De-industrialization and comparative advantage in the global value chain

ABSTRACT

We investigate the causes of de-industrialization and potential for re-industrialization using trade-linked input–output data from WIOD. By introducing a new global value chain measure of comparative advantage, we relate a sector's share in domestic final demand to that in production and separate the direct effect of trade on its income share. This method identifies the declining share of manufacturing value added in domestic final expenditures to be the main cause of de-industrialization. Differences in comparative advantage between countries do matter, especially in the case of employment shares, but have a limited impact via the direct trade effect on value added. The findings point to a peculiar paradox of industrial policy: precisely when it is successful in raising competitiveness and hence productivity growth of manufacturing, it also furthers the global decline of relative prices in manufacturing. In contrast to the national objectives of re-industrialization, effective industrial policies accelerate de-industrialization in the global economy.

KEYWORDS:

• Industrial policy
• de-industrialization
• structural change
• global value chains

Acknowledgments

Our research has benefited from many discussions on industrial policy at various events, most notably the 23rd conference of the International Input Output Association (IIOA) in Mexico City, studies commissioned for the EU Competitiveness Reports and sessions led by the Austrian ‘Forschungsplattform Internationale Wirtschaft’ (FIW). We are especially grateful for feedback on earlier drafts of the paper provided by Geoffrey Hewings and Robert Stehrer as well as Kurt Kratena, Harald Oberhofer, Marcel Timmer, Gaitzen deVries, John Zysman, two anonymous referees and the editor. We also thank Anna Strauss and Astrid Nolte for their assistance and proofreading. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Smith (1776) and Hamilton (1791) are classic early examples. Kaldor's (1966, 1967) growth laws postulated a pre-eminent role of industrialization in development. Cohen and Zysman (1987) warned of the perils of de-industrialization. Rodrik (2011) championed the resurgence of a ‘new industrial policy’.

2 Aghion et al. (2011), Berger (2013), Cimoli et al. (2015), Delbono and Lambertini (2016), Farla (2015), Marsh (2012), Mause and Gröteke (2017), O'Sullivan et al. (2013), Pianta (2014), Rifkin (2012), Stiglitz et al. (2013), Stöllinger and Holzner (2017) or Van Reenen (2013).

3 Dietzenbacher et al. (2013); Timmer et al. (2015, 2016).

4 Note that one usually does not find any (marked) decline in the manufacturing share when value added is measured at constant prices (Peneder et al., 2003; Szirmai, 2012). Acemoglu and Guerrieri (2008) show how capital deepening and heterogenous factor intensities cause quantity and value to move in opposite directions between sectors.

5 For further details, see Peneder and Streicher (2016).

6 The idea goes back to Engel's law of 1895. See also Houthakker (1957), Pasinetti (1981), Kongsamut et al. (2001) or Foellmi and Zweimüller (2008) for recent general equilibrium models.

7 For example, Schettkat and Yocarini (2006) estimated budget elasticities from data on household expenditures for the USA, Germany, France, UK, Spain and the Netherlands. For manufactured (nondurable) goods, all were in the range between 0.75 and 0.82 (except France at 0.97). In contrast, the budget elasticities for services (excluding, e.g. the health sector) were all in the range between 1.38 (USA) and 1.70 (Spain). Reported earlier estimates of the price elasticity of demand for services was in the range of −0.08 (Germany) and −0.53 (the USA).

8 See Baumol (1967), Ngai and Pissarides (2007) or Acemoglu and Guerrieri (2008).

9 This implies that the utility consumers obtain from a particular good is unbundled, separating, for instance, the value added contributed from manufacturing and that from services (such as R&D, design, marketing, transport or retail). See Herrendorf et al. (2013).

10 Note that outsourcing does not make a difference in the final expenditure share of manufacturing, if sectors are defined by the alternative final consumption expenditure approach. See Herrendorf et al. (2013).

11 For obvious reasons, policy would be ill advised to strive towards lower income levels in order to mitigate the income effect, higher consumer prices to mitigate the price effect, or an inefficient division of labour (outsourcing) to slow de-industrialization.

12 In contrast, if productivity shocks are symmetric across countries, comparative advantage is not affected and only the expenditure channel matters.

13 In order to calculate a global measure of MIVAS, one can alternatively use the ratio of the first and third columns in Table 2 to the total value added.

14 In order to calculate a global measure of DIVAS, one can alternatively use the ratio of the first and third lines in Table 2 to the total value added.

15 To mention only a few pioneering contributions in this area, Hummels et al. (2001) introduced an input–output-based indicator of vertical specialization (VS), calculating the value share of imported intermediate goods that is embodied in exports. In contrast, Johnson and Noguera (2012) introduced the concept of value-added exports (VAX) by linking input–output to bilateral trade data. Koopman et al. (2014) presented an integrated formal framework for both approaches. In recent years, the construction of trade-linked input–output databases such as WIOD has prompted a major surge of new studies. Among its first applications, Timmer et al. (2013) applied the data to construct new measures of GVC-income and jobs, Foster-McGregor and Stehrer (2013) provided generalized measures of VS, and Los et al. (2015) focused on the foreign value-added share as a measure of fragmentation in production chains. For further applications, discussions and surveys of the empirical evidence, see e.g. Amador and di Mauro (2015), Baldwin and Lopez-Gonzalez (2015), De Backer and Miroudot (2014), Johnson and Noguera (2012), Streicher and Stehrer (2015) or Timmer et al. (2014).

16 Note that our measure must be sensitive to the size of a country in order to capture the size-dependent difference in the impact of trade on structural change. For other purposes the traditional RCA measures have the advantage of not being size dependent.

17 Note that according to Equation 10(10) $\begin{array}{rl}\mathrm{\Delta }{VAS}_{t-1}^t& =\mathrm{\Delta }{DIVAS}_{t-1}^t\times {TEVAS}_{t-1}\\ & +\mathrm{\Delta }{TEVAS}_{t-1}^t\times {DIVAS}_{t-1}\\ & +\mathrm{\Delta }{DIVAS}_{t-1}^t\times \mathrm{\Delta }{TEVAS}_{t-1}^t.\end{array}$(10) the contribution of changes in TEVAS also depends on the initial level of DIVAS, which explains about 10% of its overall variance across countries.

18 Bulgaria, Czech Republic, Hungary, Lithuania, Poland and Romania.

19 See, for instance, the ‘China syndrome’ by Autor et al. (2013, 2016).

20 This is in part due to the lower decrease in relative prices in German manufacturing, which from 1995 to 2009 amounted to a modest −5% (as, e.g. compared to −30% in France or −27% in the UK). This may, for instance, indicate a pay-off to Germany's specialization in less price-sensitive segments of production.

21 Taiwan, South Korea, Germany, Japan and Italy.

22 For instance, Indonesia, Turkey, the UK, France, Brazil, Canada or Australia.

23 But even if this led to a decline in relative prices, services tend to benefit more from the resulting increase in real income. The net impact on value-added shares thus depends on the relative strength of price and income effects.