MASTER THESIS Deindustrialization over time, does China shock matter?

(1)

MASTER THESIS

Deindustrialization over time, does China shock matter?

Nugraheni Dwi Utami

June 18, 2019

Abstract

This paper analyses the impact of Chinese import competition on deindustrialization measured by real value added and employment share in developed and developing countries over 1970-2010 period. By employing quantile regression estimation with instrumental variables to correct potential endogeneity bias, the results suggest that the main driver of deindustrialization in employment in developed countries is technological change. There is heterogeneous effect of China shock. In developed countries, the effect is destructive in term of both employment and real value added in the lower quantile of distribution, with the higher magnitude for the former. In the higher quantile, complementary effect outweighs detrimental impact. In developing countries, the negative effect of China’s shock on real value-added rises as the increase in the proportion of manufacturing value-added in countries. The destructive effect on employment in developing countries seems to be harder after 1990 period.

Keywords : deindustrialization, China, trade Supervisor : Tarek M. Harchaoui

Co-Assessor : Anna Minasyan Student Number : S3731588

Email : n.dwi.utami@student.rug.nl

(2)

ii

Acknowledgements

I would like to thank my thesis supervisor Prof. Tarek M. Harchaoui of the Faculty of Economics and Business at University of Groningen for valuable insights and helpful comments. He continues to guide and support me in my academic research and writing.

In addition, I am grateful to Prof. Anna Minasyan of the Faculty of Economics and Business at University of Groningen for her work as a co-assessor.

Moreover, I'd like to thank Era Dabla-Norris, Alun Thomas, Rodrigo Garcia-Verdu and Yingyuan Chen for sharing their dataset with me.

(3)

iii

List of Figures

FIGURE 1—THE EVOLUTION OF EXPORTS FROM CHINA TO THE WORLD OVER 50 YEARS ... 13

FIGURE 2—THE IMPACT OF CHINA SHOCK ON RELATIVE SHARE REAL VA IN MANUFACTURING . 26 FIGURE 3—THE IMPACT OF CHINA SHOCK ON RELATIVE SHARE MANUFACTURING EMPLOYMENT ... 35

FIGURE A.1–STRUCTURAL TRANSFORMATION IN VALUE ADDED 1970-2010 ... 43

FIGURE A.2–STRUCTURAL TRANSFORMATION IN IN EMPLOYMENT 1970-2010 ... 44

FIGURE A.3–THE COMPARISON DEINDUSTRIALIZATION IN THE DEVELOPED AND THE DEVELOPING COUNTRIES ... 45

List of Tables

TABLE 1–DESCRIPTIVE STATISTICS OF THE VARIABLES FROM THE INTEGRATED DATASET : ... 14

TABLE 2—PANEL WITH INSTRUMENTAL VARIABLE REAL VA-MANUFACTURING SHARE ALL COUNTRY(1970-2010) ... 18

TABLE 3—QUANTILE REGRESSION WITH ENDOGENEITY CORRECTION VA-MANUFACTURING ALL COUNTRY(1970-2010) ... 22

TABLE 4—PANEL WITH INSTRUMENTAL VARIABLE MANUFACTURING EMPLOYMENT SHARE ALL COUNTRY(1970-2010) ... 28

TABLE 5—QUANTILE REGRESSION WITH ENDOGENEITY CORRECTION MANUFACTURING EMPLOYMENT SHARE ALLCOUNTRY(1970-2010) ... 31

TABLE A.1—NOMINAL GDP IN PPPSAMPLE OF COUNTRIES FOR DEINDUSTRIALIZATION REAL VA ... 40

TABLE A.2—NOMINAL GDP IN PPP OF COUNTRIES FOR EMPLOYMENT DEINDUSTRIALIZATION 42 TABLE A.3—LIST OF 61COUNTRIES FOR DEINDUSTRIALIZATION IN REAL VALUE-ADDED ... 46

TABLE A.4—LIST OF 30COUNTRIES FOR DEINDUSTRIALIZATION IN EMPLOYMENT ... 47

TABLE A.5—THE VALIDITY OF INSTRUMENTS TEST ... 48

TABLE A.6—THE CORRELATION BETWEEN INDEPENDENT VARIABLES ... 49

TABLE A.7—MULTICOLLINEARITY TEST ... 49

TABLE B.1—QR WITH ENDOGENEITY CORRECTION DEVELOPED&DEVELOPING COUNTRYREAL VALUE-ADDED (1970-2010) ... 50

TABLE B.2—QR WITH ENDOGENEITY CORRECTION DEVELOPEDCOUNTRYREAL VALUE -ADDED 1970-1989&1990-2010 ... 53

TABLE B.3—QR WITH ENDOGENEITY CORRECTION DEVELOPINGCOUNTRYREAL VALUE -ADDED 1970-1989&1990-2010 ... 55

TABLE C.1—QR WITH ENDOGENEITY CORRECTION DEVELOPINGCOUNTRYEMPLOYMENT SHARE 1970-1989&1990-2010 ... 57

TABLE C.2—QR WITH ENDOGENEITY CORRECTION DEVELOPEDCOUNTRYEMPLOYMENT SHARE 1970-1989&1990-2010 ... 60

(5)

1

I. Introduction

A vigorous political debate is in topic over the impact of globalisation on the decline in manufacturing sectors in the developed and developing world. China’s rapid expansion of manufactured exports looms large in these discussions as it has affected the pattern of trade in international markets. China’s share of global manufacturing exports grew from 0.5% in 1978 to 13.7% in 2010. One potential explanation concerning China’s export performance is her ability to provide the world with low-cost products. Consequently, it might have had an impact on global prices (Kaplinsky, 2006; Jenkins, Peters & Moreira, 2008). We argue in this paper that increased Chinese trade has induced faster deindustrialisation. A compelling piece of evidence best represented by Autor, Hanson & Dorn (2013); Acemoglu, Autor, Hanson & Price (2016); Malgouyres (2017) and Federico (2014), shows contracting in employment in low-skill manufacturing sectors in response to Chinese import competition. The contribution of this paper is to provide evidence of the impact of China shock on deindustrialisation in countries at different levels of development and to confirm the hypothesis that the main driver of deindustrialisation over time in developed countries is due to technological progress. Additionally, it is to show that China shock is affecting developing economies more than advanced economies.

Most previous findings have to be interpreted in light of two important aspects. First, existing studies focus their analysis on the negative employment effects of import penetration and rarely mentioned other causes that influence deindustrialisation. Second, the sample is limited for some countries and the time available to undertake the study was also short (Donoso et al. 2015; Federico, 2014; Dauth et al., 2014; Jenkins, Peters & Moreira, 2008; Feenstra and Sasahara, 2018). While the evidence provided in these contributions add to the existing literature by evaluating reactions to shocks on regional employment effects, they have not been utilised to examine the comparable effects or to draw a conclusion on the repercussions of Chinese’s penetration as a whole.

(6)

2

This paper presents three main results. First, while developed countries have already experienced a rapid decrease in manufacturing employment sectors over decades, Chinese import penetration has made it worse. The effect on employment deindustrialisation is harder than on real value-added, with a magnitude around 2.1 – 2.6 % points for the former and in the region of 0.5 – 1.9 for the latter. Surprisingly, in the countries that have a higher proportion of manufacturing sectors, the effect is complementary. The adverse effect of China clearly affected real value-added with a magnitude around 2.5 – 5.9% points rather than share in manufacturing employment in developing countries. Second, the negative impact of China’s shock is empirically greater in developing countries than in developed countries. The level of development also weakens the effect, implying that the wealthier the countries the lower the effect. Lastly, over the period 1990-2010, the impact of China’s shock is bigger than previous periods. It happens especially in developing countries, while in developed countries the complementary effect dominates the damaging effect.

The structure of the paper is as follows: Section II provides an extensive literature on deindustrialisation, the relation between deindustrialisation and structural transformation and trade, empirical finding on the question of China’s shock on both developed and developing countries, identifies the gap and builds the hypothesis. Section III sketches the econometrics model and discusses the data. Section IV describes the results, while Section V draws the conclusion, shows the limitations and makes further research suggestions.

II. Literature Review

1. Defining Deindustrialisation

Deindustrialisation is generally defined as a steady decline in the relative importance of the manufacturing sector. Bluestone & Harrison (1982), coined the term “deindustrialisation” to explain the occurrence of plants closing and the reduction in the manufacturing industries in the United States in the late 1970s and early 1980s. In practice, to examine whether a country has experienced deindustrialisation, we examined the trend related to manufacturing employment as a share of total employment or the proportion of manufacturing value-added in GDP (Tregenna, 2009). Since price level has an influence on the calculation of manufacturing value-added, different trends would be produced when using nominal and real value-added measures. The relative price of manufacturing in countries tends to become smaller as the level of development increases; thus, lowering the share of manufacturing value-added at current prices relative to real prices. It causes nominal manufacturing where value-added reaches a peak faster than real manufacturing value-added (Rodrik, 2016).

(7)

3

in the United States) and then dropping to about 17% in 2007 (18 years after peaking). Moreover, the pattern relating to share of employment in the manufacturing sector in China appears to behave differently. In 1987, the proportion of manufacturing employment was 15.8%. This figure decreased gradually to 11.26% in 2002. Surprisingly, there was a dramatic jump to 28.4% within a year (2002-2003), and it continued to increase progressively afterwards. It is apparent that China experienced deindustrialisation in employment from 1987-2002, but then substantial re-industrialisation after 2001.

2. Deindustrialization and Structural Transformation

Rowthorn & Ramaswamy (1997), argue that deindustrialisation in developed countries is not a negative phenomenon. Deindustrialisation is a part of structural transformation together with the progress of economic development and reallocation of economic activity across three sectors (agriculture, manufacturing and services) occurs naturally (Kuznets, 1973). Increases in the level of development have been associated with decreases in the share of employment and value-added in agriculture and increases in the share of employment and value-added in services. However, manufacturing has a different nature: its share in employment and value-added follow an inverted U-shaped – they grow for lower levels of GDP per capita, reach a peak at a certain level of development and drop constantly afterwards (Bah, 2011; Era Dabla-Norris et al., 2013; Herrendorf et al., 2014).

(8)

4

shrinking and releases labour to less productive activities such as agriculture and casual work that offers no security.

Among other things, governments worldwide are concerned about deindustrialisation. Consequently, they target the development of manufacturing sectors in their development roadmap and seek to figure out several threats to their plans, including the fear of Chinese penetration into the global market. It would be advantageous if the result of competition with China makes a country move to a sector that is more productive, such as service or skill-intensive manufacturers. However, it would be a problem if there is large scale unemployment and other sectors could not absorb this section, given that unemployment will affect economic performance in the long run. Rodrik (2016), states advanced economies place more concern on employment deindustrialisation as it is linked with losing good jobs, widening inequality, and decreasing R&D and patents which are associated with reducing potential innovation capacity in manufacturing. All of which raises public discourse in advanced economies. In terms of value-added, deindustrialisation is less noticeable as the real manufacturing value-added in advanced countries remains constant as a result of rapid growth in productivity through technological advanced in this sector. Rodrik (2016), also emphasises that differential rates in technological progress are a typical source of employment deindustrialisation in advanced countries.

(9)

5

3. Deindustrialisation and Trade

Heckscher-Ohlin’s theory explains that the driver of international trade is due to differences in factor input for production which originates. Countries tend to export goods whose production makes intensive use of factors of which they have a relatively large supply and imports goods which require large inputs of factors that are locally scarce.

A further factor that encourages countries to engage in international trade is based on differences in endowment related to technology in producing goods as modelled by Ricardian Theory. The differences in technology will affect the need for labour per unit of production in one country relative to other countries. This implies that there would be differences in factor prices which eventually cause differences in national industrial capabilities in producing goods across countries. In international trade, when trade barrier between countries fall, the production and employment structure of countries tends to become more specialised in sectors in which their factor endowments give them a comparative advantage due to production costs. The earnings of their abundant factors tend to increase relative to those of their scarce factors. Therefore, Hanson (2012) suggests that international specialisation follows the perceived patterns of a country’s comparative advantage.

Acemoglu et al. (2016), based on the Heckscher-Ohlin and Ricardo-Viner models of international trade, note that the stronger import competition with China will reduce the relative price of manufacturing goods and generates reallocation of labour and capital toward sectors whose relative prices have increased.

This view is consistent with Wood & Mayer (2010), who state that the entry of China into world markets has affected the sectoral structures of other economies. Given abundant sources of labour-intensive manufacturing resources, China concentrated on exporting labour-labour-intensive manufactured goods and imports goods in which it has a comparative disadvantage in land and resources, for instance primary commodities and skill-intensive manufactured products. The vast expansion of China’s exports and the substantial increase of imports in primary products have altered the relative prices on world markets (Kaplinsky, 2006: Mayer & Fajarnes, 2008; Fu, Kaplinsky & Zhang, 2009) and thus shifted demand functions to the left for labour-intensive manufacturers and to the right for primary goods and skill-intensive manufacturers.

4. The Question of China’s Shock A. Impact on Developed Nations

(10)

6

toward other industries with less exposure to low-wage country imports and greater capital- and skill-intensity, meaning that US manufacturing is shifting resources from comparative-disadvantage activities towards activities consistent with US comparative advantage and accelerating capital deepening across and within manufacturing industries over time.

The negative effect of Chinese penetration on employment had been examined by Autor, Hanson & Dorn (2013). They find that between 1990-2007, Chinese import penetration contributed to a one-quarter increase in unemployment in US manufacturing industries. Similarly, Acemoglu, Autor, Hanson & Price (2016), show that approximately 2-2.4 million manufacturing jobs were lost in the US between 1999-2011 as a result of the import competition from China. The relative share of employment related to the US manufacturing sector experienced a steep decline immediately after the new millennium, a question that has been addressed by (Pierce & Schott, 2016). They linked this decline to the new policy granting Permanent Normal Trade Relations (PNTR) status to China. The US had applied low tariffs on Chinese imports since 1980, but they were subject to annual renewal. The PNTR removes uncertainty with regards to tariffs changes due to annual renewals of China’s Normal Trade Relations. PNTR have benefited Chinese producers by enabling them to enter and expand into the US market. There is debate whether this new policy has benefited the US consumer. Chinese’s penetration seems to have reduced US manufacturing employment by coaxing US producers to invest in capital- or skill-intensive production technologies or less labour-intensive products that are more consistent with US comparative advantage. It can be observed that industries most affected by PNTR exhibit increases in skill intensity.

Malgouyres (2017), investigates the effect of China shock on employment in France between 1998-2008. His research establishes that import competition from China has increased the unemployment rate, and moreover, job losses are concentrated on the low and middle skill occupations. The result indicates that import competition polarised the occupational structure of employment in the manufacturing sector and stimulated skill upgrading. Federico (2014), compares the effect of import penetration from advanced countries and China. He reveals that between 1995-2007, using a panel of 230 Italian manufacturing sectors, import penetration from advanced countries tended to have insignificant effect on employment, whereas the rise of China and low-wage import penetration is related with a decrease in employment, output and the number of firms.

(11)

7

experienced large expansion. In the case of Germany, Dauth et al. (2014), maintain that the impact of China on the labour force in Germany is insignificant. Despite the damaging effect on employment, the aggregate outcome of trade competition with countries in Eastern Europe let to considerable employment gains in the German economy. Job losses are concentrated on regions specialised in import-competing industries. However, there are also substantial employment gains and lower unemployment in regions with export-orientated industries differentiation.

B. Impact on Developing Nations

The effects of Chinese export exposure may vary across countries depending on the structure of the economy of each country. Middle- or developing countries compared to developed countries, in general also have a comparative advantage in labour intensive sectors. They may experience further increase in unemployment if they fail to compete with China. However, empirical findings show that there are also benefits resulting from Chinese expansion. For example, Jenkins, Peters & Moreira (2008), emphasise that growth trade with China has been positive, though it has created winners and losers in Latin America and the Caribbean. Producers and exporters of raw materials such as agriculture, agroindustry and industrial inputs have been the winners, seeing as their exports to China increased sevenfold between 1999-2005. However, countries that specialised in commodity chains such as yarn-textile-garments, electronics, automobiles and auto parts appear to be the losers both in domestic and third markets. This is because Chinese competition caused job losses in garments and textiles export industries, plant closures and that employment had declined as a result of competition from China in the US market.

After the 2nd unbundling, the production structure became more fragmented and firms in advanced countries produce goods in other countries via labour intensive segments of their supply chain to developing countries. Thus, since China has been able to provide labour intensive resources with lower wages relative to other countries, it’s role has increased rapidly and it has been the biggest winner as the supplier of intermediates goods with roughly 11% of global intermediate exports (Baldwin & Lopez-Gonzalez, 2015). The global value chains that run through China may represent a significant opportunity as well as a threat. The negative effect is shown in some countries in East Asian and NIEs (Athukorala, 2009). Countries like Japan, Korea and Hong Kong are less competitive than China due to the high wages which then leads to the comparative advantage being lost on production lines, as an integral part of the global value chain. However, the boost in China’s processing industries has a positive spill over by increasing demand for intermediate goods from other countries in East and South Asia. The finding shows that from 1996-2011, China had a positive effect on employment in East Asia and ASEAN members by providing export opportunities to these countries (Feenstra and Sasahara, 2018).

(12)

8

Even though there is a positive effect through intra-industry trade caused by China’s rise, it look as if that the effects of the competition are much greater than the positive effects (Amann, Lau & Nixson, 2009). Those that suffer the least are the high income countries (e.g. Japan and South Korea) given that the effects of the competition are alleviated by investment through foreign direct investment (FDI) or subcontracting operations, whereas middle income countries are the main losers as they face strong competition from not only China but also low-income countries in Asia. Similarly, Giovannetti & Sanfilippo (2009), also assert that between 1995-2005. Chinese’s exports to Africa statistically reduced African exports in manufacturing products to their main developed market. Specific industries especially in textiles, clothing and footwear were displaced as they failed to compete with Chinese competition in regional markets. As a result, the reallocation effect emerged increasing competition, as the theory suggests. Iacovone, Rauch & Winters (2013), reveal that competition with China has significantly influenced production patterns regarding domestic and export markets in Mexico. The flow of exports from China challenged Mexican firms and cause firms to exit from the market and reduce their product and sales. Surprisingly, even though this shock forces some smaller and less productive plants to close down, it prompts larger plants and core products to increase productivity and even expand.

C. Gap in the Literature and Hypotheses

Overall, the large body of literature demonstrates that China has and is playing a part in deindustrialisation among developed and developing countries. However, previous studies often consider only one or a small sample of countries, whilst most only focus on one specific measure pertaining to deindustrialisation, which is employment (Donoso et al., 2015; Federico, 2014; Dauth et al., 2014; Jenkins, Peters & Moreira, 2008; Feenstra and Sasahara, 2018). Defining deindustrialisation purely in terms of employment share is conceptually limiting given that deindustrialisation processes operate not only through employment but also across output/value-added. Tregenna (2009), suggests deindustrialisation should correctly be defined in terms of a sustained decline in both the share of manufacturing employment and output.

(13)

9

goods to some developed countries. Most previous studies only consider the effect of China’s significant role within the global market. In this paper, given the lengthy time series data, we will examine the effect of China both prior to and after the second unbundling period.

Developed countries appear to have a challenge related to tackling employment deindustrialisation rather than value-added deindustrialisation. The evidence suggests that while manufacturing employment has declined continuously, value-added in real price has remained constant over decades. Better access to dynamic growth in technology in manufacturing which only requires a small amount of labour, is the main reason for labour displacement in developed nations rather than the penetration of Chinese imports (Rodrik, 2016). Moreover, the productivity differential between the manufacturing and service sectors also explains employment deindustrialisation in advanced countries (Rowthorn & Ramaswamy, 1997). The service sector may absorb the excess of labour from manufacturing, since employment tends to move to more productive sectors. However, evidence suggests that labour productivity in manufacturing has consistently outpaced that of services. The productivity growth differentials between manufacturing and services have consistently been much larger than differences in output growth between these sectors. This leads us to the first hypothesis:

H1: Developed countries face more deindustrialisation in employment rather than in real value-added more because of technological progress rather than China shock.

Given the fact that China’s products generally offer a cheaper price on the world market, countries may experience substitution of home products with an increase in import penetration, especially in similar commodities supplied by “the world’s factory”. Foreign supply of goods potentially diminishes domestic production, reducing the amount of labour that is demanded domestically. The main decrease in production and job losses are in the sector that is highly exposed to products from China (Autor, Hanson & Dorn, 2013; Bernard, Jensen & Schott, 2006; Jenkins, Peters & Moreira, 2008; Donoso et al., 2015). Moreover, competition from China could hamper the export opportunities of other developing countries to the third market, as losing competitiveness in terms of price and variety of products. This view results in the following hypothesis:

H2: the impact of China shock is more severe in developing countries than in developed countries.

China’s rapid growth and increased integration into the global economy over the past two decades have significant economic impacts on both developed and developing countries. Its export share of global trade rose from less than 1% in 1980 to around 8% in 2009, and in 2010, it became the world largest exporter ahead of the US and Germany. The debate on China’s economic impact began at the time of its accession to the WTO in 2001, as after this period, a significant surge in exports was observed by many parties and raise concern of the negative effect on economy. Thus:

(14)

10

III. Model and Data for Analysis

1. Model

To investigate the impact of China shock on structural transformation, we will develop the structural transformation model proposed by Era Dabla-Norris et al., (2013). The functional form specified at the level of sector !, country " in year # has the following structure:

$%&' = )*+ ,-ℓ/(12ℎ4"5') + ,78ℓ/(12ℎ4"5') × ℓ/(:;<==")> + )-ℓ/(:;<=="&') + )₇8ℓ/(:;<=="_&')>7+ ∑ @_A _AB_A&'+ ∑ C_' _DE_'+F_%&'………..……….(1)

Dependent variables:

The dependent variables are the real value-added and employment share in three sectors of the economy – agriculture, manufacturing and services. The value-added measure is sensitively affected by the relative price. The relative price of manufacturing in countries tends to become smaller as the level of development increases. This price effect can create a downward bias in the nominal value-added share of manufacturing. Therefore, we will use real value-added in manufacturing sector instead of nominal value-added in addition to employment shares as dependent variables.

Explanatory variables:

Changes in the value of manufacturing goods imported from China will be used as a proxy of China’s penetration following the approach outlined by Autor et al. (2013). The intuition of this proxy is because the changes in the foreign supply of manufacturing goods may replace domestic production, decreasing the amount of output produced domestically and lowering demand for labour as regards production. We also add an interaction variable between China’s penetration with GDP per capita to examine the differences in the effect of competition with China with regard to the GDP per capita of countries.

Control:

(15)

11

may affect both the supply and demand of goods. Third, the share of arable land of the total land - Arable lands is the land area in a country that can be ploughed and cultivated. Indeed, it will affect production in the agricultural sector. Fourth, a dummy variable for transition economy. It covers some Central and Eastern Europe countries and the former Soviet Union. The transition of emerging economies from a socialist to a market-based economy may affect the governmental institution. It will have an effect on the entire economy, such as the structure of the business and financial sectors which may increase the speed of structural transformation (Svejnar, 2002). Fifth, a dummy variable for island economies - Island countries may have limited access to neighbours. They have to bear the relatively higher costs of land transport (Srinivasan, 1986). Sixth, the share

of mining in total value-added. “Dutch disease” theory explains that the boom in natural resources

will have an effect on resource movement and spending effect in the economy. Labour will shift to the “boom” sector as they offer higher wages and productivity (Corden & Neary, 1982). Seventh, age dependency ratios young and old. It is the percentage of non-working young and old people of the total labour force. The age dependency ratio may have an effect on the supply of labour, savings and consumption performance which may affect the elasticity of demand for agriculture products, manufacturing goods or service (Kelley, 1973). Finally, the relationship between structural changes and level of development is likely to be curvilinear. The increase in GDP per capita is linked with the decrease in the agricultural sector and increase in services, while the manufacturing sector follows an inverted U-shape. It is expected to increase when development is at an early stage, but then decrease when the economy reaches a certain level of development. Thus, GDP per capita and (GDP per capita)2_{are included and expected to have a positive and} negative coefficient, respectively. The model also includes year dummies to account for technological change.

2. Data

(16)

12

per capita, GDP per capita PPP and GDP per capita in current international US dollars. This paper uses the former due to the comprehensiveness of the data.

The second source of data is the information on manufacturing products imported from China taken from CEPII (CHELEM) – International Trade (4 digits ISIC sectoral classification) in current Million US Dollars. This data originates from UN-COMTRADE and complementary sources (IMF, OECD, UNTAC, World Bank and National Sources) and was subsequently matched with country pairs and product category, which made it consistent and comparable across countries. This data comprises of all trade flow of multiple sectoral aggregations (production chains, stages in the production process and technological levels). Therefore, capital, intermediate and consumption goods (final demand) are included in the data.

The import data from China is merged with the baseline dataset, dropping 132 countries (6253 observations) attributable to unavailable trade with China. It leaves 84 countries (3294 observations). Then, I exclude 20 countries due to unavailable data for relative share of real value-added per sectors. Next, I omit 3 countries which have empty data concerning GDP per capita in PPP (“ppppc”) for a certain period. Therefore, 61 countries remain as regards observations for deindustrialisation of value-added (26 developed and 35 developing economies). The 61 countries exhibit around 82-94% of the world’s total GDP in PPP over 1970-2010 (see Table A.1).

The last data is taken from the Groningen Growth and Development Centre (GGDC) 10-Sector Database to calculate the share of employment per sector. Originally, this database consisted of 42 countries from five continents. However, due to the unavailability trade flow data for China, we exclude 12 countries. Thus, 30 countries were left for observation. This represents around 68-83% of the world’s total GDP in PPP over 1970-2010 (see Table A.2).

(17)

13

Figure 1— The evolution of exports from China to the World over 50 years

Source: CEPII; (in million US $)

Primary Axis: MAN: China’s export manufacturing goods; TOTAL: China’s total exports worldwide. Secondary Axis: NON-MAN: China export Non-manufacturing goods;

The descriptive statistics relating to the panel data are shown in Table 1. From the sample, Singapore has the relative minimum share of real value-added in agriculture, at 0.03% in 2010, while India is the country with the highest relative share in real value-added relating to agriculture at 39.85% in 1970. In terms of employment share in agriculture, Hong Kong is the lowest at 0.15% in 2007 and Kenya is the highest at 81.02%. Countries with the lowest and the highest relative share of real value-added in the manufacturing sector are Nigeria and Romania, at 1.37% (in 1970) and 29.45% (in 1979) respectively. Nigeria also has the lowest share of manufacturing employment in 1999, at 3.07%, while Hong Kong has the highest employment share in manufacturing in the sample, at 45.29%. Countries in the sample with the lowest and highest relative share of real value-added in-service sector are Nigeria and Hong Kong, at 11.94% (in 1972) and 90.81% (in 2010) respectively. In line with its comparative advantage, Hong Kong also has the highest employment share in service, at 95.87% in 2010, while the lowest share in service employment goes to Kenya, at 15.07% in 1970. The US is the country with the lowest and the highest manufacturing imports from China, at around 649 US dollars in 1970 and 334 billion dollars in 2010. 0.00 5000.00 10000.00 15000.00 20000.00 25000.00 30000.00 35000.00 40000.00 45000.00 0 500000 1000000 1500000 2000000 2500000 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 2015

TOTAL CHINA'S EXPORTS TO THE WORLD

(18)

14

Table 1 – Descriptive statistics of the variables from the integrated dataset1_:

Period 1970-2010

Variable N min max mean p50

va_agr 2501 0.03 39.85 7.48 4.24 va_man 2501 1.37 29.45 13.92 14.25 va_ser 2501 11.94 90.81 49.83 51.03 ea 1230 0.15 81.02 24.14 16.97 eman 1230 3.07 45.29 16.03 14.51 es 1230 15.07 95.87 58.95 62.19 man_impchina 2501 0.00 3.34 0.03 0.001 Ppppc 2501 0.18 82.36 10.99 6.66 Teco 2501 0 1 .04 0 ieco 2501 0 1 .01 0 va_mu 2501 4.62 74.65 22.96 19.70 landarea 2501 0.001 9.16 1.10 0.32 population 2501 0.002 1.22 0.05 0.17 arable 2501 0 73.39 17.56 12.61 agedep_young 2501 15.19 106.43 49.36 43.73 agedep_old 2501 4.43 35.47 13.56 12.17

Notes: va_agr = Share of agriculture in terms of relative real value added; va_man = Share of manufacturing in terms of relative real value added; va_ser = Share of services in terms of relative real value added; ea = Share of agriculture employment; eman = Share of manufacturing employment; es = Share of services employment; man_impchina = Total import of manufacturing goods from China in current U.S. dollar (in hundred billion); ppppc = GDP per capita in PPP (in thousand), from baseline dataset; Teco = Transition economy dummy; Ieco = Island economy dummy; va_mu = Share of mining in terms of relative real value added; landarea = Land area in squared kilometres (in million); population = total population (in billion); arable = percentage of arable land; agedep_young = age dependency ratio, young, as percentage of working-age population; agedep_old = age dependency ratio, old, as percentage of working-age population.

Countries with transition economy status are Hungary, Poland and Romania. The only country which is classified as an Island economy is Iceland. Hong Kong is the country with the lowest share in the mining value-added sector, at 4.62% and Saudi Arabia has the highest share in mining value-added, at 74.65%. Singapore is the smallest country in terms of land, at 670 km2, while the US is the country with the largest amount of land at around 9,161,920 km2. The most populated country is India, with a population of 1.22 billion in 2010, whereas Iceland only has a population of 204,438. Singapore is the country with zero arable land, while Bangladesh has 73.39% arable land in 1989. The age dependency ratio of young people as a percentage of working-age population is 15.19% in Hong Kong, whereas the highest is 106.43% in Kenya. The age dependency ratio of old people as a percentage of working-age population is 4.43 % in Saudi Arabia, whereas the highest is 35.47% in Japan.

(19)

15

The pattern of structural changes in real value-added appears to be similar to all countries, both developed or developing countries. The illustration of this transformation is shown in Figure A.1. We can see evidence of the structural transformation using real value-added in three sectors of the economy. Over the four decades, increasing GDP per capita (ppppc) is associated with a decrease in the real value-added share in agriculture, and increase in real value-added in services. While the manufacturing sector has a hump shaped curve in which increase from the lower levels of development reach a peak at around 16% (on average when GDP per capita reaches approximately $8,103), and subsequently decrease for higher levels of development. Figure A.2 shows the peak in manufacturing employment share is at around 22% (at GDP per capita around $10,938). Figure A.3 displays the peak in value-added in manufacturing is similar to developed countries, at around 15% (when GDP per capita range around $4,023-$8955), where manufacturing employment in developing countries reaches less than 20% nowadays, compared to developed countries at almost 30%. The striking evidence shows that the curve for the developing countries is on a downward and a leftward position of developed countries’ curve, indicating most developing countries are less specialised in this sector overtime and experienced deindustrialisation earlier in the development process compared to developed countries.

IV. Empirical Results

1. Preliminary Remarks

To examine the impact of Chinese penetration across country, we will use quantile regression estimation which allows the coefficients of the main regressor besides the explanatory variables to vary across the distribution of the dependent variable. This method can avoid sample selection bias which is not captured by pooled OLS. Before running quantile regression, we begin with pooled OLS regressions of agriculture, manufacturing and services for both value-added shares and share of employment to tease out the data.

(20)

16

Forward (Fan, 1972). After that, in 1970, China began to export to non-communist countries, especially to Japan, West Germany, France, North America and Hong Kong. In this year, China also started to import capital to increase their exports (Ho, 1980). Thus, these initial export values can be a good predictor of future trading partners for the Chinese and express the comparative advantage regarding intensive labour manufacturing industries. Second, I use the first lag of import from China as suggested by Reed (2015). Using lagged values of the endogenous explanatory variable as instruments can provide an effective estimation strategy if the lag value does not belong in the respective estimating equation and it is sufficiently correlated with the explanatory variable. The final instrument is the distance between the importing countries to China, as suggested by the standard gravity model (Frankel & Romer, 1999). The distance from China to the export market is potentially correlated with the value of imports but has no effect on the share of manufacturing value-added/employment.

The use of the instrumental variable estimator over the OLS estimator is justified only if the instruments satisfy two conditions: they must be partially correlated with the endogenous explanatory variables (“instrument relevance” condition) and must not be correlated with the disturbance process of the second-stage equation (“orthogonally” condition). The relevance of the instruments is tested using two alternative tests: The F-test relating to the joint significance of the instruments in the first-stage regression. Instrument orthogonality is tested using the Hanson J-statistic which also provides a valid test of the suitability of the overall specification of the model. Before deciding whether it is appropriate to use an IV, it is important to test whether the imports from China are endogenous. If not, the POLS estimator is still efficient and consistent (Wooldridge, 2002). The Durbin-Wu-Hausman test for endogeneity (see Table A.5) indicates that when estimating deindustrialisation measured by the share in value-added, the import from China is determined to be exogenous, since the p-value for the model is 0.63. Thus, we fail to reject the null hypothesis that the variable is exogeneous. This paper prefers using IV estimation to POLS estimation because: First, the results of both estimations remain consistent in term of sign of the coefficient, even though it appears slightly different in magnitude. Second, IV estimation produces higher R-squared and smaller root mean square error (RMSE)2_{which indicates IV estimation is}

better fit than POLS results. Lastly, intuitively, import from China may have correlation with the error term as above-mentioned. The second measure of deindustrialisation which is the manufacturing employment share indicates that imports from China should be treated as endogenous. For the model deindustrialisation measured by employment share, the instrument, value of manufacturing imports in 1970 is dropped due to a lack of power in the first stage of regression. Therefore, two instruments are left for this model. Table A.5 shows that the benchmark regressions comfortably pass these tests for an econometric based on cross-country panel data. The results of the first stage reported in the bottom panel indicate that our instruments are strong, with F-stat statistics well above levels of conventional threshold (F-stat>10). The Hansen J stat with p-values: 0.51 and 0.236 fails to reject the null hypothesis, which signifies the model is correctly

(21)

17

specified and over-identifying restrictions are valid. The Sargan-Basman test also indicates the same result and concludes that the combinations of instruments satisfy the instruments relevance and instrument orthogonality conditions.

To test the possibility of multicollinearity, I use a correlation matrix (Table A.6). The result suggests that only age dependency ratio variables have high correlation. But, the Variance Inflation Factor (VIF) test (Table A.7) indicates that the level of multicollinearity still below the limit (<10) to consider multicollinearity. Thus, we can safely ignore multicollinearity.

To estimate the effect of Chinese import penetration on the entire distribution, quantile regression (QR) will be run afterwards. The advantage of QR over OLS is that it exhibits a richer characterisation of the data, allowing us to consider the impact of a covariate on the entire distribution of dependent variable, not merely its conditional mean as in OLS. Therefore, the coefficient and the standard error will be different when using QR as compared to OLS. As a result of endogeneity in Chinese imports, the conventional quantile regression could not be used as it does not accommodate the presence of instrumental variables in the model. Quantile regression with endogeneity correction using the extended quantile regression technique, known as generalised quantile regression (see Powell, 2019) will be performed subsequently. Using this estimation, we will use Markov Chain Monte Carlo (MCMC) simulation as we have more than two independent variables as suggested by Chernozhukov & Hong (2003). This method allows the model to more closely reach the distribution of the sample matches the actual distribution requires. At the end, to conduct a robustness test, we will examine the second phase of the globalisation period, 1990-2010, where China has made a significant contribution to international trade worldwide. In the robustness test, we will compare this period and the previous period to examine the third hypothesis.

2. Analysis of the Results

A. Deindustrialization measured by share of value added. China Shock on Structural Transformation

(22)

18

Table 2 —PANEL with Instrumental Variable Real VA-Manufacturing Share ALL COUNTRY (1970-2010)

Dependent Variable: Share of Value Added per Sector

IV No Year Dummies IV With Year Dummies

(1) (2) (3) (4) (5) (6)

va_agr va_man va_ser va_agr va_man va_ser

(23)

19

Table 2

Dependent Variable: Share of Value Added per Sector

(continued)

(1) (2) (3) (4) (5) (6)

va_agr va_man va_ser va_agr va_man va_ser

1988.year_t -3.20127*** 1.45641** 1.15029 (0.53112) (0.71657) (0.87535) 1989.year_t -3.02304*** 1.41296** 0.99021 (0.52902) (0.71374) (0.87190) 1990.year_t -2.86666*** 1.21805* 1.07330 (0.52497) (0.70827) (0.86521) 1991.year_t -2.67905*** 0.94628 1.21659 (0.52257) (0.70504) (0.86127) 1992.year_t -3.00844*** 0.86616 1.43287* (0.51856) (0.69963) (0.85466) 1993.year_t -2.91962*** 0.74031 1.49090* (0.51387) (0.69331) (0.84693) 1994.year_t -2.69650*** 0.82709 1.17075 (0.51181) (0.69053) (0.84354) 1995.year_t -2.66992*** 0.91776 1.08545 (0.50972) (0.68771) (0.84009) 1996.year_t -2.27385*** 0.83568 0.83466 (0.50828) (0.68576) (0.83771) 1997.year_t -2.27597*** 0.92706 0.75934 (0.50718) (0.68428) (0.83590) 1998.year_t -2.11949*** 0.76944 0.83029 (0.50518) (0.68158) (0.83261) 1999.year_t -1.93821*** 0.87099 0.76304 (0.50430) (0.68039) (0.83115) 2000.year_t -1.87003*** 1.02658 0.65985 (0.50123) (0.67625) (0.82610) 2001.year_t -1.68339*** 0.75021 0.75518 (0.50082) (0.67570) (0.82542) 2002.year_t -1.45132*** 0.59798 0.73572 (0.49916) (0.67345) (0.82268) 2003.year_t -1.17655** 0.49722 0.63247 (0.49549) (0.66851) (0.81664) 2004.year_t -0.89702* 0.54977 0.29961 (0.49223) (0.66411) (0.81127) 2005.year_t -0.75468 0.52342 0.11956 (0.49015) (0.66130) (0.80783) 2006.year_t -0.48406 0.52605 -0.23937 (0.48845) (0.65901) (0.80503) 2007.year_t -0.34178 0.53684 -0.44246 (0.48734) (0.65751) (0.80321) 2008.year_t -0.08761 0.31422 -0.41107 (0.48682) (0.65681) (0.80234) 2009.year_t -0.15278 -0.33052 0.43016 (0.48722) (0.65734) (0.80300) _cons 49.48564*** 7.15289 23.12540** 70.62667*** 8.00335 7.60688 (6.91256) (8.29538) (10.27110) (8.32067) (11.22607) (13.71358) Obs. 2501 2501 2501 2501 2501 2501 R-squared 0.83071 0.33469 0.86651 0.86712 0.33990 0.87108

Year Dummies NO NO NO YES YES YES

Number of Countries 61 61 61 61 61 61

IV YES YES YES YES YES YES

Standard errors are in parenthesis *** p<0.01, ** p<0.05, * p<0.1

(24)

20

China’s shock will shift countries to the agricultural sector (lower productivity level) instead of the manufacturing sector which may therefore, force countries to go to structural transformation in the wrong direction.

In line with the theory, the increase in GDP per capita tends to increase the value-added in the manufacturing sector, though it then decreases after a certain point of development. It is shown in the positive sign in variable GDP per capita and negative sign in the GDP per capita square. The negative correlation between GDP per capita and VA in the agricultural sectors also incorporate the prediction of structural changes. When countries become wealthy, they will reduce the number of agricultural sectors which generally comprise low productivity and move to higher productive sectors. The service sector will always increase following the path of development, as seen in the positive sign in GDP per capita in-service sector.

The interaction between GDP per capita and imports from China on the agriculture sector is negative, meaning importing from China encourages the country to move to higher productive sectors over the growth of GDP per capita. In the manufacturing sector, the level of GDP per capita growth will be a counterbalance to the deindustrialisation of the effect of China’s penetration. Additionally, imports from China appear to have no effect on the service sector if we include year dummies.

The status of economic transition may have a positive impact on the structural changes, as it significantly decreases value-added in the agricultural sectors and increases value-added in manufacturing, although it has no effect on service sectors.

Island economies tend to benefit agricultural and manufacturing sectors, as a country might depend on domestic agriculture and manufacturing production instead of importing from other countries. This is due to high trade costs as countries that are islands cannot be easily connected to other countries. Trade is limited due to high transport costs, which may induce a spread rather than agglomeration (Baldwin, 1998), whereas the service sectors might have less of an advantage as it will be more difficult in terms of transportation or telecommunication due to the existence of a sea barrier.

(25)

21

is a potential asset with regard to producing agricultural output. Similarly, higher productive land tends to reduce the value-added in the manufacturing sector as well. The reason is because if they already have a lot of productive land which provide high return, labour might move to this sector. Population has a positive relationship with value-added in manufacturing as the growth in population will increase demand for manufacturing goods. However, population growth also represents efficiency, so that in the agricultural sector, increasing population growth will reduce net output in agriculture.

Both the dependency of young and old people tends to decrease real value-added in manufacturing sectors as it might reduce the willingness to spend money on manufacturing goods (Kelley, 1973). Whereas for the agricultural and service sectors, the higher the dependency of young people will increase the value-added because demand for food and services will increase. The increase in dependency of old people tends to decrease the VA in agriculture because they could affect the supply of labour or demand for output from the agricultural sector.

China Shock on Deindustrialization

(26)

22

Table 3—Quantile Regression with Endogeneity Correction VA-Manufacturing ALL COUNTRY (1970-2010)

Dependent Variable: Share of real Value Added in Manufacturing Sector

IV NO Trend Dummy IV WITH Trend Dummy

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

va_man va_man va_man va_man va_man va_man va_man va_man va_man va_man

(27)

23

Table 3 (continued)

IV NO TD IV WITH TD

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

(28)

24

Table 3 (continued)

IV NO TD IV WITH TD

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

y1999 2.020*** 1.665*** 1.499*** -0.072* 0.385*** (0.055) (0.092) (0.083) (0.040) (0.050) y2000 1.951*** 1.669*** 1.432*** 0.037 0.403*** (0.084) (0.076) (0.067) (0.058) (0.024) y2001 1.663*** 1.583*** 1.281*** -0.149*** -0.078*** (0.064) (0.047) (0.072) (0.048) (0.029) y2002 1.824*** 1.408*** 0.961*** -0.168*** -0.117*** (0.060) (0.073) (0.101) (0.034) (0.039) y2003 1.423*** 1.130*** 0.706*** -0.234*** 0.186*** (0.055) (0.078) (0.094) (0.047) (0.018) y2004 0.907*** 0.906*** 0.810*** -0.314*** 0.651*** (0.062) (0.064) (0.098) (0.059) (0.026) y2005 0.608*** 0.867*** 0.692*** -0.172*** 0.606*** (0.064) (0.091) (0.073) (0.058) (0.015) y2006 0.499*** 0.639*** 0.667*** 0.176*** 0.509*** (0.076) (0.066) (0.064) (0.033) (0.016) y2007 0.235*** 0.675*** 0.591*** -0.013 0.586*** (0.064) (0.068) (0.104) (0.073) (0.022) y2008 -0.020 0.593*** 0.445*** 0.427*** 0.394*** (0.086) (0.123) (0.091) (0.031) (0.020) y2009 0.023 -0.199** -0.031 -0.800*** -0.587*** (0.078) (0.089) (0.064) (0.030) (0.017) _cons -26.242*** 22.285*** - 1.163 19.191*** 21.881*** 22.716*** - -3.683*** 6.618*** 23.672*** 20.186*** (1.363) (7.097) (14.117) (2.858) (1.407) (0.333) (0.504) (0.702) (0.450) (0.154) Obs. 2501 2501 2501 2501 2501 2501 2501 2501 2501 2501 R-squared - - - -

Year Dummies NO NO NO NO NO YES YES YES YES YES

Number of

Countries 61 61 61 61 61 61 61 61 61 61

IV YES YES YES YES YES YES YES YES YES YES

(29)

25

Table B.1 (Appendix) shows the juxtaposition of the effect of China shock on different quantiles, both in developed and developing countries. From the coefficient manufacturing import from China, developing countries are negatively affected in all quantiles, while in developed countries most quantiles suffer from China shock except countries in quantile Q.75. Developing countries also suffer greater losses as compared to developed countries. In developed countries, it looks as if the higher the quantile the lower the impact. For example, countries that belong to quantile .10 such as Hong Kong, Luxembourg, Norway and the US experience greater losses at 1.9 percentage points compared to countries in quantile .90 such as Germany, Japan and Switzerland at 0.01 percentage points. However, the damage increases owing to the increase in GDP per capita as the interaction between GDP per capita and manufacturing import from China show opposite signs. In contrast to developed countries, all developing countries in all quantiles suffer considerably at around 2.5 – 5.9 percentage points as compared to developed countries that only decreased by less than 1.9 percentage points if their import of manufacturing goods from China increased by 1% point. The trend also looks different, the higher the share in real manufacturing value-added, the greater the loss. In sum, it proves the second hypothesis that states the impact of China shock is greater in developing countries than in developed countries. Interestingly, if we examine the coefficient captured by time dummies, developed countries have a negative sign but they maintain positive growth in real manufacturing VA over time (at least from 1970-20093_{) in all quantiles (Table B.1 column 2-6). In quantile .10 for example, the increase in real}

value-added is 8.35 percentage points (-0.46 – (-8.808)), while in quantile .90, the growth is 0.67% points (0.031- (-0.64)) in 2009 relative to year 1970. However, in developing countries, even though they have positive coefficients in almost all periods of study, they exhibit negative progress. In all quantiles, the decline is around 6.89 – 8.57 percentage points in 2009 than in 1970 (Table B.1 column 7-11). This finding clearly shows the evidence that developed countries have done well in real manufacturing value added, whereas developing countries suggest otherwise. This evidence confirms the first hypothesis that the rapid technological progress does reasonably to account for real manufacturing value added trend in developed countries.

Measuring China Shock

To measure the average effect of China’s penetration on the pattern of manufacturing value-added against GDP per capita in all countries, we will compare the estimation of real value-added in manufacturing without and with the existence of China’s penetration.

Real value-added in manufacturing without China Exposure:

!_"#$#%,'( = *₊+ *_-ℓ/(123445_'() + *₇8ℓ/(123445_'()97+ ∑ ;_< _<=_<'(+ ∑ >₍ _?@₍+A_B'(………….(2)

Real value-added in manufacturing with China Exposure:

!_"#$#%,'( = *₊+ C_-ℓ/(DEℎG5H₍) + C78ℓ/(DEℎG5H() × ℓ/(123445)9 + *-ℓ/(123445'() + *₇8ℓ/(123445_'()97+ ∑ ;_< _<=_<'(+ ∑ >₍ _?@₍+A_B'(………..………..…(3)

(30)

26

Contrasting the results of equation (2) and equation (3), we predict the effect of China’s shock on the share of manufacturing value-added, as illustrated in Figure 2. From this graph, we are able to see that most contractions due to China’s shock are in the lowest level of GDP per capita, seeing that as the level of development increases, the competition effect will be lower. After the intersection point, when GDP per capita is at approximately 26,903 US dollars, Chinese penetration supplements the manufacturing sector, thus it does not accelerate deindustrialisation.

Figure 2—The impact of China Shock on Relative Share Real VA in Manufacturing

The effect of China’s shock on real VA in the manufacturing sector in the presence of GDP per capita: JK_LMN,OP

JQN(RKSTOU)_VP = −X. Z[\ + ^. _\[ QN(`abcOeeeeeeeeeeeeeeeee dP)

From this equation, the effect of China Shock is negative, however, owing to the increase in the GDP per capita, the effect will be diminished. Using mean of GDP per capita for all countries. This is even more apparent in the countries which have the highest GDP per capita, given that China shock will supplement the real value-added in their countries. Thus, this result is in accordance with the illustration in Figure 2. 5 10 15 20 R e la ti ve Sh a re R e a l V A i n Ma n u fa ct u ri n g 4 6 8 10 12 ln_ppppc

Linear prediction Linear prediction

predicted no China Shock predicted with China Shock

(31)

27

Robustness Check: Second Phase of Globalization (1970-1989 vs 1990-2010)

In the second phase of Globalisation, the world economy is becoming more profoundly integrated due to the decrease in transport costs and substantial effect of technological changes. Prior to 1990, China’s exports were dominated by clothing, shoes, children’s toys, food and agricultural goods. In 2008, Hanson (2012), documented that the composition of China’s exports changed dramatically as the highest share of exports to GDP is with regards to electronic products. Additionally, other goods also experienced substantial growth, such as metals, chemicals, machinery and transport equipment, whereas clothing, shoes and food exports show a declining trend. This might be true due to the dynamics of the Chinese market and the increased sophistication of Chinese exported goods (Rodrik, 2006). Moreover, Baldwin & Lopez-Gonzalez (2015) and (Sturgeon, 2010) highlight that production for global markets is primarily concentrated in China. This is reflected in the intermediate goods trade flow from high income countries to China. At the same time, China is also a country that supplies intermediate goods for various developed countries.

If we relate the robustness test of the impact of China on real VA in the period 1970-1989 vs. 1990-2010 in developed countries (see Table B.2), the results are in accordance with dynamic production structure in China. From 1970-1989, developed countries in quantile .25 such as Australia, Canada, Netherlands and Sweden experienced a decline in real manufacturing value-added as China increased its imports, but the effect diminishes as the increase in GDP per capita is reflected in the interaction between GDP per capita and manufacturing imports from China. However, between 1990-2010, countries in this quantile benefit from manufacturing goods obtained from China. The opposite story happens in countries belonging to quantile .50, such as Portugal, Spain and the UK. From 1970-1989, they took advantage of China’s manufacturing goods, although from 1990-2010, they experienced negative effect because of a flood of “made in China” goods.

(32)

28

goods in the region of 0.4%-6.2% in year 2009 relative to year 1990. This implies that developed countries have indeed not experienced deindustrialisation in term of real value added.

The effect of China shock on developing countries appears by means of a competition channel. Developing countries which lose competitiveness with Chinese products in both domestic and international trade will experience contraction in manufacturing sectors. This is shown in the quantile regression results for developing countries; the higher the share of value-added in manufacturing, the higher the contraction (see Table B.3). From 1970-1989, the reduction is roughly 1.61% - 3.38% points in all quantiles. However, from 1990-2010, the loss is greater at around 4.59%-8.61% points in developing countries if their imports from China increased by 1%. Countries that have suffered the most are those that belong to quantile .90 such as Brazil, Malaysia, Philippines and Thailand. This might be true as they manufacture similar products to China such as machines, electronics, textiles and clothing. This finding is also consistent with the magnitude captured by time dummies. Deindustrialisation in real value-added is approximately 1.512% - 3.071% points in 1988 relative to 1970. This decline is much larger during the second phase of globalisation (1990 onwards) at roughly 5.264% - 11.141% (Table B.3). This finding is in agreement with (Rodrik, 2016), who highlights premature deindustrialisation in developing countries.

To conclude, this result is consistent with the second and third hypotheses which are the impact of China’s penetration is felt more in developing countries than in developed countries, even more so from 1990 onwards, when China’s shock hit harder than in the previous period.

B. Deindustrialization measured as a share of employment.

Table 4—PANEL with Instrumental Variable Manufacturing Employment Share ALL COUNTRY (1970-2010)

Dependent Variable: Share of Employment per Sector

(33)

29

Table 4 (continued)

(34)

30

Table 4 (continued)

(1) (2) (3) (4) (5) (6) ea eman es ea eman es 2001.year_t -0.308 3.287*** -2.663 (1.900) (0.955) (1.872) 2002.year_t -0.166 2.894*** -2.410 (1.891) (0.950) (1.863) 2003.year_t -0.212 2.425*** -1.955 (1.866) (0.938) (1.838) 2004.year_t -0.334 2.145** -1.623 (1.843) (0.926) (1.816) 2005.year_t 0.247 1.860** -1.938 (1.833) (0.921) (1.805) 2006.year_t 0.557 1.593* -2.005 (1.823) (0.916) (1.795) 2007.year_t 0.367 1.419 -1.658 (1.815) (0.912) (1.788) 2008.year_t 0.214 1.019 -1.150 (1.811) (0.910) (1.783) 2009.year_t 0.212 0.421 -0.518 (1.815) (0.912) (1.788) _cons 250.041*** -58.414*** -90.037*** 426.001*** -116.614*** -203.357*** (22.604) (14.428) (20.437) (29.531) (14.838) (29.088) Obs. 1230 1230 1230 1230 1230 1230 R-squared 0.891 0.564 0.873 0.893 0.734 0.852

Year Dummies NO NO NO YES YES YES

Number of Countries 30 30 30 30 30 30

IV YES YES YES YES YES YES

Standard errors are in parenthesis *** p<0.01, ** p<0.05, * p<0.1

China Shock on Structural Transformation

In line with the result in structural transformation measured by value added, import penetration has a negative impact on manufacturing employment as seen in Table 4. Even though the significance disappears after controlling with time dummies, the sign remains negative. Different with the negative effect of China shock on real VA in service sector, the service employment is positively affected by this import, it could imply that as Chinese export rise, the decreasing in employment in manufacturing sector might shift to the higher level of productivity which is in service sector. The control variables also determine similar outcome as previously explained in structural transformation measured by real VA. It is noticeable that even though the sign is positive, the decrease in manufacturing employment (Table 4) is larger than the decline in real VA manufacturing (Table 2) which is captured by time dummies, around 6.37% (7.43-13.80) points in 1990 relative to year 19734_{for manufacturing employment and}

only around 0.30% (1.2-1.5) points for real manufacturing value added in the same period. The level of manufacturing of manufacturing employment in 1971 stood at around 15.04% points, however this figure drops drastically to around 1.59% points in 2006. This might be the fact of automation in manufacturing industries due to technology improvement over time. Most of ‘boring and repetitive’ industrial workplace was replaced by machinery to improve productivity and quality while increasing safety, and reducing human errors which eventually yields reliability, safety, as well as profitability which is important in manufacturing sectors (Autor, 2015).

(35)

31

Table 5—Quantile Regression with Endogeneity Correction Manufacturing Employment Share ALL COUNTRY (1970-2010)

Dependent Variable: Share of Employment in Manufacturing Sector

IV NO TD IV WITH TD

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

eman eman eman eman eman eman eman eman eman eman

(36)

32

Table 5 (continued)

IV NO TD IV WITH TD

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

(37)

33

Table 5 (continued)

IV NO TD IV WITH TD

Q10 Q25 Q50 Q75 Q90 Q10 Q25 Q50 Q75 Q90

y2000 0.031*** 0.028*** 0.022*** 0.023*** 0.028*** (0.000) (0.000) (0.000) (0.001) (0.001) y2001 0.026*** 0.022*** 0.021*** 0.020*** 0.028*** (0.000) (0.000) (0.000) (0.000) (0.000) y2002 0.020*** 0.019*** 0.016*** 0.018*** 0.024*** (0.000) (0.000) (0.000) (0.001) (0.001) y2003 0.013*** 0.017*** 0.013*** 0.016*** 0.023*** (0.000) (0.000) (0.000) (0.001) (0.000) y2004 0.016*** 0.015*** 0.013*** 0.015*** 0.018*** (0.000) (0.000) (0.000) (0.000) (0.000) y2005 0.012*** 0.013*** 0.014*** 0.012*** 0.019*** (0.000) (0.000) (0.000) (0.001) (0.000) y2006 0.011*** 0.014*** 0.013*** 0.012*** 0.015*** (0.000) (0.000) (0.000) (0.001) (0.001) y2007 0.014*** 0.012*** 0.013*** 0.012*** 0.011*** (0.000) (0.000) (0.000) (0.000) (0.000) y2008 0.007*** 0.008*** 0.009*** 0.013*** 0.009*** (0.000) (0.000) (0.000) (0.001) (0.001) y2009 -0.001*** 0.005*** 0.000 0.004*** 0.008*** (0.000) (0.000) (0.000) (0.001) (0.000) _cons -0.798*** -0.693*** -1.278*** -0.925*** -1.153*** -1.454*** -1.125*** -0.919*** -1.233*** -1.487*** (0.042) (0.060) (0.250) (0.026) (0.016) (0.001) (0.000) (0.003) (0.004) (0.002) Obs. 1230 1230 1230 1230 1230 1230 1230 1230 1230 1230 R-squared - - - -

Year Dummies YES YES YES YES YES YES YES YES YES YES

Number of Countries 30 30 30 30 30 30 30 30 30 30

IV YES YES YES YES YES YES YES YES YES YES

MASTER THESIS Deindustrialization over time, does China shock matter?

MASTER THESIS

Deindustrialization over time, does China shock matter?

Nugraheni Dwi Utami

June 18, 2019

Abstract

Acknowledgements

Table of Contents

List of Figures

List of Tables

I. Introduction

II. Literature Review

III. Model and Data for Analysis

TOTAL CHINA'S EXPORTS TO THE WORLD

IV. Empirical Results