare regime mechanisms mitigat- ences, and well-established firms ing the impact of de-industrialisa- that lead GVCs. All of these yield tion – ranging from unemployment positive reputational effects, e.g. benefits and pensions to access to ‘made in France’ for leather products health and retraining programmes – or ‘made in Germany’ for automotive are weaker and cover a lower share products. In upstream, science- of the population than in advanced intensive activities (e.g. new materi- economies. It is also unlikely that als, pharmacological research, and middle-income developing countries aerospace), middle-income develop- can respond to de-industrialisation ing countries compete with advanced by mimicking advanced economies; economies who are worried about that is, focusing on downstream, losing their dominance and with higher-value service activities (e.g. countries that challenge Western marketing and the development of dominance for industrial and geopo- creative content) or upstream activi- litical reasons, such as China. All of ties linked to science and technology. the latter heavily subsidise these In downstream activities, advanced activities in order to either keep or economies benefit from having con- capture the higher value-added sumers with higher incomes, a freer GVC access points. civil society, world class universities in the creative industries and sci-

Industrialisation and De-industrialisation in the Developing World 3 www.unu.edu

In short, although much can country studies (e.g. for Malaysia, be drawn from the research on devel- see Henderson and Phillips 2007; oped countries, caution is needed Rasiah 2011; Tan 2014; for because the effects of de-industriali- Indonesia, see Manning and sation are likely to be different in Purnagunawan 2017; for Mexico, see developing countries. Furthermore, Cruz 2014; for Chile, see Gwynne the policy implications that are pro- 1986; for Pakistan, see Hamid and moted (e.g. raising education levels Khan 2015; and for Brazil, see and freer movement of labour) are Jenkins 2015; Cypher 2015). often abstract and not sufficient in the current context of a world where Drivers of Industrialisation and industrial production is dominated De-industrialisation by fragmented chains of production spreading across numerous coun- The debate on drivers of tries. As such, policy recommenda- stalled industrialisation and prema- tions usually pay little attention to, ture de-industrialisation in middle- or make limited links with, the income developing countries tended structural global forces that will to be framed as either technology- shape the future pattern of struc- driven or trade-driven. In some ways tural transformation. Available stud- the discussion has emanated from ies also say little on the state’s role in the discussion of de-industrialisa- supporting new sources of economic tion in OECD countries. Saeger growth and job growth, and in man- (1997) tested the ‘technology or aging a country’s position in a GVC- trade’ explanations and concluded world. that both played a role, though the technology explanation seemed to In the developing world dominate in OECD countries. itself, much has been written on Rodrik (2016) hypothesised that the industrialisation, employment, and former, technology, might explain development (take, for example, de-industrialisation in the OECD Definitions Andersson and Axelsson, 2016; while the latter, trade, may better Stalled industrialisation refers Cruces et al., 2017; Newman et al., explain stalled industrialisation and to economic development 2016; Szirmai et al., 2013) and epi- premature de-industrialisation in the where manufacturing shares of sodes of ‘late’ industrialisation (e.g. developing world. employment and/or GDP have plateaued and are no longer rising. Amsden, 1989, 2001; Chang, 1994; Wade, 1990). It is now well docu- The technology-driver De-industrialisation refers mented that many middle-income explanation of stalled industrialisa- to economic development developing countries are undergoing tion and premature de-industrialisa- where manufacturing shares of employment and/or GDP have contemporary processes of stalled tion points towards the driver as begun to fall. industrialisation, de-industrialisa- labour-saving technology. More pre- tion, and tertiarisation (see, for cisely, there is a high-skill bias in Premature de-industrialisation example, Amirapu and Subramanian, GVC-participation. This is based on refers to economic development where manufacturing shares 2015; Dasgupta and Singh, 2006; Grimalda and Vivarelli (2010) who of employment and/or GDP Felipe et al., 2014; Frenkel and argue that there is a skill-biased have started to fall at a share of Rapetti, 2012; Herrendorf et al., technological change. This form of manufacturing and/or GDP 2013; Palma, 2005, 2008; Pieper, technological trade has led to an per capita lower than in earlier 2000; Rodrik, 2016; Szirmai and increase in demand for skilled industrialising countries. Verspagen, 2011; Tregenna, 2009, labour. A recent example of a study 2014). There is also a set of single- that goes further empirically is that

4 Policy Brief www.merit.unu.edu

of Kunst (2019) who finds that the are more capital-intensive, such as employment lost in stalled industri- basic and fabricated metals, evolve alisation and premature de-industri- into the dominant subsectors. Then, alisation is mostly low-skilled countries move towards more tech- employment. Furthermore, he argues nology-intensive manufactures, such that the employment losses occur in as electrical machinery and appara- the formal manufacturing sector and tus. Empirically, Tregenna and

“ ...it is not the case that manufacturing has stalled, but rather that governments have failed to kick-start the process in the same way as China and other East Asian countries have done.”

that stalled industrialisation and Andreoni (2019) (see text box) find premature de-industrialisation are that there is significant heterogene- driven by jobs vulnerable to ICT ity between and within types of adoption. Rodrik (2016) provides a manufacturing by technology-inten- similar empirical estimate and also sity – low, medium and high-tech finds that the employment reduction manufacturing – in the sense that is largely among low-skill workers. not all groups and subsectors exhibit the inverted-U pattern associated An alternative, though not with de-industrialisation. necessarily mutually exclusive view, is that stalled industrialisation and Finally, a third, non-mutu- premature de-industrialisation are ally exclusive view is that stalled driven by trade. More precisely, industrialisation and premature de- intense competition and the entry of industrialisation are due to a failure China and other countries into the of government policy to put in place lower-tech end of GVCs are the effective industrial policy (see drivers. Hence, stalled industrialisa- Dasgupta and Singh, 2006 for dis- tion and premature de-industrialisa- cussion). In short, it is not the case tion are phenomena of the that manufacturing has stalled, but lower-tech manufacturing subsec- rather that governments have failed tors, notably textiles and wearing to kick-start the process in the same apparel (e.g. Haraguchi, 2015). This way as China and other East Asian argument is predicated on the idea countries have done. (Or one could that there are several stages of manu- say that many of the industrial poli- facturing development. In a first cies used by East Asia are not com- stage, at lower income levels, low- patible with WTO membership.) tech manufacturing such as food/ Wade (2014, p. 793–4) points to the beverages, textiles, and wearing following: A ‘developmental state apparel develop as these relate to Mark I’ can no longer create an basic human needs and are labour- autonomous industrial base due to intensive. As countries move through the contemporary ownership by middle income towards high income, transnational corporations of much resource-processing industries which of the GVCs and WTO rules on

Industrialisation and De-industrialisation in the Developing World 5 www.unu.edu

traditional industrial policy. with India and note India’s services- However, a ‘development state Mark led economic development and that II’ is still possible because global the potential market for tradeable rules such as those of the WTO do services is large and not fully little to constrain certain policies, exploited to date. In contrast, a less such as competitive devaluations, optimistic view is illustrated by investment incentives, trade finance, Rodrik (2018) who notes the het- and export taxes. Thus, the state can erogeneity of services. Tourism, IT, still act strategically. and finance are highly tradeable and Recent Contributions on technologically dynamic, just like the Topic The renaissance of state- manufacturing. However, few of owned enterprises in countries fac- these sectors can generate large scale Tregenna and Andreoni (2019) ing stalled industrialisation and employment for low-skilled workers explore the heterogeneity within premature de-industrialisation and and the employment generated in manufacturing underlying de-industrialisation. They the emergence of a ‘new develop- finance or business services is likely find that there are significant mentalism’ are signs of attempts to to be for high-skilled labour. In con- differences between low-, re-industrialise in the face of stalled trast, the sectors of personal and medium- and high-tech industrialisation and premature de- retail services are non-tradeable, manufacturing and variations industrialisation (see for more dis- technologically stagnant, and usually among the subsectors within cussion Kim and Sumner, 2020). overwhelmingly occupied by infor- each of these categories. mal, low-productivity, small firms. While manufacturing as a whole exhibits an inverted-U Consequences of De-industrialisation curve of industrialisation and What do the data say? de-industrialisation, this is not The central question for Rodrik (2018) compares labour pro- the case for all activities within developing countries is whether ductivity growth and employment manufacturing. They develop modern services – tradeable, high generation in wholesale and retail a typology of countries in four productivity services – can replace trade as well as in personal and com- groups: ahead and speeding, manufacturing’s potential for eco- munity services. He finds a negative ahead but slowing, behind nomic development and large scale correlation in the sense that sectors and slowing, and behind but employment generation. An opti- with the best productivity shed speeding. mistic view is illustrated by Page labour and vice versa. De Vries’ De Vries (2019) considers a (2012) who has argued that modern (2019) (see text box) empirical broader concept of ‘modern’ services such as tourism, IT, and study broadens the concept of mod- industry in keeping with finance or ‘industries without ern sectors to include modern agri- W. Arthur Lewis. Besides manufacturing, he includes smokestacks’ and high-tech modern culture, finance, and business modern agriculture, finance, agriculture have many of the fea- services and finds that premature and business services to test the tures of manufacturing. Namely, de-industrialisation is not found for de-industrialisation thesis as they are tradeable and connected to modern sectors’ output shares. these sectors share characteristics GVCs, have economies of scale, and However, premature de-industriali- with manufacturing, such as are technologically dynamic. sation is found for employment economies of scale, participation However, other services – such as shares. In short, modern agriculture in global value chains, and retail and personal services – are and modern services might at best potential for technological change and productivity growth. non-tradeable and do not exhibit mitigate the premature de-industri- He finds that premature high levels of productivity. Ghani alisation of value-added, although de-industrialisation is not and O’Connell (2014) argue that not the premature de-industrialisa- observed for modern industry services are increasingly traded, tion of employment. Hence, modern output shares but is observed for linked to GVCs, and able to gener- agriculture and modern services can- employment shares. ate large scale employment. Ghani not replace the loss of manufactur- and Kharas (2010) compare China ing jobs.

6 Policy Brief www.merit.unu.edu

Future Policy Options could – and many appear to be try- ing already – make a strategic choice Developing countries face a to seek growth through re-industri- choice between three strategies for alisation (and thus implicitly neglect economic development that pro- services in relative terms, at least duces economic growth and employ- through policy choices). This option ment growth. Option one is the requires state activism in the form of strategic route of the promotion and direct or indirect subsidies to help expansion of higher-value-added ser- domestic companies enter GVCs vices or business services such as and/or upgrade in GVCs to higher- finance, insurance, real estate and value-added activities (i.e. through business services, and services that subsidised or directed credit, protec- are internationally tradeable. This tion, and/or upskilling labour forces) could generate much value-added, and/or international companies as though with weak employment foreign direct investors (i.e. through growth potential for low and fiscal subsidies, tax breaks, or other medium skilled workers given the incentives). The challenge with this nature of these services. This option state-led re-industrialisation is that is unlikely to ensure greater equality. it is predicated on participation in Indeed, it may even increase tensions the very part of the global economy, in terms of domestic political econ- namely lower-end GVCs, that may omy as only a small proportion of have engendered stalled industriali- Recent Contributions on the population will be working in sation and de-industrialisation in the Topic such high-value-added services. many middle-income countries. Each Schlogl and Kim (2019) consider Governments could plausibly redis- of these choices is sub-optimal and Rodrik’s (2016) conjecture tribute value-added through taxa- involves trade-offs for value-added that job de-industrialisation tion, transfers, and public goods; growth, employment growth, and real in developing countries has however, the politics may be chal- wage growth for different types of adverse political consequences. lenging if this implies a major redis- workers. They subject this conjecture tributive programme. to multiple empirical tests. Looking Ahead Using country-fixed-effects A second option for devel- panel regression of sectoral employment shares in oping countries is to pursue eco- Adding to the struggle for manufacturing on a range of nomic development strategies that development is the likelihood that political outcome variables, the expand the share of the economy of technological change will accelerate paper finds no empirical support lower-value-added services. These de-industrialisation of employment for the hypothesis that (de-) include trade/restaurants/hotels, because the kinds of jobs common in industrialisation in developing transport/storage/communication, developing countries – such as routine countries is systematically government services, and commu- manufacturing work – are substan- related to (de-)democratisation. nity/social/personal services. As tially more susceptible to automation. However, a persistent, though weakening, cross-sectional noted, these sectors have higher Ultimately, if more manufacturing jobs relationship of manufacturing potential for employment growth for are automated, workers will continue employment with levels of low or medium skill workers (albeit to move into the service sector, leading democracy raises the question of often in informal, low wage services) to a bloating of service-sector employ- what time-varying factors have although they are much more limited ment and wage stagnation (see Schlogl been moderating the relationship in value-added potential. and Sumner, 2020). between employment structure and political outcomes. The A third option is that gov- Full references available at: paper offers a set of explanations. ernments of developing countries https://bit.ly/2vZQqHn

Industrialisation and De-industrialisation in the Developing World 7 www.unu.edu www.merit.unu.edu

INSIDE: Policy Brief

The United Nations University – Maastricht Economic and Social Industrialisation and Research Institute on Innovation and Technology (UNU-MERIT) is a research and training institute of United Nations University based De-industrialisation in Maastricht in the south of the Netherlands. The institute, which collaborates closely with Maastricht University, carries out research and in the Developing training on a range of social, political and economic factors that drive World economic development in a global perspective. Overall the institute functions as a unique research centre and graduate school for around The opportunities 100 PhD fellows and 140 Master’s students. It is also a UN think tank manufacturing presents addressing a broad range of policy questions on science, innovation and democratic governance. for economic development and employment creation are being exhausted earlier and with fewer of the benefits that have typically accrued to countries through manufacturing. This brief discusses the phenomenon and the policy implications for developing

countries looking ahead.

The Netherlands The

6211 AX Maastricht AX 6211

Boschstraat 24 Boschstraat

Innovation and Technology and Innovation

social Research institute on institute Research social

Maastricht Economic and Economic Maastricht Uniter Nations University - University Nations Uniter