By George Hu
Abstract: This paper examines the impact of Germany’s Hartz reforms on country and industry level trade outcomes, particularly trade balances. Utilizing product-level bilateral trade data from the UNComtrade data set along with industry-level data from EUKlems, we conduct two related analyses examining the impact of the Hartz Labor Market Reforms on German economic trade outcomes. First, on the country level, the paper finds that the Hartz reforms on average boosted German bilateral exports, net exports, and net exports as a percentage of GDP with its trading partners. Second, on the industry level, sectors with higher levels of labor productivity also tend to exhibit higher levels of bilateral exports and net exports to trading partners.
I: Introduction
Economists have extensively documented the impact of trade on various macroeconomic outcomes in advanced Western economies as free trade agreements and import competition, especially from developing economies, have multiplied over the past 20-30 years. In fact, mirroring the heated current political rhetoric surrounding both trade and growing inequality, economists have especially focused on studying the impact of increased trade and import competition on societal inequality, especially in advanced economies. Indeed, a large body of the literature, including Autor et al (2013), Autor et al (2014), and Bernard et al (2006) suggests that increased trade and import competition from developing countries have often led to higher income, wage, and wealth inequality on the regional, industry, and individual levels in both the United States and various countries in Western Europe.
However, despite the diversity and plethora of literature documenting such effects, literature surrounding (i) potential policy measures to remedy the inequality incited by increased trade and (ii) the impact of domestic firm or industry-level reform on international trade outcomes proves more uniform as well as less abundant. Specifically, many papers from the literature covering trade’s impact on societal welfare and inequality either omit discussion of policy remedies or suggest that the beneficiaries of trade compensate the shortchanged despite the political difficulties of undertaking such an action. Likewise, although research papers analyzing the impact of relevant industry, firm, and labor market-level reforms on various macroeconomic outcomes including trade do exist, economists and the economics literature remain largely divided over the effectiveness and international applicability of proposed reforms. Overall, while the economists have built a clearer understanding on how exactly the recent increase in international trade has affected levels of inequality in developed countries, a working knowledge concerning how an economy can best remedy the negative effects of increased international trade remains limited.
Aiming to address the apparent imbalance between understanding of international trade’s exacerbating impact on inequality and the understanding of how to counteract these developments, this paper seeks to begin to build an answer for the following question: how do labor market reforms and labor productivity impact country and industry level trade outcomes, specifically exports and trade balances? Considering this question, the paper utilizes German industry and product level data from 2002-2007,[1] a period in which domestic labor market reforms were gradually beginning to take effect. Focusing on Germany from 2002-2007, the paper conducts two related empirical analyses: (i) a country-level analysis which estimates the impact of the Hartz reforms on Germany’s bilateral exports and trade balances, and (ii) an industry level analysis estimating the effect of labor value added[2] on bilateral industry exports and trade balances. In the two analyses, we find evidence that the Hartz reforms increased both exports and the German trade balance, and that industries with higher labor productivity tend to more frequently export and display more favorable (i.e. positive) trade balances. The results may suggest that productivity-boosting labor market reforms may help developed countries mitigate certain negative consequences associated with increased trade on the firm and industry level, but the question of whether such reforms can help remedy trade-induced increases in societal inequality on the individual/worker level requires further research.
The remainder of the paper is organized into four sections. Section II first provides a historical and technical background on the Germany economy and the Hartz reforms. It then proceeds to review two largely independent strands of the economics literature that the paper seeks to link: (i) the impact of the Hartz reforms on German labor market and macroeconomic outcomes, as well as (ii) the relationship between firm/industry productivity and trade outcomes. Section III details the data collected for both country and industry level analysis, before specifying the empirical models that we test/run. Section IV presents the paper’s regression results for both country and industry level analyses, before considering potential direct extensions to our analyses. Lastly, section V concludes and considers the paper’s external validity and policy implications, while also proposing ideas for further related research.
II: Background and Literature Review:
II.1: The Hartz Reforms: Background
While the German economy has stood out as one of Europe’s and the OECD’s best performers over the past decade after the Great Recession, from 1990-2005 the country’s poor economic performance earned it the nickname “the sick man of Europe” (Dustmann et al, 2014). Whereas today metrics including the 3.4% Unemployment Rate, 64.1% government debt to GDP ratio, and annual GDP growth of 2.2% (2017) underline the robust health of the country’s economy, during the 1990s and early 2000s the unemployment rate in Germany remained persistently above 8%, while GDP grew anemically and consistently below 2% annually.[3] Though at first most economists attributed the country’s poor economic performance in the 1990s to the costs associated with reunification and reintegration of the former GDR (Gaskarth, 2014), the persistence of economic stagnation through the introduction of the European Monetary Union into the early 2000s reflected more serious structural issues in the economy and its regulatory institutions. Specifically, politicians and economists alike questioned the continued viability of the German welfare state and its social market economy, which continued to rely on an export-oriented manufacturing industry, in the face of increased import competition and accelerating global economic integration (Odendahl, 2017).
In response to the structural economic challenges that continued to hinder the country’s economy in the early 2000s, German policymakers and politicians resolved to overhaul the regulations and institutions which oversaw the country’s labor market. In particular, the Bundestag, under pressure from then-Chancellor Gerhard Schröder, passed the Hartz reforms, four separate measures which constituted the primary portion of Agenda 2010, a government initiative designed to restore economic growth and lower unemployment (Bauer and King, 2018). As the brainchild of the government’s Committee for Modern Services in the Labor Market chaired by the Volkswagen personnel director Peter Hartz, the reforms were gradually enacted into law from 2003-2005 (Woodcock, 2018).
The Hartz I and II reforms came into effect on January 1, 2003. These first two reforms focused on subsidizing and further incentivizing unconventional employment and entrepreneurial activities by introducing a new federal grant to entrepreneurs (Ich-AG) and offering tax relief to individuals employed in part-time roles, or “Minijobs” (Odendahl, 2017). Next, the government proceeded to implement Hartz III on January 1, 2004. Complementing the subsidization of unconventional work, the 3rd reform aimed to reform the “Job Centers” which assisted the unemployed in reentering the labor market (Woodcock, 2018). Finally, Hartz IV, the most substantial and significant of the four reforms, passed into law on January 1, 2005. Whereas the prior three reforms provided financial support to unconventional employment arrangements and vocational (re)training programs, Hartz IV reduced the scale of the German welfare state by cutting back the government’s financial assistance to the unemployed. In particular, the reform lowered the long-term unemployment benefit (Arbeitslosenhilfe) payment rate from 53-57% of an individual’s prior salary to the welfare benefit (Sozialhilfe) payment rate of €382 per month (Odendahl, 2017). Furthermore, the reform reduced the time frame during which a worker could apply for full unemployment benefits (Arbeitslosengeld I) worth 60-67% of prior salary from 12-36 months to 12 months (Woodcock, 2018). That is, after 2005, a recently unemployed worker could receive Arbeitslosengeld I for only 12 months after his/her loss of employment.
Although the economic impact of the Hartz reforms remains debatable amongst both policymakers and economists, the passage of the Hartz reforms undeniably coincided with an upswing in Germany’s macroeconomic fortunes and net export performance. As Appendix Figure 1 exhibits, the German unemployment rate, after reaching a high of 11.7% in 2005, fell to 6.4% by October 2008 and only reached a high of 8.3% in February 2009 during the depths of the ensuing Great Recession (Federal Office of Statistics, 2018a). Likewise, as shown in Appendix Figure 2, after running a trade deficit for much of the 1990s, the German trade surplus grew from under 2% of GDP in 2002 to 5.6% in 2008. Finally, as Gaskarth (2014) notes, German GDP growth gradually began to increase from its average of 1.6% annually from 1995-2001, recording growth of 3.7% in 2006 and 3.3% in 2007 (OECD, 2018). While it is not immediately clear whether the Hartz reforms’ implementation played a role in driving these changes, the reforms were quite immediately followed by an economic upswing.
Nonetheless, despite the relative economic success that occurred during the Hartz reforms’ aftermath, the economic effectiveness of the reforms rightly remains an important subject of both political and economic debate. Even before economists had a chance to evaluate their macroeconomic impact, the Hartz reforms proved politically contentious. In August 2004, over 100,000 people took part in “Monday Demonstrations” protesting the impending implementation of the Hartz IV reform across over 100 municipalities in the country. Furthermore, according to political pundits, discontent with the reforms may have also played a role in inciting Chancellor Schröder and the SPD’s defeat to Angela Merkel and the CDU in the 2005 elections (Landler, 2005).
In all, two contrasting political narratives about the reform’s impact on the German economy emerged over the following years. On the one hand, the view championed by Angela Merkel, her finance minister Wolfgang Schäuble, and the CDU/CSU maintains that the reforms contributed significantly to subsequent German economic success by increasing the country’s economic “competitiveness” in an era of increasing world trade (Odendahl, 2017). According to most mainstream politicians, by following a Fördern und Fordern (Support and Demand) policy of aiding the unemployed on the condition that they quickly return to work, Hartz IV enabled market mechanisms to lower wages and related labor costs key to sustaining output and export growth in key industries including manufacturing. On the other hand, critics of the reforms abound on the political extremes, particularly from the left. Beyond the expected sharp political claims that the reforms represent a neoliberal measure designed to favor large corporations at the expense of the worker, critics of Hartz IV claim that the reforms replaced chronically high unemployment with increasing inequality and worker dissatisfaction with their new and often low-paying jobs. To critics, the country’s emergence from stagnation to growth stemmed mostly from changing global import demand and macroeconomic trends, not the reforms, while the decrease in the unemployment rate only overshadowed increases in inequality and the rise of workers employed in low-paying part-time roles.
II.2: The Hartz Reforms: Literature Review
The economics literature surrounding the Hartz reforms has primarily focused on their impact on German labor market outcomes, and findings corroborate elements of the two political narratives pursued by supporters and critics of the reforms respectively. First, economic research has indicated that the Hartz reforms helped lower the unemployment rate by improving labor markets’ efficiency, dynamism, and flexibility. To begin, building off and confirming earlier work completed by Fahr and Sunde (2006), Hertweck et al (2012) find that labor market inflow rates have substantially increased since the introduction of the Hartz reforms in the mid-2000s, and their analysis proceeds to use an empirical matching function to show that the reforms had improved “matching efficiency” by around 20%. Complementing Hertweck et al (2012), Bauer and King (2018) also claim that the reforms brought down the country’s structural unemployment rate by more efficiently and effectively reallocating labor to employers and tasks which vocational training had better prepared individuals to undertake. Lastly, Burda (2016) notes that alongside increased matching efficiency and effectiveness, the reforms also spurred employment growth by complementing the already increasing degree of wage flexibility across industries as well as encouraging growth in part-time work.
Alongside research that appears to confirm the Hartz reforms’ positive impact on reducing unemployment, encouraging entrepreneurial job creation, and increasing labor “competitiveness”, a complementary strand of economic research has also found that the reforms have likely helped to exacerbate domestic inequality levels, particularly by worsening worker’s post-unemployment wages. Utilizing a confidential data set from the German Social Security administration, Detragiache et al (2015) claim that after the reforms, the wages of workers reentering the labor market after short-term unemployment relative to the wages of workers reentering from long-term unemployment fell by around 10 percent. In effect, Detragiache et al’s result suggest that the reforms, due to its lowering of long-term unemployment benefits, may have induced the short-term unemployed to accept less than optimal jobs, lest they remain unemployed for over 12 months and are forced to receive the meager Sozialhilfe welfare payments. Woodcock (2018) generally reaffirms Detragiache et al’s findings in a study utilizing data from the German Institute for Employment Research (IAB). In his paper, Woodcock (2018) specifically finds that the Hartz reforms substantially reduced the post-employment wages of both short and long-term unemployed workers. He proceeds to demonstrate that this reduction in wages arose from two main factors: (i) the selection of displaced workers into low-wage firms and (ii) the sorting of displaced workers into lower quality “matches” with employers.
Although the economics literature concerning the Hartz reforms mostly considers their direct impact on the German labor market, some more recent research has extended this literature to investigate the effect of the reforms on broader macroeconomic outcomes. Complementing the analysis of Detragiache et al (2015) and Woodcock (2018), Fredriksen (2012) claims in an OECD report that the Hartz reforms helped to induce pervasive wage moderation on the lower end of the income distribution, resulting (rather atypically) in increased labor earnings inequality driven by changes in the distribution’s left tail. While Fredriksen (2012) considers the reforms’ impact on macroeconomic measures of country-level inequality, Odendahl (2017) examines the Hartz reform’s impact on a wide range of macroeconomic variables including the Unemployment Rate, GDP growth, exports, and government debt. Though he appears to favor the explanation that accelerating global economic growth and import demand drove much of Germany’s turnaround in the mid-2000s, he does also mention that the reforms could have induced increased exporting not only through increased industry-level “competitiveness” but also through the hindering effect of wage moderation on domestic consumption and investment.
II.3: Productivity and Trade Outcomes
Alongside the economic research analyzing the policy effectiveness of the Hartz reforms, this paper, especially the industry-level analysis, builds off another area of the literature which examines the relationship between productivity and trade. To begin, there exists a substantial literature covering the impact of increased import competition on firm innovation and productivity levels, especially in the context of advanced economies facing competition from developing countries. The effect of competition on innovation at large has remained difficult to generalize as well as contentious within the economics literature (cf. Aghion et al, 2005), and likewise no consensus amongst economists exists regarding the impact of import competition on firm productivity and innovation. On the one hand, Bloom et al (2016) find that increased Chinese import competition boosted European firms’ productivity, IT spending, and levels of innovation in the early 2000s. On the other hand, Autor et al (2016) reach the opposite conclusion when working with patent data from the United States.
On the other side of the literature, research considering the relationship between intrinsic firm and industry productivity characteristics and export/trade outcomes generally concur that more productive firms and industries tend to exhibit higher exports. Although their paper mainly focuses on the effect of exporting on firm innovation incentives, Aghion et al (2018) note that exporters within an industry very often overlap with the industry’s major innovators. Likewise, Bernard et al (2006) find that higher-productivity firms prove more likely to become exporters, especially when trade barriers are lowered. Thus, whereas the literature on the effect of import competition on overall firm productivity remains divided, there exists a clear consensus that firms with higher levels of productivity and innovation are more likely to export.
III: Data and Empirical Strategy:
To conduct the empirical analyses, this paper’s data, recorded on an annual basis, focus on German bilateral trade outcomes as well as various industry measures from the period 2002-2007. As the Hartz reforms became law over the period 2003-2005, the period 2002-2007 serves as an ideal 6-year window into analyzing the reforms’ effectiveness. Extending the timeframe for analysis further forward would result in the inclusion of data from the Great Recession of 2008/2009, while pushing the start of the timeframe backwards would also include macroeconomic shocks such as the establishment of the European Monetary Union (EMU) in 1999 and the entry of China into the WTO in 2001. Two separate data sets are collected for country and industry level analyses.
III.1: Country Level Analysis
For the country-level analysis, we take bilateral trade data on Germany, Austria, the Netherlands, and Sweden from the UNComtrade database.[4] Observations in these bilateral trade data are unique on the reporter-partner-commodity-year level,[5] where the reporter represents the country whose bilateral net exports are reported (i.e. either Germany, Austria, the Netherlands, or Sweden in this data set), the partner represents the reporter’s bilateral trading partner, and the commodity represents the type of good being traded. Appendix Figure 3 provides a more detailed image of our data’s structure. Regarding commodity classification, the bilateral trade data is aggregated at the HS2 1996 AG2 level; that is, all traded goods are classified into 98 distinct categories. As the data contain the trade value (in current USD) of both the bilateral export flow (export_id) and bilateral import flow (import_id) between a reporter-partner-commodity group during a given year, we also generate a bilateral net exports variable (netexports) which quantifies the annual net export flow within each reporter-partner-commodity group. Lastly, we add data on each reporter’s annual GDP (in current USD) and generate a net-exports to GDP (NXGDP) variable representing each bilateral net exports variable as a fraction of reporter GDP. Additional reporter-specific variables were created for the export, netexports, and NXGDP variables, and the relevant summary statistics lie below in Table 1:
Table 1: Summary Statistics for Country-level Data
Before specifying the details of our empirical strategy, we shall note that the temporal trends displayed in our data do indeed show a positive turn in German trade fortunes, especially after the 2003-2005 period. As exhibited in Appendix Figures 4a-4c, Germany’s average bilateral exports, average bilateral net exports, and average bilateral net export to GDP ratio all grew over the 2002-2007 period, with growth during the post-Hartz IV 2005-2007 period appearing especially prominent. However, one should clearly consider these coinciding linear trends with caution. Though these trends may well reflect the impact of the Hartz reforms on trade outcomes, they could just as easily result from increased domestic GDP growth or increased global import demand, amongst other things. Therefore, a more detailed empirical analysis appears necessary to more accurately estimate the reforms’ effect.
Ultimately, for our country-level analysis, we employ a fixed-effect/difference-on-difference empirical model, the full form of which is denoted below:
Of note, our main treatment variable is a dummy variable that equals 1 when a given bilateral trade observation has Germany as the “reporter” and occurs during or after 2005, when the Hartz IV reform took effect. We test the effect of our treatment on three outcome variables (mentioned earlier in this section): bilateral exports (i.e. export_id), bilateral net exports (i.e. netexports), and the bilateral net export to reporter GDP ratio (i.e. NXGDP), which track different aspects of a given reporter’s trade outcomes with a specific commodity to a specific trading partner in a specific year. Lastly, since the UNComtrade bilateral trade data do not provide many viable potential control variables, we include a litany of fixed effects (FEs) in place of explicit controls in our empirical framework. Specifically, our FEs include reporter FEs, commodity FEs, trading partner FEs, partner-commodity FEs, partner-year FEs, and year FEs. To consider some factors which FEs can capture, the year FEs allow us to effectively control for global macroeconomic trends such as growth and inflation (since all trade values are recorded in USD), commodity FEs enable us to account for the relative tradability of different commodities in international markets, and country FEs (i.e. partner & reporter FEs) effectively track the constant (time-invariant) differences across countries. Building off these foundational FEs, the partner-year FEs account for temporal shocks to foreign countries’ aggregate import demand as well as individual trading partners’ annual GDP trends, while partner-commodity FEs account for time-invariant differences in specific commodity demand across countries.
III.2: Industry Level Analysis
For analysis at the industry level within Germany, this paper builds off the UNComtrade data set used for the country level analysis. To begin, in order to effectively apply the gradual introduction of the Hartz reforms to industry-level analysis, we restrict the UNComtrade data to the bilateral data containing Germany as the reporter.[6] After restricting the UNComtrade data, we proceed to merge them with German industry level data from the EUKlems. From the EUKlems data the paper utilizes various useful metrics on industry level productivity, including the annual contributions in value added growth derived from (i) labor hours worked, (ii) non-ICT capital investments, (iii) ICT capital investments, and (iv) Total Factor Productivity (TFP) growth.[7] In addition, the paper also utilizes EUKlems data on total labor compensation as well as average per hour labor compensation by industry in Germany. The summary statistics of the EUKlems variables of interest are presented below in Table 2:
Table 2: Summary Statistics for Industry-Level data
German industry level data in EUKlems are aggregated using a relatively disaggregated version of the ISIC rev. 3 classification system. That is, the relevant macroeconomic measures of productivity, growth, compensation, and output are disaggregated into 38 “industry” categories, thus forming industry level data. Since data in the EUKlems data set are unique at the industry year level, whereas the data in UNComtrade remain unique at the reporter-commodity-partner-year level, we construct a crosswalk matching HS2 1996 AG2 commodity codes to ISIC rev. 3 industry codes before linking the two data sets. We utilize an m:1 merge on the two data sets, as industry level metrics from EUKlems will not vary by trading partner unlike the UNComtrade data on bilateral trade flows.
Before settling on a specific empirical strategy, a preliminary review of linear time trends indicates that the upswing in Germany’s bilateral exports and bilateral trade balances coincided with a period of growth for labor-hours’ contribution to value added growth (to be denoted LHVA), as demonstrated by Appendix Figure 5. Again, the positive correlation between the increase in bilateral exports/net exports and the increase in LHVA should be treated with restraint. As Appendix Figures 6a-6c show, although TFP contribution to value added growth (TFPVA) and ICT capital contribution to value added growth (ICTVA) display no clear linear trend from 2002-2007, the increase in bilateral exports and net exports from 2002-2007 also coincides with an increase in non-ICT capital contribution to value added growth (NICTVA) during the same period, so it remains difficult to tell whether the increase in bilateral exports/net exports stemmed from an increase in LHVA or ICTVA, or from another factor. Thus, a more rigorous empirical method is required to disentangle to true impact on LHVA and labor productivity on industry level exports and trade balances.
This paper’s industry level analysis utilizes both a fixed effect/panel data as well as an instrumental variables (IV) approach to estimate the impact of LHVA, a proxy for industry labor productivity, on commodity-level bilateral exports and trade balances. The specification for the fixed effects regression is listed below:
The main outcome variables, commodity-level bilateral exports and net exports, appear unique at the commodity-partner-time level, whereas the main treatment variable LHVA and explicit controls (NICTVA, ICTVA, and TFPVA) appear unique at the industry-year level.[8] The explicit controls attempt to account for the presence of other sources of value-added growth, which due to firm-heterogeneity in productivity remain correlated with LHVA and thus present a source of omitted variable bias. We also add a control for bilateral imports (import_id) when we utilize bilateral industry exports as our outcome variable. Furthermore, similarly to the methods presented in our country-level analysis, we utilize industry FEs to account for potential differences in the tradability of commodities across industry, while we utilize partner and partner-year FEs to account for potential differences in commodity tastes and country-level import demand shocks over the 2002-2007. Lastly, since commodity-level and product-level observations in the sample are restricted to Germany, changes in within-country macroeconomic trends, such as GDP growth and inflation, should be covered by the year FEs. Nonetheless, as the literature on productivity and exporting/trade balances (e.g. Bloom et al 2016, Aghion et al 2018) mentions, there remains a potential issue of mutual endogeneity between productivity and exports that our fixed effect strategy cannot completely capture.
To remedy the potential of reverse causality and mutual endogeneity becoming a confounding factor in our fixed effect estimates, we also conduct the following IV/2SLS regression which estimates the impact of LHVA on bilateral exports and net exports below:
Of course, we add an explicit control for bilateral imports when bilateral industry exports are used as our outcome variable. The first-stage regression for our 2SLS approach is as follows:
We utilize two separate instrumental variables for our analysis: total industry labor compensation (TLC) and average per-hour labor compensation by industry (PHC). NICTVA, ICTVA, TFPVA, and year FEs are kept to account for any threats to the instruments’ exogeneity, and both instruments are found to be strong/relevant (see full IV analysis in Tables A4-A5 of the Appendix). To justify the explicit controls, the literature on firm inequality (cf. Autor et al, 2017) show that “superstar” firms with higher overall productivity and a higher level of exports tend to pay a wage premium, thus suggesting a correlation between our instrument and productivity which must be controlled for.[9]
IV: Results:
IV.1: Country Level Analysis
Table 3: Impact of Hartz IV on Germany’s Bilateral Trade Outcomes
Table 3 (above) displays the results obtained from country-level analysis. Columns (1)-(3) of the table display the results of a partially specified version of our model which excludes partner-commodity and partner-year FEs, while Columns (4)-(6) display the results obtained by the fully specified model. Likewise, columns (1) and (4) specify the impact of the reforms on bilateral exports by commodity (export_id), and columns (2) and (5) specify the impact of the reforms on bilateral net exports by commodity (netexports). However, since regression analysis of the reforms’ impact on bilateral gross exports and bilateral net exports fails to effectively control for reporter GDP growth, we utilize a new variable, bilateral net exports by commodity as a percentage of reporter GDP (NXGDP), as a third treatment variable which intrinsically accounts for GDP in columns (3) and (6).[10] Appendix Tables A1-A3 display a more complete set of empirical specifications, including OLS and weighted OLS regressions, for each of the three treatment variables. As Table 3 displays, coefficient estimates of the Hartz reforms’ impact on bilateral exports (export_id) and bilateral net exports (netexports) prove positive and significant at the 1% level, while the estimate for the reforms’ impact on NXGDP remains positive and significant at the 10% level. Interpreting the estimates of our preferred specifications in columns (4)-(6), the legal implementation of the Hartz IV reform led to an approximately $174.4 million increase in annual commodity-specific bilateral gross exports, a $38.5 million increase in annual commodity-specific bilateral net exports, and a 5.44*10-6 percentage point increase in the NXGDP ratio.
The drop in the significance of estimates utilizing NXGDP as the main outcome variable suggests that GDP growth within reporting countries may indeed display a nontrivial correlation or play a nontrivial role in increasing German bilateral export and net export outcomes after the enactment of the Hartz reforms. After all, the accounting link between GDP and overall net exports would suggest that GDP growth would correlate with higher bilateral exports and net exports, though further analysis would be required to determine the extent of this correlation, as well as whether it represents a causal effect. Also, the substantial increase in R2 values when partner-commodity and partner-year FEs are introduced in columns (4)-(6) of Table 3 lends credence to the argument that rising foreign import demand (cf. Odendahl 2017) played a nontrivial role in explaining the rise in German exports and its trade balance through the mid-2000s, though our regression estimates do still suggest that the Hartz reforms played a role.
Lastly, as a robustness check, we also utilize an adjusted version of our treatment variable (adjh4) such that it equals 1 when a given observation has Germany as the reporting country and occurs during or after 2006. Changing the time constraint on the indicator variable serves to account any temporal lags that might have occurred before the full effects of the Hartz reforms, especially Hartz IV, took place.[11] Given that labor markets may take months, even years, to fully adjust to new regulations and institutional frameworks, adjusting the treatment variable as a robustness check appeared to be an appropriate precaution. Table 4 displays the results under the adjusted treatment variable:
Table 4: Impact of Hartz IV Reform on Germany’s Trade Outcomes, Accounting for Potential Lagged Effects
As Table 4 indicates, adjusting the main treatment variable as a robustness check has a relatively small effect on the magnitudes of our original estimates in Table 3.[12] Furthermore, whereas some estimates in Table 3 only proved significant at the 10% level, all coefficient estimates of the impact of the reforms (captured by adjh4) on bilateral exports, bilateral net exports, and bilateral net exports as a percentage of reporter GDP appear significant at the 5% or 1%. Though the magnitudes of the estimates may not have changed by much, their increased significance could perhaps lend credence to our initial intuition that the full effects of the Hartz reforms would become more clearly felt at a lagged rate, though further empirical analysis is obviously required to more definitively prove the validity of this intuitive belief.
IV.2: Industry Level Analysis
Table 5: Impact of LHVA on Bilateral Net Exports on an Industry Level
Table 5 (above) displays the first portion of industry level analysis, where bilateral net exports (netexports) is used as the outcome variable. Columns (1)-(4) present OLS/Fixed Effects estimates of the impact of LHVA (measured in percentage points) on bilateral net exports (measured in current USD) for a given industry in Germany, whereas columns (5)-(6) present our 2SLS estimates when utilizing TLC and PHC as our respective instruments. Columns (1)-(3) represent a partial representation of our full fixed effects model, while column (4) displays estimates of the fully specified model. In turn, column (5) presents a 2SLS estimate using total labor compensation by industry (TLC) as an instrument. However, although the instrument proves quite strong/relevant (cf. Appendix Table A4 or A5), the lack of partner or partner-year FEs may prove a threat to exogeneity, as industries facing higher foreign import demand may indeed hire more workers and thus raise TLC in response.[13] To account for this threat, we run an alternative 2SLS regression using average per hour labor compensation by industry (PHC) as our instrument, which should prove less sensitive or correlated with foreign import demand shocks, while increases in German GDP/demand should be captured by the year FEs.[14] Thus, interpreting the coefficient of our preferred empirical specification in column (6), we observe that a one percentage point increase in the contribution of labor hours to valued added growth (LHVA) appears to incite an approximately $45 million increase in the industry’s average bilateral net exports concerning each of its associated commodities.[15]
Table 6: Impact of LHVA on Bilateral (Gross) Exports on an Industry Level
Since regressing LHVA on bilateral net exports may not fully account for increases in factor demand experienced by high-productivity industries which are likely to export, imports of commodities related to productive industries’ factor inputs may lead regression results of LHVA on industry bilateral net exports to misrepresent the true impact of labor productivity/LHVA on industries’ “success” in trade outcomes. Thus, we run a separate specification using industry bilateral gross exports as the outcome variable. As Table 6 (above) displays, changing the main treatment variable from bilateral net exports to bilateral gross exports does little to change the sign and significance of most of our results. As mentioned before, we include a control for bilateral imports (import_id) in the Table 6 regressions to account for the differing volumes of trade flows across both commodities and industries. Despite the relative consistency of sign and significance, the magnitudes of the estimates in Table 6 do appear to be slightly but consistently larger than estimates presented in Table 5. These slightly larger estimates appear to support our earlier intuition that related factor imports may have also increased in response to gains in industry labor productivity/LHVA, though exports increased by a much larger factor.
IV.3: Potential Extensions to Empirical Analysis
While our regression approaches appear to indicate that the Hartz reforms and its potential resulting effect on labor productivity, further analysis may prove helpful in refining our results. Overall, both our country and industry level analyses relied on commodity and industry level data that proved relatively aggregated at the HS2 and ISIC rev. 3 levels respectively. As UNComtrade provides commodity level data at more disaggregated levels such as HS6 and HS8 while EUKlems provides industry level data at more disaggregated measures of the ISIC rev. 3 system, performing empirical analyses using more disaggregated data could help (hopefully) strengthen the validity of our results. Furthermore, the increased number of observations that results from more finely disaggregated data will likely provide more power to statistical analyses, which could be useful to determine whether our regression coefficient estimates that were less significant accurately denote a meaningful linear relationship. The following paragraphs will discuss potential extensions and improvements specific to our country-level and industry-level analyses respectively.
First, our country-level analysis can be extended or improved in various ways. By employing fixed effects differences-on-differences regression framework, our country level analysis relies substantially on a “parallel trends” assumption between our treatment and control groups, represented by Germany vs. Austria/Sweden/the Netherlands here respectively. We selected the control group of Austria, the Netherlands, and Sweden off circumstantial evidence of their relative macroeconomic similarity to Germany.[16] However, constructing a “synthetic Germany” as a control group with a more scientific approach would undoubtedly refine the validity of the analysis. Furthermore, as hinted at in section IV.1, when regressing on bilateral exports and net exports, the lack of control for GDP growth in the reporting country could present a source of OVB, while regressing on NXGDP could produce a slightly misleading result if in fact the reforms also positively contributed to GDP growth as well as exports.[17]
Likewise, further examination could extend or improve our industry-level analysis in various ways. First, there remains some ambiguity concerning the level of policy variation in our industry level study, thus leading to some hesitancy over how to best cluster standard errors. In particular, it remains unclear whether the Hartz reforms mainly affected labor productivity levels at the EUKlems-denominated industry or the UNComtrade commodity level.[18] In our analysis in IV.2, we cluster standard errors at the industry level, and in a robustness check we also cluster at the HS2 commodity level, which produces the same estimated coefficients but smaller standard errors, presumably due to the larger number of commodity groups. Beyond standard error choice, there remains the issue of mutual endogeneity (i.e. simultaneous causality). While we attempt to utilize 2SLS/IV analysis with PHC (average per-hour labor compensation) as an instrument to mitigate these concerns, there remain determinants of bilateral exports and net exports, such as geographic clustering of industries,[19] which remain uncontrolled for and could be correlated with PHC, thus threatening its exogeneity. Ultimately, construction of a more certainly exogenous instrument could help clarify our results, as well as more formal empirical tests on how the Hartz reforms affected industry level PHC and TLC.[20] Lastly, given the winner-take-most nature of many industries and the usually intense level of competition in international markets, assuming a linear relationship between labor productivity (proxied by LHVA) and bilateral exports/net exports may not be so realistic. Thus, complementary models such as a quadratic/cubic fit model may (i) help us better understand the true relationship/return rates of LHVA on various bilateral trade outcomes and (ii) more accurately estimate LHVA’s causal effect on these outcomes.
V: Conclusion and Further Extensions:
This paper links bilateral trade data from UNComtrade with industry level data from EUKlems to investigate the impact of the Hartz reforms and the resulting increases in German labor productivity on the country’s exports and trade balances. Employing a panel data differences-on-differences strategy, the paper finds that the Hartz reforms’ implementation led to positive and mostly significant increases in Germany’s bilateral exports, net exports, and net exports as a percentage of (German) GDP during the 2002-2007 period. Likewise, specifying both a panel data and instrumental variables approach, the paper demonstrates that within Germany from 2002-2007, higher levels of labor productivity tend to induce higher bilateral exports and net exports on the industry level. While extensions to the paper’s collection of data and experimental design would further clarify its empirical analysis, the paper nonetheless still yields important insights on the impact of labor market reform and labor productivity on advanced economies’ outcomes in international economic competition and trade.
Beyond its immediate results, the paper implicitly raises an intriguing policy question over the effectiveness of labor market reform in mitigating or even counteracting the negative ramifications of increased import competition that many advanced economies currently face. Whether imposed upon them by the Troika during the Eurozone Sovereign Debt Crisis or adopted willingly by their own politicians, various Eurozone economies, especially along the monetary union’s southern periphery, have begun to enact wide-ranging labor market reforms modeled on the Hartz reforms in the hope of regaining “competitiveness” in world markets. While most research concerning the impact of the Hartz reforms as well as the results presented in this paper may suggest that labor market reforms and labor productivity gains can help mitigate the effects of increased import competition and trade, the external validity and applicability of such results to other countries and economies remains contestable. As Dustmann et al (2014) correctly point out, certain features of the German economy, including the persistent preeminence of its manufacturing industry, its relatively high reliance on low-wage labor, and relative autonomy in matters of wage bargaining differentiate it from other Eurozone economies. Thus, implementing Hartz-inspired reforms in countries such as France, Italy, or Spain, which possessed highly nationalized unions and centralized bargaining processes, smaller manufacturing sectors, and a lesser proportion of unskilled workers, may prove both more difficult politically and less effective in remedying the negative consequences of international trade (Dustmann et al, 2014).
Nonetheless, given the paper’s results and potential policy implications, there exist various possible extensions to our research. First, building off the analysis of Detraigiache et al (2015), Woodcock (2018), and Fredriksen (2012), further research on the direct short and medium-term impact of the Hartz reforms on country-level inequality would presumably provide an insightful complement to this paper’s focus on trade outcomes. Especially since over a decade has passed since the reforms’ implementation, future research could potentially examine the longer-term impact of the reforms on low-wage workers’ average wages and duration of employment. Likewise, as numerous Southern European governments have instituted Hartz-inspired labor market reforms over the past five years, further research on the impact of these more recent reforms on country-level trade outcomes could confirm whether research results concerning the Hartz reforms prove externally valid. Lastly, a more direct analysis on the relationship between labor market reform and the posterior intensity and magnitude of import competition can help clarify labor market reform’s effectiveness as a tool to counteract the negative side effects of trade. In all, the aforementioned ideas represent only a few of many research directions, and we leave it to future research to continue to advance the literature.
References:
Aghion, P., A. Bergeaud, M. Lequien, and M. Melitz, 2018, “The Impact of Exports on Innovation: Theory and Evidence,” NBER Working Paper 24600 (Cambridge: National Bureau of Economic Research).
_____, N. Bloom, R. Blundell, R. Griffith, and P. Howitt, 2005, “Competition and Innovation: an Inverted-U Relationship,” The Quarterly Journal of Economics 120(2), pp. 701-728.
Autor, D., D. Dorn, G. Hanson, 2013, “The China Syndrome: Local Labor Market Effects of Import Competition in the United States,” American Economic Review 6(103), pp. 2121-2168.
_____, D. Dorn, G. Hanson, and J. Song, 2014, “Trade Adjustment: Worker Level Evidence,” The Quarterly Journal of Economics 129(4): 1799-1860.
_____, D. Dorn, G. Hanson, G. Pisano, and P. Shu, 2016, “Foreign Competition and Domestic Innovation: Evidence from U.S. Patents,” NBER Technical Report 22879 (Cambridge: National Bureau of Economic Research).
_____, D. Dorn, L. Katz, C. Patterson, and J. Van Reenen, 2017, “The Fall of Labor Share and the Rise of Superstar Firms,” NBER Working Paper 23396 (Cambridge: National Bureau of Economic Research).
Bauer, A., and I. King, 2018, “The Hartz Reforms, the German Miracle, and Labor Reallocation,” European Economic Review 103(2018), pp. 1-17.
Bernard, A., J. Bradford Jenson, and P. Schott, 2006, “Trade Costs, Firms, and Productivity,” The Journal of Monetary Economics 53(2006), pp. 917-937.
Bloom, N., M. Draca, and J. Van Reenen, 2016, “Trade Induced Technical Change? The Impact of Chinese Imports on Innovation, IT, and Productivity,” The Review of Economic Studies 83(1), pp. 87-127.
Burda, M., 2016, “The German Labor Market Miracle, 2003-2015: An Assessment,” SFB 649 Discussion Paper 2016-005 (Berlin: SFB 649, Humboldt Universität zu Berlin).
Detragiache, E., N. Engbom, and F. Raei, 2015, “The German Labor Market Reforms and Post-Unemployment Earnings,” IMF Working Paper 15/162 (Washington: International Monetary Fund).
Dustmann, C., B. Fitzenberger, U. Schönberg, and A. Spitz-Oener, 2014, “From Sick Man of Europe to Economic Superstar: Germany’s Resurgent Economy,” Journal of Economic Perspectives, vol. 28, no. 1, pp. 167-188.
Eurostat, 2017, International Trade in Goods Data 1995-2016 (Brussels: European Commission).
Fahr, R., and U. Sunde, 2006, “Did the Hartz Reforms Speed up Job Creation? A Macro-Evaluation Using Empirical Matching Functions,” IZA Discussion Paper No. 2470 (Bonn: Institute for the Study of Labor).
Federal Statistics Office, 2018a, ILO Labor Market Statistics (Wiesbaden: Federal Statistics Office).
_____, 2018b, National Accounts Database (Wiesbaden: Federal Statistics Office).
Fredriksen, K., 2012, “Income Inequality in the European Union,” OECD Economics Department Working Paper No. 952 (Paris: Organization for Economic Cooperation and Development).
Gaskarth, Glyn, 2014, “The Hartz Reforms… and their Lessons for the UK,” Centre for Policy Studies Discussion Paper (Surrey: Centre for Policy Studies).
Hertweck, M., and O. Sigrist, 2012, “The Aggregate Effects of the Hartz Reforms in Germany,” The German Socio-Economic Panel Study at DIW-Berlin Working Paper 532-2013 (Berlin: German Socio-Economic Panel Study).
IMF, 2018, World Economic Database (Washington: International Monetary Fund).
Landler, M., 2005, “German Leader Gambles in Call for Early Election,” New York Times (New York: New York Times).
Odendahl, C., 2017, “The Hartz Myth: A Closer Look at Germany’s Labor Market Reforms,” Center for European Reform Discussion Paper (London: Center for European Reform).
OECD, 2018, National Accounts Dataset (Paris: Organization for Economic Cooperation and Development).
Woodcock, S., 2018, “The Effect of the Hartz Labor Market Reforms on Post-unemployment Outcomes, Sorting, and Matching,” Simon Fraser University Working Paper (Burnaby: Simon Fraser University).
[1] Product level data from the Netherlands, Austria, and Sweden are also used for a portion of our empirical analysis. [2] Labor Value Added from Hours Worked serves as the closest proxy for labor productivity that the data provides. [3] Statistics obtained from German Federal Office of Statistics, Eurostat, and the OECD. [4] UN Comtrade trade data concerns international trade in goods, not so much in services. We include data from Austria, the Netherlands, and Sweden to use as a “control group” for our differences-on-differences and panel data empirical analysis. [5] Technically, the observations are unique at the reporter-partner-commodity-year-tradeflow level, as the original data set split import and export figures into separate observations. However, in the process of creating a bilateral net exports variable, we created export_id and import_id variables which recorded each reporter-partner-commodity-year specific export and import trade value for both the “Export” and the “Import” observations, effectively rendering the data set unique at the reporter-partner-commodity-year level. The “double counting” of observations should have no effect on regression analysis, since each effective observation will be equally “double-weighted”. [6] This paper relies upon the existing economics literature which has generally concluded that the Hartz reforms affected labor productivity and value added by further increasing the rate of “wage moderation” over the mid-2000s. This conclusion will play a significant role in our instrumental variables analysis. Given that the reforms directly affected wage pressures only within Germany and not within neighboring/similar European economies, we necessarily limited our industry level analysis to German data to study how the Hartz reforms indirectly impacted industry level trade outcomes by affecting wages and labor productivity. [7] Non-ICT capital includes machinery, equipment, and nonresidential buildings, factors often associated with more traditional capital. ICT capital includes communication, hardware, and software infrastructure often associated with IT products and other more recently developed types of capital. [8] The definitions of the main treatment and control variables are defined earlier in this section. [9] We shall discuss the dropping of partner and partner-year fixed effects and continued inclusion of year FEs in more detail in the results section of the paper. [10] We alternatively attempted directly controlling for reporter GDP in our export_id and netexports regressions, from which we obtained positive but insignificant results. [11] Centering our focus on the Hartz reforms on the Hartz IV reform specifically appeared reasonable to us, as the Hartz IV reform clearly proved to be the largest and most expansive of the four reforms. [12] Interpretation of the elements can be found in the prior paragraph detailing Table 3 results. [13] Inclusion of these FEs may help solve exogeneity concerns, but they decrease the TLC’s relevance as an effective instrumental variable. [14] Additional threats to exogeneity of PHC shall be addressed in section IV.3. [15] By “associated” we mean the commodities linked to the industry via our crosswalk. [16] Such macroeconomic similarities include a relatively high export-to-GDP ratio, the existence of social market economies and their regulatory/institutional frameworks, as well as factors deriving from regional proximity. [17] Regressions including GDP as an explicit control in regressions on exports and net exports turned out positive but insignificant. Understanding why such regressions turned out insignificant results while a regression without the GDP control on NXGDP produced results significant at the 10% level could extend our understanding and analysis. [18] Since the Hartz reforms changed labor market laws, they would most directly be implemented at the firm level; however, since we do not possess firm level data, we assumed that firms could be more easily aggregated by industry than by commodity, which explains why industry-clustered standard errors were ultimately chosen. [19] For example, industries with higher PHC could happen be clustered around cities along the sea or a major river (i.e. the Rhine in Germany’s case), and an industry’s relative geographical proximity to urban areas and/or water could help determine its exports/net exports due to the associated lower costs of transport in these areas. [20] Our IV analysis is designed with the assumption that the Hartz reforms drove down/moderated average wages across all industries. This assumption is backed up in existing literature (i.e. Detraigiache et al (2015) and Woodcock (2018)), but no formal empirical analysis of the Hartz reforms’ effect on average wages and labor compensation occurs in this paper.
Comments