University of Southern California Dissertations and Theses

The impact of minimum wage on labor market dynamics in Germany

(USC Thesis Other)

The impact of minimum wage on labor market dynamics in Germany

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE IMPACT OF MINIMUM WAGE ON LABOR MARKET DYNAMICS IN GERMANY
by
Clément Boullé
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2023
Copyright 2023 Clément Boullé
Acknowledgements
I am grateful to my advisors, Professors Kurlat and Kahn, for their guidance and unwavering support
throughout my Ph.D. journey. Their expertise and mentorship have been invaluable.
I am also profoundly grateful to my family, especially my beloved wife Parama. Her unwavering
belief in me, endless encouragement, and outstanding support have been the driving force behind my
perseverance. Her patience, understanding, and sacrifices during the challenging phases of this journey
have been instrumental in my success. I am grateful for her love, understanding, and unwavering presence
by my side, making this achievement a shared one.
Furthermore, I am thankful to my classmates and fellow researchers for creating a stimulating and
collaborative environment. Our shared experiences and discussions have enriched my academic journey.
Lastly, I extend my appreciation to all those who have contributed to my research and to the partici-
pants who generously shared their time and knowledge.
To everyone mentioned above and to others who have supported me along the way, I offer my heartfelt
thanks. Your contributions have played a significant role in my personal and professional growth.
With gratitude,
Clément Boullé
ii
TableofContents
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2: The impact of minimum wage on labor market dynamics in Germany . . . . . . . . . . 2
2.1 The minimum wage in Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Female workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Low education workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.3 Low tenure workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Part-time workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Theoretical discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Chapter 3: Heterogeneity in earnings losses following job displacement . . . . . . . . . . . . . . . 25
3.1 Data and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.1.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.2 Measuring job displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.3 Constructing a sample of displaced workers and a control group . . . . . . . . . . 28
3.1.3.1 Baseline restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.3.2 Propensity Score Matching . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.4 Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.1 Impact of job displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.2 Heterogeneity in earnings losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.2.1 The impact of job loss on young workers . . . . . . . . . . . . . . . . . . 31
3.2.2.2 The impact of job loss on educated workers . . . . . . . . . . . . . . . . 32
3.2.2.3 The impact of job loss on manufacturing workers . . . . . . . . . . . . . 33
iii
3.3 Further research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5 Table and figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Chapter 4: Measuring professional cycling teams relative performance using fixed effects regression 46
4.1 Organization of world cycling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Point systems and data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.1 Data source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.2 PCS Points and UCI points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.3 The distribution of PCS points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3 Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.6 Figures and tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
iv
ListofTables
3.1 Summary statistics for treatment and control group at time of job displacement . . . . . . 38
4.1 List of professional cycling teams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 Moments of the distribution of PCS points scored by individual riders . . . . . . . . . . . . 56
4.3 Summary of the best individual seasons in terms of PCS points . . . . . . . . . . . . . . . . 59
v
ListofFigures
2.1 The impact of minimum wage on labor market dynamics for median wage workers . . . . 19
2.2 The impact of minimum wage on labor market dynamics for low wage workers . . . . . . 20
2.3 The impact of minimum wage on labor market dynamics for low wage female workers . . 21
2.4 The impact of minimum on labor market dynamics for low wage workers without college
education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 The impact of minimum wage on labor market dynamics for low wage workers with low
job tenure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 The impact of minimum wage on labor market dynamics for part-time workers . . . . . . 24
3.1 Impact of job displacement on total earnings . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Impact of job displacement on days employed . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Impact of job displacement on wages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4 Impact of job displacement on total earnings by age at time of displacement . . . . . . . . 40
3.5 Impact of job displacement on days employed by age at time of displacement . . . . . . . . 40
3.6 Impact of job displacement on wages by age at time of displacement . . . . . . . . . . . . . 41
3.7 Impact of job displacement on total earnings by education level . . . . . . . . . . . . . . . 41
3.8 Impact of job displacement on days employed by education level . . . . . . . . . . . . . . . 42
3.9 Impact of job displacement on wages by education level . . . . . . . . . . . . . . . . . . . . 42
3.10 Impact of job displacement on total earnings by industry . . . . . . . . . . . . . . . . . . . 43
3.11 Impact of job displacement on days employed by industry . . . . . . . . . . . . . . . . . . 44
vi
3.12 Impact of job displacement on wages by industry . . . . . . . . . . . . . . . . . . . . . . . 45
4.1 World Tour teams’ relative performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2 World Tour teams’ relative performance, measured on top riders only . . . . . . . . . . . . 60
vii
Abstract
I use the introduction of a federal minimum wage in Germany in 2015 to estimate the impact on market
regulation of labor market dynamics. I exploit large administrative data to identify the response of aggre-
gate labor markets. Using an event study analysis, I show that the separation rate for low-wage workers
in exposed sectors increased when the minimum wage was announced, but not when it became binding,
suggesting firms proactively adjusted their workforce. However, I find no evidence that labor markets
became less dynamic over time because of the introduction of the minimum wage. Additionally, I find
evidence of firms substituting full-time workers with part-time workers only in the very short run, but
not in the medium run. My results are consistent with a theory where only larger firms can exert market
power in local labor markets.
viii
Chapter1
Introduction
ch:introduction
Despite its title, this dissertation will discuss topics that are sometimes loosely related to the study of mini-
mum wage policies. A better way to think about the overall consistency and internal structure is to see this
work as a collection of statistical studies using the tools provided by labor economics to analyze real-world
applications. In the first chapter, which is the main part of the dissertation, I estimate the impact of the
introduction of a federal minimum wage in Germany in 2015 on the dynamics of labor markets across re-
gions and industries. This exercise provides policy evaluation for a widely used but controversial measure
and widen our knowledge and understanding of how price floors impact economic markets. The second
chapter uses similar data from the German Social Security administration to investigate the heterogeneity
in the cost of job loss, a concept for which there are still debates surrounding measurement and definitions.
Finally, the third chapter stray further away from labor markets to evaluate relative team performance:
the ability for a team to have individuals perform better than they would in other teams. I use a fixed-
effect regression traditionally used to estimate wages to build a measure of relative performance within
professional road cycling team, a collective sport where individual results are easily identifiable.
1
Chapter2
TheimpactofminimumwageonlabormarketdynamicsinGermany
ch:min_wage
Minimum wage is one of the most used regulatory policy in developed countries. In 2022, all OECD coun-
tries implement some kind of wage control. However, despite a large empirical literature
∗
, the political
debate around minimum wage is not always well informed. Key arguments against minimum wage of-
ten relies on the idea that out-of-work individuals will struggle to access jobs, highlighting the potential
adverse impact on labor market flows.
In this paper, I provide new estimates of the impact of minimum wage on labor market dynamics using
the introduction of a federal minimum wage in Germany in 2015. The German context provides several
advantages. First, previous estimates of the response of labor market flows often relies on small subgroups
of workers. For example, Dube and al. [24] use US administrative data to study the response of teens and
restaurant workers, two groups that are historically highly affected by minimum wage. Similarly, Brochu
and Green [5] focus on teenage workers using Canadian survey data while Portugal and Cardoso [42] do
so using reforms in Portugal in the 1980’s. Examining only small subgroups of workers leads to estimates
that are more precisely identified but that may not be representative of aggregate labor market conditions.
By studying a large federal policy and its impact on national labor markets, I provide estimates that are
better fitted to inform political debates [3, 39].
∗
See Neumark and Wascher [41] and Card and Krueger [11] for reviews
2
Nonetheless, this large natural experiment allows me to explore potential heterogeneity between work-
ers. Specifically, I estimate how labor market dynamics for workers with relatively lower wages are affected
differently by the introduction of the minimum wage. Following Ahlfeldt and al. [2], I focus on workers
who fall in the bottom 15% of the wage distribution in their sector and study how exposure to minimum
wage generates different responses. Exposure is measured by the share of workers who earned less than
the future minimum wage on June 30th 2014, six months prior to the reform. Identification relies on the
fact that relatively low paid workers in high paying sectors are less exposed to minimum wage because
their wage is comparatively higher in absolute terms. On the other hand, low paid workers in sectors that
pay relatively lower wages should be more exposed to the reform. Hence, any effect of the reform should
be more pronounced for this second group of workers.
In an event study analysis, I interact my measure of exposure with time dummies around 2015 to es-
timate how labor market flows in sectors that are relatively more exposed react differently than in less
exposed sectors. In addition, this framework allows me to disentangle short-term effects from long-term
ones by providing timed estimates of labor market dynamics’ responses. This approach builds on Card
[8] and a recent literature examining German minimum wage to provide new insights about the macroe-
conomic impact of minimum wage reforms. Namely, I provide new estimates of the impact of minimum
wage on labor market dynamics at the national level in a large developed economy.
My results show that the introduction of minimum wage did not have substantial long term impact on
labor market dynamics for low paid workers in more exposed sectors. The share of workers transitioning
from employment to unemployment increased slightly when the minimum wage was announced, but not
when it became legally binding, suggesting firms adapted proactively to the policy. However, this effect
quickly fades away after the first quarter of 2015. This is consistent with studies that find limited increase in
unemployment rates associated with minimum wage [2, 6]. In the meantime, hiring rates did not decrease
in more affected sectors, suggesting that unemployed workers did not face harder conditions to find jobs.
3
Additionally, the share of workers who change jobs without going through unemployment (job-to-job
transition) remained constant. These results differ from previous estimates obtained from restricted groups
of workers [5, 24, 40]. I interpret this discrepancy as evidence that estimates obtained from limited samples
may not be representative of aggregate labor market response to policies, but instead provide valuable
information about how specifically vulnerable workers may be affected. In this spirit, my results are in
line with those by Cengiz and al. [12], who find limited effect on labor market flows in the aggregate using
US data. I provide evidence that these patterns also exist in Germany, despite the different institutional
settings between both countries.
Additionally, I estimate how transitions in and out of part-time employment were impacted by the
introduction of minimum wage. I find that the share of workers who transition from full-time employment
to part-time employment did not increase in sectors that are more exposed to the policy, neither before
or after it became legally binding. The hiring rate of part-time workers also remained stable, suggesting
that firms did not adapt to the new regulatory constraint by reallocating their workforce towards workers
with fewer hours worked.
This paper relates to several strands of literature. First, I contribute to the recent work investigating
the policy impact of the minimum wage in Germany [7, 25] by providing estimates of labor market flows
responses. In this regard, I contribute to the small literature which estimated these responses [5, 24, 40,
12] by investigating a national policy change in Germany. Hence, my estimates differ by examining the
impact of a significantly larger reform on aggregate labor markets rather than on targeted group of highly
exposed workers. My results are consistent with those by Cengiz and al. [12] who also identify the effect
on a larger subset of exposed workers. Finally, I also contribute to the large empirical literature regarding
minimum wage [41, 11] by providing new insights of the policy in Germany.
4
In the rest of the paper, section 2.1 provides some institutional background about the minimum policy
in Germany. Section 2.2 describes the empirical strategy and section 2.3 provides the main results. Sub-
sequently, section 2.4 examines in more detail the labor market dynamics of part-time workers. Finally,
section 2.6 concludes.
2.1 TheminimumwageinGermany
sec:germany
2.1.1 Background
Until recently, Germany was one of the few European countries without a generalized minimum wage.
Historically, wages were negotiated at the sectoral level between labor unions and firms while codetermi-
nation at the firm-level ensured ample flexibility [33]. However, with the acceleration of globalization in
the 21st century, this model proved to be insufficient: more and more firms were avoiding collective bar-
gaining agreements while codetermination was helpless against majority votes coming from shareholders.
As a result, inequalities steadily rose [10] and low-wage workers conditions deteriorated. This situation
got amplified by the financial crisis and the subsequent Eurozone instability triggering many unions to
push for a federal minimum wage level. This lead the left-wing SPD party to make minimum wage one
of their main economic promise during the 2013 election. After winning enough seats to force a coalition
with Angela Merkel’s CDU, the SPD was able to bring the topic in front of the German parliament in early
2014. Wide popular support and a strong political majority made for a quick adoption in June 2014. The
new reform a federal minimum wage of 8.50 euros per hour, applicable to almost all workers in Germany.
†
It is worth noting that this new wage floor was set at remarkably high-level for a first-ever policy, sitting
at roughly 40% of the median wage in the country. Studies from the OECD confirmed that the German
minimum wage was one of the highest in Europe in relative terms, behind only France and Luxembourg.
†
Some exceptions were included for sectors were minimum wages already existed, allowing for extended period of time to
align sectoral agreements with the federal regulation. See [7] for an in-depth discussion.
5
At the time of the reform on June 31st 2014, around 15% of workers were earning less than the future
minimum wage. This share, which I will refer later on as the bite of the minimum wage was even higher
for women (21%). However, because the measure applies equally to all jobs in Germany, there is extensive
variation in how actually strict the minimum wage is across sectors and regions. Specifically, sectors
and regions that pay higher wages exhibit a lower minimum wage bite as less workers earn below the new
regulatory constraint. For example, the bite of the minimum wage in the finance and insurance was around
6%, while it climbed all the way to 30% in the agricultural sector. Geographically, former East Germany
regions were generally more exposed to the reform as the level of wages there tends to be lower [7].
2.1.2 Data
I use data from the social security system in Germany, provided by the Research Data Centre (FDZ) of the
German Federal Employment Agency (BA) at the Institute for Employment Research (IAB). The Sample of
Integrated Labour Market Biographies (SIAB) offers data on all notifications to the social security system
for a 2% random sample of all individuals who have ever been registered in the German social security
system. It consists of complete day-to-day information on earnings and time worked in each employment
spell, along with basic demographic characteristics including education, as well as information on occu-
pation, industry and receipt of unemployment benefits. In addition, this panel contains information about
the establishment that hires each workers, such as branch of industry, the location of an establishment,
and the number of employees liable to social security. I follow procedures detailed by [15] and [26] to
clean the spell data and create a quarterly panel of workers, spanning from 2008 to 2019.
6
2.2 Empiricalstrategy
sec:empirics
In this section, I focus on workers’ transition rates in and out of employment. I take advantage of the
fact that the SIAB panel tracks workers when they receive unemployment benefits to define the following
variables:
• A separation rate which measures for each industry in each region the share of full-time workers at
time t that receives unemployment benefits in t + 1. I interpret this as the share of workers who
flow from employment to unemployment.
• Ahiringrate which measures the share of full-time workers at timet who were benefit recipients at
timet− 1
• A job-to-job rate which measures for each industry in each region the share of full-time workers at
timet who work for another establishment in timet+1
In the following analysis, I compute these rates for distinct groups of workers. My objective is to study
how workers with different characteristics were affected differently by the introduction of the minimum
wage. Specifically, I want to estimate whether workers who are relatively more exposed to the minimum
wage are impacted differently than workers whose earnings were higher. To do so, I regress labor market
transition rates on exposure to the reform, as measured by the bite of the minimum wage, using an event
study analysis:
y
g
irt
=α +
8
X
k=− 4
β k
I(t =Q1
2015
+k)× bite
ir
+π t
+X
irt
+ϵ irt
(2.1)
eq:dynamics eq:dynamics
where I refers to the indicator function and y
g
irt
represents the transition rate of workers in group g in
industryi in regionr at timet. To account for seasonality in industry specific business cycles, X
irt
includes
season*industry fixed effects, while pi
t
controls for aggregate time trends. The coefficient of interest β captures how labor market flows react to the introduction of the minimum wage around the time of the
7
policy, controlling for the level of the bite in each sector. For example, in an industry where 20% of workers
earn less than minimum wage,β k
=− 0.1 means that the transition rate falls by 0.1∗ 20 = 2 percentage
points in the concerned quarter. To put this number into perspective, recall that the overall separation is
close to 5.5%. A 2 percentage point decrease would bring it down to 3.5%, which represents a 36% drop.
With more than 40 million workers in Germany, this represents more than 800,000 individual transitions
in a single quarter.
Following [2], workers are grouped together based on where they fall in the industry-specific income
distribution. I consider workers earn a relative low income if they earn in the bottom 15% of their industry
wage distribution. This threshold varies across industries, meaning they may not earn low incomes in
absolute terms, especially in high-paying sectors. Hence, the model in equation 2.1 compares low income
workers in low-paying sectors to low income workers in high-paying sectors. The former are very likely
to be directly affected by the minimum wage while the latter may not. This variation allows me to identify
the differential impact of minimum wage across sectors and regions.
As with any difference-in-differences estimation, identification relies on the assumption that the in-
troduction of the minimum wage is the only relevant economic change before and after January 1st 2015.
Because I rely on the interaction between the time dummy with the bite of the minimum wage, any po-
tential omitted variable would need to be affect different sectors in proportion with their exposure to the
policy. Because the estimation computes timely responses, I can investigate the existence of pre-policy
trends using coefficients β k
for which k < 0. In all cases, I find little evidence of pre-policy impact,
limiting the concern over omitted variables.
2.3 Results
sec:result
In this section, I estimate the impact of minimum wage on aggregate labor market dynamics for low income
workers. The main objective is to assess whether the introduction of a minimum wage had large aggregate
8
impact, something that was at the heart of the political debate surrounding the issue. I proceed to estimate
the model in equation 2.1 for all low income workers. The identification relies on the comparison between
workers in highly exposed sectors and workers in sectors that are less exposed. The main hypothesis
is that workers whose sectors are more exposed to the new regulatory constraint are more likely to be
impacted by the policy.
To ensure the specification in equation 2.1 makes sense, I reproduce the results for workers who fall
around the median of their industry-specific wage distribution. Because these workers earned more than
the minimum wage prior to the reform, even in low-paying sectors, I do not expect the policy to have
a strong impact on the labor market dynamics of median-wage workers. These results are displayed in
figure 2.1. This is no surprise to see that minimum wage have no significant impact on the labor market
prospects of median-wage workers. Because these workers’ income are often well above the new regula-
tory constraint, their exposure to the minimum wage is low and the policy does not impact them directly.
More interestingly, I now report the results obtained on workers who lies in the bottom of their industry
specific wage distribution. In low-wage sectors, these workers will often become minimum wage workers
once the policy is implemented making them particularly exposed to the new wage floor. In high-wage
sectors where the bite of the minimum wage is low, even workers in the bottom 15% of the distribution earn
more than the minimum wage in absolute terms. Hence, the exposure to the minimum wage decreases
and so should any direct impact on labor market dynamics. The results are displayed in figure 2.2.
The separation rate shows little response after the introduction of minimum wage. However, there
is a positive effect at the announcement of the policy, suggesting some firms proactively adjusted their
workforce before the minimum wage became legally binding. Quantitatively, I find the separation rate
increased by 3 percentage points in industries with a 20% bite, roughly a 50% increase in the frequency
of separation. Over a quarter, this represents around 400,000 additional separations across the countries.
Because this increase is proportional to the bite, these separations are more occurring more frequently in
9
highly exposed sectors. Nonetheless, this effect quickly fades away and remains null up to two years after
the reform. Hence, it seems that low wage workers were not flowing into unemployment more often after
the reform. On the other hand, the hiring rate also does not seem to be affected by the introduction of
minimum wage. Results remains seasonal, despite the inclusion of industry-season fixed effects. However,
this seasonal cycle is literally identical in 2014 and 2015. The cycle seems to be less volatile in 2016,
suggesting that hires became potentially less seasonal over time. Taken together, these results imply that
low-wage workers were not more likely to be separated from work, nor less likely to be hired once the
minimum wage was binding. This is consistent with previous studies that found the minimum wage in
Germany had little causal impact on the level of unemployment [7].
In addition, the job-to-job transition rate shows literally no movement around January 2015. I find that
low wage workers were not more likely to change jobs. Note that this can be consistent with results from
[25] which find that workers moved from small low-paying firms to large high-paying firms. Keeping the
number of flows constant, worker reallocation across firms can happen if flows from small firms to large
firms become relatively more important. I do not look at the direction of the flows here, only the proportion
of workers who did change jobs. Hence, this reallocation happened without a jump in the frequency of
flows, but rather due to a change in the direction of job-to-job transitions.
These results suggest that the introduction of minimum wage had no systematic impact on labor mar-
kets dynamics for low-wage workers in the aggregate. This is reassuring in the sense that it softens con-
cerns regarding the entrenchment of labor markets. Indeed, I find no evidence that workers on-the-job
were more likely to remain employed, and more positively, no evidence that low-wage workers were less
frequently hired due to higher wages. These findings differ from previous estimates in the literature. In the
US, both [24] and [40] report negative estimates while [5] find similar results in Canada. The scope of these
studies was however limited since analysis was restricted to small groups of workers (teens and restaurant
workers). Contrary to these studies who look at relatively small samples of highly exposed workers, my
10
estimates rely on more aggregated comparisons of workers. Using machine-learning techniques, [12] were
also able to expand the sample of workers belonging in their treatment group and report results that are
similar to mine.
Accordingly, one might still be concerned that some smaller groups of workers might be adversely
affected, but that this effect disappear in the aggregate. To address this concern, I now proceed to explore
potential heterogeneity behind my results. Following the literature [7], I identify three categories of work-
ers that are likely to be more affected by the minimum wage: female workers, workers without college
education and workers with low job tenure. I turn my attention to each group, providing more detailed
insights in the response of labor market dynamics to minimum wage.
2.3.1 Femaleworkers
Female workers in Germany usually earn less than men. This is also true for low-wage workers. Indeed,
10.4% of male workers earned less than 8.50 euros in June 2014. This share rises to 21.8% when considering
women. Hence, women were more exposed to the minimum wage, and as such were more likely to be
affected by the new regulation. In this section, I estimate the impact of minimum wage on labor market
dynamics for female workers. As above, I estimate equation 2.1 but I compute transition rates on female
workers who earned relatively low wage in their industry. Thus, I compare low-wage female workers in
low-paying sectors to low-wage female workers in high-wage sectors. I display the results in figure 2.3.
The separation rate for women displays a stronger response than the one estimated on the full universe
of low-wage workers. Specifically, I find a large increase in the separation rate in the second half of 2014,
when the minimum wage was announced but not yet legally binding, Quantitatively, a coefficient of 0.2
represents a 6 percentage point increase in the separation rate in an industry where the bite is 30%, or
equivalently a doubling of the frequency of separations. This effect is larger that the one obtained on the
full sample, suggesting that women were more affected by the introduction of minimum wage than men.
11
This is consistent with the fact that women were more exposed to the policy. The effect disappears over
time, meaning that once firms adjusted to the new regulatory framework, female workers were not more
likely to be separated from their job because of minimum wage.
Similarly to the results obtained in the full sample, I find that the hiring rate remains seasonal but
that the seasonal variations are not affected by the introduction of the minimum wage. In addition, the
job-to-job rate shows no response to the policy. In this regard, female workers do not differ from the full
sample. Overall, these results are consistent with the fact that women were more exposed to the minimum
wage: I find larger positive effects when they exist and I find similar zeros than in the full sample.
2.3.2 Loweducationworkers
In this section, I compute the labor market dynamics for workers who do not have any college education, a
group that the literature often identifies as more likely to be impacted by the minimum wage. Indeed, the
link between education levels and absolute income is a well-documented fact. This is also true in Germany
where the bite of the minimum wage rises to 45% among workers with no college education nor vocational
training. I proceed to estimate the model in equation 2.1 for low education workers only. This allows me
to investigate whether workers with lower educational levels were relatively more affected by the policy.
The results are displayed in figure 2.4.
Contrary to other exposed workers, I do not find evidence that the separation rate for workers without
college education increased when minimum wage was announced. There is a positive response at the time
when the minimum wage became binding, suggesting firms adapted only later. This effect is significant
at the 90% level, but fades away in the following quarters. Overall, it seems that the separation rate for
low-wage workers without college education was less affected (!) by the introduction of minimum wage
than for other low-wage workers.
12
One potential explanation might come from the underlying sorting of workers across industries. In-
deed, equation 2.1 compares low-wage workers in low-wage sectors to low-wage sectors in high-wage
sectors. Hence, if entry into high-wage sectors is conditional on educational achievements, it may be
that there are very few workers without college education among these sectors. As such, it would make
the comparison specifically difficult which could explain not only the unconventional result, but also the
relatively high standard errors obtained on these results.
As with the full sample, I find no evidence of aggregate impact of the minimum wage on the hiring rate
or the job-to-job rate for low education workers. Hence, there is no evidence that workers without college
education were relatively more impacted by the introduction of the minimum wage despite the fact that
they were more exposed to the policy.
2.3.3 Lowtenureworkers
As pointed out by [34], job tenure in Germany is strongly associated with job security: workers in their
first year on-the-job are more likely to be separated than others. To explore whether this effect interacts
with the impact of minimum wage, I estimate how labor market dynamics for low tenure workers respond
to the wage floor. I proceed to estimate the model in equation 2.1 for workers whose job tenure is less
than a year. By doing so, I compare low-wage workers with low tenure in low-wage sectors to low-wage
workers with low tenure in high-wage sectors. As above, the identification relies on the idea that low-
wage workers in high-wage sectors are less exposed to minimum wage because even though their wage
is relatively low, it is still higher than minimum wage in absolute. Results are displayed in 2.5. Note that
there is no hiring rate, since all new hires have no job tenure by definition.
As was the case for women, it seems that the separation rate of low-wage workers with low job tenure
reacted more strongly to the introduction of minimum wage than for other low-wage workers, even though
13
the confidence interval is large. This is consistent with the fact that job tenure is also correlated with in-
come, making low-tenure workers more exposed to the regulatory constraint. However, it does not seem
that low job tenure add another component to this effect, by judging the size of the estimated coefficient.
Similar to the full sample, this depicts that firms adapted their workforce when minimum wage was im-
plemented, but that there was no long-lasting effect on the separation rate. This provides some evidence
that job security for newly hired low wage workers did not deteriorate significantly because of minimum
wage.
On the other hand, the job-to-job transition rate does not seem to be impacted by the introduction of
the minimum wage. This is consistent with results described above. As pointed out by [34], changing jobs
is a frequent way to climb the job ladder for newly hired workers. My results suggest that this channel did
not react to the introduction of minimum wage for low wage workers, and that the job ladder remained a
consistent path to higher earnings.
2.4 Part-timeworkers
sec:parttime
In this section, I focus on the role of part-time work in firms’ adaptation to the new regulatory framework.
Indeed, rather than laying off workers to limit the increase in labor costs, firms can convince workers to
reduce their working hours. [7] indeed report that hours worked have decreased following the introduction
of minimum wage. To study this mechanism, I compute the following labor market dynamics:
• A full-time to part-time transition rate which measures the share of full-time workers at time t in
each industry in each region that work part-time int+1
• A part-time hiring rate which measures the share of part-time workers at time t that received un-
employment benefits in time t− 1
14
• A part-time separation rate which measures the share of part-time workers at time t that receives
unemployment benefits in time t+1
My goal is to study whether firms substituted full-time workers with part-time workers in order to
reduce potential increases in labor costs following the introduction of minimum wage. I estimate the
response of part-time workers market dynamics using the model in equation 2.1. As before, identification
relies on the comparison between low-wage part-time workers in highly exposed sectors and low-wage
part-time workers in less exposed sectors. Results are displayed in figure 2.6.
The transition rate from full-time to part-time does not seem to be affected by the introduction of
minimum wage. Given previous results regarding hours worked [7], this might be surprising. this result
mainly highlights one of the limitation of the data. Indeed, I do not observe hours worked directly. Instead,
I only observe what firms report about their workers. To be classified as part-time worker, one need to
work less than 18 hours per week. Hence, I cannot observe small reductions in hours worked that do not
cross this threshold (for example, from 40 hours to 35 hours per week).
On the other hand, the hiring rate for part-time workers did increase both when minimum wage was
announced and when it became legally binding. This suggests that firms were more likely to use part-time
workers to solve short-term labor needs in more exposed sectors than in high exposed sectors. However,
these effects quickly disappear, suggesting that there was no long-lasting effect. This provides some evi-
dence that firms substituted part-time workers to full-time workers in the short run, but it is unlikely that
they did so in the long run.
Finally, the separation rate for part-time workers increase slightly when the minimum wage is im-
plemented. The effect remains smaller than the one obtained on full-time workers, suggesting that firms
were more likely to adjust the size of their workforce by separating from full-time workers. Since part-time
workers have fewer hours, the wage increase imposed by the minimum wage is relatively smaller than for
full-time workers, making it easier for firms to absorb the subsequent cost increase. Thus, it is logical than
15
part-time workers were separated less often. However, there is no sustained effect on the separation rate
over time, similarly to results on full-time workers.
Overall, these results suggest that firms did use part-time workers as a short-term solution to absorb
the increase in labor cost associated to the rise of the minimum wage. However, I find no evidence that
workers were more likely to be hired part-time rather than full-time in the medium run.
2.5 Theoreticaldiscussion
In this section, I interpret my results under the light of two competing theories of labor markets: the
classical model of competitive markets and the more recent approach that considers firms’ market power.
Regardless of the shape of competition, when a worker and a firm are matched in a job, it creates a surplus
S which depends on workers characteristics, firm productivity and the match-specific attributes. Not
entering wage setting discussions,
‡
I will assume the wagew is the share of this surplus that accrues to
the worker: w = ϕ × S where ψ is often referred to as the worker’s bargaining power. Symmetrically,
firms receives the remaining share of the surplus (1− ψ )× S. In this setting, an increase in the level of
minimum wage does not affect the value of a worker-firm match, but rather how the value of this match
is allocated between the two parties. Everything else equal, minimum wage regulations behaves as an
external force rising the value of low-wage workers’ bargaining power.
If labor markets are competitive, firms will maximize their profits by equalizing their share of the
surplus with their marginal revenue. Hence, when the bargaining power of workers increase, the share of
the surplus received by the firm drops below its marginal revenue. This match is no longer profitable for
the firm. As a result, the firm will separate from this worker to find better matches whose total surplus
S is higher. In terms of labor market dynamics, this model predicts a peak in the separation rate when
the minimum wage hits as firms separates from workers whose matches become unprofitable. The model
‡
See David Card’s Nobel lecture for an in-depth discussion of this topic [9]
16
also predicts a decrease in the hiring rate: because firms require a higher value of S for the match to be
profitable, fewer matches fit these expectations and it becomes less likely for workers to enter one.
If firms possess some market power on the labor market [30, 49], the value of the surplus’ share accrued
to employers will be larger than the marginal revenue: (1− ϕ )× S≤ MR. When the bargaining power of
workers increase, some matches will remain profitable so long as firms reduce their markdown on wages.
In this sense, the introduction of minimum wage regulations can revert existing distortions on the labor
market [39]. In this case, the introduction of minimum wage should not impact the separation rate nor
the hiring rate. Matches that were profitable prior to the reform still are and workers will face similar
opportunities. What changed is how the value created by these matches is shared between employers and
employees.
My results do not perfectly confirm any of these two competing theories. Indeed, I do find that the
separation rate peaked when the minimum was implemented suggesting that labor markets were fairly
competitive prior to the reform. However, I do not find evidence that the hiring rate dropped in the long
run which is predicted when firms do have market power. These two apparently opposite facts can be rec-
onciled once we recognize that firms may differ in the amount of market power they exert. Indeed, larger
more productive firms are often able to markdown wages more than smaller ones [4]. Hence, when mini-
mum wage becomes binding, low-wage workers in smaller less productive firms are likely to be terminated
which leads to the observed peak in the separation rate. However, these workers will have opportunities
to find work in larger firms who reduce their markdown. The reallocation of workers from small firms
towards large one is also expected, as shown by [25].
2.6 Conclusion
sec:conclusion
In this paper, I use the introduction of a federal minimum wage in Germany to provide new estimates of
the impact of market regulations on labor market dynamics. Contrary to previous studies, I am able to use
17
a large administrative dataset to identify the impact of minimum wage on aggregate labor markets rather
than small targeted groups of workers. I show that minimum wage did not have a long-term significant
impact on labor market dynamics, suggesting that low-wage workers in exposed sectors were not more
adversely impacted than those in sectors that were less exposed. This suggests that the adverse effect of
minimum wage on labor market dynamics were limited.
Figures
18
Figure 2.1: The impact of minimum wage on labor market dynamics for median wage workers
fig:median
These graphs display the impact of minimum wage on labor market dynamics for workers whose earnings are
between the 40th and 60th percentile of their industry wage distribution, as estimated by 2.1. Vertical error bars
represents the 95% confidence interval. Standard errors are clustered at the regional sector level.
19
Figure 2.2: The impact of minimum wage on labor market dynamics for low wage workers
fig:main
These graphs display the impact of minimum wage on labor market dynamics for workers whose earnings fall below
the 15th percentile of their industry wage distribution, as estimated by 2.1. Vertical error bars represents the 95%
confidence interval. Standard errors are clustered at the regional sector level.
20
Figure 2.3: The impact of minimum wage on labor market dynamics for low wage female workers
fig:women
These graphs display the impact of minimum wage on labor market dynamics for female workers whose earnings fall
below the 15th percentile of their industry wage distribution, as estimated by 2.1. The vertical error bars represents
the 95% confidence interval. Standard errors are clustered at the regional sector level.
21
Figure 2.4: The impact of minimum on labor market dynamics for low wage workers without college
education
fig:educ
These graphs display the impact of minimum wage on labor market dynamics for workers who do not have college education
and whose earnings fall below the 15th percentile of their industry wage distribution, as estimated by 2.1. The vertical error bars
represents the 95% confidence interval. Standard errors are clustered at the regional sector level.
22
Figure 2.5: The impact of minimum wage on labor market dynamics for low wage workers with low job
tenure
fig:tenure
These graphs display the impact of minimum wage on labor market dynamics for workers with less than one year in job tenure
and whose earnings fall below the 15th percentile of their industry wage distribution, as estimated by 2.1. The vertical error bars
represents the 95% confidence interval. Standard errors are clustered at the regional sector level.
23
Figure 2.6: The impact of minimum wage on labor market dynamics for part-time workers
fig:part_time
These graphs display the impact of minimum wage on labor market dynamics for part-time workers, as estimated by 2.1. The
vertical error bars represents the 95% confidence interval. Standard errors are clustered at the regional sector level.
24
Chapter3
Heterogeneityinearningslossesfollowingjobdisplacement
ch:jobloss
The rise of unemployment during the Great Recession led to a renewed discussion about the costs of job
loss. In particular studies have found that losses in earnings for displaced workers are large and persistent
over time [48, 14].
∗
A lot of attention has been devoted to undercover some of the causes behind these ef-
fects. Davis and al. [21] point out to the role of labor markets conditions, showing that earnings losses are
larger when workers are displaced during recessions when unemployment is higher. Consistently, Fallick
and al. [27] show that the length of the subsequent unemployment spell is a strong predictor of the mag-
nitude of long-term earnings drops. Another strand of literature shows that displaced workers move from
high-wage firms to low-wage firms [43, 36] and that these moves explain the majority of losses. Finally,
others focused on how workers could mitigate these effects by switching industries [44] or enrolling in
re-training programs [37].
Despite this extensive literature, little is known about how the cost of job loss differs across individuals.
This paper investigates how demographic characteristics prior to displacement can predict the impact of
job loss. Hence, the exercise is to identify workers who are exposed to higher risk of large losses. I use
administrative data from Germany to follow workers along their career and explore how job displacement
affects different kind of workers differently. This exercise is important for several reasons. First, knowing
which workers face higher risk prior to displacement can help designing insurance schemes that reflects
∗
For example, [45] find an average reduction of roughly 20% in total earnings 10 years following job displacement.
25
these differences. Low-risk workers who experience only short drops in earnings may not need the same
kind of insurance than those who face high-risk of persistent losses. Second, this issue relates to larger
trends in the society, since the cost of job loss correlates with political polarization [22, 23], rise in inequal-
ities, and local labor market conditions [17]. As a result, investigating which individuals are most affected
by job loss can help explain larger trends that have drawn a lot of attention recently [19].
I follow the literature [32, 43] by estimating the impact of job displacement using an event study
framework. Replicating previous studies, I find displaced workers suffer a 5% drop in total earnings up to
10 years following job loss. Then, I perform similar analysis on specific subsamples to investigate how this
effect vary along certain demographic characteristics. I find that earnings losses are concentrated among
young workers who do not have a college degree. Decomposition by industry shows that manufacturing
workers are also more affected, which is consistent with [16] who show that the cost of job displacement
is higher in industries exposed to import competition.
The rest of the paper is organized as follows. Section 3.1 describes the German administrative data and
the empirical methodology used to identify the impact of job loss. Section 3.2 compiles my findings on the
cost of job loss for the entire sample of displaced workers and how this cost varies across age, education
level and industries. Finally, section 3.3 discusses how I plan to build on these findings in further research.
3.1 Dataandmethodology
sec:data
This section describes the German administrative data provided by the IAB and describes the methodology
used to define job displacement and build a satisfying control group.
26
3.1.1 Data
I use data from the social security system in Germany, provided by the Research Data Centre (FDZ) of
the German Federal Employment Agency (BA) at the Institute for Employment Research (IAB). The Sam-
ple of Integrated Labour Market Biographies (SIAB) offers data on basically all notifications to the social
security system for a 2% random sample of all individuals who have ever been registered in the German
social security system. It consists of complete day-to-day information on earnings and time worked in
each employment spell, along with basic demographic characteristics including education, as well as in-
formation on occupation, industry and receipt of unemployment benefits. In addition, this panel contains
information about the establishment that hires each workers, such as branch of industry, the location of
an establishment, and the number of employees liable to social security. I follow procedures detailed by
[15] to clean the spell data and create a yearly panel of workers.
3.1.2 Measuringjobdisplacement
I follow the existing literature based on administrative data and define job displacement as an event when
a worker with at least two years of tenure leaves a stable job at his main employer in the course of a mass-
layoff [32]. The analysis of workers leaving stable jobs has several advantages. It focuses on workers who
in all likelihood expected to remain in their job in the absence of a mass layoff, and thus were likely to be
surprised by being displaced. Moreover, given the steep reduction in job mobility with even a few years
of job tenure in Germany, very few of these workers were likely to have moved voluntarily. This reduces
the measurement error in the definition of job displacement and helps establish a counterfactual outcome,
since most of these workers would likely have remained in covered employment absent a job loss.
Following [43], I work with two common definitions of a mass-layoff event: mass-layoffs where em-
ployment declines by at least 30 percent, without returning to this level in the following two years; and
27
permanent establishment closings. To make these definitions meaningful, I follow the literature and con-
sider only workers whose employers had at least 50 employees in the year prior to the employment drop
and did not have large employment fluctuations in the year before. Smaller establishments are subject to
such greater employment fluctuations, so that these measures of mass-layoff are less meaningful.
†
Analy-
ses in [48] and [47] have shown that in the U.S., estimates are robust to the restrictions on job tenure, and
moderate variations in the restrictions on firm size and size of the mass-layoff. To ensure data availability
following job displacement, event timing is restricted to be between 1982 and 2008.
By focusing on job separations of high-tenured workers during mass-layoffs at medium-sized to large
employers, I obtain a very clean measure of job displacement that is comparable with the existing literature.
It is important to bear in mind that these definitions exclude many potential job losers, and this study does
not intend to capture the experiences of these other job losers.
3.1.3 Constructingasampleofdisplacedworkersandacontrolgroup
3.1.3.1 Baselinerestrictions
I construct my analysis sample in two steps. First, I denote the year of displacement as the “baseline year”
c and I choose for each baseline year all workers that satisfy the following restrictions on June 30th for that
year: the individual is male, is between age 24 and 50, and works full time at a West German establishment
with at least 50 employees, and has at least 3 years of tenure.
‡
I define an individual as displaced (between
year c and c + 1) if a) the individual leaves the establishment between c and c + 1 and b) the establishment
has a mass-layoff (or plant closing) between year c and c + 1.
I focus the main analysis on men for two reasons. First, to facilitate comparisons with the earlier
literature that has typically focused on men and, second, since the higher labor force attachment of men
†
[28] show that at least in the U.S., at a 30% employment loss the majority of workers leaving the firm are laid off rather than
voluntary quitters. Thus, this cutoff further helps in reducing measurement error from the presence of voluntary movers among
the displaced.
‡
These restrictions follow largely [43]. I drop workers younger than age 24 since they may not have fully entered the labor
force; and workers older than age 50 who had access to partial retirement programs in Germany during that period.
28
leads to less selection issues between in and out of the labor force, simplifying the interpretation of the
results. In addition, given we have data on the national labor market, the sample includes individuals
moving within Germany, and I do not impose restrictions on their presence in the labor market after job
loss. Moving to a new region might be an efficient way to recover from job loss, and hence should be
considered as a mitigating factor for earnings losses. There is a fraction of individuals that permanently
drops out of our sample, among others because they stop working, work in self-employment, work in
government jobs or move abroad.
§
3.1.3.2 PropensityScoreMatching
Displaced and non-displaced workers may differ in ways that make them difficult to compare. In particular,
because I impose a tenure requirement to be part of the treatment group, displaced workers tend to have
longer tenure in their current job. As wages increase with tenure, displaced workers are also more likely
to have increasing wages prior to displacement, a trend that might not exist for other workers. Hence, any
estimation strategy would have to account for differences in the level of wages but also for differences in
trends before the baseline year.
To alleviate these concerns, I use propensity-score matching to obtain a comparison group that pro-
vides appropriate counterfactual earnings trends for the displaced workers. Starting with the baseline
restrictions, I use a two step-matching estimator where I match within baseline year based on a number of
matching variables. Specifically for each baseline year, I estimate the propensity of being displaced using
age, job tenure, education, industry, establishment size in year c and past wages as predictors. For each
displaced worker I assign a single comparison worker, using the non-displaced worker with the closest
propensity score (without replacement). This yields a group of displaced workers and a very comparable
set of non-displaced workers working at similar firms. Note that there is no restriction that workers in the
§
To prevent potential selection bias due to attrition out of the labor market, a future version will keep workers in the sample
with 0 earnings when they disappear from the sample.
29
comparison group have to stay at the same establishment between year c and c + 1, nor that they cannot
be displaced in future years.
Table 3.1 displays demographics characteristics for workers in the treatment and the control group. As
desired, workers in the control are similar to displaced workers on many observable characteristics. The
right panel displays characteristics for observations of workers that are neither in the treatment nor the
control group. As expected, displaced workers have larger tenure and higher wages than the rest of the
workers. These differences dissapear when we look at workers in the control group, suggesting that our
matching procedure yields satisfying results.
3.1.4 Empiricalstrategy
Estimates of earnings losses following job displacement are obtained using an event study analysis. I follow
[43], and estimate the following equation:
y
itc
=
10
X
k=− 3
δ k
× I(t =c+k)× Disp
i
+
10
X
k=− 3
γ k
× I(t =c+k) (3.1)
eq:event eq:event
+π t
+α i
+X
it
β +ϵ itc
y
itc
is the outcome variable for personi with baseline yearc in yeart.Disp
i
is an indicator variable for
whether this individual was displaced or belongs to the control group. The main coefficients of interest are
δ k
which measure the change in earnings of displaced workers with respect to the baseline year (c), relative
to the evolution of earnings among nondisplaced workers captured by year effects ( π t
). It is important to
control for “year relative to baseline year” fixed effects (coefficients γ k
), since the tenure restriction in
the baseline year leads to hump-shaped earnings profiles around the baseline year even for the control
group that cannot be captured by year fixed effects alone. In addition, I control for individual effects α i
30
and time-varying control variables (X
it
), chiefly worker age. Since the matching procedure implies that
worker characteristics in the treatment and control groups are very similar at baseline, the inclusion of
both the worker fixed effects and the X
it
should make little difference to the estimates.
3.2 Empiricalresults
sec:emp_result
3.2.1 Impactofjobdisplacement
Figure 3.1 displays the evolution of earnings for displaced workers relative to individuals in the control
group. I find that displaced workers earn on average 5% less than individuals in the control group, up to 10
years following job displacement. These results are similar to the ones found in prior literature, confirming
the existence of a long unemployment scar. Surprisingly, I find no evidence of a decrease in employment,
as shown by figure 3.2. It seems that the only relevant channel for earnings losses is through wages, which
exhibits the same drop of around 5% (figure 3.3). Hence, it seems that displaced workers do succeed to find
a new job, but that this job pays lower wages in average. [43] found similar results by documenting that
displaced workers move from high-wage firms to low-wage firms.
3.2.2 Heterogeneityinearningslosses
In order to explore how earnings losses differ across indivduals, I estimate equation 3.1 on different sub-
samples of the population. Namely, I study how the scarring effect of job loss changes according to the
age at the time of displacement, worker’s education level and industry.
3.2.2.1 Theimpactofjoblossonyoungworkers
I decompose the sample of displaced workers into two groups, based on the workers’ age at the time of
displacement. I call young workers those who were less than 40 years old at the baseline year. There are
some reasons to believe that young workers may not be affected as much as others by job displacement.
31
For example, [46] explains how job changes are en essential source of earnings gains for workers at the
beginning of their career. Figure 3.4 displays how earnings are impacted by job displacement for both
young and old workers. Surprisingly, I find that young workers seem to be more affected than old workers.
Indeed, even though the initial loss is similar across age groups, older workers recover almost fully after
5 years while young workers suffer from losses for up to 8 years. As before, this effect is entirely driven
by wages, as shown in figure 3.6. Hence, even though young workers are more likely to switch jobs, these
changes are beneficial for workers so long that they are voluntary. When displaced, young workers do
not benefit from job switches, as they seem to find jobs that pay lower wages than their pre-displacement
level.
It is worth noting that the relatively fast recovery of older workers might hide some selection bias. If
older workers who fail to bounce back following job displacement exit the labor market, they disappear
from my sample. Hence, the effect estimated by equation 3.1 could overestimate the actual earnings of
older workers, as only those who manage to find a satisfying job remains in the sample. Further research
on the attrition of old workers following job displacement is required to address this potential issue.
3.2.2.2 Theimpactofjoblossoneducatedworkers
Figures 3.7, 3.8 and 3.9 replicates the analysis, focusing on workers’ educational level at the time of dis-
placement. I split the main sample into two groups: those who have a college degree prior to displacement
and those who do not. I find that the earnings losses for college graduates are close to 0, even in the first
year following displacement. This suggests that workers with higher education are more successful in
adapting to their job loss. Several factors can explain this tendency. First, college graduates may have
skills that more easily transferable between sectors. Hence, when hit by a shock, it might be easier for
them to switch to another firm or another industry, relative to workers with industry-specific skills. A
measure of skill transferability by occupation would be helpful to dwell more into this question. Second,
32
several studies found that college graduates tend to move across regions more than less educated individ-
uals [38]. Following job displacement, moving to another region where labor markets prospects are better
can be an effective strategy to limit the earnings losses. In this case, this higher mobility rate could help
explain the discrepancy between the trajectory of educated workers compared to less educated ones.This
concern is primarily important in Germany where the mobility rate is relatively low.
On the other hand, workers without a college degree suffer from persistent earnings losses, up to 10
years following job displacement. This result raises questions about the importance of these scarring effects
on inequalities. Indeed, less educated workers are more likely to earn lower earnings to begin with and
they seem to suffer from higher more persistent losses following job displacement. Estimating precisely
the contribution of this channel to the level of inequalities is beyond the scope of this paper, but it might
be an important channel to explain lifetime earnings discrepancies. In the US, [35] find that people at the
bottom of the lifetime earnings distribution are 7 times more likely to have spend one year unemployed
than people at the top of the distribution. As far as I know, this result has not been replicated in Germany
so far. Documenting heterogeneity in earnings losses along the income distribution would be an insightful
exercise and a natural extension to the present paper.
3.2.2.3 Theimpactofjoblossonmanufacturingworkers
I study the heterogeneity in earnings losses across industries by separating manufacturing workers from
others. This distinction is really rough. Manufacturing in Germany hides more diversity than it does in
the US, as some industrial sectors grew consistently over the last 40 years, helping Germany maintain
a positive trade balance with the rest of the world. I suppose that a more precise distinction between
declining sectors and growing sectors would produce more contrasting results.
Regardless, this rough distinction is enough to document substantial differences between workers.
Figure 3.10 shows that manufacturing workers experience higher losses following job displacement, even
33
though the difference attenuates over time. This is consistent with evidence that earnings losses vary with
the economic cycle, being stronger during recessions [21, 13]. Indeed, manufacturing jobs in Germany
are declining as a percentage of total employment [18]. Hence, when facing arguably tougher economic
conditions, workers in declining sectors suffer larger losses. In this spirit, [16] found that workers in
sectors particularly exposed to import competition suffer from higher scarring effects, while workers in
exporting sectors tend to suffer lower losses. Sectoral economic conditions seem to be an important driver
of the heterogeneity in earnings losses.
3.3 Furtherresearch
sec:research
This section explores several ways to improve the quality of the results I described above, and how one
can go further to ask policy-relevant questions regarding the disparate effects of job loss.
Improvingtheexistingresults
First, I want to explain the small modifications I plan to do to improve the quality of my results. As one
can see on figures 3.1, 3.4 and 3.10, the pre-trend patterns raises doubts about the quality of identification
in the current analysis. To help with this issue, I plan to define the baseline year as the year prior to job
displacement rather than the year of displacement itself. Indeed, in a yearly panel, when one is displaced
in yearc, the observation for this year already includes some time where the worker was displaced. This
means that earnings and employment are mechanically lower in year c than in year c− 1. By defining
baseline in yearc− 1, I can avoid this issue, while the full effect of job displacement would be reached only
2 years following the baseline. I expect this to also increase the magnitude of the earnings losses, from
around 5% to around 8%, making them more in line with the existing literature.
The second technical point I want to explore deals with the potential attrition of old workers into pre-
retirement programs or out of the labor force. As mentioned above, this selection problem potentially bias
34
my results and explain why I find old workers recover faster than young workers. If attrition is important,
my current results underestimate the real losses experienced by older workers. One way to explore this
issue is to keep all workers in the sample for an arbitrary number of years, giving them 0 income in
years they were out of the sample. If attrition is high, this will introduce many observations with really
low earnings, removing the observed rebound in figure 3.4. Based on future results on this point, it may
also be interesting to explore why and how old workers decide to exit the labor market following job
displacement.
Finally, I need to explore how the results change when I keep women into the sample of displaced
workers. Existing literature has contradictory results: [43] find that results on full sample are not sig-
nificantly different than results found on male workers only while [31] documents a large gender gap in
earnings losses.
Exploreheterogeneityfurther
One straightforward way to build further research on the heterogeneity of earnings losses following job
displacement is to explore differences along characteristics that I did not use so far. First, I can easily build
finer age bins to estimate more precise results along the life-cycle. When looking at older workers, this
can also be informative about potential attrition effects. For example, if workers exit the labor market
by selecting into pre-retirement programs, I should observe that attrition begins earlier for workers aged
45-50 than for workers aged 40-45. Second, I can explore the heterogeneity along the current job tenure of
displaced workers. Because wages increase with tenure and because high-tenured workers are more likely
to have invested into firm-specific human capital, high-tenured workers are likely to suffer from higher
earnings losses. It may also be possible that the 3-year tenure requirement imposed in the identification
strategy removes these effects, if they are relevant only for the beginning of one’s spell in a given job.
35
Moving this tenure requirement to different values could be another helpful way to check how robust my
results are.
It can be interesting to explore how the scarring effects of job displacement vary along the income
distribution. In the US, [29] found that workers are both extremes of the distribution were the most affected.
Such results have not been replicated for Germany, where insurance programs for disadvantaged workers
are more developed than in the US. Finally, [20] detail new techniques using machine-learning tools to
explore potential heterogeneity along dimensions researchers may have looked upon. Applying this tool
to this issue has the potential to raise new facts and questions.
Labormarketsfrictions
This paper connects naturally with the existing literature on labor market frictions. Job displacement has
such strong long-lasting effects because workers fail to find a similar job to the one they had before, moving
from high-wage firms to low-wage firm [43]. However, the causes behind this pattern remain unclear. It
can be insightful to explore what strategies people use to adapt to job loss, and how it impacts future
earnings. For example, I would like to explore the behavior of displaced individuals to document whether
they switch industries, and why they might fail to do so. As discussed above, skill transferability might be
an important factor to ease this transition, while it might be more difficult for workers who invested into
sector-specific human capital. In this regard, the role of training is crucial. The German social security
system conditions part of the unemployment benefits to participation into training programs designed to
provide skills and ease transitions to new jobs for displaced workers. The efficacy of such programs might
be a key parameter to help people facing higher income risk following job displacement.
On the other hand, sectoral moves might also be tied to geographical moves, if industries are rela-
tively clustered across cities. A low geographic mobility rate might then impede workers’ ability to switch
careers. Public policies can be double-edged: helping distressed places makes sense to provide relief in
36
the short run but might become counterproductive in the long run if it provides incentives to stay within
regions with lower economic growth. In this sense, place-based subsidies can become a friction by reduc-
ing geographic mobility of workers. The creation of institutions promoting geographic mobility can help
alleviate this concern (see for example [38]).
Finally, one can explore the role of the minimun wage as a market friction. If this price floor binds and
constraint employment in low-wage industries, the effect of job loss might be higher due to the relative
difficulty of finding another job. Germany recently increased their minimum wage. This change in policy
can be used to identify the impact of minimum wage on earnings losses following job displacement. The
rise of the minimum wage changed the composition of employment in Germany, inducing workers to shift
towards higher-wage establishments while forcing low-wage firms to close. If these plants were the main
source of re-employment for displaced workers, the rise of the minimum might shut down this channel,
creating potentially high risk for displaced workers. Exploring this channel would provide a new vision
of the long-lasting debate about minimum wage and its impact on the labor market.
3.4 Conclusion
In this paper, I looked at the discrepancies in the impact of job displacement of future earnings. This
exercise allowed me to have a first idea about which workers face the highest risk of long-lasting negative
effects. Simple demographic characteristics such as age or education seem to have a powerful impact in
this regard. The last section of this paper documented several ways I can go further with this research and
how I can contribute to the existing literature on Germany which looked into many issues: the impact of
globalization, the impact of automation or the effect of local labor markets on political outcomes.
3.5 Tableandfigures
37
Table 3.1: Summary statistics for treatment and control group at time of job displacement
Characteristics at baseline year
Treatment group Control group Others
N Mean St. Dev. N Mean St. Dev. N mean st. dv
Age 7998 38.897 6.909 7609 38.735 9.488 15406083 39.156 12.417
Education 7957 2.003 0.502 7609 2.008 0.502 13561125 1.912 0.521
Tenure (in days) 7998 3551.9 2056.6 7609 3470.9 2521.29 13830349 2158 2212
Establishment size 7998 555.64 1308.83 7609 629.12 2457.51 13755410 962.11 4180
Earnings (in log) 7989 4.843 0.418 7601 4.856 0.471 14237810 4.199 0.847
Wage (in log) 7989 10.699 0.459 7601 10.738 0.494 14237810 9.988 0.979
Days employed 7998 352.29 38.76 7609 359.56 25.65 15406083 316.36 105.9
table:baseline
This table displays summary statistics for demographic characteristics and labor market outcomes. Panel a displays this infor-
mation for displaced individuals at the time of displacement. Panel b provides similar data for individuals in the control group at
the baseline year. Panel c displays information for all the observations in the panel that are neither in treatment group nor in the
control group.
Figure 3.1: Impact of job displacement on total earnings
fig:earn_full
38
Figure 3.2: Impact of job displacement on days employed
fig:emp_full
Figure 3.3: Impact of job displacement on wages
fig:wage_full
39
Figure 3.4: Impact of job displacement on total earnings by age at time of displacement
fig:earn_age
Figure 3.5: Impact of job displacement on days employed by age at time of displacement
fig:emp_age
40
Figure 3.6: Impact of job displacement on wages by age at time of displacement
fig:wage_age
Figure 3.7: Impact of job displacement on total earnings by education level
fig:earn_educ
41
Figure 3.8: Impact of job displacement on days employed by education level
fig:emp_educ
Figure 3.9: Impact of job displacement on wages by education level
fig:wage_educ
42
Figure 3.10: Impact of job displacement on total earnings by industry
fig:earn_ind
43
Figure 3.11: Impact of job displacement on days employed by industry
fig:emp_ind
44
Figure 3.12: Impact of job displacement on wages by industry
fig:wage_ind
45
Chapter4
Measuringprofessionalcyclingteamsrelativeperformanceusingfixed
effectsregression
ch:cycling
Ranking teams is the core of all professional sports. However, results alone often lack subtleties leaving
many fans debating further. One specific point lies in the inability to disentangle team’s success from the
intrinsic abilities of their athletes. Since the best athletes often compete for the best teams, it is hard to say
which teams perform better relative to the quality of their roster. In this paper, I use two-way fixed effect
regressions to disentangle athletes’ abilities and team’s success. This provides a measure of relative team
performance, estimating teams’ ability to achieve higher results with athletes of similar strengths.
I use this method to rank professional road cycling teams over the 2011-2022 period. Road cycling
has one advantage over other professional team sports. Even though collective tactics are key to a team’s
success, riders are ranked individually at the end of each race. This provides a reliable measure of individual
performances within teams, something which often remain elusive in other collective sports. I measure this
individual performance using the points system created by the procyclingstats.com website. This ranking
accounts for race prestige and competitiveness more accurately than the official point system developed
by the Union Cycliste Internationale (UCI).
46
I build a measure of relative team performance by estimating teams’ ability to make their riders per-
form better than they would in another team. To do so, I regress riders’ individual point totals on ob-
servable characteristics, a rider fixed effect and a team fixed effect. This regression is similar to the one
first introduced by Abowd and al. [1] to estimate workers’ wages. I interpret the team fixed-effect as a
measure of relative performance: it measures each team contribution to their riders’ score, controlling for
rider individual characteristics. This provides some insights about which team is able to make their rider
consistently perform over their expected level.
Using this method, I rank every current World Tour teams in terms of relative performance. This
ranking looks radically different to the ranking in absolute performance, as measured by the amount of
points scored by each team. This suggests than some of the best teams in the world are indeed strong
because the best riders compete for them, while others are better at achieving results with less dominant
athletes. This measure should not be thought as an alternative way of measuring performance. In the end,
winning the best races remains every team’s objective, and guiding strong riders towards these wins should
be considered as success. Instead, I believe that this measure of relative performance is a complement to
absolute performance, providing a more complete assessment of team success than results alone.
The remaining of the paper is organized as follows. Section 4.1 provides a quick overview of the
institutional structure of professional road cycling. Section 4.2 discusses the different point systems that
coexist in the sport, and gives a brief overview of how points are allocated between riders. Section 4.3
details the empirical strategy used to measure teams’ relative performance, and section 4.4 displays the
results. Finally, section 4.5 concludes.
4.1 Organizationofworldcycling
sec:orga
Since 2011, road cycling is organized around a system called the World Tour. This system is managed by
The Union Cycliste Internationale (UCI) based in Switzerland. The UCI grants 18 World Tour licenses to
47
teams based on financial accounts and past performances. These licenses are granted for a 3-year cycle, so
long as teams remain financially stable. World Tour licenses grant teams invitations to the most prestigious
races of the year, called World Tour races. These includes the three Grand Tours (the Tour de France, the
Giro d’Italia and the Vuelta a España), the five Monuments of cycling (Milano-San Remo, the Ronde van
Vlaanderen, Paris-Roubaix, Liège-Bastogne-Liège and Il Lombardia) and other major races across the year.
Most teams relies on private sponsorships to finance their operations. As such, competing and performing
well in these races is crucial since they offer much better media coverage than smaller races. This is
especially true for the Tour de France which benefit from media attention even outside the typical sports
outlets.
Teams that do not receive a World Tour license compete at the Continental level, which constitute
the second tier of road cycling. It is common for race organizers to invite one or two Continental teams
from their home country, but these teams rarely compete in the entire World Tour calendar throughout
the year. Instead, they rely on smaller races, classified as .1 or .2. .1 races represents the second tier in the
race ranking, .2 the third tier.
The only race of the year that is not raced by usual teams is the World Championships, where national
teams compete instead. Winning this race crowns a rider as official world champion, and he receives a
distinctive rainbow jersey which he can wear for the entire season. In addition to the prestige of being
world champion, wearing a specific jersey provides significant media attention for sponsors. Hence, teams
often try to have their best riders in competitive form for the race, even though national teams take over.
48
4.2 Pointsystemsanddata
sec:points
4.2.1 Datasource
I personally collected data from the procyclingstats.com website. This website is widely recognized as the
main source of information regarding race results, riders’ affiliation and teams’ success. For every year
from 2011 to 2022, I gathered information on riders’ age, points scored and the teams they were riding
for. The data also contains the number of wins and podiums achieved by each rider. The final data set
contains information on 2050 different riders, spread across 36 teams. Every team that has been at World
Tour level over the period is represented in the sample, along with the top Continental teams. I decided
to include Continental teams that consistently received invitations to the top World Tour races over the
period, or who are expected to compete at World Tour level in 2023 (namely, Team Arkéa-Samsic and
Alpecin-Deceuninck). A special case was made for Euskatel-Euskadi which was a World Tour team until
its disbanding at the end of the 2013 season. The team was rebuilt in 2018 and reached Continental status
in 2020. Despite the same name, these two versions are treated as two different teams in the analysis.
The complete list of teams in the sample can be found in table 4.1, which displays the most recent name
for each teams in the sample. Changing names is common for teams that rely on sponsorship for survival,
but does not change the identity of the team. In practice, managers and riders keep their positions inside
a team despite the frequent name changes. For example, Team Sky and Ineos-Grenadiers are considered
to be the same team despite the change in naming at the end of the 2019 season.
4.2.2 PCSPointsandUCIpoints
The UCI provides a point system to measure how well riders perform. For each race they compete in,
athletes score points based on the result they achieve. Finishing first provides the most amount points, with
a decreasing scale for riders finishing second, third, etc. To account for different levels of competitiveness
49
between races, the UCI scale differs across races. World Tour results are worth more than results in lower
tier races. However, the UCI points system is flawed for three reasons. First, very few teams pay attention
to their UCI ranking and rather focus on performing as best as possible in the biggest races. This was clearly
shown in the 2022 season. The UCI announced in 2020 that a relegation system would be implemented
at the end of the 3-year cycle based on the UCI points scored by each team. Despite this being public
information, many teams were caught by surprise when their points total put them in danger of relegation.
Second, the UCI scale changed several times over the last 10 years, making comparisons over time more
difficult. Finally, and more importantly, the scales used by the UCI are biased towards small one-day races.
For example, winning a stage in the Tour de France is worth 120 UCI points, while finishing second in
worth 50. In the meantime, winning a 1.1 race is worth 125 points while finishing second is worth 80.
To avoid this bias, I instead use the points system created by procyclingstats.com, called PCS points.
This scale was created with the explicit intention to correct the bias of the UCI points system. In the PCS
points scale, a 1.1 win is worth 80 points while a Tour de France stage win is worth 100 points. This
measure of performance is more in line with the actual competitiveness of races, giving more points to
races that teams care about more.
∗
All points values given in the remaining of the paper will be in PCS
points, rather than UCI points.
4.2.3 ThedistributionofPCSpoints
By design, the point system rewards heavily riders who consistently achieve high results throughout the
year. Because teams often use less strong riders to help their leader win, the distribution of points is
highly skewed to the left. Many riders score few points, with a median score equal to 109 points. On the
other hand, high performers can score more than this in a single day. Descriptive statistics of the points
distribution are displayed in table 4.2. The entire distribution of points is displayed in figure 4.6. To better
∗
Details on how PCS points are computed for each race can be found online, at
https://www.procyclingstats.com/info.php?s=point-scales
50
visualize the right tail, I also plot the distribution of points for riders who scored more than 750 points in
figure 4.6. Top riders score much more than the average rider and account for most of the team success.
Hence, it is more valuable for teams to have one leader who perform exceptionally well rather several
slightly better than average athletes.
To give more meaning to these numbers, table 4.3 displays the 10 best individual seasons in terms of
PCS points. Without surprise, this list includes the top stars of the sport such as Tadej Pogačar, Alejandro
Valverde or Peter Sagan.
4.3 Empiricalstrategy
sec:maths
My goal is to provide a measure of teams’ performance that controls for the quality of riders in the squad.
For example, UAE Team Emirates has been one of the highest scoring team in the last few seasons. In
the meantime, their main rider Tadej Pogačar posted two of the three highest scores in the last ten years.
How to identify which part of UAE Team Emirates success is due to the team performance and which part
needs to be attributed to Pogačar’s exceptional performances? On the other spectrum, smaller teams may
be outperforming their expectations, and yet fail to score as much since they fight for smaller races or
podiums rather than wins.
To address this issue, I build a measure of relative performance that accounts for riders intrinsic qual-
ities. I do by using a two-way fixed effects regression, as described by equation 4.1.
y
ift
=α +βX
ift
+π i
+ν f
+ϵ ift
(4.1)
eq:akm eq:akm
The left-hand variable is the amount of points scored by rider i in team f in year t. On the right hand
side, X
ift
controls for rider and team characteristics that are observables. Namely, I include a quadratic
polynomial of rider’s age and a dummy for World Tour status. The team fixed effect ν f
estimates team
51
contribution to the amount of points scored by each rider. Crucially, this team contribution does not
depend on how strong a rider is, since this would be captured by the rider fixed effect pi
i
. Hence, I can
interpret the coefficient ν f
as a measure of relative performance. It evaluates how good a team is at making
their riders score more than what is expected of them.
The identification of teams relative performance relies on riders who change teams during their career.
To score well in this measure of performance, a team needs to consistently make their rider score more
than they would have in another team. Italian sprinter Elia Viviani is a good example. He scored his best
two seasons in terms of PCS points in 2018 and 2019, riding for Deceuninck-Quick Step. In these years,
he scored respectively 2,237 and 1,928 points. Outside these two years, riding for other teams, his best
seasons are 2017 (1,381 points) and 2013 (852 points). In this case, the model specified in equation 4.1
needs to assign really strong positive team effect to explain Viviani’s performances in 2018 and 2019. This
supports the argument that Deceuninck-Quick Step is a good team in relative terms, by making Viviani
outperform compared to his performances at other teams.
4.4 Results
sec:results
I estimate the model in 4.1 and report the value of teams fixed effects, measuring relative performance. I
ranked the 18 current World Tour teams, and displayed the results in figure 4.1. The height of the light
blue bars represents each team relative performance, while standard errors are depicted in black brackets.
I find that the best team in cycling during the last 10 years was Deceuninck-Quick Step. This should
come to no surprise to cycling fans, as the team is famous for its reliance on collective strengths and
excellent use of team tactics. On average, this team is able to make their riders score 250 points more than
they would if riding for Euskatel-Euskadi, the worst team in the sample.
†
It is also not surprising that most
†
Finding that Euskatel-Euskadi, in its pre-2013 version, is the worst team in the 2012-2022 period is also a positive sign that
the measure of relative performance is accurate. Indeed, the team was disbanded at the end of the 2013 season partly because it
was struggling to compete at the higher level.
52
teams are really close to each other. There are no secret recipe for success, and most teams performs as
good as they can with the riders they have.
A more unexpected result is to find UAE Team Emirates and Jumbo-Visma, arguably the two best teams
in the world these last few seasons, in the bottom three in terms of relative performances. Even though
these teams perform well and win prestigious races, it appears that they have a limited ability to make their
riders overperform their expectations. Instead, it seems teams who currently dominate World Tour cycling
have the ability to grow promising riders and make them achieve their full potential. This ability remain
extremely valuable, as home grown riders are often cheaper to acquire than already established superstars.
On the other hands, teams with smaller budget like Intermarché-Wanty Gobert or Lotto-Soudal who were
fighting in the bottom of the ranking in 2022 appear under a more positive light in relative performance,
sitting respectively third and fifth.
Because the distribution of points is heavily skewed towards top scorers, I also measure teams’ ability
to make their star riders better. To do so, I restrict the sample to riders who ever scored at least once 570
points or more in a season. This threshold corresponds to the 90th percentile of the PCS points distribution.
Hence, I keep in my sample only rider who scored in the top 10% at some point in their career. Results on
this limited sample are displayed in figure 4.2.
The ranking of teams remains stable compared to the full sample regression, even though the gap
between the best teams and the worst teams have widened. This suggests that the best teams in relative
performance top this ranking because they are able to exploit top riders better than others. This is a
reassuring sign for the validity of this measure of team relative performance.
4.5 Conclusion
sec:conclusions
This paper provides a new way of measuring sport teams performance, controlling for athletes’ individual
characteristics. I use this technique to rank professional road cycling teams, a sport where it is easier
53
to identify individual performances within teams. Interestingly, this ranking differs dramatically from
the one provided by results alone suggesting that some teams are better at guiding strong riders towards
prestigious wins while others manage to consistently make their riders perform higher than what would
be expected of them. This ranking provides a different perspective of how we assess team success and
complements measures of absolute performance.
4.6 Figuresandtables
54
Table 4.1: List of professional cycling teams
List of teams PCS Points in 2022
Current World Tour teams
Ineos Grenadiers 9,233
Jumbo-Visma 8,794
UAE Team Emirates 8,730
Bora Hansgrohe 7,263
Bahrain-Victorious 7,214
Intermarché-Wanty Gobert 6,672
Quick-Step Alpha Vinyl Team 6,471
Groupama-FDJ 5,998
Movistar Team 5,599
Cofidis 5,002
AG2R Citroen 4,854
Trek-Segafredo 4,818
Lotto-Soudal 4,551
EF Education-EasyPost 4,534
Team BikeExchange - Jayco 4,309
DSM 4,067
Israel - Premier Tech 3,949
Astana Qazaqstan Team 3,168
Major Continental teams
Alpecin-Deceuninck 6,007
Team Arkéa Samsic 4,705
TotalEnergies 3,944
B&B Hotels - KTM 1,907
Sport Vlaanderen - Baloise 1,509
Drone Hopper - Androni Giocattoli 1,240
Bardiani-CSF-Faizanè 1,141
Caja Rural - Seguros RGA 1,047
Euskatel-Euskadi (2020-2022) 493
Past World Tour teams
CCC Team
Team Qhubeka NextHash
Team Katusha Alpecin
IAM Cycling
Tinkoff
Euskatel-Euskadi (2011-2013)
HTC HighRoad
Liquigas-Cannondale
Vacansoleil-DCM
tab:teams
List of professional teams included in the sample. The sample includes every team that competed at World Tour level during the
2011-2022 and continental teams that consistently receive invitations to World Tour races and score more than 1000 PCS points
per season. An exception was made for Euskatel-Euskadi, due to the previous World Tour status of the team before 2013.
55
Table 4.2: Moments of the distribution of PCS points scored by individual riders
tab:stats
Minimum
10th
Percentile
Q1 Median Mean Q3
90th
percentile
Maximum
Value 0 10 41 109 226 271 573 3413
[t]
fig:dist
56
[t]
fig:dist_top
57
Figure 4.1: World Tour teams’ relative performance
fig:indexWT
This graph displays current World Tour teams relative performances, as measured by equation 4.1. These coefficients measure
how many additional points are scored by riders, compared to the reference team. For convenience, the reference team is Euskatel-
Euskadi in its pre-2013 version, which is the worst team in the sample.
58
Table 4.3: Summary of the best individual seasons in terms of PCS points
Rank Name Year Team PCS Points Major wins
1 POGAČAR Tadej 2022 UAE Team Emirates 3413 Tirreno-Adriatico, Il Lombardia, UAE Tour
2 SAGAN Peter 2016 Tinkoff 3315 Ronde van Vlaanderen, World Championships, GP de Québec
3 POGAČAR Tadej 2021 UAE Team Emirates 3305 Tour de France, Il Lombardia, Liège-Bastogne-Liège
4 GILBERT Philippe 2011 Omega Pharma-Lotto 3031 Liège-Bastogne-Lièege, World Championships, Amstel Gold Race
5 VALVERDE Alejandro 2015 Movistar Team 3022 Lièege-Bastogne-Lièege, Spain National Champion, La Fleche Wallonne
6 VALVERDE Alejandro 2014 Movistar Team 2959 La Fleche Wallonne, Clasica San Sebastian
7 VALVERDE Alejandro 2018 Movistar Team 2915 World Championships, Volta a Catalunya
8 VAN AERT Wout 2022 Jumbo-Visma 2862 E3 Saxo Bank Classic, Omloop Het Nieuwsblad, Bretagne Classic
9 VAN AERT Wout 2021 Jumbo-Visma 2820 Amstel Gold Race, Gent-Wevelgem
10 ROGLIČ Primož 2019 Jumbo-Visma 2818 Vuelta a España, Tirreno-Adriatico, Tour de Romandie
tab:top
59
Figure 4.2: World Tour teams’ relative performance, measured on top riders only
fig:index10
This graph displays current World Tour teams relative performances, as measured by equation 4.1. Only riders who ever scored
more than 570 points in a season are considered. These coefficients measure how many additional points are scored by riders,
compared to the reference team. For convenience, the reference team is Euskatel-Euskadi in its pre-2013 version, which is the
worst team in the sample.
60
Bibliography
[1] John M Abowd, Francis Kramarz, and David N Margolis. “High wage workers and high wage
firms”. In: Econometrica 67.2 (1999), pp. 251–333.
[2] Gabriel M Ahlfeldt, Duncan Roth, and Tobias Seidel. “The regional effects of Germany’s national
minimum wage”. In: Economics Letters 172 (2018), pp. 127–130.
[3] Dale Belman and Paul J Wolfson. What does the minimum wage do? WE Upjohn Institute, 2014.
[4] David Berger, Kyle Herkenhoff, and Simon Mongey. “Labor market power”. In: American Economic
Review 112.4 (2022), pp. 1147–93.
[5] Pierre Brochu and David A Green. “The impact of minimum wages on labour market transitions”.
In: The Economic Journal 123.573 (2013), pp. 1203–1235.
[6] Marco Caliendo, Alexandra Fedorets, Malte Preuss, Carsten Schröder, and Linda Wittbrodt. “The
short-run employment effects of the German minimum wage reform”. In: Labour Economics 53
(2018), pp. 46–62.
[7] Marco Caliendo, Linda Wittbrodt, and Carsten Schröder. “The causal effects of the minimum wage
introduction in Germany–an overview”. In: German Economic Review 20.3 (2019), pp. 257–292.
[8] David Card. “Using regional variation in wages to measure the effects of the federal minimum
wage”. In: Ilr Review 46.1 (1992), pp. 22–37.
[9] David Card. “Who set your wage?” In: American Economic Review 112.4 (2022), pp. 1075–90.
[10] David Card, Jörg Heining, and Patrick Kline. “Workplace heterogeneity and the rise of West
German wage inequality”. In: The Quarterly journal of economics 128.3 (2013), pp. 967–1015.
[11] David Card and Alan B Krueger. “Myth and measurement”. In: Myth and Measurement. Princeton
University Press, 2015.
[12] Doruk Cengiz, Arindrajit Dube, Attila Lindner, and David Zentler-Munro. “Seeing beyond the
trees: using machine learning to estimate the impact of minimum wages on labor market
outcomes”. In: Journal of Labor Economics 40.S1 (2022), S203–S247.
61
[13] Kenneth A Couch, Nicholas A Jolly, and Dana W Placzek. “Earnings losses of displaced workers
and the business cycle: an analysis with administrative data”. In: Economics Letters 111.1 (2011),
pp. 16–19.
[14] Kenneth A Couch and Dana W Placzek. “Earnings losses of displaced workers revisited”. In:
American Economic Review 100.1 (2010), pp. 572–89.
[15] Wolfgang Dauth and Johann Eppelsheimer. “Preparing the sample of integrated labour market
biographies (SIAB) for scientific analysis: a guide”. In: Journal for Labour Market Research 54.1
(2020), pp. 1–14.
[16] Wolfgang Dauth, Sebastian Findeisen, and Jens Suedekum. “Adjusting to globalization in
Germany”. In: Journal of Labor Economics 39.1 (2021), pp. 263–302.
[17] Wolfgang Dauth, Sebastian Findeisen, and Jens Suedekum. “The rise of the East and the Far East:
German labor markets and trade integration”. In: Journal of the European Economic Association 12.6
(2014), pp. 1643–1675.
[18] Wolfgang Dauth, Sebastian Findeisen, and Jens Suedekum. “Trade and manufacturing jobs in
Germany”. In: American Economic Review 107.5 (2017), pp. 337–42.
[19] H David, David Dorn, and Gordon H Hanson. “The China syndrome: Local labor market effects of
import competition in the United States”. In: American Economic Review 103.6 (2013), pp. 2121–68.
[20] Jonathan Davis and Sara B Heller. “Using causal forests to predict treatment heterogeneity: An
application to summer jobs”. In: American Economic Review 107.5 (2017), pp. 546–50.
[21] Steven J Davis and Till M Von Wachter. Recessions and the cost of job loss. Tech. rep. National
Bureau of Economic Research, 2011.
[22] Christian Dippel, Robert Gold, and Stephan Heblich. Globalization and its (dis-) content: Trade
shocks and voting behavior. Tech. rep. National Bureau of Economic Research, 2015.
[23] David Dorn, Gordon Hanson, Kaveh Majlesi, et al. “Importing political polarization? The electoral
consequences of rising trade exposure”. In: American Economic Review 110.10 (2020), pp. 3139–83.
[24] Arindrajit Dube, T William Lester, and Michael Reich. “Minimum wage shocks, employment flows,
and labor market frictions”. In: Journal of Labor Economics 34.3 (2016), pp. 663–704.
[25] Christian Dustmann, Attila Lindner, Uta Schönberg, Matthias Umkehrer, and Philipp Vom Berge.
“Reallocation effects of the minimum wage”. In: The Quarterly Journal of Economics 137.1 (2022),
pp. 267–328.
[26] Johanna Eberle and Alexandra Schmucker. “Creating cross-sectional data and biographical
variables with the Sample of Integrated Labour Market Biographies 1975-2014 - programming
examples for Stata.” In: Institute for Employment Research, Nuremberg. FDZ-Methodenreport
06/2017 (2017).
62
[27] Bruce Fallick, John C Haltiwanger, Erika McEntarfer, and Matthew Staiger. Job Displacement and
Job Mobility: The Role of Joblessness. Tech. rep. National Bureau of Economic Research, 2021.
[28] Henry S Farber. “Employment, hours, and earnings consequences of job loss: US evidence from the
displaced workers survey”. In: Journal of Labor Economics 35.S1 (2017), S235–S272.
[29] Fatih Guvenen, Fatih Karahan, Serdar Ozkan, and Jae Song. “Heterogeneous scarring effects of
full-year nonemployment”. In: American Economic Review 107.5 (2017), pp. 369–73.
[30] Boris Hirsch, Elke J Jahn, and Claus Schnabel. “Do employers have more monopsony power in
slack labor markets?” In: ILR Review 71.3 (2018), pp. 676–704.
[31] Hannah Illing, Johannes F Schmieder, and Simon Trenkle. The Gender Gap in Earnings Losses after
Job Displacement. Tech. rep. National Bureau of Economic Research, 2021.
[32] Louis S Jacobson, Robert J LaLonde, and Daniel G Sullivan. “Earnings losses of displaced workers”.
In: The American economic review (1993), pp. 685–709.
[33] Simon Jäger, Shakked Noy, and Benjamin Schoefer. “The German model of industrial relations:
Balancing flexibility and collective action”. In: Journal of Economic Perspectives 36.4 (2022),
pp. 53–80.
[34] Gregor Jarosch. Searching for job security and the consequences of job loss. Tech. rep. National
Bureau of Economic Research, 2021.
[35] Fatih Karahan, Jae Song, Serdar Ozkan, et al. “Sources of Inequality in Earnings Growth over the
Life Cycle”. In: 2018 Meeting Papers. 313. Society for Economic Dynamics. 2018.
[36] Marta Lachowska, Alexandre Mas, and Stephen A Woodbury. “Sources of displaced workers’
long-term earnings losses”. In: American Economic Review 110.10 (2020), pp. 3231–66.
[37] Pauline Leung, Zhuan Pei, et al. Further Education During Unemployment. Tech. rep. 2020.
[38] Jens Ludwig and Steven Raphael. The Mobility Bank: Increasing Residential Mobility to Boost
Economic Mobility. Hamilton Project, Brookings Institution, 2010.
[39] Alan Manning. “The elusive employment effect of the minimum wage”. In: Journal of Economic
Perspectives 35.1 (2021), pp. 3–26.
[40] Jonathan Meer and Jeremy West. “Effects of the minimum wage on employment dynamics”. In:
Journal of Human Resources 51.2 (2016), pp. 500–522.
[41] David Neumark and William L Wascher. Minimum wages. MIT press, 2008.
[42] Pedro Portugal and Ana Rute Cardoso. “Disentangling the minimum wage puzzle: an analysis of
worker accessions and separations”. In: Journal of the European Economic Association 4.5 (2006),
pp. 988–1013.
63
[43] Johannes Schmieder, Till von Wachter, and Jörg Heining. The costs of job displacement over the
business cycle and its sources: evidence from Germany. Tech. rep. Technical report, UCLA, Mimeo,
2018.
[44] Robert Schoeni and Michael Dardia. Estimates of earnings losses of displaced workers using
california administrative data. Tech. rep. PSC Research Report, 2003.
[45] Daniel Sullivan and Till Von Wachter. “Job displacement and mortality: an analysis using
administrative data”. In: The Quarterly Journal of Economics 124.3 (2009), pp. 1265–1306.
[46] Robert H Topel and Michael P Ward. “Job mobility and the careers of young men”. In: The
Quarterly Journal of Economics 107.2 (1992), pp. 439–479.
[47] Till Von Wachter, Elizabeth Weber Handwerker, and Andrew KG Hildreth. “Estimating
the’True’Cost of Job Loss: Evidence Using Matched Data from California 1991-2000”. In: US Census
Bureau Center for Economic Studies Paper No. CES-WP-09-14 (2009).
[48] Till Von Wachter, Jae Song, and Joyce Manchester. “Long-term earnings losses due to mass layoffs
during the 1982 recession: An analysis using US administrative data from 1974 to 2004”. In:
unpublished paper, Columbia University (2009).
[49] Chen Yeh, Claudia Macaluso, and Brad Hershbein. “Monopsony in the US labor market”. In:
American Economic Review 112.7 (2022), pp. 2099–2138.
64

Abstract (if available)

Abstract I use the introduction of a federal minimum wage in Germany in 2015 to estimate the impact on market regulation of labor market dynamics. I exploit large administrative data to identify the response of aggregate labor markets. Using an event study analysis, I show that the separation rate for low-wage workers in exposed sectors increased when the minimum wage was announced, but not when it became binding, suggesting firms proactively adjusted their workforce. However, I find no evidence that labor markets became less dynamic over time because of the introduction of the minimum wage. Additionally, I find evidence of firms substituting full-time workers with part-time workers only in the very short run, but not in the medium run. My results are consistent with a theory where only larger firms can exert market power in local labor markets.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on narrative economics, Climate macrofinance and migration

PDF

Essays on applied microeconomics

PDF

Three essays on macro and labor finance

PDF

Essays in labor economics: demographic determinants of labor supply

PDF

Essays on Japanese macroeconomy

PDF

Essays on the economics of climate change adaptation in developing countries

PDF

Migration, location, and economic opportunity

PDF

Essays in development economics

PDF

The impact of economic shocks on firm behavior: insights from three studies

PDF

Three essays on heterogeneous responses to macroeconomic shocks

PDF

Trade, training, employment, and wages: evidence from the U.S. manufacturing industry

PDF

Essays in climate change adaptation: role of market power in incentivizing adaptation behavior

PDF

Essays in environmental economics

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays on innovation, human capital, and COVID-19 related policies

PDF

The impact of digital transformation on urban communities, welfare, and distributional outcomes

PDF

The EITC, labor supply, and child development

PDF

Essays on the platform design and information structure in the digital economy

PDF

Three papers on dynamic externalities and regional growth

PDF

Taxi driver learns dynamic multi-market equilibrium

Asset Metadata

Creator Boullé, Clément (author)

Core Title The impact of minimum wage on labor market dynamics in Germany

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2023-08

Publication Date 07/10/2023

Defense Date 07/07/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economics,labor market,minimum wage,OAI-PMH Harvest

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kurlat, Pablo (committee chair), Kahn, Matthew (committee member), Nix, Emily (committee member), Quach, Simon (committee member)

Creator Email cboulle@usc.edu,clement.boulle55@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113262990

Unique identifier UC113262990

Identifier etd-BoullClmen-12041.pdf (filename)

Legacy Identifier etd-BoullClmen-12041

Document Type Dissertation

Format theses (aat)

Rights Boullé, Clément

Internet Media Type application/pdf

Type texts

Source 20230710-usctheses-batch-1064 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu