Thesis M.Sc. Economic Development & Globalization
The Effect of Chinese Import Competition on Manufacturing Employment
Growth in the European Union
Student: Justin Hooimeijer s3164810 j.a.f.hooimeijer@student.rug.nl Supervisor: Dr. Laméris
Abstract
This thesis analyses the effect of Chinese import competition between 1998 and 2018 on EU15 manufacturing employment by instrumenting for changes in EU imports from China using changes in other economies’ imports from China. Ricardo’s comparative advantage serves as theoretical base for this thesis. The results suggest a positive association between China’s revealed comparative advantage and Chinese import competition in the EU. The main specification shows that increasing Chinese import competition is negatively associated to the growth rate of EU15 manufacturing employment. The results also reveal that the negative effect of Chinese import competition on EU15 manufacturing employment was weakened during the China Shock.
Key Words: Chinese import competition, international trade, EU manufacturing employment
1 1 Introduction
During the last two decades, a new economic powerhouse has emerged in the Far East that is tippling the global balance of power (Wang 2012). China joined the World Trade Organization (WTO) in 2001, paving the way for a boom in Chinese exports. Autor, Dorn & Hanson (2016) coined the term ‘China Shock’ to describe this surge of Chinese exports at the turn of the millennium. Across the West, the emergence of China as the manufacturer of the world was, at least by some parties, identified as the culprit for declining domestic employment and production. Most notably, US-President Donald Trump and his Republican Party have blamed China for decreasing domestic manufacturing, resulting in a Sino-American trade war in 2018. A priori, it seems reasonable to assume that the China Shock has affected the USA and the European Union in the same way. Yet, both popular and academic literature on the consequences of the China Shock appear to be US-centric. The elaborate academic research on the effects of increasing Chinese exports on US labour markets leads to the question of what the impact of increasing Chinese exports has been on EU labour markets.
This thesis identifies Ricardo’s theory of comparative advantage as the theoretical foundation for explaining international trade flows. Krugman (2002) argues in favour of the suitability of the Ricardian trade theory in today’s world. Eaton & Kortum (2002) quantify comparative advantage and empirically show that comparative advantages can lead to gains of trade. Changing global trade flows can increase the degree of competition a country experiences from abroad. Auer & Fischer (2010) show that imports from low-wage countries strongly affect prices in the USA, providing evidence that import competition affects domestic industries. The effect of Chinese import competition on regional US labour markets is highlighted in Autor, Dorn & Hanson (2013b). The authors use an instrumental variable approach to determine the degree of regional Chinese import competition in the US. Acemoglu et al. (2016) use input-output linkages in the US to identify the negative employment response to increasing Chinese import competition.
Although research on the effect of Chinese import competition in the EU is mostly conducted on national levels, some authors focus on the EU as a whole. Bloom et al. (2011) show how Chinese import competition affects European firms. The authors find that lower-tech firms tend to exit the market or decrease their number of employees. Flückiger & Ludwig (2015) analyse Chinese competition in export markets to determine how China’s growing manufacturing sector affects EU export-oriented manufacturing. The authors find that Chinese export competition in the EU has led to a decline in manufacturing output and, consequently, to a decrease in industry-specific employment between 1995 and 2008. Research on national levels provides more detailed analyses of the impact of Chinese import competition on the European manufacturing sectors. Malgouyres (2016) finds that Chinese import competition negatively affects wages in the French manufacturing sector. Regional import competition in Norway is associated to a decline in manufacturing employment (Balshik 2014). In Denmark, manufacturing and textile industries experienced a decline in employment, largely because of Chinese import competition (Keller & Utar 2019).
Previous research has shown how changing international trade flows can affect local labour markets and increase unemployment. Moreover, there is very strong evidence linking
2
Chinese import competition to declining manufacturing sectors in the US and in several European countries. These findings motivate research at a European level. It stands to reason that the negative effect of Chinese import competition is pervasive across the EU. Considering that the EU’s tariff and external trade policy is shaped at a supranational level, this thesis investigates the impact of Chinese import competition at a European level. The novelty of this research lies in its broader scope and focus on sectoral labour markets across nations, instead on regional labour markets within nations. Contrary to Flückiger & Ludwig (2015), Chinese import competition in the EU is analysed rather than Chinese competition in export markets. This thesis contributes to the existing literature by analysing the following research question: What is the effect of Chinese import competition on manufacturing employment in the European Union?
To provide more empirical substance to the theoretical foundation of the Ricardian theory, this thesis first investigates the effect of China’s comparative advantage on Chinese import competition in the EU using an Ordinary Least Squares (OLS) regression. The Balassa Index is constructed with UN Comtrade export data and serves as a measurement of revealed comparative advantage (Balassa 1965). Chinese import competition is based on UN Comtrade import data. Statistically significant evidence is found that a unit increase in China’s revealed comparative advantage is associated with a 0.82 percentage point increase in Chinese import competition (Table 2). This result positively links China’s industries that enjoy a comparative advantage to increases in EU imports from China. Ricardian theory predicts that countries specialize according to their comparative advantage in order to export more in these specialized industries. Hence, the finding that China’s comparative advantage is positively linked to exports growth to the EU suggests that Ricardo’s comparative advantage seems to be a useful theoretical foundation. Following Autor, Dorn & Hanson (2013b), this thesis uses a two-stage least squares (2SLS) instrumental variable approach to investigate the effect of Chinese import competition on the growth rate of EU manufacturing employment. EU manufacturing employment is based on the Eurostat Labour Force Survey (LFS). Using Chinese import competition in other developed economies as instruments allows this thesis to control for EU-specific market shocks. As such, the 2SLS regression estimates the effect of supply-driven Chinese import competition. This thesis finds empirical evidence that Chinese import competition negatively affects EU manufacturing employment. The results suggest that a percentage point increase in Chinese import competition decreases the growth rate of EU manufacturing employment by 0.047 percentage points (Table 4).
This thesis is structured as follows. Section 2 reviews the relevant literature and provides the theoretical foundation of this thesis. The research on Chinese import competition in the US and European nations is discussed and this thesis’ hypotheses are presented. Section 3 provides the methodological approach and discussed the data and data construction. Moreover, this section provides the empirical specifications and equations used in this thesis. Section 4 provides descriptive statistics, empirical results, and robustness checks. A discussion of the results and possible caveats is also provided in section 4. Section 5 concludes this thesis.
3 2 Literature Review
2.1 Theoretical Background
The two last centuries have shown a great number of theories and models on international trade. A prominent theory of international trade is the Heckscher-Ohlin model which focusses on the factor endowments of two nations. The model considers traded commodities as bundles of factors of production such as land, labour and capital (Leamer 1995). It predicts that capital abundant nations export more capital abundant goods while importing more labour abundant goods. Consequently, the model expects convergence of prices due to international trade. The model is extended by the Stolper-Samuelson Theorem with the prediction that relative wages across nations should also converge due to international trade. Despite the theoretical appeal of the Heckscher-Ohlin-Samuelson model, empirical evidence for it seems difficult to obtain. The Linder hypothesis refutes the Heckscher-Ohlin’s implication that capital abundant (i.e. wealthier) nations are expected to trade mainly with labour abundant (i.e. poorer) nations. Choi (2002) finds empirical evidence in support of the Linder hypothesis. The Stolper-Samuelson-Theorem also lacks supportive evidence. Magee & Oppenheimer (1994) find that, amongst other issues, the a priori assumptions concerning full (global) labour mobility leave any empirical proof unrealistic.
Globalizing businesses led to research of international firms and their role in trade. Melitz (2003) drops the stylized assumption that countries are the relevant players. Instead, the author investigates the role of heterogeneous firms in intra-industry trade. He finds that trade reallocates resources and labour towards more productive firms, which in turn increases aggregate industry productivity. In the “Melitz-model”, firms’ production decisions are determined by heterogeneity in terms of firm productivity. 1 Helpman (2004) complements the Melitz-model by also accounting for firms’ possible decision to engage in FDI. Alvarez & Lopez (2005) provide empirical evidence supporting the assumption that firm heterogeneity affects firms’ export decisions. The gains from trade caused by intra-industry trade are discussed in Melitz and Tefler (2012). These authors focus on gains from trade associated with firm heterogeneity and find that differences in productivity lead to gains. Although the Melitz-model (and its extensions) provide interesting new insights in the dynamics of international trade, it will not serve as a useful theoretical base for this thesis. The setup of Melitz-type models and research is different than the scope of this thesis. Instead of narrowing and focussing on intra-industry trade effects, this thesis addresses a broader perspective and aims to determine the overall industry trade effect on employment.
Despite being one of the oldest theories on international trade, David Ricardo’s theory of comparative advantage provides useful insights concerning the mechanics of international trade (Golub & Hsieh 2000). The concept is simple and can be illustrated by an example. There are two countries that both produce the same goods and do not impose trade restrictions on each other. Trade is balanced between the two nations. There is only one factor of production, namely labour, which is mobile within countries but immobile internationally. There is full employment in both nations. The model considers the case when one country does not enjoy an absolute advantage in either goods (i.e. it is less productive). Why would these countries still trade if one
1Production decisions: (1) exit the market, (2) produce for domestic market or (3) produce for both domestic and export markets
4
country can produce both goods more efficiently? Ricardo provides the reason by accounting for opportunity costs and showing that countries are better off by shifting their production to the good in which the nation achieves a lower opportunity cost than the other nation. By trading, both nations can benefit from their respective lower opportunity costs.
Although this textbook model obviously is simplified and an abstraction of reality, it allows to illustrate the strength of Ricardo’s theory. It explains the concept of opportunity cost and how it paves the way for mutually beneficial international trade. The elementary yet sturdy theory shifts international trade theory away from absolute advantages and introduces the concept of comparative advantage. Even in today’s modern globalized world, the theory provides insights in international trade patterns, particularly in situations where nations specialize without enjoying an absolute advantage.
2.1.1 The merits of the Ricardian model
During its two centuries of existence, the theory of comparative advantage has not remained unscrutinised. Many objections focus on the basic assumptions needed to construct the stylized model. Schumacher (2013) critically reviews the model’s premises and argues that the assumption concerning balanced trade is unrealistic and therefore invalid. Moreover, he also argued that the assumption of full employment leaves the reason for international trade (i.e. welfare improvements) obsolete. He challenges the assumption of full employment and argues that in cases of unemployment, nations gain more from attaining full employment and boosting domestic production while abstaining from international trade.
Krugman (2002), however, defends the theory and discusses the suitability of the model to the modern world. He argues that much of the criticism is caused by failing to understand the model and its implications. As such, Krugman (2002) points out that the assumption of full employment is a reasonable approximation if one realizes that international trade is a long-run process. He points out that economies have a self-correcting tendency towards full employment and that central banks’ policies aim towards achieving this goal. Similar arguments are made to defend the assumption of balanced trade. There is a self-correcting tendency when trade is unbalanced (i.e. an unbalanced current account) by means of the capital account. Moreover, Krugman (2002) also points out that governments and central banks can adopt certain policies (e.g. using exchange rates) to balance trade. Lastly, it is also pointed out that unbalanced current accounts are no reason for alarm.
2.1.2 Empirical support for the theory of comparative advantage
Krugman (2002) argues in favour of the suitability of the classical Ricardian theory to our modern, globalized economy. He states that the basic theorem can be used as simple explanation for contemporary trade flows that go beyond simple commodity trade. His arguments about the canonical Ricardian theory are supported by empirical findings from Golub & Hsieh (2000). The authors find that relative productivity and labour costs help explain trade patterns using data on US trade flows.
Extending Krugman’s (2002) arguments, Eaton & Kortum (2002) update the model by quantifying comparative advantage and incorporating geographical features. Moreover, their model also accounts for the importance of intermediate goods in modern trade. Quite suitable to
5
this thesis, Eaton & Kortum (2002) also focus on manufacturing by using data on bilateral trade in manufactured goods. The aim of the Eaton & Kortum model is to estimate bilateral trade flows using, amongst other variables, comparative advantage as independent variable for international trade flows. The authors measure comparative advantage by using data from 19 OECD countries on technology, research & development, years of schooling and prices as proxies for competitiveness. The paper empirically supports that comparative advantage can lead to gains from trade, but also highlights the negative effect of geographic barriers.
Chor (2010) complements the Eaton & Kortum (2002) model by investigating the interaction of country and industry characteristics in predicting industry trade flows. The author argues that industries may benefit from varying institutional conditions. As such, comparative advantage is achieved by the match of country-specific conditions and industry-specific requirements. Chor (2010, page 153) finds “strong evidence for the importance of factor endowments, financial development, legal institutions, and labour market regimes as sources of comparative advantage”. These endowments and institutional factors are summarized in the independent variable “Ricardian comparative advantage” or “Ricardian productivity”. The author finds that this proxy for comparative advantage is positively associated with the volume of industry trade flows between countries. This suggests that comparative advantages are predictors of international trade.
Eaton & Kortum (2002) and Chor (2010) use proxies such as technology and prices to measure comparative advantage. Alternatively, Costinot & Donaldson (2012) use an approach that circumvents the use of proxies. Instead of focussing on the manufacturing sector, Costinot & Donaldson (2012) use data from the agricultural sector to investigate the theory of comparative advantage. Productivity and efficiency can be measured more easily in the agricultural sector due to scientific knowledge on how essential inputs (like water) are used for outputs (such as crops). The authors test Ricardo’s theory by comparing predicted output levels to observed output levels across countries and crops. Their analysis shows that Ricardian predicted levels have significant explanatory power with respect to actual output levels. This alternative method therefore affirms the relevance of comparative advantages in explaining international trade patterns.
The previous section highlights the importance of Ricardo’s theory of comparative advantage when discussing international trade flows. It ranges from the Ricardian model to the more sophisticated models and alternative empirical methods. The empirical findings affirm the merit of the theory of comparative advantage as useful indicator of international trade flows. Both the theory and the empirical evidence for the role of comparative advantages in international trade highlight that sectoral comparative advantage is likely to be associated with the degree of import competition experienced in home markets.
2.2 Import competition and labour market effects
At the end of the 2000s, academic literature did not seem to have fully caught on to the game-changing developments caused by the emergence of China as economic powerhouse. In that period, the bulk of research on trade-induced (labour) market shocks seems to have focussed
6
on low-wage import competition in general, rather than specific Chinese import competition. In this research, it is found that high-income countries export higher-quality goods (Amiti & Khandelwal 2013) and that income countries import relatively more from other high-income countries (Hallak 2006). The effect of tariffs on the quality of export goods is also investigated. Amiti & Khandelwal (2013) find that “lower tariffs discourage quality upgrading for products distant from the [quality] frontier”. This tendency, however, does not seem to apply to China as its exports have increased in quality over time despite joining the WTO.
The overall effect of import competition from low-wage countries on prices in the US has also been investigated. Auer & Fischer (2010) investigate 325 manufacturing industries in which the U.S. imports from nine low-wage countries, including China. The authors observe that when aggregate output in low-wage countries (i.e. labour abundant countries) increases, exports to the U.S. increase in labour-intensive industries and much less so in capital-intensive industries. Hence, as emerging economies grow, they will benefit from their comparative advantage caused by abundant cheap labour and focus their exports on labour-intensive goods. This suggests that this paper’s indicators of comparative advantage in China are expected to be positively correlated with the volume of imports to the EU. Auer & Fischer (2010) find evidence that imports from low-wage countries to the US strongly affect prices in the US. They find that industries in which imports from low-wage countries capture 1% of total market share, prices experience a downwards pressure of 2.35%. This paper’s aim is to investigate the effect of import competition from China on the manufacturing sector in Europe. The results from Auer & Fischer (2010) provide first evidence for a negative association between the two factors.
2.3 The China Shock as source of import competition
During the 2010s, academic literature seemed to have grasped the role of China in trade shock that occurred since the 2000s. As such, this increase in import competition was commonly referred to as the ‘China Shock’, the massive increase of Chinese exports since it joined the World Trade Organisation (WTO) in 2001. Prior to its entry into the WTO, China already enjoyed low ‘most-favoured-nation’ tariffs from the US. 2 These tariffs, however, needed yearly approval by US congress, which is why China’s accession to the WTO significantly reduced uncertainty about the tariffs (Pierce & Schott 2016). Through integrating itself in global trade, China has been changing the balance of power in the world and has emerged as a manufacturing and trading powerhouse (Wang 2012). The ‘China Shock’ effect was amplified because imports from low-wage countries to the US and EU have historically been small (Krugman 2000). In the early 1990s, imports from China to both the EU and USA accounted for only 2% of total imports. By 2007, this value had grown to almost 11% of all EU and US imports (Bloom 2011). China has been particularly successful in gaining a comparative advantage in labour-intensive industries and to shift away from exporting basic commodities and materials towards increasingly sophisticated manufactured goods (Amiti 2010). This process of specialization in manufactured goods and leveraging its comparative advantage has been at the centre of China’s trade growth.
7
Academic literature seems to have been focussed on the effect of the China Shock on the United States, overlooking the consequences of the shock in the European Union. Research on the latter has mainly been done at national levels and has been limited to mainly Western European countries. This thesis has a broader scope and will focus on EU-wide effects of import competition from China. The US-centric academic literature on the China shock, however, can provide a useful blueprint for this thesis’ research, which is why this aspect will be reviewed first. Then, detailed national literature from the EU will be consulted for some regional preliminary results. Also, the applicability of the US approaches to the EU will be discussed.
2.3.1 The China Shock and the USA
The 2000s can arguably be summarized as an era of employment and income polarization. Autor (2015) investigates the relationship between mean wages and occupational skill percentile between 1979 and 2012. He finds that the bottom and top echelon in terms of skill-level have experienced significantly larger increases in mean wages compared to the population segment between the 20th and 80th percentile of skill level. Moreover, Autor (2015) also finds that the employment share of the population between the 25th and 75th percentile in terms of skill level has decreased between 1989 and 2012 in the US. Alvaredo et al. (2013) show that the income share of top earners has increased in much of the western world since the 1980s. These findings raise the question why both the income distribution and employment distribution can be described as U-shaped, indicating stagnation of the middle segment in terms of skill level. Autor (2015) indicates that increasing import competition has had far-reaching implications for US workers by “altering competitive conditions for US manufacturers” (Autor 2015, page 22). This thesis aims to investigate how import competition has affect the middle-class in terms of skill level. Autor, Dorn & Hanson (2013a) explore two prominent explanations for this recent trend. The authors aim to disentangle the effects that rapid technological progress and increasing international trade have had on employment. They find that regional exposure to technological change is mostly uncorrelated to regional exposure to trade competition. This finding indicates that the effect of trade on labour is pervasive and identifiable in the USA.
The findings by Autor, Dorn & Hanson (2013a) have paved the way for further research on the impact of trade. As such, the same authors also specifically have investigated the labour market effects of Chinese import competition in the US. They point out that 89 percent of growth in imports from low-wage countries to the US between 2000 and 2007 is caused by increasing imports from China. Using data on US local labour markets and import exposure between 1990 and 2007, Autor, Dorn & Hanson (2013b) find that increasing imports leads to higher unemployment in local labour markets. Local labour markets are defined by commuting zones (CZs). Industry import competition from China is measured using data on industry labour shares and change in US imports from China in that particular industry. The authors use the data to compute the yearly change in industry-specific imports per worker per region which is used as proxy for Chinese import competition. Next, Autor, Dorn & Hanson (2013b) isolate the supply-driven components of US imports from China by using an instrumental variable approach. Using data on Chinese exports to other developed countries, the authors can identify the supply-driven import competition, while accounting for the unobserved demand-driven component. The analysis shows that Chinese import competition negatively affects local labour markets through higher unemployment and declining wages. This is, of course, in line with the findings of Autor,
8
Dorn & Hanson (2013a). The methodology of isolating the supply-driven import competition will also be applied in this thesis.
After the initial findings of the strong impact of Chinese import competition on employment in the manufacturing sector in the US, academic literature increasingly investigated this phenomenon. Autor, Dorn & Hanson (2015) untangle the effect of import competition and technological progress. The authors used a similar methodology to the approach of Autor, Dorn & Hanson, 2013b regarding the data on import competition. They also account for routine-biased technological change, as both globalization and technical progress have led to changing skill demands and job polarization. The authors reaffirm that local labour markets exposed to Chinese trade competition experience decreasing employment across all major occupational groups.3 Technological progress, however, has a neutral overall effect on employment. The results show that technological progress does lead to shifts in employment type and to substantial decreases in routine-task intensive employment. The occupational shift increases unemployment in the manufacturing sector.
Alternative methods to research the effect of increasing competition from China on employment have also been published. Acemoglu et al. (2016) base their research on input-output linkages in the US to identify China’s growth and increase in import competition using data on industry imports as well as industry exports and the change of Chinese imports over time. The authors find negative employment responses and few offsetting employment gains in unaffected industries due to increasing Chinese competition. Employment in nonexposed industries are negatively affected due to spill-over effects. Pierce & Schott (2016) take yet another approach to investigate the effect of import competition. The authors use micro data on plants in the United States and base their regression on estimates of hypothetical tariffs. They investigate how the permanent low US tariffs on Chinese imports since 2001 have affected the US labour market. Pierce & Schott’s (2016) find that the increase in import competition has negatively affect employment at plant level and therefore confirm the results of previously mentioned research. The authors also investigate the effect of permanent low tariffs on Chinese imports in the EU by comparing employment before and after 1980, the year the EU granted China permanent low tariffs.4 The authors do not find evidence for similar employment responses to the US in the EU.
Research on Chinese import-competition in the United States is mainly focussed on the its implications for manufacturing employment. However, the impact of increasing import-competition can also be explored through different channels. Pierce & Schott (2018) take a novel approach and use a difference-in-difference analysis to investigate investment responses to trade liberalization and increased import competition from China. The authors find that industries exposed to import competition are associated with lower relative investment and higher establishment exits. Feenstra & Sasahara (2018) take a broader perspective and expand their analysis beyond the China Shock and manufacturing employment. The authors also investigate how increasing US exports have affected employment not only in manufacturing, but also in
3 Including managerial, professional and technical jobs.
4 Pierce & Schott (2016) refer to the European Union in 1980, while it formally was called the European Economic Community (EEC) until 1993.
9
resource industries and services. More importantly, however, the authors also investigate the effect of Chinese imports on US employment. Decreased demand for domestically produced goods has led to a decrease of 1.4 million jobs in the manufacturing sector and 0.6 million jobs in services. Despite taking a slightly different approach or emphasis, both Pierce & Schott (2018) and Feenstra & Sasahara (2018) reaffirm that increasing competition from China has negatively affected US manufacturing and, as a consequence, its employment rates.
2.3.2 The China Shock and the EU
Academic research on the China Shock in the EU has mostly been done on a national level. This thesis first investigates literature focussed on European-wide effects and then zooms in on the extensive literature on national responses to increasing Chinese import competition.
Bloom et al. (2011) explore the impact of Chinese imports on firms’ productivity and technology across 12 European countries. The authors also include non-manufacturing firms in their analysis and use data on patents, IT, R&D and productivity to achieve a broader analysis.5 Their paper shows that firms exposed to Chinese import competition but are not forced out of business, tend to upgrade their productivity and technologies. The results also show that lower-tech firms tend to exit the market and/or decrease employment, while higher-lower-tech firms stay in business. The paper therefore finds evidence that import competition from China is associated with increasing unemployment in certain sectors in Europe. Terzidis et al. (2019) conduct a meta-study and find that the manufacturing employment in high-income countries has declined since the 1980s. The paper also affirms that trade and globalization are more likely to benefit high-skilled labour and to affect lower-tech industries negatively. Lastly, the paper points out that the negative effects of increasing trade are more noticeable in the US compared to the EU. This may explain why research on import competition and trade shocks are more US-centric.
Although the papers by Bloom et al. (2011) and Terzidis et al. (2019) indicate that Chinese import competition negatively affects labour conditions in the EU, they do not focus on the European manufacturing sector. Flückiger & Ludwig (2015), however, narrow their analysis down and investigate the effect of Chinese competition on European exports and labour in the manufacturing industry. On average, manufacturing industries in Europe export 51% of their total production, meaning that shocks in exports substantially affect the overall output of industries. Flückiger & Ludwig (2015) measure Chinese export competition by using data on the home country’s trade partners. The authors compute the trade partners’ share of imports originating from China and use this fraction as measurement of export competition. The paper finds that Chinese import competition in Europe’s exports markets is strong enough to cause a decline in total manufacturing output in the home country. Moreover, the decrease in manufacturing output has also led to significant lower industry-specific employment. Despite the authors’ emphasis on export competition, this paper provides additional indicators that the China Shock has negatively affected EU manufacturing employment.
The research on EU-wide effects of the China Shock seems to confirm that increasing Chinese competition has negatively affected European manufacturing, resulting in lower output
5 Data on R&D is based on self-reporting by firms. Data on productivity is based on employment, capital, materials, wages and sales (Bloom et al. 2011)
10
and employment. Academic literature on national levels also provides insightful information on how increasing competition has affected manufacturing industries across Europe. Bugamelli et al. (2014) explore the effect of increased imports from China on Italian manufacturers’ output prices using sectoral data on (import) competition and data on firm specifications.6 The authors find that increasing imports from China are negatively associated with the growth of Italian firms’ output prices. More importantly for this thesis, however, the authors also find evidence for a contraction in employment and wages.
Comparable to Bugamelli et al. (2014), Branstetter et al. (2019) use firm-level data to explore the effect of Chinese import competition on manufacturing in Portugal. Branstetter et al. (2019) focus on manufacturing labour markets and find that firms more exposed to Chinese competition resort to letting more employees go. The authors show that between 1995 and 2000 firms primarily adjust to increasing competition through exiting the market, whereas firms’ primary adjustment mechanism in the period 2000 to 2007 was decreasing their number of employees. Donoso et al. (2014) investigate the probability of becoming unemployed as a result of increasing Chinese import competition in Spain. The authors find an increase of the probability of becoming unemployed by a value between 9% and 44%. Contrary to Bugamelli et al. (2014), the authors do not find evidence for changes in wages in Spain as a result of increasing import competition. Malgouyres (2016) complements the literature by exploring the effect of Chinese import competition on French labour markets and also accounting for spill-over effects. The author shows that not only local manufacturing labour markets are negatively affected, but also local nonmanufacturing labour markets through the spill-over effect. Malgouyres (2016) also finds that Chinese import competition negatively affects wages and increases wage polarization in France.
Research specifically on German labour markets has been more extensive. Dauth et al. (2014) investigate the effect of increasing imports from Eastern Europe and China between 1988 and 2008 on employment. The authors use the methodology from Autor, Dorn & Hanson (2013b) to estimate import competition in Germany, which reaffirms the suitability of this approach for this thesis. Due to the fact that Dauth et al. (2014) incorporate both Eastern Europe and China in their analysis and concentrate on the pre-China Shock period, China’s role may be underestimated. The authors find no significant evidence for trade exposure from China. Dauth et al. (2017) expand their analysis by including data on the service and agricultural sector and by investigating the transition process from employment in manufacturing to services. Surprisingly, the authors do not find evidence that increasing trade with Eastern Europe and China is associated with higher unemployment in manufacturing. Also, the results show that unemployed manufacturing workers do not seem to find new employment in the service sector easily. Employment in services is mainly driven by new labour market entrants. Dauth et al. (2019) delve deeper into the mechanics of German labour markets by focussing on how German workers adjust to increasing trade from Eastern Europe and China. They find that low-skilled industry-specific employees are most severely affected by increasing foreign competition. The authors conclude that Germany is a vastly different case compared to the US due to its high trade surplus
11
and the fact that Germany mainly trades with other high-income countries. This (re)affirms the need for research on the Germany and the EU in general.
Balsvik et al (2014) explore the effect of increasing Chinese import competition on labour markets in the “Nordic model”. The authors define the Nordic model as the social and economic system of Norway and the other Nordic countries. The authors reaffirm the usage of the approach by Autor, Dorn & Hanson (2013b) and expand this methodology by also including a measurement of export market exposure. Using data on Norway, Balsvik et al. (2014) find that regional import competition exposure is associated with a decrease in employment share in manufacturing. Keller & Utar (2019) complement the findings by focussing their analysis on employment in Denmark. Comparable to Autor’s (2015) and Malgouyres’ (2016) findings regarding the US and France, Keller & Utar (2019) show evidence of job polarization in Denmark in terms of wages and employment. The authors show that Danish middle-income manufacturing and textile workers experienced the strongest decline in employment in the period 2000-2009. This decline is, to a large degree, caused by rising import competition from China.
Despite the literature’s focus on national trends, an overall response to increasing Chinese import competition can be summarized. Based on the findings of the discussed European countries, it has become evident that the China Shock is likely to have negatively affected the European manufacturing labour market. Although Germany seems to stand out as least affected country, the overwhelming evidence from other EU countries motivate further research to overall EU-wide consequences of increasing Chinese competition.
2.4 Hypotheses
Prior to delving into the effect of the China Shock on manufacturing employment in the EU, this thesis takes a step back and briefly investigates the suitability of its academic foundation. Theoretically, the concept of comparative advantage seems to provide an insightful and applicable explanation for China’s sudden growth. In order to provide empirical proof for the link between comparative advantage and import competition, this thesis tests the following hypotheses.
H0a: The revealed comparative advantage per industry of China’s manufacturing industry is not associated with Chinese import competition per industry in the EU.
H1a: The revealed comparative advantage of China’s manufacturing industry is positively associated with Chinese import competition in the EU.
This thesis’ main research question is on the effect of increasing sectoral Chinese import competition on sectoral EU manufacturing labour. Based on the findings of the studies cited above, the following hypothesis is tested:
H0b: Chinese import competition per industry in the EU is not associated with manufacturing employment in the EU.
H1b: Chinese import competition per industry in the EU negatively affects manufacturing employment in the EU.
12
In addition to investigating the overall effect of Chinese import competition on EU manufacturing employment, this thesis also accounts for economic shocks. As such, the following (sub)hypotheses is tested:
H0c: The effect that Chinese import competition per industry has on EU manufacturing employment per industry is not significantly affected by the China Shock and/or Great Recession.
H1c: The effect that Chinese import competition per industry has on EU manufacturing employment per industry is larger due to China’s accession to the WTO.
H2c: The effect that Chinese import competition per industry has on EU manufacturing employment per industry is affect by the Great Recession.
3 Methodology and Data
3.1 Revealed Comparative Advantage 3.1.1 Methodology
Theoretical evidence suggests that a country’s comparative advantage is reflected in the degree of import competition that trading partners experience for that country. This thesis aims to test said theorical evidence by investigating the effect of China’s comparative advantage on Chinese imports and import competition in the EU. First, a measure of comparative advantage needs to be determined. The presented research uses the Balassa-index as measurement of comparative advantage. The Balassa index essentially is a normalized export share. Simply put, the Balassa-index is computed by comparing the share of industry i in country c’s total exports to the share of industry i in the total exports of reference nation n (van Marrewijk, 2009). The Balassa index for China is constructed according to the following equation:
𝐶𝐵𝐼𝑖,𝑡 = 𝑋𝐶,𝑖,𝑡 𝑋𝐶,𝑡 ∑𝑋𝑛,𝑖,𝑡 ∑𝑋𝑛,𝑡 (1)
where CBIit denotes the Balassa-index per industry i in China in year t. 𝑋𝐶,𝑖,𝑡 𝑋 𝐶,𝑡
⁄ is Chinese exports X in industry i as a share of total exports X from China in year t. ∑𝑋𝑛,𝑖,𝑡
∑𝑋𝑛,𝑡
⁄ is the sum of exports X in industry i from countries n as a share of the sum of total exports X from all nations n in year t. Based on this equation, this thesis computes yearly Balassa indexes for nearly 80 industries. Indices that are larger than one indicate that an industry enjoys a revealed comparative advantage (Hinloopen & van Marrewijk, 2006). The econometric analysis shows the effect of this measurement of comparative advantage on EU import competition from China.
An Ordinary Least Squares (OLS) regression is used to find evidence for a positive association between China’s comparative advantage and EU imports from China and Chinese import competition in the EU. The period between 1998 and 2018 is examined in order to include the period before China’s accession to the WTO. The effect China’s comparative advantage
13
across all industries is investigated. This means that the first model estimates the overall effect of the comparative advantage that China enjoys across all its exports is investigated. In this model, the effect of China’s Balassa index per industry per year (𝐶𝐵𝐼𝑖,𝑡 ) on the EU’s imports from China per industry i per year t (IMEU,i,t) is estimated. The aim is to investigate whether
industries in which China enjoys a high comparative advantage are positively associated with the EU industry imports from China. Additionally, the effect of China’s overall comparative advantage on other developed nations’ imports per industry i per year t from China (IMEU,i,t) is
investigated.7 Exploring the effect of China’s comparative advantage on other nations’ imports serves as robustness check.
This thesis main concern is the European manufacturing sector. Hence, a second model narrows the sample down to only the manufacturing industries outlined inthe appendix in Table A on the right-hand side. These manufacturing industries also are the industries of interest in the main analysis outlined in section 3.2. The fact that this model only accounts for China’s comparative advantage in manufacturing industries is denoted by adding a subscript m: 𝐶𝐵𝐼𝑚,𝑡 . The dependent variable is modified to only account for EU imports from China in manufacturing, which is reflected by the subscript m in the new variable IMEU,m,t. This secondary
model investigates whether Chinese manufacturing industry’s comparative advantage is positively associated with the EU imports from China per manufacturing industry.
The first and second model outlined above investigate whether there is a positive association between China’s comparative advantage and EU imports from China. This research, however, seeks to investigate the effect of Chinese import competition in the EU. As such, the effect of China’s comparative advantage on Chinese import competition in the EU is also investigated. Autor, Dorn & Hanson (2013b) examine the impact of US import competition from China on regional labour markets in the USA. The authors define import competition as change in imports from China per year per industry. This thesis will use a slightly modified measure of import competition. The variable Chinese import competition per manufacturing industry in the EU is constructed as follows:
𝐼𝐶𝐸𝑈,𝑚,𝑡 =
△𝐼𝑀𝐸𝑈,𝑚,𝑡
𝐼𝑀𝐸𝑈,𝑚,𝑡 (2)
where ICEU,m,t denotes Chinese import competition per manufacturing industry m in the EU.
∆𝐼𝑀𝐸𝑈,𝑚,𝑡 denotes the annual change in EU manufacturing imports from China in manufacturing industry m in year t. This value is then taken as share of imports from China in manufacturing industry m in year t to the EU. This thesis bases the variable ICEU,m,t on percentage changes of
imports per industry as industry sizes are not relevant when estimating the degree of competition an industry experiences. The variable ICEU,m,t can also be considered the annual growth rate of
Chinese imports to the EU.
3.1.2 Data
In order to construct the Balassa-index, data on China and a set of countries which serves as reference group is needed. The reference group consists of the top 10 goods exporters in 2018
14
and the BRIC nations.8 Considering that China is the 2018 largest goods exporter, it is sensible to use the largest goods exporters as reference group. The BRIC countries are also incorporated into the reference group to broaden the degree of economic development in the reference group and to also account for emerging economies. This reference group accounts for both the largest exporters and largest emerging economies. The data on industry exports is obtained from the UN Comtrade database and is classified according to the Standard International Trade Classification (SITC), Revision 3. Data on yearly total exports also is obtained from the UN Comtrade database. All data is in nominal US dollars and ranges from 1998 until 2018.
The data for EU imports from China is also obtained from the UN Comtrade database and classified according to the SITC, Revision 3. The values are in current US dollars and based on EU28 imports from China between 2000 and 2018.9 The data from 1997,1998 and 1999 is constructed by aggregating national data from the EU28 countries on Chinese imports. One of the challenges of researching the EU is that the European Union in itself is not constant over time. For this thesis, data on EU imports are always based on the EU28 countries for three reasons. The first reason is rather practical: there simply is more accessible data based on the EU28 countries. Secondly, it is not an option to use evolutive data (i.e. data that includes EU expansions over time), as these datasets would include sudden jumps in EU imports caused by new nations joining the EU. Thirdly, it also is more sensible from a logical and theoretical point of view. The data on EU imports from China is solely used as basis to measure import competition. This thesis in interested in the effect of Chinese import competition across the European Union, which means that the import competition that e.g. Poland experienced prior to joining the EU in 2004 is also relevant. As a result, import data on EU28 countries is deemed suitable to measure Chinese import competition in the EU.
The data on other countries’ imports from China are also obtained from the UN Comtrade database. The data is from 1997 to 2018 and in current US dollars. The data on the EU and on all other countries is classified according to the SITC, Revision 3. It includes all two-digit industry codes, which allows this thesis to investigate the overall of the Balassa index. In order to focus on manufacturing industries, a smaller scope of industries is constructed in the appendix in Table A. The list of industries is based on whether these two-digit industries classify as part of the manufacturing sector. First, the industries themself were investigated by checking the one-digit division they fell under. Then, the three- and four-one-digit subindustries of each industry were also checked for more detailed information about the products produced in each industry. The divisions and subindustries are used to identify the manufacturing industries according to their SITC classification.
The EU has developed a classification system according to economic activities called NACE.10 While SITC identifies industries according to product categories, NACE classifies industries according to the economic activity itself. This thesis also uses NACE, revision 1.1 and NACE, revision 2 to classify data according to manufacturing industries in the EU. Although the differences are very limited, Table B in the appendix provides a conversion table from NACE,
8Top exporters according to the World Bank: (China,) USA, Germany, Japan, Rep. of Korea, France, the Netherlands, Hong Kong SAR, Italy, UK and Singapore. BRIC countries: Brazil, Russia, India (and China). 9 The EU28 is the abbreviation of European Union (EU) prior to the UK’s withdrawal from the EU.
15
rev 1.1 to NACE, rev. 2. It shows that the manufacture of food and beverages (code 15 NACE, rev. 1.1) are split into two independent categories in NACE 2. The largest conversion from NACE, rev. 1.1 to NACE, rev. 2, however, concerns the manufacture of computer, electronic and optical products (code C26 NACE, rev. 2) and the manufacture of electrical equipment (code C27 NACE, rev. 2). According to NACE, rev. 1.1, these industries were split up into three independent industries (code 30, 31 and 32 NACE, rev. 1.1). Presumably due to modernisation and increasing difficulties to differentiate between the three industries, NACE, rev. 2 classifies these industries differently. By examining the subindustries and their conversion tables, this thesis has matched the two-digit industries as shown in the appendix in Table B.
Both the SITC and NACE methods of classifying industries have useful benefits. The SITC method classifies industries according to the goods produced. Hence, data on international trade is often classified according to the SITC. The NACE method, however, allows a classification based on economic activity, which also includes non-traded and intangible goods (e.g. education, health, etc.). This thesis uses data classified according to both SITC, rev. 3 and NACE, rev. 2, which is why a conversion table is needed. Table A in the appendix matches the list of SITC, rev. 3 industries classified as manufacturing industries with their corresponding NACE, rev 2 classification. Contrary to the service sector, manufactured goods are generally tangible and traded. This means that it is mostly clear which manufacturing activity (i.e. NACE) produced which manufactured good (i.e. SITC). An example of these unambiguous matches would be the manufacture of wearing apparel (code C14 NACE, rev. 2) and articles of apparel and clothing (code 84 SITC, rev. 3). Straightforward one-to-one matches between the two classification systems are also checked by examining the more detailed three-digit classifications. The three-digit descriptions provide more detailed information concerning the specifics of each classification.
When converting industry codes from SITC, rev. 3 to NACE, rev. 2, one-to-one matches between the classifications are ideal cases. The two-digit SITC, rev. 3 classification is more detailed, which meant that some SITC, rev. 3 industries were aggregated to match one NACE, rev. 2 classification. These aggregations and matches were based on the subindustries (three and four-digit) classifications and descriptions. That information provided insights into what each two-digit industries consists of. Admittedly, the matches provided in Table A in the appendix are the best approximation of a SITC-to-NACE conversion table. There is no academic literature on real, empirically proven conversion tables between the two methods of industry classification.
3.2 Import Competition from China 3.2.1 Methodology
This thesis’ main empirical analysis is on the effect of Chinese import competition on EU employment in manufacturing industries. The main hypothesis is tested using a panel regression with the variable Chinese import competition in the EU as main independent variable. The dependent variable is based on employment in manufacturing in both the EU15 and EU28. The reasons for testing the models on both EU15 and EU28 are rather practical: the available data on employment in manufacturing in the EU28 ranges from 2002 until 2018. In order to
16
account for the China Shock, data that precedes 2001 is also needed. As such, a dependent variable based on manufacturing employment in the EU15 is included. This dataset ranges from 1997 until 2018. This thesis specifies the percentage change in EU15 manufacturing employment ∆𝐿𝑖,𝑡,15 as its main dependent variable. The dependent variable can be interpreted as the annual growth rate of manufacturing employment in the EU15. The percentage change in EU28 manufacturing employment ∆𝐿𝑖,𝑡,28 is used as secondary dependent variable and serves as a robustness check for the results based on the EU15.
The panel regression is first tested to determine whether a Fixed Effects or a Random Effects regression is more appropriate. Having determined which type of model is suitable, a baseline analysis is conducted. First, only the effect of Chinese import competition and of the control variables is estimated on the growth rate in EU manufacturing employment. In the second specification, time dummies are added for the China Shock and the Great Recession to test this thesis’ sub-hypotheses. For the China Shock, a dummy is added that equals 1 for the period from 2002 until 2007, and is 0 otherwise.11 This thesis follows the approach by Pierce & Schott (2016) and Acemoglu et al. (2016) in identifying 2007 as the end of the China Shock. The Great Recession is generally identified as the period from 2007 until 2010, which is why the Great Recession dummy equals 1 for that period and 0 otherwise. The last large addition to this thesis’ empirical analysis are two interaction terms between the main independent variable Chinese import competition and the two time dummies for the China Shock and Great Recession. The interaction terms allow to identify how the China Shock and Great Recession affect the relation between Chinese import competition and the change in EU manufacturing employment. This thesis will first conduct the empirical analysis outline above using the dependent variable 𝐼𝐶𝐸𝑈,𝑚,𝑡 as outlined in equation 2.
Basing this thesis’ measurement for import competition (equation 2) on realized changes in EU imports is problematic in the main econometric analysis. In this model, the variable ICEU,m,t
is likely to be endogenous, which leads to biased and inconsistent estimates. There is an omitted variable bias, because both the variable ICEU,m,t and the dependent variable ∆𝐿𝑖,𝑡 are affected by domestic EU shocks. This means that the variable ICEU,m,t may change due to changes in overall
demand in Europe, while supply-driven Chinese competition remains unchanged. European demand shocks are correlated with this thesis’ dependent variables, as changes in overall demand would also result in changes in domestic EU production and consequently also in employment per industry. Hence, the effect of Chinese import competition may be understated as EU employment and imports could be correlated with EU demand shocks. This thesis therefore needs to differentiate between supply-driven Chinese competition and European demand fluctuations in order to identify the causal effect of changes in Chinese imports on EU employment in manufacturing.
Therefore, this thesis follows Autor, Dorn & Hanson (2013b) and instruments Chinese import competition in the EU using measurements of Chinese import competition in economies similar to the EU in terms of economic development. A good instrument should be strongly correlated to the endogenous variable (here: ICEU,m,t), while being uncorrelated to the omitted 11 China joined the WTO in December 2001, which is why the year 2001 is not counted as part of the China Shock dummy.
17
variable that is captured in the main regression’s error term (i.e. EU market conditions). The instrument should affect the dependent variable only and exclusively through the instrumented variable. This means that the instrument should not be associated to the dependent variable through other channels than the instrument’s association to the instrumented variable. This thesis’ instrumental variables fit these requirements as other nations’ imports from China are highly correlated to supply-driven Chinese competition. It is very unlikely that industries in China push their goods to certain countries or regions differently, which means that it can be assumed that the instruments are strongly correlated with the endogenous variable. It can also be assumed that there is very limited association between domestic demand shocks in the countries that serve as instruments and domestic demand shocks in the EU. As such, the instruments are uncorrelated to the error term. Chinese import competition in the set of developed countries is not considered to affect EU labour markets in ways other than through its association to the instrumented variable ICEU,m,t. Besides the association caused by supply-driven Chinese
competition, it is very unlikely that the instruments affect EU manufacturing employment through other channels.
This thesis instruments the variable ICEU,m,t using percentage changes in imports from
China to a group of similar other economies in terms of economic development.12 The instruments are measured as follows:
𝐼𝐶𝑛,𝑚,𝑡 =
△𝐼𝑀𝑛,𝑚,𝑡
𝐼𝑀𝑛,𝑚,𝑡 (3)
where ICn,m,t is measured by the relative change in other nations’ n imports in manufacturing
industry m in year t. In order to test the instruments’ relevance, this thesis uses a F-test of joint significance for the first stage regression.
The empirical analysis includes a set of control variables. The aim of these control variables is to account for observable variables that are likely to affect the dependent variable. The effect of these variables is controlled for by adding them to the regression and estimating their coefficients. Without control variables, the impact of the controls would be omitted and therefore affect the coefficients of the main independent variables. This thesis controls for GDP growth, labour force growth, manufacturing value added growth and final consumption expenditure growth. The reasons to control for GDP and labour force growth are straightforward. GDP growth shows the overall economic growth and labour force growth indicates how the supply of labour changes over time. The growth of manufacturing value added is controlled for as it serves as an indicator of manufacturing output, while final consumption expenditure growth serves as a proxy of final demand. These variables therefore control for the EU manufacturing market conditions and allows this thesis to isolate the true effect of Chinese import competition.
3.2.2 Data
The dependent variables are based on the Eurostat EU Labour Force Survey (LFS). The LFS provides two datasets. One dataset is based on the EU15 and ranges from 1997 until 2018. The other dataset is based on the EU28 and is limited to the period from 2002 until 2018. The data shows yearly employment in thousands of persons according to the EU NACE economic
18
activity classification. From 1997 until 2008, data is classified according to NACE Revision 1.1. The later period is classified according to NACE Revision 2, which means that a conversion table is needed to merge the two periods. This thesis constructs a conversion table based on the Eurostat conversion table and detailed descriptions of both revision 1.1 and revision 2 classifications. The conversion table is discussed in section 3.1.2 and is found in the appendix in Table B. The Eurostat LFS is used to construct the annual growth rate in EU15 manufacturing employment ∆𝐿𝑖,𝑡,15 and the annual growth rate in EU28 manufacturing employment ∆𝐿𝑖,𝑡,28.
The data for the main independent variable is discussed in section 3.1.2. As discussed in section 3.1.2, the data on the imports from China are classified according to the SITC, Rev. 3. Due to the fact that the variables are classified according to different classifications, a conversion table is needed. The construction of this conversion table is based on the descriptions and definitions of both the NACE and SITC classifications at a three-digit level. The conversion is discussed in more detail in section 3.1.2. Table A in the appendix provides the conversion table.
The control variables are obtained from the World Bank database and are constructed through aggregating EU national data. The variables are not on sectoral levels and measured as annual percentage changes in GDP, manufacturing value added and final consumption expenditure. Due to the fact that that the data is aggregated from different national accounts, data is deflated to constant 2010 U.S. dollars. Manufacturing value added growth is the change in net output of the manufacturing sector after subtracting all intermediate goods. Manufacturing value added growth is included as control variable to capture the annual growth or decline the EU’s manufacturing sector as a whole. Final consumption expenditure growth is the change in the market value of households’ spending on all goods and services. Consumption expenditure growth is accounted for to include a variable for consumers’ final overall demand in the EU. The data on EU labour force growth is based on people of aged 15 year or older who are currently employed and who are unemployed but looking for a job It also includes individuals that have entered the working age and are first-time job searchers. EU labour force serves as an indicator for the EU labour market developments. It is expected that a growing labour force would make is easier to hire new workers. This means that a growing labour force is expected to be reflected in growing manufacturing employment. The data is made available by the International Labour Organization (ILO).
3.3 Main Empirical Specifications 3.3.1 Revealed Comparative Advantage
This thesis’ first empirical analysis is on the relation between China’s industry Balassa indexes and EU imports from China. An Ordinary Least Squares (OLS) regression is used to analyse the relation between the two variables:
𝐼𝑀𝐸𝑈,𝑖,𝑡 = 𝛽0+ 𝛽1𝐶𝐵𝐼𝑖,𝑡+ 𝑒 (4) This specification tests the overall association between China’s Balassa indexes and EU imports from China. The simple empirical test does not aim to prove a causal effect; instead it only aims to prove a positive association between the two variables. A positive and statistically significant
19
correlation between the two variables suffices as empirical evidence for the explanatory power of China’s comparative advantage on EU imports from China. A Breusch-Pagan test is conducted to test the model for heteroscedasticity. The effect of China’s comparative advantage on other nations’ imports from China is added as a robustness check.
Due to this thesis’ focus on the EU manufacturing industry, the effect of China’s Balassa indexes on EU imports from China will also be analysed when limiting the sample to only manufacturing industries This yields the following OLS regression:
𝐼𝑀𝐸𝑈,𝑚,𝑡= 𝛽0+ 𝛽1𝐶𝐵𝐼𝑚,𝑡+ 𝑒 (5) This specification tests whether there is a positive association between China’s Balassa indexes withing the manufacturing industry and EU manufacturing imports from China.
Lastly, the effect of China’s Balassa indexes on Chinese import competition in the EU is analysed:
𝐼𝐶𝐸𝑈,𝑚,𝑡 = 𝛽0+ 𝛽1𝐶𝐵𝐼𝑚,𝑡 + 𝑒 (6) where the dependent variable is limited to Chinese import competition in the manufacturing industry in the EU. The aim is to test whether the China’s comparative advantage is not only reflected in the amount of EU manufacturing imports from China, but also in the degree of Chinese import competition experienced in the EU.
3.3.2 Import Competition from China
This thesis’ main econometric analysis is on the effect of increasing import competition from China on sectoral employment in the EU. A panel regression appears to be most suitable to the research question and dataset. It is important to specify whether these models are fixed effects or random effects regressions. The difference between the models is outlined first. A fixed effects model assumes that there is time-constant unobserved heterogeneity that affects each group of observations equally over time. The fixed effect correlates with the explanatory variables. In this thesis, a fixed effects model would mean that there is time-constant effect in each manufacturing industry that the model controls for. A random effects model assumes that the time constant-effect does not correlate with the explanatory variables. In this thesis, this would mean that the heterogeneity of each manufacturing industry does not correlate with any of the explanatory variables in all periods. A Hausman test is conducted to determine whether a Fixed Effects (FE) or Random Effects (RE) model should be applied. This method formally tests for differences in the coefficients of the time-varying independent variables (i.e. 𝐼𝐶𝐸𝑈,𝑚,𝑡).
The first empirical analysis is a baseline model without using the instrumental variable approach discussed in section 3.2.1. The baseline model will solely estimate the effect of Chinese import competition on the dependent variables ∆𝐿𝑖,𝑡based on the EU15 and EU28. The first two models are specified as follows:
∆𝐿𝑖,𝑡,15= 𝛽0+ 𝛽1𝐼𝐶𝐸𝑈,𝑚,𝑡+ 𝜷𝒏𝑿⃑⃑ + 𝑒 (7a)
20
This thesis also investigates the effect of the China Shock and of the Great Recession on (changes in) EU employment in manufacturing. The effect of these macroeconomic shocks is estimated by adding two time dummies for the period of the China Shock and the Great Recession specified in section 3.2.1. The dummy variables are ‘turned on’ during these periods. This yields the following specification:
∆𝐿𝑖,𝑡,15= 𝛽0+ 𝛽1𝐼𝐶𝐸𝑈,𝑚,𝑡+ 𝛽2𝑆 + 𝛽3𝑅 + 𝜷𝒏𝑿⃑⃑ + 𝑒 (8a)
∆𝐿𝑖,𝑡,28= 𝛽0+ 𝛽1𝐼𝐶𝐸𝑈,𝑚,𝑡+ 𝛽2𝑆 + 𝛽3𝑅 + 𝜷𝒏𝑿⃑⃑ + 𝑒 (8b) where S denotes the dummy for the China Shock and R the dummy for Great Recession. In this specification, the dummies only provide estimates for effect of the China Shock and the Great Recession on (changes in) EU employment in manufacturing. In order to estimate how these macroeconomic shocks affect (changes in) EU employment in manufacturing through Chinese import competition, an interaction term is needed:
∆𝐿𝑖,𝑡,15= 𝛽0+ 𝛽1𝐼𝐶𝐸𝑈,𝑚,𝑡+ 𝛽2𝑆 + 𝛽3𝑅 + 𝛽4𝐼𝐶𝐸𝑈,𝑚,𝑡∗ 𝑆 + 𝛽5𝐼𝐶𝐸𝑈,𝑚,𝑡∗ 𝑅 + 𝜷𝒏𝑿⃑⃑ + 𝑒 (9a)
∆𝐿𝑖,𝑡,28= 𝛽0+ 𝛽1𝐼𝐶𝐸𝑈,𝑚,𝑡+ 𝛽2𝑆 + 𝛽3𝑅 + 𝛽4𝐼𝐶𝐸𝑈,𝑚,𝑡∗ 𝑆 + 𝛽5𝐼𝐶𝐸𝑈,𝑚,𝑡∗ 𝑅 + 𝜷𝒏𝑿⃑⃑ + 𝑒 (9b) where 𝐼𝐶𝐸𝑈,𝑚,𝑡𝑆 and 𝐼𝐶𝐸𝑈,𝑚,𝑡𝑅 denote the interaction terms. The interaction terms show whether the effect of Chinese import competition on (changes in) EU employment in manufacturing is amplified of dampened through the China Shock and Great Recession. In addition to that, the set of control variables discussed in section 3.2.1. is also added to models 9a and 9b. The control variables are summarized in equations 9a and 9b as vector X.
In order to isolate the effect of Chinese import competition on European (labour) market conditions and shocks, this thesis will follow a modified approach from Autor, Dorn & Hanson (2013b) where the authors use an instrumental variable to estimate Chinese import competition in the US. The two-stage least square (2SLS) instrumental variable method is applied in this thesis because the independent variable 𝐼𝐶𝐸𝑈,𝑚,𝑡 is considered endogenous. The IV approach is focussed on a model based on EU15 manufacturing employment as this model allows for the period 1998-2018. The results based on the EU28 manufacturing employment is provided in the appendix as robustness check. The starting point of the 2SLS instrumental variable method is the same as in the fixed effects model. The model is specified according to equation 7a. Results of estimating equation 7a are expected to be biased and inconsistent as 𝐼𝐶𝐸𝑈,𝑚,𝑡 correlates with the error term e. To deal with the endogeneity, the variable 𝐼𝐶𝐸𝑈,𝑚,𝑡 is estimated using a set of instruments 𝐼𝐶𝑛,𝑚,𝑡. This yields the first-stage regression where the endogenous variable 𝐼𝐶𝐸𝑈,𝑚,𝑡 is on the left-hand side and the set of instruments 𝐼𝐶𝑛,𝑚,𝑡 and the exogenous vector X on the right-hand side:
𝐼𝐶𝐸𝑈,𝑚,𝑡 = 𝛼0+ 𝛾𝑛𝐼𝐶𝑛,𝑚,𝑡+ 𝜶𝒏𝑿⃑⃑ + 𝑢 (10) The error term u is uncorrelated to the set of instruments 𝐼𝐶𝑖,𝑛, which means that the instruments are exogenous. By estimating the first-stage model, the following fitted value is obtained:
𝐼𝐶
