MASTER THESIS
The ’China Shock’, Exports and German
Employment: A Global Input-Output-Analysis
Robin Konietzny
June 19, 2018
Abstract
This study quantifies the employment effect of the China trade shock on the German labor market, initiated by China’s WTO accession in 2001, us-ing global input-output analysis and data provided by WIOD. By applyus-ing a labor-demand driven model, it is computed that in the period from 1995-2011 additional German exports in final and intermediate goods to all trading part-ners led to an increase in labor demand of 7.72 million jobs. In the same period imports from China reduced aggregate German labor demand by 1.15 million jobs. The net effect in merchandise sectors amounts to a positive employment effect of 5.22 million jobs.
Keywords: Input-Output-Analysis, China Trade Shock, German Employment
Supervisior: Asst. Prof. Dr. Tristan Kohl, University of Groningen Co-Assessor: Prof. Dr. Krisztina Kis-Katos, University of G¨ottingen Student number: S3426017 / 11602906
Email: r.konietzny@student.rug.nl
Acknowledgements
I would like to thank my thesis supervisor Asst. Prof. Dr. Tristan Kohl of the Faculty of Economics & Business at the University of Groningen. Whenever I had questions about my research and writing he provided valuable insights and helpful comments. In addition he continues to guide and support me in my academic pursuits.
Furthermore, I am grateful to Prof. Dr. Krisztina Kis-Katos of the Faculty of Economic Sciences at the University of G¨ottingen for her work as a co-assessor.
In addition, present thesis benefited from input and comments of Prof. Bart Los of the Faculty of Economics & Business at the University of Groningen. Further, I would like to express gratitude to Prof. Dr. Marc-Andreas M¨undler of the Department of Economics at the University of California, San Diego, for sharing his data and code on the share of manufacturing employment.
Contents
Glossary iv
List of Figures v
List of Tables v
1 Introduction 1
2 Background and Literature Review 2
2.1 Employment structure and the dual role of trade integration . . . 2
2.2 Findings for the U.S. . . 7
2.3 Findings for European countries other than Germany . . . 10
2.4 Findings for Germany . . . 11
2.5 Hypotheses . . . 14
2.6 Theoretical Model. . . 15
3 Data and Methodology 15 3.1 Data Source . . . 15
3.2 Structure of WIOD . . . 16
3.3 Employment Effect of Trade Integration . . . 18
3.3.1 Employment Effect of Export Expansion . . . 18
3.3.2 Employment Effect of Import Expansion, by Assumption . . . 19
3.3.3 Employment Effect of Import Expansion, by Estimation . . . 22
3.4 Remarks & Limitations . . . 23
4 Empirical Analysis 25 4.1 Quantifying the Employment Effect of Trade Integration . . . 25
4.1.1 Employment Effect of Exports . . . 25
4.1.2 Employment Effect of Imports . . . 28
4.1.3 Net Employment Effect. . . 31
5 Conclusion 33
Bibliography 36
Glossary
ADH Autor, Dorn & Hanson.
DFS Dauth, Findeisen & Suedekum. IV Instrumental variable.
LATE Local average treatment effects. LHS Left-hand side.
R&D Research & development. RHS Right-hand side.
ROW Rest-of-the-world.
List of Figures
1 Employment Share of Manufacturing in Germany and the U.S., 1977-2013 . 4
2 Manufacturing and Services Employment in Germany, 1993-2014 . . . 5
3 China’s share of world manufacturing activity, 1990–2012 . . . 6
List of Tables
1 Employment Effect of German Exports, 1995-2011 (million workers) . . . 26
2 Decomposition of Employment Effect of German Merchandise versus Service Exports, 1995-2011 (million workers) . . . 27
3 Employment Effect of Imports from China while Adjusting German Produc-tion, 1995-2011 (million workers) . . . 29
4 Decomposition of Employment Effect of German Merchandise versus Service Imports from China, while Estimating German Production, OLS, 1995-2011 (million workers) . . . 30
5 Net Employment Effect of German Total Exports and German Imports from China, 1995-2011 (million workers) . . . 32
1
Introduction
The structure of employment is subject to a process of constant change in which jobs are created or destroyed. In advanced economies, structural change mainly manifests itself in a shift of employment from manufacturing towards the services sector (Acemoglu et al.
2016). Economists identified two main drivers of this trend, namely technological change and trade. Especially the reallocation of manufacturing production from developed economies to economies such as China has attracted the attention of the scientific community and is dis-cussed in recent publications. Declining manufacturing employment has led to a stream of economic literature that aims at identifying the channels through which jobs are affected by trade (e.g. Autor, Dorn & Hanson 2013; Dauth, Findeisen & Suedekum 2016). While trade can lead to both job creation and destruction in the domestic labor market, literature for the U.S. is mainly focused on the import channel and identifies negative employment effects.
The findings for the U.S. have to be interpreted in light of two important aspects. Firstly, existing research focuses its analysis on the employment effect of imports and the related negative employment effects of import penetration. Secondly, most approaches base their econometric strategy on trade in final goods and do not fully account for extensive backward and forward linkages through trade in intermediates. Thirdly, the U.S. trade balance is characterized by a significant trade deficit that contributes to a great extent to negative employment effects. In sum, the empirical strategies applied in contributions such as Autor, Dorn & Hanson (2013,2015), henceforth referred to asADH, limit their analytic value. While the evidence provided in these contributions add to the existing literature by evaluating regional employment effects and reactions to shocks, they should not be utilized to assess aggregate effects or to draw a conclusion on the repercussions of trade as a whole.
this event is often called the ’China trade shock’ or simply ’China shock’ in trade economics. Based on the identification of this shock Feenstra & Sasahara (2017) evaluate the impact of trade and Chinese imports on the U.S. labor market in the period 1995-2011 using a model proposed by Los, Timmer & De Vries (2015). Their labor-demand driven model makes use of data provided by the World Input-Output Database (WIOD). In order to quantify the German employment effect, this paper replicates the approach by Feenstra & Sasahara (2017) for the German economy. By conducting global input-output analysis for Germany, which has a considerably different trade structure than the U.S., this thesis contributes to the existing literature.
The remainder of this thesis is structured as follows: In section2different types of employ-ment are differentiated that will facilitate the identification of winners and losers in distinct industries. In addition, another possible driver of structural change, namely technological change, will be presented briefly. Findings for the U.S., several European economies and Ger-many will be discussed to frame the analysis of this study. Section 3 introduces the WIOD
data utilized, presents the global input-output framework and documents the econometric technique that is implemented to calculate employment effects. Section4provides the results that were computed and contrasts them with existing findings. Conclusively, section 5 sum-marizes the findings, evaluates the method used and its limitations, and makes suggestions for future research.
2
Background and Literature Review
2.1
Employment structure and the dual role of trade integration
In order to evaluate China’s impact on the domestic labor market of trading partners, a closer inspection of employment in general is necessary. The current debate that was mainly started by the economic scholars ADH, focuses on negative impacts on manufacturing employment in the U.S. (e.g. Autor, Dorn & Hanson2013,2016). The empirical method adopted byADH
builds upon the idea that an economy as a whole can be depicted as a collection of small open economies. Differences between these small economies can be exploited to evaluate the local response to a trade shock. The authors segmented the U.S. economy into several regions based on commuting zones and evaluated to what extent differences in Chinese import exposure led to different labor market impacts.
Thirdly, utilizing regional disparities and evaluating local reactions to shocks is a sound empirical strategy that is methodologically similar to difference-in-differences approaches. While this type of estimation is able to identify cross-regional disparities it has a tendency to overestimate negative regional impacts when characteristics such as local industry structure are omitted and therefore should be interpreted with caution (Muendler 2017).
Even though publications that make use of this type of empirical strategy identify impor-tant aspects of trade integration and labor market impacts, they highlight only one aspect of trade, namely the import channel, and its repercussions. It is important to evaluate and compare different contributions and findings against the backdrop of their context and to identify the channel that is inspected by specific publications. While export expansion is mainly linked to increased domestic labor demand, rising import penetration has poten-tially negative impacts on labor markets driven by the displacement of domestic production. In a global economy that is characterized by fragmentation of value chains and extensive input-output linkages, trade shocks occur in a complex system of relations and dependencies (Johnson & Noguera 2012). Specific economic methods, using various empirical approaches, can only provide answers to particular questions. Thus, when comparing the findings of various economic scholars, the research design, and econometric strategy should be consid-ered and carefully assessed. Since the employment effect depends on the trade channel, the industry, and other determinants, a conclusive approach requires a precise definition of the setup.
Employment structure Among other things, the structure of employment is of great in-terest and will provide a tool to differentiate between those that are better and worse off. Employment can generally be split into two groups, manufacturing employment and services employment. While the latter is on the rise in developed economies due to increased income and non-homothetic demand1, decreases in manufacturing employment are prominently dis-cussed in media and scientific research (e.g. Acemoglu et al. 2016; The Economist 2018b). Manufacturing is widely seen as having special characteristics2 that are used as justification for policy recommendations (Helper, Krueger & Wial 2012).
From 1995 to 2011 the U.S. experienced a decline in manufacturing employment from 17 million to circa 12 million (Bureau of Labor Statistics 2018). For Germany, the number of people employed in the manufacturing sector decreased from 8 million in 1995 to 7 mil-lion in 2010 (Statistisches Bundesamt 2017). While the loss of manufacturing employment is less pronounced in Germany, Dauth, Findeisen & Suedekum (2017) underline that a
dif-1The concept of non-homothetic demand describes the observation that the share of goods and services
expenditure varies with aggregate income. On average, individuals with higher income spend a higher share of their income on services compared to individuals with a lower income. In other words, the income elasticity of demand is unequal to one.
ferentiation between exporting and importing manufacturing sectors has to be made. The German exporting manufacturing sector was more resilient to trade integration than sectors competing with imports from other countries.
Figure 1: Employment Share of Manufacturing in Germany and the U.S., 1977-2013
0
.1
.2
.3
Employment Share of Manufacturing
1977 1983 1989 1995 2001 2007 2013 Calendar year
Germany United States
Source: Adapted from Muendler 2017.
Figure1depicts the employment share of manufacturing in total employment in Germany and the U.S. in the period 1977-2013. The German employment share of manufacturing amounted to about 30% in the mid 1980s and declined to circa 20% in the 2010s. For the U.S., the employment share of manufacturing has always been lower compared to Germany. Starting in the late 1970s, manufacturing made up about 10% of employment. Over the course of the following years, figure 1 displays a slow but steady decline that led to a share of circa 7% in 2013.
Figure 2: Manufacturing and Services Employment in Germany, 1993-2014
Source: Dauth, Findeisen & Suedekum2017.
share. The main gains in employment were realized in the services sector that rose to more than 1.1 times the level of 1993.
Drivers of structural change Firstly, technological change leads to a decreasing demand for labor since the same level of output can be produced with lower levels of labor input, due to efficiency gains. This development is empirically confirmed when looking at the share of manufacturing in real output and the manufacturing share in total employment. Baily & Bosworth (2014) looked at the two variables from 1960 to the present and found that while the share in real output remains almost constant, the share of manufacturing in total employment is declining. In this case, technological progress may partly explain a decreasing employment share.
Secondly, rising trade and further globalization lead to value chains that span borders, with a complex network of import and export relations. Declines in manufacturing employ-ment in rich countries may then be explained by increased import competition in markets for labor-intensive goods. Economies such as China are said to be relatively labor-abundant, with relatively low wage levels (Krugman2008). Thus, countries like China have a compara-tive advantage in the production of labor-intensive goods which are usually produced within the manufacturing sector. Taking the differences in endowment into account, the production of labor-intensive goods may be shifted from rich, capital-abundant high-wage countries to labor-abundant, low-wage countries (Wood 1995).
Figure 3: China’s share of world manufacturing activity, 1990–2012
Source: Autor, Dorn & Hanson2016.
available year, China’s share of world manufacturing value added amounted to almost one quarter and its manufacturing exports share was measured at circa 18%. Consequently, when analyzing the impact of trade in manufactured goods of a given economy, China’s impact on the domestic manufacturing sectors is significant in most cases.
Trade integration When a country opens up to the world market by entering the WTO, decreasing its trade barriers unilaterally or implementing trade agreements, markets of part-ner countries are affected in two ways. On the one hand, the domestic market is penetrated by foreign imports. On the other hand, an additional foreign market can be served by ex-porting final and intermediate goods. When focusing on the impact of imports, a further distinction can be made by differentiating between domestic production and consumption. Producers generally experience rising import competition while consumers have access to a broader range of product varieties, often in combination with lower prices, due to increased competition. These positive and negative effects reflect the dual role of a country’s trade integration. While individual studies underline negative impacts for the U.S., caused by increased import competition (e.g. Pierce & Schott 2016), it is crucial to acknowledge and quantify positive effects that were identified for the German labor market (e.g. Dauth, Find-eisen & Suedekum 2017) to obtain a more balanced picture. In order to evaluate the impact of trade and give substantiated policy recommendations, the underlying mechanisms have to be fully understood. In addition, it is crucial to define and analyze the channels different studies are looking at.
Neverthe-less, with further trade integration, distributions to specific actors may decrease, leading to a shrinking of individual margins. Trade then creates winners and losers and is skill- and factor-biased3 (Berman, Bound & Machin 1998; Krugman 2000). It is most important to
recognize the existence of globalization processes that create an unequal distribution of gains from trade. Policy measures have to ensure that these losers are protected accordingly and can benefit from gains of trade. Thus, before recommendations can be made, it has to be clear who is losing and who is wining. Traditional trade models such as the Ricardian or the Heckscher-Ohlin approach identify gains from trade based on differences in technology or en-dowment (Muendler 2017). Underlying assumptions of their approaches are perfect mobility and friction-less reallocation of laid-off workers. In practice, several frictions and immobility lead to a situation in which specific actors of a trading economy can be worse off. One pos-sible outcome of increased trade integration is rising wage and employment volatility due to foreign direct investment, which decreases economic security on the worker-level (Scheve & Slaughter 2004).
2.2
Findings for the U.S.
One way of analyzing the impact of China’sWTO entry is taking China and a trade partner and assessing the impact on the partner’s market. The effect of Chinese import competition on U.S. production is especially present in the economic literature as well as in public discus-sion (e.g. The Economist 2018a; The New York Times 2018). In this field of research ADH
have published a series of papers analyzing the consequences of rising Chinese import compe-tition for the U.S. labor market (Autor, Dorn & Hanson 2013,2015; Autor, Dorn, Hanson & Song 2014). They find that in the period from 1990 to 2007 an increase in import exposure distinguished for different geographical regions led to rising unemployment, decreasing labor force participation and a reduction in local labor market wages. Note that the following three contributions by ADH are built upon the analysis of import competition and its effects on labor markets. They do not evaluate employment effects of exports.
The first paper of the series, Autor, Dorn & Hanson (2013), focuses on local labor markets, defined by U.S. commuting zones. Individual commuting zones display initial differences in industry specialization. Differences in specialization, in turn, result in varying levels of im-port exposure. The authors link local imim-port exposure to changes in employment, earnings, and transfer payments. ADH find that regions specializing in industries directly compet-ing with Chinese imports experienced stronger declines in manufacturcompet-ing employment, riscompet-ing unemployment and increasing rates of labor force non-participation. The results show that specific regions seem more vulnerable to Chinese import competition than others.
Evalu-3In developed economies demand for skills is generally biased towards high-skilled labor coinciding with a
ating these effects jointly the authors estimate that they explain up to 25% of the loss of manufacturing employment within the U.S. One main limitation of this paper is due to its empirical strategy. The approach by ADH makes use of repeated cross-sections of specific regions and cannot evaluate the adjustment processes individuals experience as a reaction to trade shocks.
The following study by the Autor, Dorn, Hanson & Song (2014) builds on their 2013 paper and extends it by looking at individual workers. It complements the first study by pairing trade data with longitudinal data on individual earnings. This empirical strategy allows for a close inspection of worker-level adjustments to trade shocks. ADH identify the exposure to Chinese import competition by determining the worker’s initial industry of employment and then evaluating his life-time earnings. Individuals who experienced higher initial import exposure face lower life-time earnings, have a higher likelihood of obtaining social security payments, spend less time working for their initial employer or industry and switch out of manufacturing industries more often. Workers employed in high-wage positions that are exposed to import competition react to trade shocks by switching employers with minimal earning losses, overall mobility is higher in this case. Additionally, high-wage workers are more likely to switch to a job outside of manufacturing. Overall, labor adjustment costs differ significantly based on skill-level and initial employment. The identification strategy used in this approach assumes that Chinese exports to other economies are productivity-driven and not by changes in demand or technology. Thus, Chinese exports to countries other than the U.S. are used as an instrument to model Chinese import growth to the U.S.
In a third paper Autor, Dorn & Hanson (2015) use their approach and identification strategy to contrast the impact of trade with the influence of technological change. As already discussed earlier, economists have identified both technology and trade as two of the main determinants of labor market developments. ADH differ between routine vs. non-routine tasks and evaluate their future prospects in the light of increasing trade integration. The analysis of trade and technology effects is conducted separately since both measures are largely uncorrelated. They find that with China’s accession to the WTO in 2001 and the resulting import penetration of the U.S. labor market the impact of trade on manufacturing has become stronger while the impact of technology seems to shift from manufacturing to non-manufacturing.
The relevance of the Chinese import shock for the U.S. is echoed in further findings. Regions experiencing higher exposures to import shocks suffer an impaired provision of local public goods (Feler & Senses2017), higher mortality rates (Pierce & Schott2016) and higher rates of single motherhood and child poverty (Autor, Dorn & Hanson2017). When evaluating import exposure from a political perspective, scholars have found that the higher exposure, the more likely a Congress representative is to vote in favour of protectionist trade bills (Feigenbaum & Hall 2015). Further, the likelihood that an extreme right- or left-leaning representative is elected increases with rising import penetration (Autor, Dorn & Hanson
2016). The same paper by ADHfound that in regions with higher penetration contributions to the 2016 Trump campaign were higher.
The above-mentioned contributions mostly focus on negative effects that arise with Chi-nese trade integration. Nevertheless, a full assessment of trade and its impact on welfare needs to evaluate both positive and negative effects that result in a net impact. One can argue that the first positive effect of Chinese trade integration is a price reduction in manufactured goods. Amiti et al. (2017) estimate that China’s WTO accession reduced manufactured goods prices in the U.S. by 7.6% in the period from 2000 and 2006. Further, firms from high-wage countries, such as the U.S. or Germany, are able to offshore production to China. Off-shoring may then raise domestic productivity since workers can focus on higher value-added activities. Additionally, the quantity of intermediate goods that domestic firms can access and use is increased while prices tend to decrease (Autor 2018).
De Vries (2015), and the evaluation of both imports and exports, the global input-output framework put forward by Feenstra & Sasahara (2017) will serve as an empirical basis of this thesis.
2.3
Findings for European countries other than Germany
While the impact of the ’China trade shock’ on the U.S. labor market represents the most discussed country pair, economic scholars have also evaluated the impact of increased Chinese trade participation on other countries. This section will discuss the findings for several European countries.
Balsvik, Jensen & Salvanes (2014) evaluate local labor market effects of increased exposure to Chinese import competition for Norway. The authors use an approach that is able to differentiate between low- and high-skilled labor. They find that especially low-skilled labor is affected by increased import competition. Workers in this section tend to become unemployed or decide to quit the labor force. At the same time, Balsvik, Jensen, and Salvanes cannot verify any wage effects. All in all their approach can explain up to 10% of the decrease in manufacturing employment which corresponds to half the effect found by Autor, Dorn & Hanson (2013). Balsvik, Jensen & Salvanes (2014) argue that the reaction of the Norwegian labor market can be explained by the Nordic employment model, which combines flexibility in the labor market with an extensive social security system. In Norway, wages remained stable while manufacturing employment decreased.
Similar results are obtained when analyzing the labor market reaction in Spain. Donoso, Mart´ın & Minondo (2015) establish a connection between import competition and the prob-ability of becoming unemployed. By linking industry-level import data and employment histories of specific manufacturing workers they find that higher exposure to import compe-tition is positively associated with a rise in the individual unemployment probability. Put into numbers an increase in Chinese import competition of one standard deviation results in an increase in the unemployment probability between 0.8% and 3.5%. In accordance with the findings for Norway, the Spanish manufacturing wage level seems not to be affected by increased exposure to import competition from China.
in Chinese imports (Ashournia, Munch & Nguyen2014). Further, jobs involving high-routine tasks suffer wage losses whereas occupations with a higher content of non-routine tasks gain in terms of average wages . These findings are supported by results reported by Utar (2014). For Denmark, the author finds that in low-wage, low skilled sectors, rising trade with China increases the likelihood of unemployment and reduces expected life time earnings for affected workers. Utar underlines the importance of adjustment costs and the different speeds with which specific types of workers recover from the shock. In particular, less-educated, older workers that received a training that is specific to their task have troubles adjusting to Chinese competition.
Additional studies focusing on European countries other than Germany evaluate the rela-tionship between trade exposure and political outcomes. Malgouyres (2017) analyses French presidential elections in the light of trade shocks. The author looks at the period 1995-2012 and examines the relation between exposure to low-wage country import competition and voting outcomes of the far-right party Front National. He finds that an increase in imports-per-worker leads to an increase in the voting share of the Front National, which is statistically significant. Another recent political shock, that allows for an evaluation of the impact of global trade on political outcomes, is the Brexit voting. By using disaggregated referendum returns and individual-level data Colantone & Stanig (2016) show that in regions experiencing higher rates of exposure to Chinese imports, people tended to choose the Leave option. The authors argue that this outcome is mainly driven by labor displacement that leads to a group of globalization losers that are not effectively compensated.
2.4
Findings for Germany
The presented studies for the U.S., especially by ADH, highlight the negative impacts that came along with Chinese trade integration. It is crucial to evaluate the results in the light of the U.S. trade balance. Generally speaking, the U.S. tends to have a negative trade balance which means that the economy imports more goods than it exports. In contrast to that, the German economy tends to export more than it imports. Since China’s WTO entry in 2001, the U.S. trade balance has always been negative while the German trade balance has always been positive. For the latest year 2016, the net trade in goods and services for the U.S. amounts to a deficit of roughly 504 billion USD. Germany had a trade surplus of circa 273 billion USD in the same year (World Bank 2018).
shock, using input-output analysis.
In a series of papers the German economist Dalia Marin presents a number of reasons why the China shock may have had a different impact on the German labor market com-pared to the U.S. She gives three main reasons. Firstly, Germany was able to react to Chinese trade integration in a different manner because the economy’s import competing sectors had already adapted to competition by firms from Eastern Europe. Secondly, the ge-ographic proximity of the Eastern European economies offered attractive export opportunities for German firms. Thirdly, German export supply matched with Chinese import demand for high-quality products that could be satisfied by German supply (Marin 2010a; Marin, Schymik & Tscheke 2015). Additionally, the importance of Eastern European economies for German trade integration is captured by contributions looking at value added trade. Kaplan, Kohl & Mart´ınez-Zarzoso (2018) evaluate trade along fragmented supply-chains against the backdrop of the 2004 European Union enlargement. They find that European trade inte-gration created a sizeable number of jobs in entrant economies and at the same time had a positive impact on the labor markets of incumbents.
Dauth, Findeisen & Suedekum (2014) use an empirical approach that is similar to the one used by Autor, Dorn & Hanson (2013) to evaluate the impact of increased exposure to import competition from China. In addition, they also consider trade integration of Eastern European economies due to their importance for German trade. Their findings differ sub-stantially from the results for the U.S. Dauth et al. find that import exposure from the East, which is China and the Eastern European economies in this case, has a negative causal effect on manufacturing employment. Thus, firms that are competing with imports from the East generally experience a negative impact of trade. Nevertheless, the losses in manufacturing employment are on average balanced out by a rise in export exposure. Increases in export exposure seem to have a positive causal effect on manufacturing employment by providing new opportunities for exporting to the newly integrated markets. Taking the negative effects of import exposure and the positive effect of export exposure into account, Dauth et al. es-timate a positive aggregate effect. They come to the conclusion that the trade integration of China and the Eastern European economies preserved up to 442,000 full-time equivalent jobs in Germany over the period 1988-2008. Another interesting contribution of their study is the comparison of both the Chinese trade shock as well as the East European trade integration. The authors argue that trade with Eastern Europe had a bigger impact on local labor mar-kets in Germany than Chinese trade integration. They explain this result by differences in import structures.
some workers to leave affected industries. Many of these workers then move into industries that are less affected by imports. The relatively high mobility on the German labor market mitigates the negative impact of increased import exposure. Yet, in the long run, lifetime earnings are reduced by the shock. Even tough export oriented industries should attract workers from import-competing industries, employment transition does not occur friction-lessly. Push-effects in negatively affected industries are found to be stronger than pull-effects of positively affected industries. Optimally, there should be no difference between both ef-fects.
A third study by Dauth, Findeisen & Suedekum (2017) specifically evaluates the impact of trade on German manufacturing employment. In accordance with the findings for other developed, high-income countries the German economy experiences a structural transforma-tion of employment. While services are on the rise, manufacturing employment is decreasing. The authors assume that this general trend is mainly driven by technological change. The interesting finding of DFS is the effect of trade. In contrast to other cases like the U.S. rising trade with low-wage countries did not increase the speed of manufacturing decline but slowed it down. The authors calculate that rising trade exposure with low-wage countries has increased the probability to move into manufacturing by 0.130 percentage point for labor market entrants. For returnees out of unemployment the probability of entering the manu-facturing sector has increased by 0.252 percentage point. On aggregate, trade exposure to China and Eastern Europe secured between 128,000 and 259,000 manufacturing jobs that would have been pulled to the service, public or agricultural sector.
The analytic value of contributions such as Dauth, Findeisen & Suedekum (2014, 2016,
2.5
Hypotheses
In sum, a large body of literature has focused on negative labor market impacts in the U.S. On the one hand, low-wage, low-skilled manufacturing seems to be negatively affected. On the other hand, U.S. firms that compete on the exporting market were able to increase their production by entering the Chinese market. One main determinant of the U.S. experience is the economy’s large trade deficit with China. In total, the U.S. economy imports more than it exports in this bilateral trade relationship. In contrast to the situation in the U.S., Germany’s bilateral trade deficit is less pronounced. Additionally, the type of goods and services that are traded is different. Another factor that differentiates the German situation is the impact of Eastern European countries. After the fall of the iron curtain German firms built up relations with their East European counterparts, resulting in rising trade integration. Based on these differences we expect a different reaction of the German labor market to the Chinese trade shock.
Countries that enter trade relations with partners that are able to provide similar com-modities at labor price may experience substitution of domestic production with increased import penetration. Foreign supply of goods and services potentially replaces domestic pro-duction, reducing the amount of labor that is demanded domestically (Acemoglu et al. 2016; Autor, Dorn & Hanson 2013, 2016; Dauth, Findeisen & Suedekum 2014; Pierce & Schott
2016). This leads us to the first hypothesis:
H1: Increased exposure to Chinese imports had a negative impact on German aggregate employment.
Using an input-output approach that evaluates backward-and-forward linkages as well as deliveries of intermediates to the domestic market, suggests a positive aspect of import expansion from the perspective of the domestic market. Better and cheaper access to a broader range of intermediates reduces production costs and increases competitiveness on global markets. Further, low value-added activities can be offshored to cheaper locations enabling the headquarters to focus on high value-added activities (Marin 2010a; Marin, Schymik & Tscheke 2015). This perspective results in the following hypothesis:
H2: Increased exposure to Chinese imports had a positive impact on German aggregate employment.
New market opportunities in foreign economies are mainly taken by the most productive firms of an economy. Firms differ with respect to their productivity so that export expansion will positively affect an isolated group of firms while the least productive may be forced to exit the market (Melitz 2003). If the reallocation of employment from the least productive to the most productive firms is impaired by frictions, negative employments effect may be the result (Dauth, Findeisen & Suedekum 2016). Thus:
If an economy can seize new market opportunities, especially in case of a large foreign trading partner such as China, domestic exporting firms increase their output, generating employment. Especially, if the domestic import support matches with domestic export sup-ply, additional labor is demanded domestically to satisfy the demand for goods and services abroad. The German exporting sector that has its main competitive advantage in indus-tries such as automotive and machinery, is then the main driver of increases in employment (Dauth, Findeisen & Suedekum2014; Muendler2017). This idea is summarized in subsequent hypothesis:
H4: Rising exports to China and other trading partners had a positive impact on German aggregate employment.
The need to thoroughly study employment effects due to theoretical ambiguity can be illustrated using domestic industries importing intermediates from foreign suppliers. Allowing for trade in intermediate goods in addition to trade in final goods requires an augmentation of classical trade models. The relatively low price of imported intermediate reduces production costs. The industries themselves have the opportunity to increase their competitiveness and benefit from growing export markets. At the same time, parts of production may be offshored to lower cost locations. While the first effect leads to a positive employment effect, the latter has a negative impact. The net impact then depends on both the positive and negative effects that go along with trade integration.
2.6
Theoretical Model
The impact of Chinese trade integration on the German labor market will be evaluated using global input-output analysis. The empirical model will follow the one presented by Feenstra & Sasahara (2017). Feenstra & Sasahara rely on a demand-driven labor market model proposed by Los, Timmer & De Vries (2016). This model is able to evaluate both direct and indirect effects of trade by capturing inter-industry as well as inter-country linkages. First, the demand levels are calculated which are then used to identify the amount of labor required to satisfy the given demand, thus the term demand-driven. The chosen framework allows for a full accounting of underlying input-output linkages between industries and trading economies. Further, the impact of import competition and export expansion can be evaluated separately using trade data on the bilateral relation between Germany and China.
3
Data and Methodology
3.1
Data Source
of WIOD is utilized because it allows for a precise replication of Feenstra & Sasahara’s findings for the U.S. This precise replication will guarantee the validity of the computation for German employment and thereby increases robustness. Moreover, China’sWTOaccession took place in 2001. In order to study the employment effects of the accession, the additional information contained in the second release for the period 2012-2014 is not required.4 The
2013 release provides data on 40 different countries as well as an estimated model for the rest-of-the-world (ROW) in the period from 1995 to 2011. Further, individual economies are subdivided into 35 sectors. The basis for these tables are national accounts that are used to construct national input-output tables. In addition, international trade statistics are exploited to capture international trade in final and intermediate goods and services.
Information on employment is extracted from WIOD’s Socio Economic Accounts, using the July 2014 release. Based on the availability of data on Chinese employment, the number of persons engaged (EM P ) is utilized. Measuring employment in total hours worked by persons engaged (HEM P) would be a more precise representation of employment but the data for China is not available.
3.2
Structure of WIOD
A world input-output table (WIOT) contains information on all transactions in the global-ized, interconnected world economy. It summarizes both transactions between intermediate and final users. In general, the goods and services supplied by an individual industry are captured in a row. Columns reflect the production technology used by a specific industry. They contain intermediate deliveries per industry and the value added by the input factors labor and capital. Besides intermediate deliveries, WIOTs give information on final demand per industry. Based on an accounting identity gross output, that captures all inputs used in an industry, has to equal total use, which is the sum of all intermediate and final deliveries of an individual industry (Timmer et al. 2015).
TheWIOTas published byWIODcontain data on 40 countries and the rest of the world (N = 40 + 1) for 35 individual sectors per country (S = 35). The (N × S) × (N × S) matrix Z denotes the value of intermediate deliveries. An individual element zir,js indicates the intermediate deliveries from industry r in country i to industry s in country j. The value of final good demand is captured in the (N × S) × (N × K) matrix F . WIOD differentiates between five different categories (K) of final demand. Since this decomposition is not needed for the following calculations, industry specific final demand is summed up country-wise for final demands, resulting in a reduction of F to the dimensions (N × S) × (N ). An individual element of this matrix fir,j displays the value of final goods delivered from industry r in country i to country j. The last important element included in a WIOTis the total value of sector-specific output xir which is the sum of the value of intermediate and final deliveries.
Lastly, sub-matrix W contains information on primary inputs such as labor and capital. These primary inputs are summarized in row vectors of the dimension (N × S) × (1). The primary input vector of interest for this analysis is the employment vector, denoted by h.5
Within the input-output framework the following equation has to hold: xir = X s X j zir,js+ X j fir,j. (1)
Simplifying equation (1) by dropping the country and sector subscripts yields:6
x = Z + f . (2)
In order to evaluate the input that is needed to produce one unit of value in gross output, input coefficients are calculated. They are defined as:
air,js≡ zir,js
xjs
. (3)
The global input coefficient matrix is calculated as follows:7
A ≡ Z ˆx−1. (4)
Dividing intermediate deliveries by gross output for all elements of the intermediate input matrix Z results in the global input coefficient matrix A that has the dimensions (N × S) × (N × S). Using the reduced form of the final demand matrix F by summing row-wise over each industry yields the final demand vector f .
Solving equation (4) for Z and inserting this definition of Z into equation (2) yields:
x = Ax + f . (5)
Rewriting equation (5) we obtain:8
x = (I − A)−1f . (6)
In a next step, we define M ≡ (I − A)−1 which describes the Leontief inverse. The Leontief inverse captures both the direct absorption of final goods as well as all stages of intermediate production that are needed to satisfy final demand (Leontief 1936).
The last piece of information needed to evaluate the impact of the ’China trade shock’ on the German labor market using a global input-output framework is employment data.
5For a schematic outline of a world input-output table see tableA.
6Henceforth italicized lower case letters will indicate scalars, bold lower case letters will indicate vectors
and bold upper case letters will indicate matrices. For matrix operations and manipulations country and industry sub-scripts will be included when needed.
7The use of a hat indicates a diagonal matrix.
Analogously to the calculation of the input coefficient we are interested in the amount of labor that is required to produce one unit of value in gross output. Information on labor is summed up in the (N × S) × (1) employment vector h. The employment vector provides the number of persons engaged in a specific industry. Labor input coefficients are obtained as follows:
b ≡ h0xˆ−1. (7)
Finally, labor demand for a given final demand can be calculated by:
L = ˆbx = ˆb(I − A)−1f . (8)
Starting from equation (9) superscripts will be added to indicate the year t.
3.3
Employment Effect of Trade Integration
3.3.1 Employment Effect of Export Expansion
The following approach is based on initial work by Los, Timmer & De Vries (2015) and adapts the methodology of Feenstra & Sasahara (2017) for the German economy.
Final Demand Additional labor demand due to German export expansion based on changes in the level of exports of final products is quantified using:9
˜
L11,X1 = ˆb11(I − A11)−1f11− ˆb11(I − A11)−1f˜11,X. (9) Equation (9) is used to solve for the additional labor that is required to satisfy final demand due to increased exports. The first term on the right-hand side (RHS) captures the actual employment in 2011. The second term describes the counterfactual situation in which German exports to other economies remained at the 1995 level. The resulting differential between the first and the second term yields the employment effect generated by increased imports in the period from 1995 to 2011. The differential quantifies the creation of employment, measured in number of persons engaged, if ˜L11,X1 > 0, or jobs that are lost and no longer needed, in case ˜L11,X1 < 0.
The crucial modification of the final demand vector ˜f11,X is conducted in the rows within the final demand matrix that describe the consumption of German final goods in all countries except Germany itself. Foreign demand for German final production is substituted by the demand levels of 1995. German demand for German final production remains at the level of 2011. This modification is accomplished using the vector manipulations defined in equation
A1.
9Next to the year superscript, X will be added to indicate modifications to evaluate the effects of export
Final & Intermediate Demand The next specification extends equation (9) by addi-tionally incorporating export expansion in intermediates:
˜
L11,X2= ˆb11(I − A11)−1f11− ˆb11(I − ˜A11,X)−1f˜11,X. (10) Similar to the procedure that was applied to the final demand vector, the input coefficient matrix A is modified. Intermediate demand levels for German intermediates in all countries except Germany are set to the levels of 1995. German demand for German intermediates remains at the level of 2011. The modified input coefficient matrix ˜A11,X is defined by equation A2.
The two labor differential vectors for trade in final goods, ˜L11,X1, and trade in final and intermediate goods, ˜L11,X2, contain the employment effects over all sectors and countries with dimensions (N × S) × (1). The sub-vector relevant to our analysis, ˜L11,XDEU, describes the aggregate effect of export expansion on the German labor market. In order to evaluate the impact of trade on the resource (primary), manufacturing (secondary) and services (tertiary) sector, the employment effect is decomposed using the following aggregation:
˜ L11,X,ResDEU = 3 X s=1 ˜
L11,XDEU, L˜11,X,M anuDEU = 16 X
s=4 ˜
L11,XDEU, L˜11,X,ServDEU = 35 X
s=17 ˜
L11,XDEU. (11)
The three specifications defined in equation (11) summarize the resource sector (WIOD
sectors 1-3), the manufacturing sector (WIOD sectors 4-16) and the services sector (WIOD
sectors 17-35). The aggregate effect on the German labor market is obtained by summing over WIOD sectors 1-35. Throughout this paper, the decomposition of employment effects will be based on this classification.10
3.3.2 Employment Effect of Import Expansion, by Assumption
Final demand Corresponding to the technique used in section3.3.1the employment effect of import expansion can be computed by adapting the underlying final and intermediate demands to capture the impact of the ’China trade shock’ on German labor demand. Firstly, the employment effect of import expansion for final products is calculated by:
˜
L11,M 1 = ˆb11(I − A11)−1f11− ˆb11(I − A11)−1f˜11,M. (12) Equation (10) corrects for German exports, setting the demand for final and intermediate goods to to the levels of 1995. This is achieved by manipulating the elements of A and f that contain information on foreign demand for German production. The analysis of the bilateral trade relation between Germany and China requires a manipulation of both the German demand for German production and the German demand for Chinese production. This
adaption is necessary due to possible import substitution. If countries trade commodities of a similar type domestic production can potentially be substituted by foreign imports. The counterfactual situation is established by reducing Chinese imports for final demand purposes to the levels of 1995.
In the final demand vector, ˜f11,M, foreign final demand for Chinese production remains at the levels of 2011 while German final demand for Chinese production is set to the lev-els of 1995. In addition, German domestic production for German domestic final demand,
˜ f11
DEU,DEU, is adapted. Using the counterfactual model approach, we now have to correct for decreased Chinese imports that are substituted by German production. Since the ’pass-through’ parameter11 from Chinese imports to German domestic production is unknown, we have to make assumptions about the German production that would have occurred if Chinese imports remained at the 1995 levels. In addition to making assumptions about the amount of German production and the related market share, hypothetical domestic market shares of domestic producers can also be estimated using regression analysis. This type of estimation approach will be applied in section 3.3.3 but first we evaluate the employment effect based on assumptions.
Equation A3 specifies the manipulations needed to address a reduction in imports from China while correcting for domestic production.
In a next step, the hypothetical domestic final demand for domestic production is cal-culated, assuming three different functional forms. The first functional form, defined in equation (13), assumes that the there is perfect substitution between Chinese imports and German domestic production. The difference between Chinese imports in 2011 and 1995 is added to German production for domestic final demand.
Functional form 1: f˜DEU,DEU11 = fDEU,DEU11 + [fCHN,DEU11 − f95
CHN,DEU]. (13)
Functional form 2, defined in (14), assumes a proportional relationship between domestic production and total German final demand. German domestic production in 2011 is com-puted by calculating the market share of German production in German final demand in 1995 and multiplying the market share in 1995 with total German final demand in 2011. This functional form has the characteristic that imports from countries other than China are incorporated in the formula, being accounted for in the denominator PN
i=1f 95 i,DEU.
Functional form 2: f˜DEU,DEU11 = f 95 DEU,DEU PN i=1f 95 i,DEU × N X i=1 fi,DEU11 . (14)
The third specification, functional form 3 defined in equation (15), is a further adaption of functional form 2. Within the share term in the denominator only imports form China
11The term ’pass-through’ parameter refers to the substitutional relationship between foreign imports and
remain at the 1995 levels while imports from economies other than China are evaluated at the 2011 levels. The resulting market share of German producers is multiplied with total German final demand in 2011.
Functional form 3: f˜DEU,DEU11 = f 11 DEU,DEU f95 CHN,DEU+ PN i6=CHNf 11 i,DEU × N X i=1 fi,DEU11 . (15)
Final & Intermediate Demand Secondly, the employment effect of import expansion incorporating both changes in final and intermediate goods is computed by:
˜
L11,M 2= ˆb11(I − A11)−1f11− ˆb11(I − ˜A11,M)−1f˜11,M. (16) In order to incorporate changes in intermediates deliveries into the analysis of the em-ployment effect of import expansion, the global input-coefficient matrix A11 is modified. The sub-matrices that are of relevance are Chinese intermediate production for German interme-diate demand, ˜A11
CHN,DEU, and German intermediate production for German intermediate demand, ˜A11
DEU,DEU. ˜A11,M is defined inA4.
Analogously to the modification of the final demand matrix, Chinese intermediate pro-duction for German final demand is set to the levels of 1995. It is important to note that Chinese production for the sub-matrix ˜A11CHN,DEU in 1995 is evaluated at the output levels of 2011:
˜
A11CHN,DEU ≡ ZCHN,DEU95 (ˆx11DEU)−1. (17)
Since German intermediate production will substitute for decreased Chinese intermediate imports, German intermediate production has to be adapted. At first, this is accomplished by assuming three different functional forms and in section 3.3.3 by estimating German producers’ domestic market share.
˜
A11DEU,DEU ≡ ˜ZDEU,DEU11 (ˆx11DEU)−1. (18) The interpretation of the three functional forms of ˜ZDEU,DEU11 is equivalent to the interpre-tation of the three functional forms of German final demand for German production specified in equations (13), (14) & (15).
Functional form 1: Z˜DEU,DEU11 = ZDEU,DEU11 + [ZCHN,DEU11 − Z95
CHN,DEU]. (19)
Functional form 2: Z˜DEU,DEU11 = Z 95 DEU,DEU PN i=1Z 95 i,DEU × N X i=1 Zi,DEU11 . (20)
3.3.3 Employment Effect of Import Expansion, by Estimation
In section 3.3.2we made assumptions on the relationship between Chinese imports and Ger-man production for domestic deGer-mand purposes. This was accomplished by defining three functional forms that differ with respect to the substitution reaction of German producers. In this section, we will use the annual country-sector-specific data to estimate the relation-ship between Chinese imports and domestic producers market share. Market shares will be estimated for both final and intermediate production.
Final demand The change in labor demand, based on import expansion and the estimation of German production for domestic final consumption purposes is computed using the labor differential equation:
ˇ
L11,M 1= ˆb11(I − A11)−1f11− ˆb11(I − A11)−1fˇ11. (22) The final demand vector that comprises the estimation of ˇf11 is built as defined in A5. Firstly, we specify the market share estimation of German production for domestic final demand purposes as:
f(DEU,r),DEU11 PN i=1f(i,r),DEU11 = αs0+ α1 f(CHN,r),DEU11 PN i=1f(i,r),DEU11 + u11s . (23)
The left-hand side (LHS) of the regression model specified in equation (23) defines the dependent variable, the market share of German producers in the domestic final good market, f11
(DEU,r),DEU/ PN
i=1f 11
(i,r),DEU. The first term on the RHS denotes sector fixed effects. The second term, α1, is the ’pass-through’ parameter we are interested in. The market share of Chinese imports on the German final good market serves as the independent variable.
Using the estimated ’pass-through’ parameter, estimated sector fixed effect and a hy-pothetical Chinese import market share based on the imports levels of 1995, the German producers’ domestic market share is predicted as follows:
ˇ f share11(DEU,r),(DEU )≡ ˇαs0+ ˇα1 f95 (CHN,r),DEU f95 (CHN,r),DEU + PN i6=CHNf 11 (i,s),DEU + ˇu11s . (24) In a final step the predicted market share is used to calculate the counterfactual German production for German final demand purposes:
ˇ
f(DEU,r),(DEU )11 ≡ ˇf share11(DEU,r),(DEU ) N X
i=1
f(i,r),DEU11 . (25)
Intermediate demand For intermediate deliveries labor demand is computed using (28), (29) and the adapted input-output matrix A6.
ˇ
ˇ
ADEU,DEU11 ≡ ˇZ11DEU,DEU(ˆxDEU11 )−1. (27) Equivalent to the procedure applied to final demand purposes, intermediate production in Germany is estimated by relating Chinese intermediate imports and German producers’ domestic market share. While equation (23) estimates the ’pass-through’ parameter for final goods only, the following regression model estimates intermediate goods market ’pass-through’ parameters for the first 34 sectors as defined by WIOD:12
z11 (DEU,r),DEU PN i=1z 11 (i,r),DEU = αr,s0 + αs1 z 11 (CHN,r),DEU PN i=1z 11 (r,s),DEU + u11r,s. (28)
Using the estimation results, predicted hypothetical market shares can be calculated for individual sectors as follows:
ˇ zshare11(DEU,r),(DEU,s) ≡ ˇαr,s0 + ˇαs1 z 95 (CHN,r),(DEU,s) z95 (CHN,r),(DEU,s)+ PN i6=CHNz 11 (i,r),(DEU,s) + ˇu11r,s. (29)
Finally, these market shares are used to calculate hypothetical German production for German intermediate demand purposes:
ˇ
z(DEU,r),(DEU,s)11 ≡ ˇzshare11(DEU,r),(DEU,s) N X
i=1
z(i,r),(DEU,s)11 . (30)
3.4
Remarks & Limitations
The limitations of approaches such as Dauth, Findeisen & Suedekum (2014) or Autor, Dorn & Hanson (2013) are attributable to their empirical strategies. Muendler (2017) identifies two main constraints, the first one related to the instrumental variable (IV) strategy and the second one to difference-in-differences estimation. Firstly, the instrument that is used by Dauth, Findeisen & Suedekum (2014) and Autor, Dorn & Hanson (2013) is Chinese trade with other high-income countries other than the country of interest. This is a valid and clever identification strategy, when we assume that Chinese trade relations depict conditions internal to the Chinese economy, such as productivity. Secondly, when looking at local labor markets or individual regions, the resulting estimation captures so called local average treatment effects (LATE). It is important to realize that these estimations describe local impacts and should not be used to evaluate the general average effect. In this context, a LATE IV estimation is able to isolate the impact of the Chinese trade shock given the conditions that were identified earlier. Nevertheless, this kind of empirical approach can lack external validity. This lack of validity is due to the context-specific nature of the estimator. Local determinants such as productivity, industry heterogeneity or labor market flexibility
12Regression analysis is not conducted forWIOD sector 35 since the required information is not given in
cannot be segregated from the causal effect of the China trade shock. Predicting future developments is specifically problematic when using LATE IV estimation (Muendler 2017).
Consequently, input-output approaches enrich the academic discourse by describing com-plex inter-industry and inter-economy relations. Modelling the entire global economy and its interconnections, effects for single economies can be derived and used to capture ag-gregate effects. As opposed to LATE estimations these models allow for the estimation of economy-wide effects. Further, input-output models capture the pattern of specialization that is contained in global supply chains.
Although the approach presented in this paper is a powerful tool to assess a globalized world economy and its impact on local labor markets, it has some limitations and the results should be interpreted in light of the assumptions made.
Firstly, it is important to note that the approach by Feenstra & Sasahara (2017), that is adapted in this paper, computes hypothetical labor demand levels. A general equilibrium is not derived. Thus, labor demand may increase or decrease but whether the demand is met by a supply that leads to market clearing cannot be discussed. Extending input-output analysis based on quantitative general equilibrium models is one option to integrate an equilibrium condition.
Secondly, as a consequence of country- and industry-aggregation, input-output analysis can compute aggregate effects as well as differentiate between industries. Regional differences and local labor market impacts should be assessed using LATE IV estimation as described by Muendler (2017) and implemented by Autor, Dorn & Hanson (2013) or Dauth, Findeisen & Suedekum (2017).
Thirdly, the functional forms defined in equations (13)-(15) & (19)-(21) rest upon certain assumptions. The underlying assumptions can and always should be subject to discussion. The functional forms as proposed by Feenstra & Sasahara (2017) determine specific substi-tutional effects that are initiated when trade with China is evaluated in a counterfactual sit-uation. These definitions are used to estimate German domestic production. It is disputable whether the substitutional relationship between Chinese imports and German production is as direct as assumed in the sense that German demand may have been satisfied by other imports. Particularly in light of Germany’s trade relations with Eastern Europe that were already in place and use before the significant increase of trade with China, the functional forms are arguably not as sensitive as they should be.
export of medium- to high-quality products has the potential to displace demand for Ger-man exports in third markets, potentially resulting in negative domestic employment effects (Garc´ıa-Herrero, Marbach & Xu 2018). In sum, the sensitivity of the functional forms could be increased by accounting for substitutional sourcing relationships outside the country-dyad that is analyzed.
Fifthly, the July 2014 release ofWIOD’s Socio Economic Accounts does not contain data on the the numbers of hours worked for China. While information in hourly employment is available for Germany, a more precise estimate of employment effects should be constructed using more detailed employment data. Future releases of WIOD promise to increase the specificity of labor used in the production of final and intermediate goods and services.
4
Empirical Analysis
While section3describes the empirical techniques that are used to derive the employment ef-fect of trade on the German labor market, this section will evaluate the findings and integrate the resulting effects into existing literature.
4.1
Quantifying the Employment Effect of Trade Integration
4.1.1 Employment Effect of Exports
Employment effects can be the outcome of both export expansion and increased import penetration. The first part of this section is used to closely inspect the findings for German export expansion in the period 1995-2011.
The results from (9) and (10) are presented in Table 1. Through final good exports only 2.88 million jobs were created in the period from 1995-2011, corresponding to 7.7% of total employment in 1995. Of those 2.88 million jobs 1.13 million were added in the manufacturing sector and 1.46 million in the services sector. For manufacturing, only this equates to a 15.1% increase in manufacturing jobs relative to the level of 1995. Taking into account both final and intermediate exports form Germany to trading partners, a total of 7.72 million jobs were created, corresponding to 20.5% of the total employment level of 37.6 million in 1995. Aggregating manufacturing and services sector we see that those industries added 3.03 million jobs and 4.24 million jobs to total labor demand. Table 1indicates that exporting industries experienced a large positive employment effect within the chosen time period. The finding in itself is not surprising, based on the export orientation of the German economy. As opposed to the U.S., the German trade balance is positive, driving increased demand for labor.
Table 1: Employment Effect of German Exports, 1995-2011 (million workers) Through final good
exports only
Through final good and intermediate exports Employment in 1995 Manufacturing 1.13 3.03 7.5 (15.1%) (40.4%) Resource 0.30 0.46 2.2 (13.6%) (20.9%) Services 1.46 4.24 27.9 (5.2%) (15.2%) All sectors 2.88 7.72 37.6 (7.7%) (20.5%)
Notes: The employment effect, measured in millions of persons employed, is given by the numbers without parentheses. Positive numbers indicate increased labor demand while negative numbers indicate reduced labor demand. The ratio of the individual employment effect relative to the total employment in the benchmark year 1995 is stated in parentheses. WIOD sectors are aggregated using sector definitions given in (11). The last column states the total employment for the three broad sector definitions while the last row reports the sum of employment effects in all 35 sectors.
sectors only. Table 2 summarizes the numbers for exports from merchandise and services sectors separately. Almost the entire amount of additional labor demand due to final and intermediate exports in the manufacturing sector originates from merchandise sectors. For services, the calculations show that from a total of 4.24 million additional jobs 2.73 million arise in consequence of exports from merchandise sectors, while 1.51 million jobs were created by exports from services sectors. Additional demand for employment is driven by trade in merchandise sectors while added labor demand in services relies both on exports from services and merchandise sectors. In accordance with the approach by Feenstra & Sasahara (2017) the effect of interest is the direct effect of exports from merchandise sectors. Thus, column 3 of Table 2is used to calculate net employment effects at a later stage.
highlighted when looking at the ratio of the employment effect relative to total employment in the benchmark year 1995. From 1995-2011 exports alone generated additional employment which is equivalent to 20.5% of German employment in 1995. U.S. exports led to employment growth that corresponds to only 4.9%.
Table 2: Decomposition of Employment Effect of German Merchandise versus Service Exports, 1995-2011 (million workers)
Impact of final good and intermediate
exports from all sectors
Impact of final good and intermediate
exports from merchandise sectors
Impact of final good and intermediate exports from service
sectors
Manufacturing 3.03 3.00 0.026
Resource 0.46 0.45 0.006
Services 4.24 2.73 1.51
All sectors 7.72 6.19 1.54
Notes: The employment effect is measured in millions of persons employed. Positive numbers indicate increased labor demand while negative numbers indicate reduced labor demand. See Table A1 for the sector definitions.
This first finding is mainly driven by the pronounced trade surplus of Germany. Naturally, if you produce more domestically than you consume as an economy, you can export to other countries and thereby increase your output, generating employment within the economy. Another important driver of this outcome is the country-specific trade structure. Especially export-oriented industries, that specialized in penetrating foreign markets, were the main winners of export expansion. By serving new markets and increasing the amount of products and services sold to other markets these industries experienced strong employment gains and were able to reduce unemployment (Muendler 2017). Further, Marin, Schymik & Tscheke (2015) argue that the fundamental determinant of Germany’s success on the world market was based on its organizational performance. German producers fragmented their production and offshored parts to Eastern Europe or China. By doing so, the economy was able to remain competitive in spite of rising wages and a stronger EURO. In addition, after offshoring low value-added activities abroad, German exporting firms were able to focus on research and development (R&D), improve their organizational structure and increase product quality even more. Germany’s generally high product quality matched with Chinese demand for high-quality products especially in the automotive and machinery sectors. Thus, the German export supply found a compatible counterpart in Chinese import demand (Marin 2017).
4.1.2 Employment Effect of Imports
The second channel leading to labor market reactions is the import channel. Through in-creased import of goods and services competing with output generated by domestic firms, negative labor market impacts are expected. An extensive evaluation of the impact of trade on German employment requires a quantification of the jobs that were lost due to this mech-anism.
It is crucial to underline that the paper by Feenstra & Sasahara (2017) equates the net employment effect by contrasting German total exports with German imports from China. This approach is also used in present analysis. The following manipulations of sub-matrices evaluate the bilateral trade relationship between Germany and China to capture the effects of Chinese import penetration on the German labor market. Consequently, data adaptions are conducted for this country pair. As pointed out by Feenstra & Sasahara (2017) an analysis of the employment effect of import expansion that does not correct for domestic production leads to the counter-intuitive effect that imports from China may lead to a positive employment effect on the domestic market. Using extensions taken from Johnson & Noguera (2012) Feenstra & Sasahara show that omitting domestic demand leads to misleading results. Since the hypothetical domestic market share of domestic producers cannot be observed it has to be estimated using different methods. The first method makes assumptions about the production that would have occurred if China had not started to serve the German demand. The second method makes use of the information contained in the time series 1995-2011 and predicts domestic production.
Imports, by assumption The analysis of the employment impact of Chinese imports on the German labor market based on three different substitutional assumptions is summarized in Table 3. Computing the employment effect of imports using the three distinct functional forms specified in equations (13)-(15) for final demand and equations (19)-(21) for interme-diate leads to three panels. Equivalent to the findings of Feenstra & Sasahara (2017) panel B predicts the largest impact on employment while panel C leads to the smallest demand reduction.
Functional form 1, assuming that increased German production perfectly compensates for reduced Chinese imports, predicts that through final good imports only 380,000 thousand jobs were lost in the period 1995-2011. The reduction in labor demand by 380,000 jobs was due to 200,000 lost in manufacturing sectors and 170,000 in services. Evaluating both final and intermediate imports, the total loss increases to 870,000 of which 410,000 jobs were lost in the manufacturing sector and 440,000 in services.
Table 3: Employment Effect of Imports from China while Adjusting German Production, 1995-2011 (million workers)
Panel A: Functional form 1 Through final good imports
only
Through final good and intermediate imports
Manufacturing -0.20 -0.41
Resource -0.015 -0.022
Services -0.17 -0.44
All sectors -0.38 -0.87
Panel B: Functional form 2 Through final good imports
only
Through final good and intermediate imports
Manufacturing -0.48 -1.41
Resource -0.16 -0.44
Services -0.72 -2.06
All sectors -1.36 -3.91
Panel C: Functional form 3 Through final good imports
only
Through final good and intermediate imports
Manufacturing -0.062 -0.15
Resource -0.008 -0.012
Services -0.067 -0.22
All sectors -0.137 -0.38
Notes: The employment effect is measured in millions of persons employed. Positive numbers indicate increased labor demand while negative numbers indicate reduced labor demand. See Table A1 for the sector definitions.
sector. Additionally taking into account the impact of intermediate goods, the predicted reduction in labor demand amounts to 3.91 million jobs, with 2.06 million lost in services and 1.41 million lost in manufacturing.
Functional form 3, that only holds constant the Chinese market share from 1995 and that evaluates the remaining shares at the levels of 2011, quantifies the total number of jobs lost due to Chinese imports as 137,000 for final demand only and 380,000 for both final and intermediate demand.
