Experiencing Globalization in Germany: The Implications of Trade for Job Satisfaction and Job Security

(1)

Robin Konietzny Research Master Thesis

January 31, 2020

Abstract

This paper adds to the discussion on the repercussions of economic globalization by an-alyzing the effect of import exposure on non-pecuniary aspects of employment. I study the effect of occupation-level import exposure on job satisfaction and job security worry in Germany using individual-level data from the German Socio-Economic Panel. To this end, I propose two channels through which imports affect workers, a competition-increasing supply and a competitiveness-enhancing use channel. I find that rising exposure to Chi-nese imports reduces job security by increasing competition in final and intermediate goods through the supply channel. The negative effect of import exposure is mainly concentrated on individuals in occupations that typically require lower skill-levels and less specialization. The effect on job satisfaction through both the supply and use channel is only moderate. My results augment the growing literature on the labor market repercussions of trade as well as the literature on the effect of trade on political outcomes. In this context, I provide an additional explanation for the anti-globalization backlash recently observed in developed economies.

Keywords: Import competition, job satisfaction, job security, occupations Supervisors: Prof. Dr. Bart Los, University of Groningen

Asst. Prof. Dr. Milena Nikolova, University of Groningen Student number: S3426017

Email: r.konietzny@rug.nl

Study program: Research Master in Economics & Business Faculty: Economics & Business

(2)

1 Introduction

A recent survey indicates that the acceptance of globalization paradoxically reaches its limits in countries profiting the most from globalization in the last decades. For Germany, the country of interest for this paper, almost two out of three economic experts agree that the wider population is no longer in favor of globalization (ifo Institute 2019).

In order to identify the potential drivers of this backlash, one should differentiate between several dimensions of globalization, such as economic, social or political globalization. This paper focuses on economic globalization. I contribute to the discussion on economic globalization, by exploring novel channels through which increased import penetration affects individuals. To this end, I analyze the non-pecuniary consequences of trade integration using import exposure at the occupation-level. Specifically, I focus on changes in job satisfaction and perceived job security which capture effects that go beyond traditional mechanisms such as wage changes or lay-offs. Thus, I answer the following research question: How do job satisfaction and the perception of job security change if an occupation faces higher degrees of import exposure?

My contribution is threefold. First, this paper is the first to analyze job satisfaction and worry about job security in a framework that differentiates between a use and supply channel of import exposure. As use imports, I define inter-industry imports in intermediate goods. As supply imports, I define intra-industry imports in both final and intermediate goods. Second, in contrast to previous literature focusing on industry-level exposure to imports, I explicitly consider import penetration at the occupation-level. Third, I differentiate between imports from Eastern Europe and China to Germany to compare the trade integration process with Eastern Europe with the effects stemming from Chinese trade integration.

(3)

intermediate products provided by foreign suppliers. Using a cheaper and broader variety of inter-mediate inputs should result in a competitiveness-enhancing effect at the industry-level. Increased competitiveness, in turn, should affect non-pecuniary labor market outcomes. Thus, I should be able to identify effects even for occupations that are associated with industries mainly producing nontradable goods.

In addition to differentiating between a supply and use import channel, I set up a measure of occupational exposure to imports following the concept proposed by Ebenstein et al. (2014). This relates to the literature on trade-in-tasks (Grossman & Rossi-Hansberg2008) that explicitly focuses on tradable tasks and the effects of fragmented global value chains. When economies trade tasks in global value chains, evaluating economic globalization at the industry-level does not account for the idea put forward by the trade-in-tasks framework. Trade repercussions affect specific occupations in which individuals perform certain tasks. Since I am interested in non-pecuniary labor market outcomes at the individual level, the occupation-level perspective is the more appropriate approach to pick up the heterogeneous effects of changes in imports.

(4)

China as a dominant supplier (Dauth, Findeisen & Suedekum 2014).

I find that imports affect job satisfaction and job security worry mainly through the supply channel. This effect is especially salient for imports from China. While there is evidence for a negative impact of Chinese supply imports on job security, the effect on job satisfaction is not statistically significant at conventional levels. When I differentiate between different occupations, I find that individuals working in occupations that are typically associated with lower skill levels and less specialization fare worse than individuals working in more skill demanding occupations. The results are robust to a varying length of import exposure and still hold when I implement an IV regression to account for potential endogeneity.

Putting these findings in perspective, Dauth, Findeisen & Suedekum (2017) show for Ger-many that trade has slowed down the structural transformation that usually results in decreasing manufacturing and increasing service employment. Exports to new trade markets drive this find-ing by stabilizfind-ing employment in manufacturfind-ing industries. Lookfind-ing at earnfind-ings, positive effects due to rising export opportunities stem from two sources. Especially high-skilled individuals have experienced earnings gains both on-the-job and by switching to another company that operates in the same industry. At the same time, low-skilled individuals that have accumulated a high stock of industry-specific human capital have been most negatively affected by imports shocks (Dauth, Findeisen & Suedekum 2020, forthcoming).

(5)

produced by the domestic industry itself. I depart from this perspective, by setting up a supply and use import channel which differentiate both between inter- and intra-industry trade and final and intermediate good imports. While Geishecker, Riedl & Frijters (2012) only focus on competition-increasing effects, my contribution provides a more nuanced perspective by simultaneously allowing for competitiveness-enhancing effects. Colantone, Crin`o & Ogliari (2019) find that increased import competition leads to higher levels of distress for UK workers. Using a group of 31 countries, Dluhosch & Horgos (2019) provide evidence for a significant impact of several globalization dimensions on non-pecuniary employment outcomes at the workplace. With changes in these outcomes, additional economic and non-economic domains are affected.

A number of studies relate changes in job status to other economic variables, such as wage bargaining. If workers expect to lose their job in the near future, they earn substantially less (Blanchflower 1991). To account for a potential job loss, workers may consume less and engage in precautionary saving (Stephens 2004). Job loss fears also relate to more general dimensions of subjective well-being. Clark, Knabe & R¨atzel (2010) show that negative externalities of local unemployment impact employed individuals through increases in job loss fear. Furthermore, the perception of globalization is relevant because of its influence on recent political developments. Autor et al. (2016) show that local labor markets in the U.S. experiencing higher trade exposure exhibit more support for extreme political ideologies. In the U.K., regions experiencing a higher exposure to economic globalization voiced more support for the Leave option in the Brexit vote (Colantone & Stanig 2018a). For Germany, Dippel, Gold & Heblich (2015) study the effect of trade integration with China and Eastern Europe in the period 1997-2009. They find that the electoral success of extreme right-wing parties rises with import competition and declines with export opportunities. Thus, while the effect of trade integration differs when comparing Germany with the U.S. (Dauth, Findeisen & Suedekum 2014vs. Autor, Dorn & Hanson2013), the effect on political outcomes shows similar features (Autor et al. (2016) vs. Dippel, Gold & Heblich 2015). I bridge the gap between these two research areas by assessing additional channels through which individuals may be affected.

(6)

presents the estimation results and discusses the outcomes. Section VI offers a summary, derives policy conclusions and suggests future research.

2 Trade & Well-Being

2.1 Trade in Final & Intermediate Goods

On aggregate, trade results in significant improvements in welfare (Costinot & Rodr´ıguez-Clare 2018; Feenstra 2018). Yet, fragmented value chains and trade in intermediates make quantifying welfare effects more complex. Following the rise of global value chains, the imports of a trading economy not only comprise final but also intermediate goods that domestic industries use as inputs. This trade in intermediates is of increasing importance with value chains becoming more and more fragmented (Johnson & Noguera 2017; Timmer et al. 2014), providing incentives to allocate each production process to the location with the lowest costs, even if the production stage is taking place in a remote location. As a consequence, most countries exhibit rising specialization in a specific production stage, so called vertical specialization (Hummels, Ishii & Yi2001) and value-added by a particular production stage is of increasing importance (Koopman, Wang & Wei2014). In a typical global value chain, a producer purchases several inputs, adds his own value and subsequently delivers the good to the next stage which may be situated at home or abroad. This next stage may be final consumption or intermediate consumption.

A tool to account for the aforementioned aspects of trade and to analyze trends in global supply chains are world input-output tables (WIOT). To understand WIOTs, it is helpful to first look at national input-output tables (NIOT). A national input-output table provides information on the production interdependencies between different industries of an economy. Typically, the entries of a column represents the inputs of an industry while the rows describe deliveries of an industry to each industry of the economy. By connecting NIOTs with bilateral international trade flows, one can compile WIOTs that summarize all transactions within the global economy. What makes WIOTs especially useful for the purpose of this study is the differentiation of imports based on the country and industry of origin.

(7)

supply and use imports of domestic industry j.1

Supply channel First, there is a direct competition effect for German industry j due to imports from foreign industry j∗_{, both in final and intermediate products. Suppose the foreign industry}

j∗ supplying the product experiences a productivity increase which leads to an increase in export supply. As imports directly compete with the supply of intermediate and final goods by German industry j on the domestic market, the increased supply of imports potentially decreases demand for domestic producers. If domestic industries face a higher exposure to foreign competition the conditions on the labor market change. In developed economies mainly manufacturing industries seem to be negatively affected through this channel (see Autor, Dorn & Hanson 2013; Pierce & Schott2016for the U.S. and Dauth, Findeisen & Suedekum2017for Germany). On the occupation-level, Ebenstein et al. (2014) show that in the U.S. mainly blue-collar production workers with a high degree of routineness associated with their occupation, have experienced the largest wage decreases.

Use channel Second, the supply chain perspective allows for an assessment of imports of in-termediate inputs.2 Engaging in trade increases the number of available intermediate inputs and decreases prices (Amiti & Konings 2007; Goldberg et al. 2009). Thus, the availability of inter-mediate goods imports may have a competitiveness-enhancing effects for German industry j. I define competitiveness-enhancing imports as the sum of intermediate input imports to j supplied by all foreign industries except j∗. Antr`as, Fort & Tintelnot (2017) show that U.S. manufacturing firms that make use of imported intermediate inputs increase both their operational scale and the number of inputs they source domestically. Interestingly, this mechanism seems to be especially relevant for non-manufacturing firms which mainly produce nontradable goods. There is evidence that the positive employment effects of intermediate demand imports is increasingly important for non-manufacturing firms (Wang et al. 2018). As most studies that evaluate import shocks focus on manufacturing industries, the use of intermediate demand imports in non-manufacturing is an interesting mechanism left to explore. Colantone & Stanig (2019) show that rising exposure to Chi-nese imports affect the voting behavior of private or public service workers that should be sheltered from trade shocks in manufacturing industries. By evaluating the import shock on the grounds of

1

The use of an asterisk indicates a foreign industry, e.g j∗.

2_{In setting up the use channel, I abstract from the notion of an offshorability indicator proposed by other scholars}

(8)

different occupations that are spread over industries, I account for this mechanism.

2.2 German Trade Integration with China & Eastern Europe

Following the fall of the iron curtain in 1989, Eastern European countries joined the World Trade Organization (WTO) (e.g. Czech Republic, Hungary or Poland in 1995) or signed partnership and cooperation agreements (e.g. Russian Federation in 1997). In a next step, many of these economies joined the European Union in the course of the 2004 European Union enlargement. Due to these steps, the trade barriers between Germany and Eastern Europe significantly decreased starting from the almost trade prohibiting levels in the period prior to 1989 (Dauth, Findeisen & Suedekum2014, Online appendix). 1 2 3 4 5 Im p or t vo lu m e, 19 95 = 1 1995 2000 2005 2010

Median Use 1st quartile Use 3rd quartile Use

Median Supply 1st quartile Supply 3rd quartile Supply

Figure 1: German supply and use imports from Eastern Europe & China relative to domestic value added, per industry, 1995-2011. Notes: Author’s calculations, based on WIOD data for final and intermediate demand imports. I first divide use and supply imports by the value added by the industry importing the goods. Next, I normalize the import flows to 1 in 1995.

(9)

importance of trade for the German economy. In general, I find an upward trend for all industries. Second, there are differences in trade exposure on the industry-level. In general, there is a higher spread in supply imports, depicted in grey. The industries that experienced the highest increase in imports in combination with an overall high import level are mainly manufacturing industries. The manufacturing of transport equipment (NACE 34-35) experienced an increase of approximately 440% in use imports in the period 1995-2008. In the same period, the manufacturing of machinery and equipment (NACE 29) saw a fourfold increase of imports through the use channel. When looking at the supply channel, I find that many industries experiencing the highest proportional increase in imports were manufacturing industries. The manufacturing of machinery and equipment experienced a 460% increase in supply imports from 1995-2008. For the manufacturing of electrical and optical equipment (NACE 30-33), the rise in imports amounts to 860% in the same period. Supply imports for manufacturing of transport equipment rose by approximately 770%.

To judge the relevance of Eastern European economies on their position within the interna-tional trading system, domestic productivity changes in these economies are relevant, too. Data on productivity growth provides evidence that Eastern Europe has had higher productivity growth relative to Germany. Already before the European enlargement in 2004, Eastern Europe caught up in terms of average labor productivity as measured by the total output over hours worked in manufacturing (Dauth, Findeisen & Suedekum2014, Online appendix).

The differentiation between final and intermediate demand as well as intra- and inter-industry trade, that I conceptualize as use and supply import channel, is important to frame the previous findings of Dauth, Findeisen & Suedekum (2014). Following the trade integration with Eastern Europe, Germany mainly engaged in intra-industry trade in final products. I capture this type of import through the competition-increasing use channel. After trade with Eastern Europe gained momentum, China became a dominant supplier of imports on the world market but in contrast to trade integration with Eastern Europe, German trade with China was primarily inter-industry trade. Dauth, Findeisen & Suedekum argue that Germany had already established inter-industry trade linkages with other economies before, such that Chinese imports displaced trade flows from other economies. As a consequence, the effect on German production was limited. Note that I capture this type of trade link through the use channel.

(10)

This explains their relevance as a trading partner for the German economy. But which additional aspects come into play compared to the China shock in the U.S.? First, the German economy had already adapted to competition by firms from Eastern Europe as the Chinese trade integration gained momentum. Second, the Eastern European economies not only increased competition but provided attractive new markets for German exports. Third, the German export supply matched with Chinese import demand, focusing on high-quality products (Marin 2011; Marin, Schymik & Tscheke 2015). This is also reflected in research that evaluates the European trade integration process along fragmented supply-chains. Kaplan, Kohl & Mart´ınez-Zarzoso (2018) find that the 2004 European Union enlargement led to a sizeable number of new jobs in entrant economies and at the same time had a neutral or positive impact on the labor markets of incumbents.

While the effects of German trade integration with the East on the earnings of manufacturing workers are substantial (Dauth, Findeisen & Suedekum 2020, forthcoming), non-pecuniary effects at the worker-level hinge on the perception of change. As a consequence, I expect differences when import flows originate from different suppliers. Not only should cultural distance between the importer and the supplier matter for the volume of trade (Felbermayr & Toubal 2010; Guiso, Sapienza & Zingales 2009) but it should also affect the perception of foreign competition itself. Take the example of a supplier of parts and components for the German automotive industry. Workers employed in the automotive industry may not perceive the import of a car component from Eastern Europe as a threat to their job status and not judge it as competition. This may be due to a relatively short cultural distance. In addition, the component may be sourced from a supplier within the European Union that is subject to similar labor regulation. In contrast, a Chinese import may be perceived as a threat due to higher cultural distance. A German worker in the automotive industry may see the imported good as the result of unfair labor competition from abroad. In this example, the different perception of imports from Eastern Europe and China should translate into differences in non-pecuniary effects at the worker-level.3 _{An alternative explanation}

could be the activity of multinational enterprises, headquartered in Germany, that have offshored their production to Eastern German economies. In line with this argument, Marin (2011) argues that German firms have delegated certain tasks to low-cost affiliates in Eastern Europe to increase competitiveness. While a foreign affiliate may be perceived as a threat by domestic workers, import

3

(11)

competition by an unaffiliated competitor may pose the bigger threat.

2.3 Non-Pecuniary Effects of Trade

The effect of trade on standard employment indicators is well-documented. Using changes in im-port competition from China, a number of studies show that deeper trade integration is typically associated with decreasing wages, lower labor-force participation rates and increases in unemploy-ment (Autor et al.2014; Balsvik, Jensen & Salvanes2015; Caliendo, Dvorkin & Parro2019; Dauth, Findeisen & Suedekum 2017). At the same time, a new stream of literature explores how deeper trade integration influences political outcomes (Colantone & Stanig 2018a,b, 2019; Dippel, Gold & Heblich 2015; Malgouyres2017) and non-pecuniary effects that relate to different dimensions of subjective well-being such as life satisfaction, mental or physical health.

A contribution by McManus & Schaur (2016) focuses on the link between import competition and injuries at the workplace. The authors use plant-level panel data on injuries and illnesses and connect it to industry-level changes in Chinese import competition for U.S. manufacturing indus-tries. For the period 1996-2007, they find that industries experiencing higher degrees of competition from China register a rise in injury rates. The magnitude of this effect is higher for smaller firms, with the risk of injury increasing by up to 13%. The authors argue that the increase in risk through import competition is equivalent to a 1% to 2% wage decrease for the most affected establishments.

Adda & Fawaz (2020, forthcoming) consider the impact of import competition on mental and physical health, health-related behaviour and access to health care. Using this information, they also relate import shocks to mortality. The authors build their analysis on a health survey that contains information on individuals living in the U.S. over the period 1998-2009. Adda & Fawaz find that a one billion dollar import increase leads to a rise in the mortality hazard of manufacturing workers by up to 6%. An important point the authors suggest is that the repercussions of import competition may be more pronounced than other aggregate shocks. The immobility of labor and a bias towards older individuals with lower human capital may drive this result.

(12)

suicide and related causes of death such as accidental poisoning as a consequence of drug use. Counties with higher degrees of trade exposure show higher rates of suicides and related causes of death.

Colantone, Crin`o & Ogliari (2019) document detrimental effects of the trade aspects of glob-alization on mental health in the U.K. According to their study, negative repercussions arise due to job loss, wage changes, job satisfaction and expectations about the future employment status. Both job loss and wage changes can be seen as traditional labor market indicators. Job satisfaction and expectations about the future employment status, however, are characteristics which are not covered by traditional labor market indicators. While testing four different microeconomic channels through which globalization may affect mental distress, Colantone, Crin`o, and Ogliari find that increased import competition decreases the probability that a worker reports satisfaction with his job using data taken from the British Household Panel Survey. Additional evidence is provided in an appendix that assesses four different determinants of job satisfaction which are total pay, job security, workload and the content of the job. They show that all four dimensions are negatively affected by increased import competition.

Dluhosch & Horgos (2019) study the impact of international competition on job satisfaction for a group of 31 countries. They show that both a more liberal trade policy and deeper trade integration result in a decrease in individual job satisfaction. A major shortcoming of this study is the level of aggregation. Apart from the individual control variables, all independent variables are measured at the country level. With changes in international trade and technological change being highly specific to certain industries and tasks, averaging over all industries of an entire economy may lead to misleading results. The impact analysis of import competition on work satisfaction should be conducted on the lowest possible level of aggregation.

(13)

after reemployment may be smallest for low-skilled while high-skilled individuals are more vulnerable to job loss in terms of the wage cuts when being reemployed.

In addition to effects operating through the import channel, non-monetary effects of trade integration may also originate from an increase in export activity. Hummels, Munch & Xiang (2016) show that changes in the exporting activity of Danish firms lead to an exogenous shock to labor demand. One option to meet this rise in labor demand is an increase in working hours and the intensity of work. Via this adjustment channel, the authors identify negative effects for individual workers. They find workers in export-oriented firms to have higher injury rates in general. Further, female employees exhibit higher sickness rates as a response to this exogenous shock. The authors argue that while the wage gain outweighs the ex-ante utility loss, the ex-post utility loss of workers experiencing health-related negative effects is substantial.

2.4 Hypotheses

Taking into account the characteristics of fragmented value chains leading to trade in goods and services for both final and intermediate demand purposes, the differentiation between the integration with Eastern Europe and China as well as the findings related to non-pecuniary effects of trade, I propose the following hypotheses:

Hypothesis 1 The effect of increased import competition on non-pecuniary job outcomes differs for the supply and use channel of imports. While increases in supply imports are detrimental to job satisfaction and perceived job security, increases in use imports result in positive effects.

Hypothesis 2 Supply imports from China are perceived as a higher threat than imports from Eastern Europe.

(14)

3 Data & Methods

I test my hypotheses using individual-level data from a panel data set that I merge with data on import penetration at the occupation-level. The individual-level data contains information on job satisfaction and job security worry that I use as dependent variables. The main explanatory variable is import competition at the occupation-level. I augment the analysis by various controls on the individual-level and by adding multi-dimensional fixed effects.

In the baseline regression equation the dependent variable Yi,s,k,r,t denotes either job

satisfac-tion or worry about job security of individual i, working in sector s, with occupasatisfac-tion k, in federal state r in year t.

Yi,s,k,r,t= β1× ∆SupplyImpk,t+ β2× ∆U seImpk,t+ Ii,tβ30 + αi+ αs+ αs,t+ αr,t+ αk+ i,s,k,r,t (1)

The main explanatory variables ∆SupplyImpk,t and ∆U seImpk,t capture the change in imports

from Eastern Europe and China to Germany through the supply and the use channel. Ii,tare vectors

of individual characteristics. αi, αs, αs,t, αr,t, αk are individual, sector, sector-year, state-year and

occupation fixed effects.4 The individual-level fixed effect αi controls for time-invariant differences

at the individual level related to job satisfaction and job security worry. For identification, I exploit within person variation over time. The sector fixed effect, αs, controls for time-invariant

heterogeneity between sectors. As a consequence, I only compare individuals working in the same sector. Year-specific shocks at the sector and federal state level are flexibly absorbed by sector-year and state-year fixed effects. According to Huber & Winkler (2019), one should include state-year fixed effects in the case of Germany because regions in Eastern and Western Germany have different characteristics and differ with regard to their industry specialization. αk controls for time-invariant

differences at the occupational level. The error term is given by i,s,k,r,t.

3.1 Individual-Level Data

The first data set is the German Socio-Economic Panel (SOEP) (Goebel et al.2019), a wide-ranging annual representative longitudinal study of private households. SOEP contains information on

in-4_{On the industry-level, I have trade data for 35 industries as illustrated in Table}_A1_{. On the sector-level, I aggregate}

(15)

dicators such as satisfaction with life at today, satisfaction with work, worry about the economic development or worries about job security in the period 1991-2017. In addition, the data base pro-vides a rich set of individual characteristics that can be included as controls or used for heterogeneity analysis.

I restrict the sample to individuals within working age (18 to 65 years) and for which the SOEP provides information on both the industry of employment, captured in NACE rev. 1.1, and the occupation, captured in ISCO-88.5 I drop individuals working in the primary or energy sector or employed in the armed forces. Further, I focus on respondents in full or part time employment. Since many respondents did not participate in every single wave, I work with an unbalanced panel in order to maximise the number of observations.

SOEP provides information on satisfaction with work using a scale ranging form 0 to 10, comparable to a satisfaction with life scale. Worry about job security can take three different values. Individuals are either be ’not concerned at all’, ’somewhat concerned’ or ’very concerned’.6

Including measures of job satisfaction and perception about job security into the model requires consideration regarding the exact specification. One could either include the variables in levels or in changes. In accordance with the discussion by Colantone, Crin`o & Ogliari (2019), I assume that an individual evaluates his level of job satisfaction and security relative to a reference point or what Colantone, Crin`o & Ogliari call ’usual condition’. How changes in individual circumstances affect particular types of satisfaction depends on the personal evaluation of the individual and is highly specific to the life event and the domain we assess. Significant economic shocks can change the reference point of the individual. However, this adaptation process may differ for each individual (Ferrer-i-Carbonell & Van Praag 2008). If we observe a change in life satisfaction or expectation about the future of employment, both changes in the underlying construct or changing reference points may drive it. In addition to the evaluation of the changing circumstances, individuals may differ with respect to their reporting function, i.e., the way individuals translate their happiness into responses to survey questions about happiness (Bond & Lang 2019). I resolve these issues by

5

The SOEP gives information on the industry affiliation of the individual based on NACE rev. 1.1, which corre-sponds to ISIC rev. 3, and NACE rev. 2, which correcorre-sponds to ISIC rev. 4. Data on the NACE rev. 1.1 classification is available for the survey waves 1984-2017. For NACE rev. 2 data availability starts in 2013. Since there is no un-ambiguous method to match NACE rev. 1.1 industries to rev. 2 industries at the 2 digit level, I use NACE rev. 1.1. The current occupation is codified using 4 digit ISCO-88 occupation codes with a total of 306 occupations. I decide to not use the latest version of the occupation classification, ISCO-08, due to data availability. I rely on educational attainment captured by ISCED 97 to differentiate between different groups of workers.

6

(16)

including the job satisfaction and job security variables in levels while controlling for individual fixed effects. The individual fixed effects account for time-invariant differences in reference point formation and reporting behavior across individuals. This leads to an identification strategy that focuses on short-run deviations of the outcome variables from their within-individual mean (Ferrer-i-Carbonell & Frijters 2004; Frey & Stutzer 2002).

3.2 Industry-Level Data

To compute trade exposures I use the World Input-Output Database (WIOD) (Timmer et al. 2015). Although other data initiatives provide WIOTs, the tables published by WIOD have been built to assess developments over time. Industry level data is drawn from the 2013 release of WIOD covering the period 1995-2011. The database contains information on 35 ISIC rev. 3 industries in 40 countries and a model for the rest of the world. Based on WIOD, I compute two import exposure indicators.7 _{In order to add to the findings by Dauth, Findeisen & Suedekum (}₂₀₁₄_,₂₀₁₇_{), Marin}

(2011) and Dippel, Gold & Heblich (2015), this study focuses on trade exposure stemming from Eastern Europe8 _{and China.}

3.3 Model Specification 3.3.1 Supply Channel

For the supply channel, I define the degree of competition from abroad that industry j faces in year t as the proportional increase of intermediate and final product imports over the last n years. This proportional increase is lagged by 1 year. Thus, if n = 2 the import shock captures the proportional change in imports between t − 3 and t − 1. The supply channel is exemplified using China as trading partner:

∆SupplyImpj,t= w_j,t−1−nSupplyImp×

ImpCHN →DEU_j,t−1 − ImpCHN →DEU j,t−1−n

ImpCHN →DEU j,t−1−n

!

. (2)

7_{WIOD provides information in US Dollars. Since I am interested in non-pecuniary aspects of import competition,}

I do neither deflate the trade flows nor convert them into Euro. To identify the impact of imports on job satisfaction and perceived job security, I rely on the variation of trade exposure, not on the absolute level.

8

(17)

The numerator ImpCHN →DEU

j,t−1 − ImpCHN →DEUj,t−1−n captures the change in imports of final and

intermediate products from China to Germany over the period n. Note that the import exposure of German industry j stems from the Chinese industry j∗ _{that falls within the same industry}

classification, i.e., there is intra-industry trade. Chinese industry j∗ provides imports for final demand purposes and competes with final products German industry j supplies on its domestic market. At the same time, there is competition in intermediate products. A German firm searching for inputs may choose a foreign, e.g. Chinese, supplier or a domestic supplier. As a result, the German and Chinese supplier compete in intermediate product deliveries on the German market.9 The denominator contains the imports of final and intermediate products from China to Germany in the initial period ImpCHN →DEU

j,t−1−n . I use factor w

SupplyImp

j,t−1−n to weight the proportional increase in

imports. As a first option, which is in line with the import competition indicator by Colantone, Crin`o & Ogliari (2019) and Huber & Winkler (2019), I assume that

w_j,t−1−nSupplyImp= 1. (3)

Using a weighting factor of 1 implies that there is a direct relationship between the proportional increase in Chinese imports and job satisfaction/job security worry. However, the change in China’s imports should be related to their relative importance for the individual. A first option to account for this, could be a factor w that relates the amount of Chinese industry j∗’s imports to the amount of German industry j’s deliveries to German users. This factor would capture the market share of Chinese imports and attach a higher weight to industries that face a sizeable Chinese market share.

While the share of gross value of sales may be especially relevant to a firm operating within industry j, I assume that the individual worker is more interested in how this relates to his labor income. What may be more important for an individual worker is the labor compensation or wage bill of his industry j that is embodied in the sales of final and intermediate products on domestic markets. This gives an indication of what is ’at stake’. If the import exposure of industry j is relatively small compared to the wages that depend on domestic sales, workers may not be aware of the exposure. However, if imports are relatively large compared to the wages that depend on

9

(18)

domestic sales, this may have a sizeable impact on job satisfaction/job security worry:

w_j,t−1−nSupplyImp= Imp

CHN →DEU j,t−1−n

LabCompDEU →DEU_j,t−1−n . (4)

The specification of the supply channel may raise concerns about the endogeneity of trade exposure. Unobserved domestic demand factors may impact both the independent variables and the dependent variable. I use the instrumental variable strategy suggested by Autor, Dorn & Hanson (2013) to address this concern. Specifically, I instrument the change in imports from China to Germany with changes in imports from China to other high income countries within the same industry classification. This instrument should not be affected by shocks that are idiosyncratic to the German economy.10 _{Changes in exposure to Chinese imports in other high income countries}

should be driven by Chinese productivity growth and the reduction in trade costs. Thus, the identification strategy builds upon the assumption that the within-industry, exogenous component of rising import exposure in Germany is a result of China’s comparative advantage and a decrease in trade costs (Autor & Dorn 2013; Dauth, Findeisen & Suedekum2014). Accordingly, the import penetration indicator becomes:

∆SupplyImpIV_j,t = wSupplyImp_j,t−1−n × Imp

CHN →HIGH

j,t−1 − ImpCHN →HIGHj,t−1−n

ImpCHN →HIGH_j,t−1−n

!

. (5)

I follow Dauth, Findeisen & Suedekum (2014) who provide an extensive discussion of the potential instrument group for Germany. The authors argue that suitable candidates for the instru-ment are countries with a similar income level. At the same time, they should not share a border with Germany and not be part of the European Monetary Union to be relevant instruments. They suggest an instrument that consists of Australia, Canada, Japan, Norway, New Zealand, Sweden, Singapore, and the U.K. Due to data availability, my instrument comprises Australia, Canada, Japan, Sweden, and the U.K.

As discussed by Colantone, Crin`o & Ogliari (2019) the choice of the import exposure length n is subject to a trade-off regarding individual vs. industry characteristics. While imports are rather volatile, their impact on variables such as job satisfaction or worries about job security may only

10

(19)

come into effect over longer time periods. This would be an argument to use longer lags. At the same time, the analysis should account for confounding factors on the individual- and industry-level that change rather slowly. These underlying long-term characteristics may be correlated with import competition and choosing a longer lag could reduce the precision of the estimation. For the baseline specification I choose a length of n = 2. Contributions focusing on standard labor market repercussions typically opt for longer lags. To come up with the baseline specification, I rely on the time structure that is applied in contributions that relate political outcomes, such as voting, to trade shocks (e.g. Colantone & Stanig 2019). As an additional robustness check, I vary the shock length to see whether the results still hold when individuals experience the trade shock over a longer period.

3.3.2 Use Channel

Next, I define an indicator that focuses on intermediate imports and their competitiveness-enhancing effect. Imports are potentially competitiveness-enhancing because German industry j can make use of a cheaper and broader set of intermediate inputs. While the supply channel indicator builds upon imports from the Chinese industry j∗ to German industry j, I compute the use channel based on imports from all Chinese industries except j∗ that German industry j uses as inputs:

∆U seImpj,t = wU seImp_j,t−1−n×

P

g6=j ImpCHN →DEUj,t−1 − ImpCHN →DEUj,t−1−n

P

g6=j ImpCHN →DEUj,t−1−n

(6)

The numerator P

g6=j ImpCHN →DEUj,t−1 − ImpCHN →DEUj,t−1−n captures the change in imports of

intermediate products from China to Germany over the period n. The import exposure of German industry j stems from the supply of all Chinese industry except j∗_{. All Chinese industries g}∗_{6= j}∗

provide intermediates that German industry j uses as inputs. The denominator contains the imports of intermediate products from China to Germany in the initial period ImpCHN →DEU

j,t−1−n . Factor

wU seImp_j,t−1−n weights the proportional increase in imports. Again, I first assume that

wU seImp_j,t−1−n= 1. (7)

(20)

period. In addition to intermediate deliveries from all other countries, I include the value added by capital and labor to relate the change in Chinese intermediate inputs to the change in the use of all other inputs: w_j,t−1−nU seImp = P g6=j ImpCHN →DEUj,t−1−n P

g6=j ImpALL→DEUj,t−1−n + V ADEUj,t−1−n

. (8)

Analogously, the instrumented use channel indicator builds upon the same factor in combina-tion with intermediate imports from China to other high income countries:

∆U seImpIV_j,t = w_j,t−1−nU seImp× P

g6=j ImpCHN →HIGHj,t−1 − ImpCHN →HIGHj,t−1−n

P

g6=j ImpCHN →HIGHj,t−1−n

(9)

3.3.3 Import Exposure at the Occupation-Level

Since I am interested in the relationship between job satisfaction/worries about job security and import competition at the occupation-level, I follow Ebenstein et al. (2014) to set up an import exposure measure for each occupation. First, I take the total number of workers occupied in occupation k in industry j at the beginning of each period under consideration, t-1-n. This number is then related to the total number of workers within an occupation at time t-1-n. To arrive at an occupational exposure of final demand import competition, I compute the import penetration at time t-1-n for occupation k for the supply channel as:

∆SupplyImpk,t= J X j=1 Lk,j,t−1−n Lk,t−1−n ∆SupplyImpj,t (10)

Analogously, for the use channel, I arrive at:

∆U seImpk,t= J X j=1 L_{k,j,t−1−n} Lk,t−1−n ∆U seImpj,t (11)

(21)

distri-bution of occupations would change drastically in the period during which individuals experience the import exposure, the occupation-level indicator would not adequately capture this. This thought further justifies choosing shorter lengths of import exposure. The notion that the distribution of occupations remains constant also hinges on comparable barriers to switching for all occupations.

4 Descriptives & Preliminary Results

TableA3 compares the mean, standard deviation and number of observations for the entire SOEP sample of workers with the estimation sample that I use to conduct my analysis. Since the SOEP sample is a representative study of German households, the estimation sample should have a similar composition to arrive at conclusions that hold for the population of workers. When looking at the mean and standard deviation, the SOEP worker sample and the estimation sample have a similar composition. As a consequence of the conditions that I impose on the SOEP worker sample, the major differences in means appear for employment related, individual-level variables. Most notably, the means of annual working hours and the years of unemployment experience differ. Individuals in the estimation sample work around 100 hours more than individuals in the larger SOEP worker sample. Another difference is the unemployment experience, measured in years. The mean unemployment experience of individuals in the SOEP worker sample is roughly one month higher. But how do individuals fare at their job?

Mean job satisfaction is around 7, on a scale from 0 to 10, with a standard deviation of 2. When asked about their worry about job security, individuals report a mean worry of roughly 1.7, on a scale ranging form 1 to 3, with a standard deviation of 0.7. Since my empirical strategy relies on within-individual variation of job satisfaction and job security worry, I evaluate the between-and within-variation in Table 1. In total, I base my analysis upon 103,417 observations of 18,137 individuals, which leads to 5.7 observations per individual on average.11 _{The overall mean and}

standard deviation presented in Table 1 correspond to the values in Table A3. In addition, I provide the between and within component of the standard deviation.12 _{For both job satisfaction}

11

Ideally, I would like to work with a balanced panel but I opt for an unbalanced panel to maximize observations.

12_{The minimum and maximum value of the within component capture an individual’s deviation from their own}

(22)

and job security worry, I find relatively larger differences between individuals and smaller differences over time within individuals. The finding that the variation of subjective well-being measures across sections is larger than the variation over time, is in line with previous studies (e.g. Deaton 2008). At the same time, the difference between the between and within component are not sizeable such that there is substantial variation over time that remains to be explained.

Table 1: Descriptive statistics, job satisfaction & job security worry.

Mean Std. Dev. Min Max N/n/T-bar Job satisfaction overall 7.019 1.966 0 10 103,417

between . 1.534 0 10 18,137

within . 1.365 -1.623 14.419 5.702 Job security worry overall 1.694 0.708 1 3 103,417

between . 0.568 1 3 18,137

within . 0.461 -0.163 3.551 5.702

Notes: Author’s calculations based on SOEP data (unbalanced panel, N = 103,417). N is the total number of observations in the estimation sample, n is the number of individuals and T-Bar the average number of observations per individual.

Figures2-3group individuals according to their educational attainment and occupation. Ed-ucational attainment distinguishes between low-, medium- and high-skilled workers based on the International Classification of Education.13 On the occupation-level, I differentiate between 9 dif-ferent major occupational groups.

4.1 Job satisfaction

In Figure 2, I plot the mean of job satisfaction for low-, medium- and high-skilled individuals in the period 1997-2011. In general, there is a sizeable variation of mean job satisfaction both within groups over time and between groups. Particularly, the overall trend of the time series for low-skilled individuals differs from those being medium- and high-skilled. Mean job satisfaction of the low-skilled reached a trough in 2006. Subsequently, the mean job satisfaction of low-low-skilled individuals increased up to a level of about 7.1 in 2008. The sizeable reduction in job satisfaction that followed this peak coincides with the 2008 financial crisis. Surprisingly, the negative impact of the financial crisis was offset by a strong rise in 2010. Comparing the mean job satisfaction of the medium- and high-skilled, I find that both time series followed similar dynamics, albeit high-skilled individuals

13

(23)

reporting a mean job satisfaction which is slightly higher in most years. Most noticeably, the overall peak of job satisfaction of the medium- and high-skilled seems to have coincided with the economic boom of the year 2000. Both medium- and high-skilled people reported their highest average job satisfaction in 2001. For both groups, the 2008 financial crisis also appears in the time series. Both groups experienced a trough. On average, medium-skilled individuals were the least satisfied in 2009, while high-skilled individuals already reported rising job satisfaction in that year, following a trough in 2008. Similar to the strong increase in mean job satisfaction of low-skilled individuals, both the medium- and high-skilled quickly recovered from the negative impact of the crisis.

6. 6 6. 8 7 7. 2 2000 2000 2000 2005 2010 2005 2010 2005 2010

Low-skilled Medium-skilled High-skilled

M ea n J ob S at is fa ct io n Survey Year

Figure 2: Mean Job Satisfaction by educational attainment, 1997-2011. Notes: Author’s calculations based on SOEP data (unbalanced panel, N = 103,417). I evaluate educational attainment by ISCED 97. Low-skilled corresponds to ISCED 1-2, medium-skilled corresponds to ISCED 3-4 and high-skilled corresponds to ISCED 5-6.

Figure2 depicts mean job satisfaction over the years 1997-2011 for 9 different major occupa-tional groups. Three main observations stand out. First, mean job satisfaction is rather stable over time for each occupational group.14 Second, the lower the skill level and specialization required to work in an occupation, the lower mean job satisfaction. Legislators, senior officials and managers are on average the most satisfied with their job while individuals working in elementary occupations are the least satisfied. Third, the negative impact of the 2008 financial crisis is most salient for

14

(24)

individuals working in occupations requiring low-levels of skill and specialization, as captured in the last row of Figure 2.

6. 5 6. 5 6. 5 7 7 7 7. 5 7. 5 7. 5 8 8 8 2000 2000 2000 2005 2010 2005 2010 2005 2010

Legislators, senior officials, managers Professionals Technicians, associate professionals

Clerks Service/shop/market sales workers Skilled agricultural/fishery workers

Craft/related trades workers Plant/machine operators, assemblers Elementary occupations

M ea n J o b S a ti sf a ct io n Survey Year

Figure 3: Mean Job Satisfaction by occupation, 1997-2011. Notes: Author’s calculations based on SOEP data (unbalanced panel, N = 103,417).

4.2 Job security worry

Job security worry is measured by the question ’Are you concerned about your job security?’, with possible answers being ’not concerned’, ’somewhat concerned’ or ’very concerned’. For every year, I compute the share of every option for three different skill-levels and depict it in Figure 4.

(25)

the answer share of ’somewhat concerned’ decreases while the share of ’not concerned’ increases. Another relevant observation is the share of individuals reporting to be ’very concerned’. The fraction of the ’very concerned’ among low and medium-skilled individuals rose from 2008-2010 with the stronger increase among the low-skilled. For the group of high-skilled, the 2008 crisis did not lead to a sizeable increase in concern.

0 .2 .4 .6 2000 2000 2000 2005 2010 2005 2010 2005 2010

Low-skilled Medium-skilled High-skilled

Not concerned Somewhat concerned Very concerned

Are you concerned about your job security?

S h ar e of an sw er Survey Year

Figure 4: Worry about job security by educational attainment, 1997-2011. Notes: Author’s calculations based on SOEP data (unbalanced panel, N = 103,417).

Figure 5 depicts job security worry by major occupational group. I find that individuals working in an occupation that requires higher skill level and specialization generally report lower job security worry, i.e., a higher fraction of individuals being ’not concerned’ and a lower fraction of individuals being ’very concerned’. This holds when I compare the first row with the last row of Figure 5. The most stable answer shares are reported by the major occupational group 2, comprised of professional working in fields such as STEM, life science and health or teaching. The impact of negative economic circumstances such as the financial crises in 2008, is most noticeable for individuals working in low-skill and low-specialization occupations, captured by an increase in the fraction of people reporting to be ’very concerned’.

(26)

extraordinary circumstances which turned ”the sick man of Europe” into an ”economic superstar” while reducing the unemployment rate from 11.1% in 2005 to 7.7% in 2010 (Dustmann et al.2014). Among other events, such as the adoption of the Euro in 1999, this period was characterized by further trade integration with Eastern Europe and the rise of China as major player in international trade. One channel through which German individuals experienced deeper trade integration is increasing import penetration. In the next section, I explore to which degree this occupation-level import competition varies between and within occupations.

0 0 0 .2 .2 .2 .4 .4 .4 .6 .6 .6 .8 .8 .8 2000 2000 2000 2005 2010 2005 2010 2005 2010

Legislators, senior officials, managers Professionals Technicians, associate professionals

Clerks Service/shop/market sales workers Skilled agricultural/fishery workers

Craft/related trades workers Plant/machine operators, assemblers Elementary occupations

Not concerned Somewhat concerned Very concerned

Are you concerned about your job security?

S h ar e of an sw er Survey Year

Figure 5: Worry about job security by occupation, 1997-2011. Notes: Author’s calculations based on SOEP data (unbalanced panel, N = 103,417).

4.3 Import shocks

(27)

fractions within an industry vary across industries and over time and lead to variation at the occupation-level.

To judge the degree to which occupation-level import penetration varies between and within occupations, I provide descriptive statistics in Table 2. The upper three variables capture sup-ply imports while the lower three variables capture use imports. Further, I differentiate between imports from Eastern Europe & China and imports from the respective country. As an example, ∆SupplyImpEAST

k,t captures supply imports from Eastern Europe & China in year t for occupation

k.

Table 2: Descriptive statistics, import penetration at the occupation level. Mean Std. Dev. Min Max N/n/T-bar ∆SupplyImpEAST k,t overall 0 1 -4.378 8.390 103,417 between . 0.510 -3.424 5.733 18,137 within . 0.901 -5.315 8.226 5.702 ∆SupplyImpCHN k,t overall 0 1 -0.332 15.660 103,417 between . 0.670 -0.305 10.833 18,137 within . 0.851 -7.996 14.521 5.702 ∆SupplyImpEEU k,t overall 0 1 -4.320 8.323 103,417 between . 0.511 -3.374 5.695 18,137 within . 0.901 -5.263 8.144 5.702 ∆U seImpEAST

k,t overall 0 1 -1.994 6.739 103,417 between . 0.698 -1.130 5.068 18,137 within . 0.763 -5.003 6.328 5.702 ∆U seImpCHN k,t overall 0 1 -2.339 5.939 103,417 between . 0.713 -1.272 4.593 18,137 within . 0.747 -5.006 5.459 5.702 ∆U seImpEEU

k,t overall 0 1 -1.440 7.167 103,417

between . 0.664 -1.298 4.996 18,137 within . 0.790 -4.263 6.765 5.702

Notes: Descriptive statistics based on standardized coefficients.

First, I evaluate the variation of occupation-level supply imports. The upper section of Table 2 reports descriptives for supply imports from the Eastern Europe and China as well as for both countries separately.15 Supply imports vary between and within occupations. Importantly, the within component of the standard deviation is substantially higher than the between component. The lower section of Table 2 contains descriptive statistics for use imports. When comparing the between and within standard deviation for use imports, I find that the within component is

15

(28)

higher than the between component, i.e., there is more variation within occupations than between occupations. At the same time, the difference between the between and within component standard deviation of use imports is not as large as of supply imports. The within variation of use imports from China is comparable to the within variation of imports from Eastern Europe.

In terms of use imports that increase competitiveness, the variation in both China and Eastern Europe imports is comparable. One could have expected a different finding, with Germany mainly relying on cheaper Chinese intermediate inputs after China’s opening up for trade. The fact that use imports did not increase as strongly as supply imports, may be explained with Germany’s trade structure at the beginning of the China shock. Dauth, Findeisen & Suedekum (2014) provide suggestive evidence that underlines the important role Eastern Europe played in moderating the rise of China as a trading partner. At the time when Chinese intermediates became available, Germany already imported goods from industries in which China proceeded to become the most important supplier.

5 Results & Discussion

5.1 Imports from Eastern Europe & China 5.1.1 Job Satisfaction

(29)

satisfaction should be interpreted in light of the reaction of job satisfaction to major macroeconomic shocks in Figures2and3. In 2009, the year following the 2008 financial crisis, mean job satisfaction of the low-skilled dropped by approximately 0.3 points. For the medium and high-skilled, the economic boom of the year 2000 increased mean job satisfaction by approximately 0.2 points.

Table 3: Job satisfaction, exposure to imports from China & Eastern Europe.

Job satisfaction (1) (2) (3) (4) (5)

∆SupplyImpEAST

k,t -0.0108 -0.00597 -0.00579 -0.00625 -0.00580 Weighting factor: Eq. 3 (0.00932) (0.00459) (0.00453) (0.00497) (0.00479) ∆U seImpEAST

k,t -0.0289*** 0.00971 0.00997 0.00605 0.00704 Weighting factor: Eq. 4 (0.0106) (0.00785) (0.00806) (0.00706) (0.00696)

R-squared 0.518 0.524 0.526 0.528 0.529

(6) (7) (8) (9) (10)

∆SupplyImpEAST

k,t -0.0376*** -0.0114 -0.0118 -0.00716 -0.00695 Weighting factor: Eq. 7 (0.00668) (0.00856) (0.00852) (0.00857) (0.00839) ∆U seImpEAST

k,t -0.0212** 0.00860 0.00904 0.00355 0.00454 Weighting factor: Eq. 8 (0.00937) (0.00676) (0.00681) (0.00782) (0.00781)

R-squared 0.518 0.524 0.526 0.528 0.529

Observations 103,417 103,417 103,417 103,417 103,417

Individual FE Yes Yes Yes Yes Yes

Sector FE Yes Yes Yes Yes

Sector x year FE Yes Yes Yes Yes

State x year FE Yes Yes Yes

Occupation FE Yes Yes

Individual controls Yes

Notes: The dependent variable is job satisfaction, on a scale form 0 to 10. 10 indicates the highest possible job satisfaction and 0 the lowest. Supply imports are the change in final and intermediate demand imports according to equation2between t − 3 and t − 1. I compute the import shock per industry and relate it to the occupation share within that industry as described in section 3.3.3. Supply imports are the change in intermediate demand imports according to equation5 between t − 3 and t − 1. Again, I relate industry-level imports to occupations. Sector FE are fixed effects for each section of NACE rev. 1.1. Sector x year FE are fixed effects for each section of NACE rev. 1.1 in each year. State x year FE are fixed effects for each state in each year. Occupation FE are fixed effects for each 3-digit occupation. Individual controls include unemployment experience, hours worked annually, a dummy for overtime, company size, the number of persons in the household, the number of children in the household, the martial status, household income from assets, a dummy for house ownership, the living space of the house and educational attainment based on ISCED 97. The standard errors are clustered at the 4-digit occupation level. ***, **, * indicate significance at the 1, 5 and 10% level.

5.1.2 Job Security Worry

(30)

and vice versa. I report regression results in Table 4. The model specification with individual fixed effects associates an increase in use or supply imports with an increase in job security worry. With additional fixed effects and individual controls, the only coefficients that remains statistically significant are the weighted supply import coefficients. In model specifications (5) − (8) of Table 4 supply imports are weighted by the labor compensation that is embodied in the sales of final and intermediate products on domestic markets. An increase in supply imports from the East by one standard deviation, leads to an increase in job security of 0.005 points. The competitiveness-enhancing imports through the use channel do not seem to affect perceived job security.

Table 4: Job security worry, exposure to imports from China & Eastern Europe.

Job security worry (1) (2) (3) (4) (5)

∆SupplyImpEAST

k,t 0.00846*** -0.000333 -0.000570 -0.000404 3.49e-06 Weighting factor: Eq. 3 (0.00302) (0.00260) (0.00258) (0.00259) (0.00260) ∆U seImpEAST

k,t 0.0218*** -0.00254 -0.00228 -0.00171 -0.00196 Weighting factor: Eq. 4 (0.00702) (0.00449) (0.00449) (0.00508) (0.00472)

R-squared 0.577 0.590 0.593 0.594 0.596

(6) (7) (8) (9) (10)

∆SupplyImpEAST

k,t 0.0152** 0.00502** 0.00488** 0.00533** 0.00532*** Weighting factor: Eq. 7 (0.00732) (0.00195) (0.00198) (0.00208) (0.00202) ∆U seImpEAST

k,t 0.0161** -0.00336 -0.00308 -0.00284 -0.00291 Weighting factor: Eq. 8 (0.00792) (0.00411) (0.00413) (0.00450) (0.00399)

R-squared 0.577 0.590 0.593 0.594 0.596

Observations 103,417 103,417 103,417 103,417 103,417

Notes: The dependent variable is job security worry, on a scale form 1 to 3. 1 indicates ’no worry about job security’ 2 indicates ’some worry’ and 3 indicates ’a lot of worry’. See notes of Table3

for other variables.

5.2 Imports from China

(31)

larger than the negative effect due to imports from Eastern Europe. Following Dauth, Findeisen & Suedekum (2014), the effect of use imports from China should not have a sizeable effect because Germany was already importing from other countries when China became a dominant supplier.

5.2.1 Job Satisfaction

Table5summarizes the estimates for the impact of trade on job satisfaction for China. For the im-port exposure measures weighted by a factor of 1 in specifications (1) − (4), the estimated coefficient of supply imports is negative and significant. Increases in supply imports from China are associated with lower levels of job satisfaction. The magnitude of the coefficients for use imports are compara-ble to the coefficients estimated for Eastern Europe and China, albeit insignificant. Weighting the import exposure by the respective use and supply factors does not change the direction of the effect, with similar effect sizes in model specifications (6) − (8) compared to the unweighted specifications (2) − (4). Yet, the supply import coefficients are no longer significant.

Table 5: Job satisfaction, exposure to imports from China.

∆SupplyImpCHN

k,t -0.0144* -0.0116** -0.0110** -0.00863** -0.00923** Weighting factor: Eq. 3 (0.00802) (0.00462) (0.00472) (0.00411) (0.00420)

∆U seImpCHN

k,t -0.0284** 0.0150 0.0151 0.0102 0.0113

Weighting factor: Eq. 4 (0.0113) (0.00965) (0.00972) (0.0102) (0.0101)

R-squared 0.518 0.524 0.526 0.528 0.529

(6) (7) (8) (9) (10)

∆SupplyImpCHN

k,t -0.0396*** -0.00976 -0.00993 -0.00571 -0.00534 Weighting factor: Eq. 7 (0.00682) (0.00880) (0.00873) (0.00881) (0.00869) ∆U seImpCHN

k,t -0.0157 0.0106 0.0110 0.00521 0.00595

R-squared 0.518 0.524 0.526 0.528 0.529

Observations 103,417 103,417 103,417 103,417 103,417

Notes: See Table3.

(32)

in Figure 6. Typically the occupations summarized in the upper groups are associated with higher skill-levels and more specialization. Even though most of the marginal effects are insignificant at the 95% level, there is evidence for a heterogeneous effect across occupations.

Legislators, senior officials and managers

Professionals

Technicians and associate professionals

Clerks

Service/shop/market sales workers

Skilled agricultural and fishery workers

Craft and related trades workers

Plant/machine operators, assemblers

Elementary occupations

-3 -2 -1 0 1

Effects on Linear Prediction

Figure 6: Marginal effects of supply imports from China on predicted job satisfaction, 1-digit occupation. Notes: The underlying regression is equivalent to specification (5) in Table5, excluding occupation fixed effects. Confidence intervals are at the 95% level.

Table6 reports the estimates for the impact of exposure to Chinese imports on job security worry. Evaluating the unweighted import exposure measures, I find a significant negative effect of increased exposure to Chinese supply imports on job security worry, i.e., the estimated coefficient is positive. The negative effect through the supply channel is robust to the inclusion of fixed effects and controls and still significant at the 1% level in the most demanding specification (5). Interpreting the estimated coefficient of model specification (5) in Table 6, I find that a one standard deviation increase in supply imports from China leads to a 0.005 increase in job security worry. When weighting the exposure to Chinese imports in specifications (6) − (10), I find that the negative impact through the supply import channel remains significant and comparable in size.

(33)

Table 6: Job security worry, exposure to imports from China.

∆SupplyImpCHN

k,t 0.00165 0.00513*** 0.00512*** 0.00561*** 0.00545*** Weighting factor: Eq. 3 (0.00284) (0.00149) (0.00151) (0.00143) (0.00134) ∆U seImpCHN

k,t 0.0329*** -0.00161 -0.00138 -0.000415 -0.000362 Weighting factor: Eq. 4 (0.0103) (0.00505) (0.00498) (0.00548) (0.00507)

0.577 0.590 0.593 0.594 0.596

(6) (7) (8) (9) (10)

∆SupplyImpCHN

k,t 0.0170** 0.00478*** 0.00463*** 0.00536*** 0.00543*** Weighting factor: Eq. 7 (0.00702) (0.00148) (0.00146) (0.00147) (0.00144) ∆U seImpCHN

k,t 0.0159** 0.00140 0.00159 0.00202 0.00184

R-squared 0.577 0.590 0.593 0.594 0.596

Observations 103,417 103,417 103,417 103,417 103,417

Notes: See Table4.

different occupations, I interact the import variables with 9 different 1-digit occupations. I report the marginal effects in Figure 7.

In total, there are two major occupational group that have significant effect size. For craft and related trades workers as well as individuals working as plant and machine operators and assemblers, the effect is significant and above zero. In addition, the number of observations per occupation is large. For individuals working in these occupations, an increase in supply imports increases their worry about job security.

(34)

Legislators, senior officials and managers

Professionals

Technicians and associate professionals

Clerks

Service/shop/market sales workers

Skilled agricultural and fishery workers

Craft and related trades workers

Plant/machine operators, assemblers

Elementary occupations

-.1 -.05 0 .05 .1

Effects on Linear Prediction

Figure 7: Marginal effects of supply imports from China on predicted job security worry, 1-digit occupation. Notes: The underlying regression is equivalent to specification (10) in Table 6, excluding occupation fixed effects. Confidence intervals are at the 95% level. I omit the marginal effects for ”Skilled agricultural and fishery workers” for better readability. The marginal effect is not significant and the 95% interval ranges from -3.8 to 1.4.

(35)

5.3 Imports from Eastern Europe

5.3.1 Job Satisfaction

Following the analysis of Chinese imports, I now assess the effect of exposure to imports from Eastern Europe. First, I report the estimates for the impact of imports from Eastern Europe on job satisfaction. Table 7 summarizes the related regressions. For the unweighted import exposure, I find that the point estimates indicate a negative impact on job satisfaction. However, these point estimates are imprecise and thus insignificant. The same holds for use imports. While the sign of the estimated coefficients is in line with hypothesis 1, the coefficients are not significant. When I weight the imports, the overall interpretation does not change. As such, the regressions in Table 7 do not provide evidence in favor of a significant impact of Eastern Europe on non-pecuniary dimensions.

Table 7: Job satisfaction, exposure to imports from Eastern Europe.

∆SupplyImpEEU

k,t -0.0104 -0.00450 -0.00434 -0.00521 -0.00470 Weighting factor: Eq. 3 (0.00898) (0.00465) (0.00461) (0.00504) (0.00481) ∆U seImpEEU

R-squared 0.518 0.524 0.526 0.528 0.529

(6) (7) (8) (9) (10)

∆SupplyImpEEU

k,t -0.0143* -0.00621 -0.00683 -0.00502 -0.00525 Weighting factor: Eq. 7 (0.00743) (0.00752) (0.00756) (0.00676) (0.00661) ∆U seImpEEU

R-squared 0.518 0.524 0.526 0.528 0.529

Observations 103,417 103,417 103,417 103,417 103,417

(36)

Proceeding with a set of regressions models that relate exposure to imports from Eastern Europe and job security worry, I find no evidence in favor of a meaningful impact. In combination with the statistically significant negative effect of increased exposure to Chinese supply imports on job security worry, this finding provides some support for hypothesis 2.

Table 8: Job security worry, exposure to imports from Eastern Europe.

∆SupplyImpEEU

k,t 0.00775*** 0.000306 7.44e-05 0.000212 0.000597 (0.00246) (0.00223) (0.00220) (0.00220) (0.00223) ∆U seImpEEU

k,t 0.0153*** -0.00462 -0.00432 -0.00392 -0.00411 (0.00517) (0.00338) (0.00343) (0.00381) (0.00356) R-squared 0.576 0.590 0.593 0.594 0.596 (6) (7) (8) (9) (10) ∆SupplyImpEEU k,t 0.00450 0.00173 0.00173 0.00123 0.000998 (0.00391) (0.00276) (0.00283) (0.00308) (0.00306) ∆U seImpEEU

k,t 0.0150** -0.00542* -0.00520* -0.00519 -0.00508* (0.00693) (0.00311) (0.00314) (0.00341) (0.00307)

R-squared 0.576 0.590 0.593 0.594 0.596

Observations 103,417 103,417 103,417 103,417 103,417

Notes: See Table4.

5.4 Discussion

(37)

channel. This mainly holds for job security worry. Increases in supply imports lead to increasing levels of job security worry. When looking at the use channel, I am not able to identify statistically significant impacts in most cases. Yet, when using weighted use imports from Eastern Europe, there seems to be a reduction in job security worry if imports increase. In contrast to other coefficients, these effects are only significant at the 10% level, merely providing suggestive evidence for a positive effect of supply imports. Even though, the direction of the effect for the supply and use channel do not hold in all specifications, they lend some support to the notion that use and supply imports have opposing effects. In sum, supply imports seem to be detrimental to job satisfaction and perceived job security which is in line with an competition-increasing effect. There is suggestive evidence that use imports are at least not detrimental job satisfaction and job security. However, I can not provide conclusive support for a competitiveness-enhancing effect in this context. This supports hypothesis 1 for the most part.