Frictions financières et marché du travail
Transféré par
anegue jeanFrictions financières et marché du travail
Transféré par
anegue jeanFrictions financières et marché du travail
Marine Sales
To cite this version:
Marine Sales. Frictions financières et marché du travail. Economies et finances. Université Paris
Saclay (COmUE), 2018. Français. �NNT : 2018SACLN041�. �tel-01952836�
HAL Id: tel-01952836
https://theses.hal.science/tel-01952836v1
Submitted on 12 Dec 2018
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Financial Frictions and ALabor Market
NNT : 2018SACLN041
Allan, à Toni
A mes parents, à mon papi
Thèse de doctorat de l'Université Paris-Saclay
préparée à l'Ecole Normale Supérieure Paris-Saclay
École doctorale n◦ 578 Sciences de l'homme et de la société (SHS)
Spécialité de doctorat : Sciences économiques
Thèse présentée et soutenue à Cachan, le 07/12/2018, par
Mme. Marine Salès
Composition du jury :
M. François Fontaine
Professeur, Paris School of Economics Rapporteur
M. Etienne Lehmann
Professeur, Université Panthéon Assas Président du jury
M. Gregory Levieuge
Professeur, Université d'Orléans Rapporteur
M. Franck Malherbet
Professeur, École Nationale de la Statistique et de l'Administration
Examinateur
Économique
M. Fabien Tripier
Professeur, Université d'Evry-Val-D'Essonne Examinateur
M. Hubert Kempf
Professeur, Ecole Normale Supérieure Paris-Saclay Directeur de thèse
1
2
A Allan, à Toni
A mes parents, à mon papi
3
4
Acknowledgements
Je remercie en premier lieu mon directeur de thèse, Hubert Kempf, pour sa con ance, sa patience et son
aide tout au long de ce processus de recherche. Il a toujours fait preuve de la plus grande compréhension et
de la plus grande bienveillance à mon égard, et ce quels que soient mes choix professionnels et personnels.
Je l'en remercie sincèrement.
Je remercie également les membres du jury qui m'ont fait l'honneur d'accepter de lire ce travail de thèse,
y apporter leur jugement et en discuter lors de la soutenance de thèse.
Je tiens à remercier Sabine Sépari pour ces belles années passées à ses côtés au sein de la préparation à
l'agrégation d'économie et de gestion. Elle a toujours été présente pour moi, m'a très souvent soutenue et
fut de précieux conseils. Je ne serais jamais arrivée là où je suis maintenant sans elle.
J'exprime également ma gratitude à Nicolas Drouhin qui m'a donné le goût et l'envie de faire de la
recherche lors de mon mémoire d'initiation à la recherche de master 1. Je garde un excellent souvenir de
ces quelques mois de recherche réalisée sous sa direction.
En n, je suis reconnaissante envers toute l'équipe du "Laplace" de l'ancien CES-Cachan, en particulier
Emmanuelle Taugourdeau pour son soutien constant, Farid Toubal, François Pannequin, Jean-Christophe
Tavanti, Nathalie Etchart-Vincent pour ses relectures attentives, et Thomas Vendryies. Toutes nos discus-
sions ont été enrichissantes et stimulantes.
Bien entendu, cette thèse doit beaucoup à mes collègues doctorants, qui sont devenus au l des années
des amis, Bastien, Elissa, Florian, Guillaume, Imen, Julien, Lenka, Maïva, Morgane, Olga et Samuel. Merci
pour nos nombreuses discussions, les moments de rire, mais aussi les moments de doutes où vous avez tou-
jours été là pour me soutenir et m'encourager.
J'ai aussi une pensée particulière pour tous mes anciens élèves de la "prépa agrég" de l'ENS Paris-Saclay.
Ils m'ont occasionné beaucoup de travail, mais quel travail passionnant ! Ils m'ont aussi aidée à prendre le
recul dont je pouvais avoir besoin sur mes travaux de recherche.
Cette année, j'ai concilié l'aboutissement de ma thèse et mes enseignements en classe préparatoire ENS
Paris-Saclay au Lycée Gaston Berger à Lille. Je souhaite remercier chaleureusement mes étudiants de pre-
mière et deuxième année, ainsi que toute l'équipe du lycée, en particulier Ariane Noiville, Cédric Canis,
Julie Saulnier et Patrick Broutin. Ils m'ont permis de nir sereinement ma thèse grâce à leur bienveillance,
leur bonne humeur communicative et leur gentillesse.
5
6
Je pense aussi bien évidemment à mes chers amis de Cachan, de Paris et de Rennes. Amélie, Emilie,
Guillaume, Thomas et Valentin, Guilhem, Léonard et Chloé, Marie, Jérôme et Lou-Anne, Marion, Morgan
et Antoine, Marion et Etienne, Morgane, Pierre-Louis, Sophie et Arnaud, et toute la "team" Liré. Merci
d'être encore et toujours à mes côtés. Vous êtes des amis admirables !
Je souhaite remercier du fond du c÷ur ma famille, tout particulièrement mes parents et mon frère,
Thomas, ainsi que mes grands-mères chéries, Tatie Nicole, Nathalie, mes cousines, Flavie et Swann, et mon
cousin, Kévin. Merci d'avoir toujours cru en moi. J'ai une chance incroyable d'avoir des parents et une
famille comme la mienne. J'ai une pensée très particulière pour mon Papi qui aurait tellement aimé tenir
cette thèse entre ses mains.
J'ai également la joie d'avoir une belle-famille exceptionnelle qui m'a accueillie les bras ouverts (et qui a
du supporter elle aussi ces longues années de thèse) : Alexis, Emma, Eden et Emgi, Andréa, Arthur, Fabiola
et Hugo, Joséphine, Marie, Damien et Emmanuel, Papi et Mamie Millet, Rémi, Véronique et Yvonnick.
En n, merci à mon mari, Toni. Merci pour ta patience, ton estime, ton soutien indéfectible et ton
amour tout au long de ces années. Je clos ces remerciements en pensant à la personne qui illumine tous les
jours de notre vie par son merveilleux sourire et sa gaîté, notre ls, Allan. Votre présence attentive et vos
encouragements sont pour moi les piliers fondateurs de ce que je suis et de ce que je fais.
Contents
Acknowledgements 5
Contents 7
General introduction 15
I Credit Imperfections, Labor Market Frictions and Unemployment: a DSGE approach 21
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Related literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1 Model overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Wholesale-good rms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Wage and hours bargaining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Intermediate and nal-good rms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.6 Monetary and scal policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.7 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Quantitative exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Appendices 59
A Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
II Credit Constraints and Labor Market: the role of Wage Bargaining Regimes 63
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.1 Model overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.2 Labor market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
7
8 CONTENTS
2.3 Households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.4 Capitalists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.5 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.6 Wage contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.7 Market clearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3 Sources of aggregate ine ciencies under right-to-manage and e cient bargaining regimes . . 80
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Appendices 83
A Borrowing constraint computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
B Binding borrowing constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
C Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
III Do Corporate Credit Conditions alter Labor Market Dynamics? A SVAR analysis in a
Transatlantic Perspective 89
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2 Empirical investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.1 VAR and SVAR methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.2 Data and SVAR de nition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.3 Identi cation of shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.1 Credit shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.2 Technological shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.3 Forecast errors variance decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4 Robustness analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5 Are credit shocks for Germany generating Schumpeterian creative destruction eects? . . . . 112
6 What drive the unemployment dynamics in the United-States and in Germany? . . . . . . . . 115
7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Appendices 121
A Data de nitions and sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
B Identi cation of short-term SVAR models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
C SVAR models speci cations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
D Cumulative impulse responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
E Robustness analysis - Impulse responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
General conclusion 141
CONTENTS 9
Bibliography 143
Summary 147
10 CONTENTS
List of Figures
1 Unemployment, Baa-Aaa spread and default rate between 1970-Q1 and 2007-Q4 for the
United-States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Labor-market tightness and Baa-Aaa spread between 1970-Q1 and 2007-Q4 for the United-
States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Timing of events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4 Private sector model overview and ows of funds . . . . . . . . . . . . . . . . . . . . . . . . . 30
5 The risk premium as a function of ω̄ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6 Vacancy posting cost as a function of ω̄ for dierent values of monitoring costs: µ = 0.15
(solid line), µ = 0.2 (dotted line) and µ = 0.25 (dashed line) . . . . . . . . . . . . . . . . . . . 42
7 IRF to positive networth shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8 IRF to positive monitoring cost shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
9 IRF to positive idiosyncratic volatility shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
10 Timing of events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
11 Sources of ine ciencies depending on bargaining regimes . . . . . . . . . . . . . . . . . . . . . 81
12 Unemployment, job vacancies and non- nancial corporations credit growth between 1952-Q1
and 2016-Q1 for the United-States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
13 Unemployment, job vacancies and non- nancial corporations credit growth between 1991-Q1
and 2016-Q1 for Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
14 Structural impulse responses to credit shock. Benchmark model. . . . . . . . . . . . . . . . . 101
15 Structural impulse responses to technological shock. Benchmark model. . . . . . . . . . . . . 104
16 Structural impulse responses to credit shock. Data period to 2007-Q4. . . . . . . . . . . . . . 110
17 Structural impulse responses to technological shock. Data period until 2007-Q4. . . . . . . . . 111
18 Structural impulse responses to credit shock. Investment added. . . . . . . . . . . . . . . . . 113
19 Structural impulse responses to technological shock. Investment added. . . . . . . . . . . . . 114
20 Structural impulse responses to technological shock. Unemployment dynamics. . . . . . . . . 117
21 Structural impulse responses to credit shock. Unemployment dynamics. . . . . . . . . . . . . 118
11
12 LIST OF FIGURES
22 Cumulative impulse responses to credit shock . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
23 Cumulative impulse responses to technological shock . . . . . . . . . . . . . . . . . . . . . . . 129
24 Structural impulse responses to credit and technological shocks for the United-States. Lag = 6.130
25 Structural impulse responses to credit shock. Unemployment ordered second. . . . . . . . . . 131
26 Structural impulse responses to technological shock. Unemployment ordered second. . . . . . 132
27 Structural impulse responses to credit shock. Output ordered rst. . . . . . . . . . . . . . . . 133
28 Structural impulse responses to technological shock. Output ordered rst. . . . . . . . . . . . 134
29 Structural impulse responses to credit and technological shocks for Germany. Unemployment
and vacancies expressed in rst-dierence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
30 Structural impulse responses to credit shock. Credit to output ratio. . . . . . . . . . . . . . . 136
31 Structural impulse responses to technological shock. Credit to output ratio. . . . . . . . . . . 137
32 Structural impulse responses to credit shock. Consumption added. . . . . . . . . . . . . . . . 138
33 Structural impulse responses to technological shock. Consumption added. . . . . . . . . . . . 139
34 Structural impulse responses to credit and technological shocks for the United-States. 1991.1
to 2016.2 data period. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
List of Tables
1 Parameters values for quantitative analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2 Forecast error variance decomposition of labor market variables - Germany and United-States
technological and credit shocks (percentage) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3 De nitions and source of data - Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4 De nitions and source of data - United-States . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5 SVAR lag order selection by selection criteria for the United-States . . . . . . . . . . . . . . . 126
6 SVAR lag order selection by selection criteria for Germany . . . . . . . . . . . . . . . . . . . . 126
7 Augmented Dickey-Fuller (ADF) tests for the United-Sates . . . . . . . . . . . . . . . . . . . 126
8 Augmented Dickey-Fuller (ADF) tests for Germany . . . . . . . . . . . . . . . . . . . . . . . . 127
13
14 LIST OF TABLES
General introduction
"The nancial crisis of 2008 has thrown open the question of the interaction between capital and
labour markets. Equilibrium matching models are built on the assumption of perfect capital
markets. The implied arbitrage equations under perfect foresight and unlimited borrowing and
lending are used to calculate a value for jobs and workers. These are good starting assumptions,
and they have yielded important results. But future work needs to explore other assumptions
about capital markets, and integrate the nancial sector with the labour market." Christopher
Pissarides, Nobel Prize Lecture (2010).
In decentralized economies, rational agents may have di culties to meet on markets, to coordinate be-
cause of imperfect or incomplete information. They are led to anticipate behaviors of other economic agents
to base their own decisions and actions. Imperfect information is at the roots of ine ciencies on dierent
markets, especially in cases of asymmetric information.
The Great Recession highlighted potential interactions between labor and credit markets. As Christo-
pher Pissarides noticed in 2010, these interactions have to be investigated deeply by economists taking into
account the existence of frictions on those markets. The increase in unemployment rates in many countries
following the Great Recession highlights the role that nancial frictions (imperfect information on repay-
ment capacity of borrowers) may play on labor markets. Labor markets are themselves subject to frictions
between labor demand and supply (imperfect information on jobs characteristics, on jobs oer...) that may
impact credit markets.
Due to imperfect information, frictions appear on credit and labor markets. These frictions would be
the source of economic ine ciencies and could interact with each other to exacerbate theses ine ciencies.
Financial frictions have been a lot discussed in the literature (Bernanke and Gertler (1989), Bernanke and
Gertler (1995), Bernanke et al. (1999), Carlstrom and Fuerst (1997), Carlstrom and Fuerst (2001), Fiore and
Tristani (2013), Gertler et al. (2010) and Kiyotaki and Moore (1997) among others). These articles have been
devoted to understand the relationship between nancial markets and overall macroeconomic performances.
Labor markets have been also considered as being frictional and thus integrated in macroeconomic models,
15
16 General introduction
but without any nancial frictions (Andolfatto (1996), Blanchard and Galí (2010), Campolmi and Faia
(2011), Christiano et al. (2016), Christoel et al. (2009), Galí et al. (2012), Gertler and Trigari (2009),
Krause et al. (2008), Lechthaler et al. (2010), Merz (1995), Thomas and Zanetti (2009), Trigari (2009),
Walsh (2005) among others).
However, as the Great Recession and Christopher Pissarides remind us, imperfect information and frictions
exist on both markets. Macroeconomic models should integrate the whole frictions and analyze the various
impacts of these frictions on macroeconomic performances. In this dissertation, I focus on a speci c causality
link, from credit markets to labor markets. My purpose is to analyze the impact of nancial frictions on
labor markets main variables, as wages, unemployment, vacancies, knowing that these labor markets are
themselves frictional.
Furthermore, markets' institutions are crucial in terms of imperfect information level. Institutions are
de ned by North (1994) as:
"the formal rules (constitutions, statute and common law, regulations...), the informal con-
straints (norms of behavior, conventions, and internally imposed codes of conduct), and the
enforcement characteristics of each."
He adds that these institutions set the incentive structure of economies and de ne the way the "game
is played." as institutions are the source of more or less information for economic agents. For examples,
labor unions bring information to workers, or national employment agencies are aimed to ease the matching
between employers and employees. Thus, they contribute to determine the equilibrium in which one economy
will stand. And institutions could also modify the way economic agents react facing imperfect information.
For example, if traditionally wages are bargained according to a right-to-manage regime, rms may use it
to adapt themselves facing asymmetric information on credit markets. On labor markets, institutions are
quite fundamental as they are often considered as a reason why labor markets are not functioning well.
Institutions on labor markets are numerous. One important institution worth to focus on is the degree
of coordination of wage bargaining (Checchi and García-Peñalosa (2008), Amable et al. (2007)). In the
literature, Trigari (2006) dierentiate two degree of coordination in the bargaining process, the so-called
e cient and right-to-manage bargaining.
Another fundamental institution on labor markets is the number and density of labor unions. In some
countries, as Germany, there is a tradition of strong labor unions. For example, in Germany, in 2017 18%
of workers were members of labor unions according to the OECD. In the United-States, in 2017 only 10%
of workers were members of unions according to the Bureau of Labor Statistics.
As labor market institutions impact the level of information in one economy, as well as the way economic
agents react to imperfect information, they in uence labor markets frictions, and as a consequence, could
interact with nancial frictions. Depending on the institutional environment on labor markets, nancial
frictions may have dierent eects on frictional labor markets.
General introduction 17
This brings up three questions:
• In an imperfect information environment, do nancial frictions have an impact on labor markets? If
so, through which mechanism does this impact take place?
• How dierent wage bargaining regimes modify the impact of nancial frictions on labor market out-
comes?
• From an empirical perspective, are there dierences in labor market dynamics to credit shocks between
Germany and the United-States (US)? How potential dierences in labor market dynamics to credit
shocks could be explained?
The imperfect information on credit and labor markets may bring to adverse interactions for the stability
of economies and to mechanisms amplifying economic shocks. These particular interactions between the
labor market and the credit market can take place through dierent channels of transmission. In particular,
a shock in the credit market aecting the borrowing capacity of rms or the price of their credit may modify
accordingly the search behavior of workers by rms and/or the level of unemployment in an economy. This
is what we study in the rst chapter of this dissertation. It is showed that nancial frictions have an impact
on the overall level of employment in an economy through a marginal cost channel. We develop a New-
Keynesian DSGE model integrating asymmetric information in the credit market à la Bernanke et al. (1999)
and a search and matching process in the labor market à la Mortensen and Pissarides (1994) associated
with a Nash bargaining process.
Asymmetric information on labor and credit markets are fundamentally linked to the institutional envi-
ronment of each economy. The bargaining regime is part of this institutional environment: laws for example
may determine which bargaining regime will take place. This wage and employment bargaining regime can
alter the degree of labor markets ine ciencies. The second chapter shows that nancial frictions create a
nancial mark-up, which can be added to another mark-up related to imperfections in the labor market
according to the existing regime of wage bargaining, either an e cient or a right-to-manage bargaining.
Finally, the last chapter challenges some previous results. It emphasizes that labor market institutions
change the way credit shocks aect labor markets. In some countries, with speci c labor market institutions,
credit shocks may have no or particular impact on labor markets. I compare the impact of credit shocks on
labor markets in Germany and in the United-Sates. I use a structural vector auto-regressive model. Em-
pirical results for the United-States are consistent with theoretical results obtained in chapter I. However,
for Germany, responses of labor markets variables to credit shocks are either not signi cant, or contrary to
those expected. I nd that the explanation could be found in the particular institutional functioning of the
German labor market.
18 General introduction
Roadmap of the dissertation
The dissertation is made of three dierent chapters on the interactions between nancial frictions and labor
markets. The two rst chapters investigate in an imperfect information environment, how nancial frictions
interact with labor markets depending on labor markets institutions. The third chapter presents an empir-
ical analysis of potential discrepancies on the way nancial frictions interact with labor markets depending
on countries, and labor markets institutions, considered.
Chapter I. Credit Imperfections, Labor Market Frictions and Unemployment: a DSGE
approach1
This chapter studies the impact of costly external nance for rms on unemployment, vacancy posting
and wages by focusing on shocks originating from credit markets. The theoretical model demonstrates the
existence of a nancial mark-up charged by nancial intermediaries, that is transmitted to labor markets
by rms via a marginal costs channel. Higher credit market frictions are the source of lower posting va-
cancies and higher unemployment level as it increases rms' marginal costs. The theoretical model is then
calibrated by using quarterly United-States data for the sample period 1960:Q1 to 2007:Q4. We nd that
employment and vacancy posting increase following positive monitoring cost, net worth and idiosyncratic
volatility shocks. Dierent channels of propagation from the nancial sphere of the economy to the labor
market are investigated and the results appear to be consistent with the theoretical model. These channels
converge all to the role of the nancial mark-up that is charged by banks to overcome agency problems. This
nancial mark-up is passed through the rest of the economy by higher marginal costs and higher in ation.
That in turn reduces the levels of vacancies posting, employment, wages and consumption, and nally the
level of output. The evolution of credit market conditions changes the opportunity cost for resources used
to create new jobs. Thus, it alters the dynamics of job vacancies and unemployment.
Chapter II. Credit Constraints and Labor Market: the role of Wage Bargaining Regimes
In this chapter, I compare two bargaining regimes, the so-called 'e cient bargaining' (EB) and the so-called
'right-to-manage' (RTM) bargaining in a search and matching model integrating a collateral constraint.
The impact of credit frictions on unemployment (extensive margin), wages and hours worked per employee
(intensive margin) is not the same depending on the way hours and wages are bargained. Especially, this
impact is modi ed through the bargaining power of workers relative to rms that is itself dependent on the
level of nancial frictions. With an EB regime, the wage splits the surplus of a match on the labor market
according to the rm's bargaining power that depends negatively on the level of collateral constraints. So,
credit frictions increase the bargaining power of workers: they extract a higher rent from the bargaining
relatively to a framework without nancial frictions. With a RTM regime, the impact of nancial frictions
1 This chapter is based on a co-written paper with Imen Ben Mohamed between 2012 and 2015 as it can be found as
hal-01082491. The version in this dissertation is a revised version of which I am solely responsible for.
General introduction 19
exists but it is mitigated by the fact that the rm is able to modify the level of hours worked by each worker.
Thus, a rm compensates the existence of nancial frictions by reducing the level of hours demanded for
each worker. So in both cases, a higher level of collateral constraints leads to an increase in the worker's
bargaining power. They ask for a bigger rent, but their capacities to extract a bigger part of the surplus
depend on the bargaining regime: a RTM regime appears to restore partly the bargaining power of rms
by giving them a higher degree of freedom in the bargaining process that is not internalized by workers.
I identify to that purpose two ine ciency gaps compared to a case without any friction, a nancial and a
wage ine ciency gap, the last one being present only under a RTM regime. Firms use intensive margins
to alleviate nancial frictions. As a consequence, the bargaining regime prevailing on labor markets may
modify the way nancial frictions impact these labor markets.
Chapter III. Do Corporate Credit Conditions Alter Labor Market Dynamics? A SVAR
Analysis in a Transatlantic Perspective
In this chapter, I investigate the eects of technological and credit shocks on unemployment and vacancies
in the United-States and Germany. I estimate structural VARs based on quarterly data, where shocks
are identi ed through short-run restrictions. Shocks are identi ed by assuming that rms need external
nancing before production is realized and sold. First, I nd a positive impact of technological shocks on
employment and vacancies in both countries. Then, a common view widespread today is to consider that
more credit in one economy will be the source of better labor market outcomes as it implies lower external
nancial constraints for rms. However, credit shocks appear to aect dierently labor market variables in
each country. In the United-States, a positive credit shock increases vacancies and decreases unemployment,
while in Germany the opposite eect is obtained for unemployment and vacancies, with an insigni cant result
for vacancies. Eects of a credit shock on labor market variables are thus ambiguous for Germany. My
empirical results suggest that the previous view can be challenged and discussed as an increase in the level
of credit in one economy does not necessarily lead to better conditions on labor markets. Finally, a credit
shock has a positive impact on output in the United-States, whereas this impact is ambiguous in Germany,
consistent with the idea that good credit conditions are not su cient to improve the economic dynamics in
this particular country. To explain this result, I consider two explanations: a 'Schumpeterian' mechanism
and a 'search for conciliation' mechanism. I nd that German rms separate from workers when credit level
is increasing in the economy. Firms adjust their wage bill when credit conditions are favorable. The role of
labor unions could explain such results as labor union are strong in a country as Germany. This argument is
nally partly reinforced by the fact that I illustrate the potentiality of non-linearity in the impact of credit
shocks on labor markets that could be investigated deeply in future research.
20 General introduction
Chapter I
Credit Imperfections, Labor Market
Frictions and Unemployment: a DSGE
approach
1 Introduction
Credit market imperfections are suspected of playing a key role in the worsening or improvement of labor
markets position1 . In recent years, especially following the Great Recession, there has been an increasing
interest for macro-economists to analyze interactions between frictional credit and labor markets. Questions
have been raised about the fact that higher credit imperfections may be the source of a slowdown of the
economy, and not its consequence. By themselves, nancial frictions could destabilize the whole economy.
This chapter aims to study the potential destabilizing eect of nancial frictions on real economy and partic-
ularly on labor markets, that were aected a lot during the previous crisis. In the United-States for example,
the unemployment rate rises from 5% in 2008 to 10% in 2009. Thus, the question raised in this chapter
is: in an imperfect information environment, do nancial frictions have an impact on labor markets? If
so, through which mechanism does this impact take place? We nd that nancial frictions have a negative
impact on labor markets situations through a nancial mark-up charged by nancial intermediaries so as
to tackle asymmetric information on credit markets.
The research tended to focus either on the impact of nancial frictions on overall macroeconomic perfor-
mances (Bernanke and Gertler (1989), Bernanke and Gertler (1995), Bernanke et al. (1999), Carlstrom and
Fuerst (1997), Fiore and Tristani (2013), Gertler et al. (2010) and Kiyotaki and Moore (1997)), either on
1 This chapter is based on a co-written paper with Imen Ben Mohamed between 2012 and 2015 as it can be found on
hal-01082491. The version in this dissertation is a revised version of which I am solely responsible for.
21
22 1. INTRODUCTION
the impact of labor market frictions on overall macroeconomic performances (Andolfatto (1996), Blanchard
and Galí (2010), Campolmi and Faia (2011), Christiano et al. (2016), Christoel et al. (2009), Galí et al.
(2012), Gertler and Trigari (2009), Krause et al. (2008), Lechthaler et al. (2010), Merz (1995), Thomas and
Zanetti (2009), Trigari (2009) and Walsh (2005)). However, a less but growing attention has been paid
to the impact of nancial frictions on labor markets, being themselves imperfect (Christiano et al. (2011),
Petrosky-Nadeau (2014), Thomas and Zanetti (2009) and Zanetti and Mumtaz (2011) among others). This
chapter aims to complement and improve previous works by using a general equilibrium approach, by mod-
eling nancial frictions in a particular way and by widening components rms are constrained to borrow.
Figures 1 and 2 shed light on the potential causal relationship that we propose to study and highlight.
Evolution of unemployment rate, Baa-Aaa spread and default rate between 1970-Q1 and 2007-Q4 for the
United-States are represented in gure 12 . A correlation is observed among these variables, especially for
the unemployment rate and the Baa-Aaa spread (0.76). The higher the unemployment rate is, the higher
the Baa-Aaa spread is and conversely. For the default rate, the correlation is less explicit, due to plausible
structural forces between 1971 and 1982 linked to the Federal Reserve monetary policy. However, some
periods of correlation still exist: 1979-Q1 until 1985-Q4 (0.6) and from 1990 (0.32).
Then, a negative correlation between the labor market tightness (vacancies3 over unemployment) and the
Baa-Aaa spread is shown on gure 2. The observed negative correlation between 1970.Q1 and 2007.Q4 is
quite huge (−0.84). It induces that the higher vacancy posting are relative to unemployment, the lower is
the Baa-Aaa spread and conversely. These very basic empirical correlations show that interactions between
frictional credit and labor markets may exist: a high risk premium on credit markets is associated with
a deteriorate labor market situation. To study these potential interactions, we construct and calibrate a
new-Keynesien general equilibrium model integrating credit and labor market frictions. We focus on the
impacts of higher credit market frictions on labor markets variables.
The model is a new-Keynesian model with asymmetric information in the credit market à la Bernanke
et al. (1999) and a search and matching process in the labor market à la Mortensen and Pissarides (1994)
associated with a Nash bargaining process. Capital spending, wage bill and vacancy posting costs are as-
sumed to be paid in advance of production and revenues are realized, requiring partial external nancing
for rms. The model, based on these features, provides an explanation of cyclical uctuations in key labor
market variables (unemployment, vacancies, hours worked per employee and wages) and in credit market
variables (risk premium and default rate). We nd that the existence of a risk premium charged by nancial
intermediaries impacts the vacancy posting decisions, the wage bill and unemployment levels in the economy,
2 The unemployment rate is the ratio of civilian unemployed persons to the civilian labor force. The default rate is the
default rate for Moody's rated US speculative-grade corporate bonds. The Baa-Aaa spread is the Moody's seasoned Baa-Aaa
corporate bond yield.
3 Vacancies are obtained from Conference Board Help Wanted OnLine data series.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 23
2.5
15
Unemployment (%), Default Rate (%)
2
Baa−Aaa Spread
10
1.5
5
1
.5
0
1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006
Quaters
Unemployment rate Baa−Aaa spread Default rate
Figure 1: Unemployment, Baa-Aaa spread and default rate between 1970-Q1 and 2007-Q4 for the United-States
30
2.5
25
2
Labor market tightness
20
Baa−Aaa Spread
1.5
10 15
1
5
.5
0
1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006
Quarters
Labor market tightness Baa−Aaa Spread
Figure 2: Labor-market tightness and Baa-Aaa spread between 1970-Q1 and 2007-Q4 for the United-States
as well as the level of in ation. When the risk premium increases, the net worth of entrepreneurs decreases.
It increases their dependence on external funds, making job posting more expensive. So, less vacancies
are posted and a higher equilibrium unemployment is obtained. More precisely, asymmetric information
in the credit market pushes up marginal costs and prices, as well as hiring costs by a nancial mark-up,
depending on the levels of monitoring cost and break-even entrepreneur-speci c productivity. The higher
are monitoring cost and break-even entrepreneur-speci c productivity, the higher is the nancial mark-up.
This nancial mark-up is made to overcome the agency problem between nancial intermediaries and rms.
It is then bypassed by rms on prices and aects their hiring behavior, as well as wages, employment and
in ation levels in the economy. As a consequence, nancial frictions have a negative impact on labor markets
through this nancial mark-up.
24 2. RELATED LITERATURE
A calibration exercise is then carried out to investigate the impacts of a net worth shock, a monitoring
cost shock and an idiosyncratic volatility shock on macroeconomic variables, such as vacancies, unemploy-
ment rate and real wages. Quarterly data for the sample period 1960:Q1 to 2007:Q4 are used. The most
striking result to emerge from this exercise is that employment rate and vacancies posting increase following
positive monitoring cost, net worth and idiosyncratic volatility shocks. Dierent channels of propagation
from the nancial sphere of the economy to the labor market are investigated and appear to be consistent
with the theoretical model. The key mechanism behind these results is that following positive shocks on
the credit market, the nancial mark-up charged by nancial intermediaries decreases, leading to lower
real marginal costs and real hiring costs paid by rms, that is passed through prices in the economy, and
inducing rms to post more vacancies. The unemployment as a consequence decreases. Furthermore, after a
positive net worth shock, a substitution eect appears between hours worked per employee and the number
of employees, that to say between intensive and extensive margins. This element is veri ed in the data as the
extensive margin is known to be always more reactive that the intensive one. This substitution eect does
not appear following a positive monitoring cost shock or a negative idiosyncratic volatility shock, resulting
in a higher positive eect for the economy compared to the net worth shock eect.
Section 2 consists of a related literature review. The theoretical model is developed in section 3. In
section 4, we outline the quantitative exercise and present the results. Section 5 concludes.
2 Related literature
This chapter is at the intersection of dierent lines of research. Firstly, a number of research papers
introduce search and matching frictions on labor markets in real business cycle (RBC) models or in new-
Keynesian (NK) models. Other articles highlight the role of nancial frictions for macroeconomic dynamics,
without taking into account search and matching frictions on labor markets. Finally, more recent studies
embody simultaneously frictions in labor and credit markets in partial equilibrium models or in dynamic
stochastic general equilibrium (DSGE) models, to study interactions and implications of these two types of
frictions.
The assumption of Walrasien labor markets is considered as a weakness of standard RBC and NK mod-
els. Indeed, these models do not take into account variations in the number of unemployed workers, the
extensive margin that never changes. They allow only to study variations in hours worked per employee,
the intensive margin. This may seem annoying to the extent that unemployment is an important indicator
of performances of the economy in its use of resources and it is a major policy issue, especially since the
Great Recession. Furthermore, this kind of models is ineective to explain the eect of various shocks on
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 25
unemployment dynamics. As a consequence, many articles have introduced search and matching frictions in
labor markets based on Mortensen and Pissarides (1994) framework in RBC models or in NK models (An-
dolfatto (1996), Blanchard and Galí (2010), Campolmi and Faia (2011), Christiano et al. (2016), Christoel
et al. (2009), Galí et al. (2012), Gertler and Trigari (2009), Krause et al. (2008), Lechthaler et al. (2010),
Merz (1995), Thomas and Zanetti (2009), Trigari (2009), Walsh (2005) among others).
Papers, as those of Andolfatto (1996) and Merz (1995), study implications of search and matching
frictions for economic uctuations in RBC models. Both models show that labor market frictions are a
mechanism of ampli cation and persistence for technological shocks. These frictions improve the empirical
performance of RBC models, compared to a standard one, even if they do not predict enough cyclical move-
ments in vacancies and output compared to data. Moreover, Andolfatto (1996), by introducing extensive
and intensive margins, nds that most of the variability of total hours worked is due to changes in unem-
ployment level rather than in hours worked per employee.
Then, several papers in the same spirit (Campolmi and Faia (2011), Lechthaler et al. (2010), Thomas and
Zanetti (2009), Trigari (2009) and Walsh (2005)) examine the role of matching frictions in NK models. For
example, Walsh (2005) develops a NK DSGE model with labor market frictions and with dierent potential
sources of persistence (habit persistence, price stickiness and policy inertia). He founds through a calibrated
model that it ampli es for US data the output response and decreases the in ation response to a monetary
policy shock, as well as it generates persistence in output and in ation as observed in data and as standard
NK models do not succeed to generate. In the same idea, Trigari (2009) considers cyclical uctuations of
output, in ation and labor market variables following a monetary policy shock. She studies the possibility
of endogenous separation between rms and workers, and distinguishes extensive and intensive margins.
Her estimated model is able to replicate well for US data the observed responses of output, in ation and
labor market data to a monetary policy shock. Using a VAR, she nds as observed in data that in a model
with labor market frictions, the response of in ation is less volatile and response of output more persistent
after a monetary policy shock than in a standard NK model.
However, these Mortensen-Pissarides search and matching models of unemployment remains unable to
match important stylized facts observed in data. In particular, these types of models are not performing
well to explain high volatility and persistence of unemployment and vacancies, as well as the relative smooth
behavior of real wages found in data. The framework of Nash bargaining appears to lead to an exaggerated
procyclical movements in wages after a positive productivity shock for example, that dampens the rm's
incentives to hire. Wages absorb much of the change in the expected bene t to a new worker induced by
uctuations in labor productivity. As a consequence, several papers try to tackle this issue by introducing
wage rigidity mechanisms (Blanchard and Galí (2010), Christiano et al. (2016), Gertler and Trigari (2009)
and Shimer (2004)) or hiring and ring costs (Lechthaler et al. (2010) for example). Firstly, Blanchard
and Galí (2010) nd that search and matching frictions modify the level of unemployment but the un-
26 2. RELATED LITERATURE
employment rate stays invariant to productivity shocks. Thus, they study alternative wage-setting (Nash
bargaining wage and more rigid real wages) and show that rigid wages enable to have ine cient uctuations
in unemployment after a productivity shock. Lechthaler et al. (2010) introduce in a new-Keynesian model
labor market frictions, through hiring and ring costs but no wage rigidity. They nd trough a calibration
exercise more persistence in output and unemployment in response to real and monetary policy shocks and
in in ation in response to real shocks, as well as a strong ampli cation eect of these shocks on unemploy-
ment and on the job nding rate.
On the other hand, frictions have been also studied on the credit market side (Bernanke and Gertler
(1989), Bernanke and Gertler (1995), Bernanke et al. (1999), Carlstrom and Fuerst (1997), Carlstrom and
Fuerst (2001), Fiore and Tristani (2013), Gertler et al. (2010) and Kiyotaki and Moore (1997)). These arti-
cles have been devoted to understand the relationship between nancial markets and overall macroeconomic
performances. Financial factors are indeed suspected to amplify and increase persistence of macroeconomic
variables responses to aggregate shocks. The idea behind is that deteriorating credit conditions could be
the source of poor economic activity and not the consequence of a declining real economy.
Bernanke and Gertler (1989), Bernanke et al. (1999), Carlstrom and Fuerst (1997) and Kiyotaki and
Moore (1997) develop the concept of a nancial accelerator in DSGE models integrating money and price
stickiness. Without credit frictions, an entrepreneur can resort to external nancing to raise capital at a
risk-free interest rate. With credit market frictions, asymmetric information appears in the form of moral
hazard between the lender and the borrower. Borrower is induced to report to the lender a lower real output
produced than its true level.
As a consequence, this type of asymmetric information leads rst to restrictions for borrowers on the
amount of external nancing available, based on the existence of collateral constraints to cover their poten-
tial inability to reimburse loans (Kiyotaki and Moore (1997)). In this framework used in the chapter II of
this dissertation, agents face endogenous credit limits determined by the value of collateralized assets.
Then, asymmetric information between a lender and a borrower can lead to a second modelization of
nancial frictions (Bernanke and Gertler (1989), Bernanke et al. (1999) and Carlstrom and Fuerst (1997)),
namely a higher cost of external nancing compared to internal nancing opportunity cost (the risk-free
interest rate), that to say an external nance premium or a risk premium. The canonical RBC model of
Carlstrom and Fuerst (1997) integrates such risk premium and enables to show that it leads the economy
to return more slowly to the steady-state after being hit by a shock. Debt arises as an optimal nancial
contract between rms and banks such that rms borrow at a premium over the risk-free rate. The nancial
contract is designed to minimize the expected agency costs. It speci es returns when bankruptcy or success
occurs and a monitoring threshold as developed in our upcoming model.
However, the previous papers assume standard Walrasien labor markets. Only few papers consider both
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 27
credit and labor markets frictions, except the ones of Christiano et al. (2011), Petrosky-Nadeau (2014) and
Zanetti and Mumtaz (2011). Labor market frictions imply that it is costly to hire new workers. The func-
tioning of frictional labor markets prevents the competitive allocation of labor resources, and thus it could
interact with nancial frictions to impact production, unemployment, investment and capital accumulation.
Those models enhance the Bernanke et al. (1999) framework with a more realistic labor market. Christiano
et al. (2011) show in a new-Keynesian model that nancial and labor markets frictions are able to change
the model dynamics in an open economy setting, and improve the forecasting properties of the model for
Swedish data, in particular for in ation. Petrosky-Nadeau (2014) considers that rms nance only their
job vacancy costs with external nancing on frictional credit markets. He nds that the easing of nancing
constraints during an expansion (a productivity shock) reduces the opportunity cost for resources allocated
to job creation (cost channel) because rms are able to accumulate net worth. Credit market frictions
generate persistence in the dynamics of labor-market tightness. Zanetti and Mumtaz (2011) demonstrate
through a Bayesian estimation that labor and nancial frictions are supported by data and that they play
together to amplify or reduce the variables' reaction to various shocks. Firms have in their model to paid
only capital in advance of production.
In our comprehensive model, we introduce both credit and labor markets frictions. First, we assume
that wages, job vacancy costs as well as capital are nanced in advance of production. To our knowledge,
no paper takes into account that the whole input costs are paid in advance in a DSGE framework. Then,
we model extensive and intensive margins of employment to obtain a more precise idea of adjustments in
the labor market. Calvo-price stickiness is also introduced in the model to observe the behavior of in ation
in a model of this type. Finally, the research to date focus on technological or monetary policy shocks.
Few paper tries to investigate direct shocks from the nancial sphere, such as monitoring cost, net worth or
idiosyncratic volatility shocks. This chapter tries to bring these gaps.
28 3. MODEL
3 Model
3.1 Model overview
The model is populated by various types of agents: households, wholesale-good rms managed by en-
trepreneurs, retailers, nal-good rms, nancial intermediaries and a government that conducts scal and
monetary policies together with a central bank. The model is build on the modeling of asymmetric infor-
mation in the credit market à la Bernanke et al. (1999) and a modeling of the labor market with a search
and matching process à la Mortensen and Pissarides (1994). Figure 3 delivers the timing of events in a
synthetic way. Figure 4 reports the ow of funds for the private sector of the economy.
The household sector consists of a continuum of identical households of length unity. Each household
is constituted of members who are either employed or unemployed searching for a job. They all supply
inelastically hours of labor, consume nal goods, rent capital to wholesale rms and save through their
deposits to nancial intermediaries.
Entrepreneurs manage and owned wholesale rms, that produce wholesale goods using a constant return-
to-scale technology using labor and capital as inputs. Entrepreneurs have nite lifetime. Following Bernanke
et al. (1999), each entrepreneur is assumed to have a given probability to survive to the next period. Surviv-
ing entrepreneurs carry their pro ts as a part of their net worth. Dying entrepreneurs consume everything.
Total hiring costs, capital spending and wages are assumed to be paid by wholesale rms managed by
entrepreneurs once capital and labor are rented, that to say before production and revenues are realized.
External nancing is required for wholesale rms. However, wholesale-good production is subject to an
idiosyncratic shock privately observed by entrepreneurs after the nancial contract arrangement, while -
nancial intermediaries need to pay a monitoring cost to check the real output produced as well as the
e ciency of the recruitment process. This agency problem will alter the marginal cost of production and
the hiring costs.
Finally, the production sector has three dierent layers in the spirit of Bernanke et al. (1999). At the
rst layer, where agency problem and search and matching frictions occur, a continuum of perfectly com-
petitive wholesale rms produce homogeneous goods using capital and labor. At the second layer, where
price stickiness arises, wholesale goods are dierentiated costlessly by a continuum of monopolistic rms.
The realized pro ts are rebated lump-sum to households. The nal good is then homogeneous and can be
used for consumption, capital accumulation and government spending.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 29
Agents are thus interacting in ve dierent markets (labor market, capital market, credit market, money
market and goods market), where the timing of events is given by gure 3.
Birth of new en- Solvent entrepreneurs
trepreneurs repay their debt and a
fraction of them die
Households choices
Renting of capital Production on consumption,
and labor of all investment and deposits
sectors
t−1 t t+1
Time
Entrepreneurs'
observation of their
Realisation of idiosyncratic shock
aggregate shocks
Financial intermediaries
bear monitoring
Optimal nancial costs for insolvent
contrat establishment: entrepreneurs and seize
each rm borrows proceeds of production
funds and complies
to its capital rental
and wage payments
Figure 3: Timing of events
30
Payments Deposits' repayment
Final-good Financial
Households
producers Final goods Deposits intermediaries
Searc
hing
Matching
and wage
bargaining
Po s ting
La
Loans
Di
Pr
od b or
vid
uc su
vaca
en
tio
n
n pp
ds
ly
Loans' reimbursment
bi
ll
cies
Ca
Payments
pi
Optimal nancial contract
ta
Retail goods
l
Entrepreneurs
Wholesale goods Wholesale-good
Retailers
producers
Payments
Figure 4: Private sector model overview and ows of funds
3. MODEL
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 31
3.2 Households
Households are seen as a large representative family represented by the unit interval consisting of a
continuum of members, either employed or unemployed searching for a job4 . As in Andolfatto (1996) and
Merz (1995), there is a full risk sharing of consumption in order to avoid distributional issues due to hetero-
geneity in incomes among family members. So, the family pools its income such that a perfect consumption
is fully insured for all members5 . The same notation is then used for the representative household and for
the consumption of each member. In any period t, the number of employed family members is nht ∈ (0, 1)
and the number of unemployed family members searching for a job is Uth 6 .
The representative household maximizes its expected discounted utility7 :
∞
" #
X H 1+τ nht
E0 βt log(Ct − hCt−1 ) − t (I.1)
t=0
1+τ
subject to a ow budget constraints sequence:
Rt−1 Dt−1 Πt Tt
Wt nht Ht + (1 − nht )b + + rtK Kth + +
Pt Pt Pt
(I.2)
Dt
= Ct + It +
Pt
and to a sequence of employment laws of motion:
nht = (1 − δ)nht−1 + q(θt )Uth (I.3)
and to a sequence of physical capital laws of motion:
h
It = Kt+1 − (1 − δK )Kth (I.4)
β ∈ (0, 1) is the household intertemporal discount factor, Ct is the household's real nal goods consumption
(as well as each family member's consumption), h is a habit persistence parameter8 , Ht are the eective
hours of work, τ denotes the inverse of the Frisch elasticity of labor supply, Wt is the real wage, b are
the real unemployment bene ts that each unemployed member receives from the government9 , Kth is the
household's real physical capital stock at the beginning of period t as the household owns a part of the
4 Full participation in the labor market is assumed. The transition between in and out the labor force is ignored.
5 The family optimally allocates the same consumption for each member, regardless their respective labor market status and
individual income. This assumption is quite strong. Some papers as the one of Iliopulos et al. (2014) are considering dierent
levels of consumption depending on the respective members employment status.
6 Only the unemployed members can search passively for a job and can be hired. Employed members are not allowed to
look for another job.
7 The form of the utility function is based on the ones used by Bernanke et al. (1999) and Gertler et al. (2008).
8 When h > 0, the model allows for habit persistence in consumption preferences to take into account the necessary empirical
persistence in the consumption process.
9 b can be interpreted as home production or as unemployment bene ts, as we do, provided by the government and nanced
by lump-sum taxes.
32 3. MODEL
capital stock of the economy, It is the household's real investment, rtK is the real renting capital interest
rate, δK is the capital depreciation rate, Dt is the nominal amount deposits the household carries to period
t + 1, Rt is the nominal risk-free interest rate, δ is the exogenous job destruction rate, q(θt ) is the probability
for an unemployed member to nd a new job and Pt is the aggregate price level. The household also makes
a nominal lump-sum transfer, Tt , to the government and receives a nominal lump-sum pro t, Πt , as the
household has a diversi ed ownership stake in retail rms.
The household decides on its level of consumption, investment and deposits at the end of the period t.
As a result, he knows all the variables of its optimization program when it takes its decisions. And it is why
h
Kt+1 and Dt are carried to the next period.
Back-and-forth between employment and unemployment for household members are carried out by search
and matching processes in the labor market. The household takes as given the probability for an unemployed
member to nd a new job, q(θt ). This probability depends on the ratio of total vacancies to unemployed
workers, θt , the aggregate labor market tightness. Furthermore, the fraction δ of employed workers of period
t−1 that are assumed to be separated from their jobs before period t is also taken as given by the household.
So the number of searching unemployed members at the start of period t is de ned as: Uth = 1 − (1 − δ)nht−1 .
The rst-order conditions of the representative household's problem are given by:
1 1
(Ct ) λt = − βhEt (I.5)
Ct − hCt−1 Ct+1 − hCt
λt+1 Rt
(Dt ) 1 = βEt (I.6)
λt πt+1
h K
(I.7)
(Kt+1 ) λt = βEt λt+1 (1 − δK ) + rt+1
Ht1+τ
(nht ) Wtn − WtU = W t Ht − −b (I.8)
(1 + τ )λt
λt+1 n U
+βEt (1 − δ)(1 − q(θt+1 ))(Wt+1 − Wt+1 )
λt
Pt
where πt ≡ is the in ation rate, λt is the Lagrange multiplier associated to the household's budget
Pt−1
constraint and Wtn − WtU corresponds to the ratio between the Lagrange multiplier to the law of motion of
nht and the Lagrange multiplier of the budget constraint λt .
Equation (I.5) de nes the marginal utility of consumption when there is habit formation. It states that the
Lagrange multiplier equals the marginal utility of consumption. (I.6) corresponds to the household choices
in terms of deposits. From equations (I.5) and (I.6), we derive the household's stochastic discount factor
βEt λλt+1
t
. Equation (I.7) corresponds to the household choices in terms of renting capital. Finally, equation
(I.8) designates the discounted net value in period t to the household of having a new employed worker in
terms of current consumption. It is the sum of the real wage earned by the new employed worker in period
t, reduced for the marginal disutility of working and for the unemployment bene ts that are foregone, plus
the expected discounted gain from being either employed or unemployed during the subsequent periods. A
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 33
worker is still employed at the period t + 1 if the match has not been exogenously destroyed before period
t + 1 with a probability (1 − δ) ; or if the match has been destroyed with the probability δ before period
t + 1, but that another matching occurs at the period t + 1 with the probability q(θt+1 ). And a worker
becomes unemployed if the match is destroyed before period t + 1 and if he or she does not nd a job in the
period t + 1, that to say with the probability δ(1 − q(θt+1 )). Finally, an unemployed worker nds a job in
period t + 1 with a probability q(θt+1 ).
3.3 Wholesale-good rms
There is a continuum of unit mass of wholesale rms indexed by i ∈ [0, 1]. They are owned and managed
by nite lived risk-neutral entrepreneurs. Wholesale-good rms need labor and capital to produce. Yitws is
the quantity of wholesale goods produced by a rm i using Nitf = nfit Hit total hours of labor and Kit units
of physical capital, according to the following constant-returns production function:
f (1−α)
Yitws = At Kit
α
Nit (I.9)
where α is the capital share in production and At is the aggregate technological shock, realized at the
beginning of each period, source of systematic risk. Physical capital, Kit , is rented from households and
other rms (as part of their net-worth as detailed hereafter) at a competitive price, rtK . Total hours worked,
Nitf , are paid to employed workers, nfit , through the wage, Wt .
Each period, wholesale rms draw an idiosyncratic shock, ωit , de ned as a productivity and management
e ciency shock, re ecting management skills, hiring e ciency and input utilization skills of rms. This
idiosyncratic shock is the source of wholesale rms heterogeneity. ωit is i.i.d. with a time-varying mean,
ωmt , a continuous distribution function, Φ(.) and a density function, φ(.), being identical across rms. ωit
is de ned over a non-negative support and Φ(0) = 0. Moreover, its variance, re ecting the shock's volatility
and the entrepreneurs' riskiness, is time-varying and its standard deviation, σωt , follows a rst-order auto-
regressive process identical across wholesale rms, given by:
iid
log(σωt ) = (1 − ρσ ) log(σ̄ω ) + ρσ log(σωt−1 ) + uσt , ρσ ∈ (0, 1) where uσt ∼ N (0, σσ2 )
where σ̄ω is the steady-state value of the standard deviation, σωt .
Entrepreneurs
Entrepreneurs and households have the same time preferences rate, β . The optimization problem of an
entrepreneurs is:
∞
X
max
e
Et β s (1 − ςt )Ci,t+s
e
(I.10)
Cit
s=0
34 3. MODEL
where Cit
e
is the consumption of the entrepreneur managing the rm i.
An entrepreneur receives pro ts that are rebated from wholesale rm i that did not go bankrupt. These
pro ts are used either for consumption, or for net worth accumulation, Xit , depending on a probability of
survival entrepreneurs ςt 10 .
With a probability 1 − ςt , a solvent entrepreneur dies in a given period. He can then consume all his net
worth just before death. With a probability ςt , a solvent entrepreneur is able to carry a net worth to the
next period as he is not dying. This accumulated net worth is carried out as physical capital, Kit
e
, to the
next period t.
Finally, as for the household, the entrepreneur decides on its level of consumption at the end of the period
t.
Matching and hiring workers
Matching technology. A job creation occurs when an entrepreneur and an unemployed worker searching for
a job meet on the labor market after search and matching processes à la Mortensen and Pissarides (1994).
Total vacancies actively posted by entrepreneurs, Vt , are lled by unemployed workers passively searching
for a job, Ut , via an aggregate constant return to scale matching function, m(Ut , Vt ) = Utρ Vt1−ρ where
ρ ∈ (0, 1) is the elasticity of matches to unemployment11 .
Hiring workers is a costly and time-consuming process for entrepreneurs. To hire a new worker, en-
trepreneurs (managing wholesale-good rms) create vacancies at a real unit cost, γ . New hired workers
in period t start working immediately12 . Then, total matches that produce in period t are assumed to be
destroyed at an exogenous rate, δ , before period t + 1. So the evolution of aggregate employment is de ned
as:
nt = (1 − δ)nt−1 + m(Ut , Vt ) (I.11)
The productive employment stock of period t corresponds to period t − 1 surviving matches from the
exogenous separation, (1 − δ)nt−1 , plus the new hires from the matching of period t, m(Ut , Vt ). As the labor
force is normalized to one, unemployment corresponds to ut = 1 − nt .
As standard in the search and matching literature, matching probabilities are q(θt ) and p(θt ), corresponding
respectively to job nding and lling rates13 .
10 The same assumption is made by Bernanke et al. (1999) and Paustian (2004). Carlstrom and Fuerst (1997) make the
dierent assumption, that consumers and entrepreneurs have dierent time-discount factors with entrepreneurs less impatient
than consumers.
11 Gertler et al. (2008) use the same speci cation. The Cobb-Douglas matching function is used in almost all macroeconomic
models with search and matching frictions to represent the aggregate ows of hires. Furthermore, the constant returns to scale
assumption (homogeneity of degree one) seems to be supported empirically according to Petrongolo and Pissarides (2001).
12 Following Blanchard and Galí (2010), Gertler et al. (2008), Krause and Lubik (2007) and Thomas and Zanetti (2009),
employed workers are assumed to be immediately productive after being hired.
13 Matching probabilities are de ned as p(θ ) ≡ m(Ut ,Vt ) and q(θ ) ≡ m(Ut ,Vt ) as the matching function is constant-return-
t Vt t Ut
to-scale. Note that q(θt ) = θt p(θt ) and ∂p(θt )/∂θt < 0, ∂qθt /∂θt > 0.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 35
Hiring workers. A wholesale rm i begins each period t with an employment stock (1 − δ)nfit−1 . To hire a
new worker, the rm has to post vacancies, Vit , at a real unit cost. Each vacancy is lled with a probability
p(θt ). The rm takes as given this probability. So the employment laws of motion for rm i is:
nfit = (1 − δ)nfit−1 + p(θt )Vit (I.12)
Optimal nancial contract
The total input costs of a wholesale rm i correspond to capital rental costs, wage payments and hiring
costs: Bit = rtK Kit + Wt Ht nfit + γVit . We assume that these costs are paid after the rental of production
factors (labor and capital) but before the observation of rm's i idiosyncratic shock and before that produc-
tion and revenues are realized. Furthermore, at the equilibrium, all workers at rm i earn the same wage
and work the same number of hours as workers are assumed to be homogeneous. Thus, matches do not
depend on any idiosyncratic component. Idiosyncratic e ciency for each rm is common across all workers
working for each rm.
To nance a part of these costs, the rm uses the net worth accumulated by its entrepreneur from the
previous period, Xit . This net worth is carried from a period to another in the form of capital:
e
Xit = Kit (1 + rtK − δK ) + W e (I.13)
where W e is a real exogenous entrepreneurial wage14 .
But the rm's internal funds are insu cient to nance all input costs. Indeed, as assumed above, en-
trepreneurs have an exogenous probability to die each period. This assumption is made to limit the size of
aggregate net worth in an in nite horizon set up15 . So the rm needs external nancing to nance its total
input costs. A nancial intermediation is realized through a large number of atomistic risk-neutral banks.
Banks are assumed to hold enough large and diversi ed portfolios to ensure perfect risk pooling for their
creditors, the households, carrying deposits to banks16 . Intra-period loans are stipulated and established
after the aggregate shock, At . To eliminate aggregate uncertainty from the lender-borrower relationship,
the aggregate technological shock is assumed to be observed by all agents in the economy and it is realized
before any loan contract is established.
14 This endowment income ensures that each rm/entrepreneur i continues to produce in each period even if it becomes
insolvent in a given period.
15 Since the rate of return on internal funds is higher than the one of external funds due to asymmetric information on credit
markets, risk neutral entrepreneurs may be willing to postpone consumption and would only accumulate net worth. The same
assumption is made by Bernanke et al. (1999) and Paustian (2004) to ensure that rms need external nancing. Carlstrom
and Fuerst (1997) make the dierent assumption, that consumers and entrepreneurs have dierent time-discount factors with
entrepreneurs less impatient than consumers.
16 In nitely-lived households are risk averse but they become risk neutral for the nancial contract. Carlstrom and Fuerst
(1997) explain this fact by the absence of uncertainty about the term of the one-period contract since the aggregate uncertainty
is realized before the contract establishment. Furthermore, by the law of large numbers as banks are nancing a continuum of
dierent entrepreneurs, households know they will receive the expected return of the idiosyncratic shock.
36 3. MODEL
Furthermore, entrepreneurs are subject to idiosyncratic shocks privately observed by them after the
nancial contractual arrangement, but not observed by banks. This private information creates a moral
hazard problem as entrepreneurs may be encouraged to under-report the true value of their production
when they have to reimburse the loan. So banks have to monitor wholesale rms which declare themselves
bankrupt after the production occurs. This monitoring is costly for banks.
Under a costly state veri cation framework, the perfectly competitive nancial intermediaries' setting
ensures that each rm-bank pair writes the borrowing contract that maximizes the expected return of the
borrower, the entrepreneur, under the constraint that the expected return to the lender, the bank, ex-
ceeds its opportunity cost, namely the risk free interest rate, Rt (participation constraint). So the optimal
incentive-compatible nancial arrangement is a risky debt contract, whose terms are the optimal solution
to a standard principle-agent problem between entrepreneurs and banks17 .
A wholesale rm i borrows a nominal amount of Pt (Bit − Xit ) from the bank at an implicit interest rate,
Rtl . The optimal short-term nancial contract speci es contractually rm's leverage ratio and a break-even
entrepreneur-speci c productivity level, ω̄it , satisfying:
Rtl Pt (Bit − Xit ) = ω̄it Ptws Yitws (I.14)
Indeed, if ωit ≥ ω̄it , the entrepreneur pays back the value ω̄it Ptws Yitws , the loan amount augmented with
interest.
If ωit < ω̄it , the rm goes bankrupt and the bank con scates the total output produced. The bank can
observe this state of nature at a monitoring cost µt ∈ (0, 1), a fraction of the total output produced by the
rm. After the realization of shocks and production occurrence, it is given by ωit Ptws Yitws .
Note that although we use the same costly state veri cation framework used in Bernanke et al. (1999) and
Carlstrom and Fuerst (1997) for examples, asymmetric information is here introduced between wholesale-
good producers, the entrepreneurs, and banks. Financial intermediaries and banks are used interchangeably
in the model. As they are operating in a competitive market, only the behavior of a representative bank is
considered below. Furthermore, note that unlike the costly state veri cation framework similar to Carlstrom
and Fuerst (1997), credit contracts are here nominal.
f
Formally, the expected return earned by the wholesale rm i, Eit , is given by:
Z
f
Eit = Ptws Yitws (ωit − ω̄it )φ(ωit )dωit (I.15)
ωit >ω̄it
17 See Townsend (1979) and Gale and Hellwig (1985).
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 37
Using the statistic properties of the random idiosyncratic shock, equation (I.15) can be developed to see
that the expected return of the wholesale rm is a fraction of its realized total output:
f
Eit = Ptws Yitws f (ω̄it ) (I.16)
Z
where f (ω̄it ) = ωit φ(ωit )dωit − ω̄it [1 − Φ(ω̄it )]. Note that f (ω̄it ) ∈ (0, 1)18 and
ωit >ω̄it
f 0 (ω̄it ) = Φ(ω̄it , ) − 1 ≤ 0. The wholesale rm's expected return is a decreasing function of ω̄it . This result is
quite intuitive since an increase of the default rate, Φ(ω̄it ), reduces the gross share of return going to the rm.
Similarly, the expected return earned by the bank, Eit
b
, is given by:
hZ Z i
b
Eit = Ptws Yitws ω̄it φ(ωit )dωit + (1 − µt ) ωit φ(ωit )dωi
ωit >ω̄it ωit <ω̄it
It is straightforward to show that lender's expected return is also a fraction g(ω̄it ) ∈ (0, 1) of wholesale rm's
output19 . Then,
b
Eit = Ptws Yitws g(ω̄it ) (I.17)
where,
g(ω̄it ) = 1 − f (ω̄it ) − µt Γ(ω̄it )
Z
with Γ(ω̄it ) = ωit φ(ωit )dωit . So, an amount of the realized rm's total output is lost due to mon-
ωit <ω̄it
itoring in cases of declaring bankruptcy. In this context, the monitoring cost spending is a synonym of
bankruptcy and it is spent in terms of currency so that bankruptcy has no direct impact on the real out-
put. As a consequence, the proportion of the total amount recovered by the bank in case of bankruptcy is
(1 − µt )Γ(ω̄it ). 1 − µt is thus interpreted as the recovery rate. Following Livdan et al. (2009) and Petrosky-
Nadeau (2014), this recovery rate is assumed to be time-varying and has the following speci cation:
1 − µt = s0t exp s1 (ωit − 1) (I.18)
s1 is the elasticity of the recovery rate to the entrepreneurial productivity and management e ciency level
and s0t is interpreted as a monitoring cost shock, following a rst-order auto-regressive process:
iid
log s0t = (1 − ρs0 ) log s¯0 + ρs0 log s0t−1 + us0
t , ρs0 ∈ (0, 1) where us0 2
t ∼ N (0, σs0 )
where s̄0 is the steady-state value of the monitoring cost shock, s0t .
18 f 0 (ω̄) ≤ 0, ∀ ω̄ ∈ [0, ∞). In addition, lim
Z = 1 and limω̄→∞ f (ω̄) = 0Z.
ω̄→0 f (ω̄) Then, f (ω̄) ∈ (0,
Z 1).
19 Since 1 − f (ω̄) ∈ (0, 1) and by de nition we have ωφ(ω)dω ∈ (0, 1) and ω̄φ(ω)dω + ωφ(ω)dω ∈ (0, 1), then
ω<ω̄ ω>ω̄ ω<ω̄
g(ω̄) ∈ (0, 1), lim g(ω̄) = 0 and ω̄→∞
lim g(ω̄) = 1 − µ.
ω̄→0
38 3. MODEL
Finally, the optimal debt contract is a solution to the maximization of wholesale rm's expected return
under the constraint that expected return to the bank exceeds its opportunity cost, the risk-free interest
rate.
The contractually rms' leverage ratio, Lit , is given thus by:
Bit 1
Lit ≡ = (I.19)
Xit 1 − Rt Sit g(ω̄it )
φ(ω̄it )
where Sit = {1−µt [Γ(ω̄it )+ω̄it h(ω̄it )f (ω̄it )]}−1 , with h(ω̄it ) the hazard rate20 de ned by h(ω̄it ) = .
1 − Φ(ω̄it )
Note that Lω̄ > 0. And for a given level of net worth Xit , a higher leverage ratio is associated with a higher
default rate. Indeed, the probability of default increases as the loan amount raises (in the spirit of Stiglitz
and Weiss (1981)).
And the contractually bankruptcy threshold, ω̄it , is given by:
Rt Sit f (ω̄it ) −f 0 (ω̄it )
= (I.20)
1 − Rt Sit g(ω̄it ) −g 0 (ω̄it )
At the optimum, the bank participation constraint is binding. Using this result with equation (I.14), the
risk premium, ∆it , de ned as the ratio of the lending rate to the risk free rate, Rtl /Rt , is given by:
ω̄it
∆it = (I.21)
g(ω̄it )
Net worth evolution and dynamic pro t function
Wholesale rms consider as given the nancial contract when maximizing their pro ts to determine their
labor and capital demand. They do not know at that time how much they will be able to produce and sell
(the idiosyncratic shock has not been revealed) ; thus they base their labor and capital demand on their
expected pro ts as developed hereafter.
Wholesale rms maximize their expected discounted value to the entrepreneurs. This value is decomposed
in two parts due to the existence of nancial asymmetric information. The intertemporal objective function
of a rm i is de ned as:
∞
X λt+1
E0 βt [Eω Πfit + Cit
e
] (I.22)
t=0
λt
First, there is the so-called 'standard' expected pro ts coming from the rm production, Eω Πfit . These
pro ts correspond to the dierence between the value of the expected total output produced by the rm
and the total input costs. However, this pro t is expected at the time decisions are made. Indeed, it occurs
20 We assume that ω̄h(ω̄, σ ) is increasing in ω̄ in order to ensure the concavity of the lender's net share of return, g(ω̄), and
ω
avoid any credit rationing at the equilibrium. This regularity condition is without loss of generality and it is satis ed by most
of the continuous probability distributions. See Bernanke et al. (1999) for details.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 39
before the idiosyncratic shock is revealed to the rm. As a consequence, this 'standard' pro t is de ned as:
Ptws ws
Eω Πfit = E(ωit ) Y − Rt Sit (Wt Nitf + rtK Kit + γVit ) (I.23)
Pt it
Firms take as given the renting capital interest rate rtK as well as the Nash wage, Wt . Furthermore,
E(ωit ) = ωmt the time-varying mean of the idiosyncratic shock is identical across rms.
In addition, the objective of the rm is to maximize the entrepreneurs' consumption constituted through
the nancial expected returns earned by the wholesale rm they own. These nancial expected returns are
used as consumption, Cit
e
, for entrepreneurs dying before period t + 1 and for capital accumulation, Kit+1
e
,
so as to constitute a net worth for next period, Xit+1 , for solvent entrepreneurs not dying before period
f
t + 1. Using equation (I.14) and equation (I.19), the expected returns earned by the wholesale rm i, Eit ,
are given by:
Ptws Yitws f (ω̄it ) = Rt Sit f (ω̄it )Bit (I.24)
Taking into account the death probability ςt for entrepreneurs, we obtain:
e
Rt Sit Bit = (1 − ςt )Cit e
+ ςt Kit+1 (I.25)
Substituting now the de nition of Bit from equation (I.19) and the de nition of Xit from equation (I.13),
we obtain a de nition for the capital laws of motion of entrepreneurs:
Rt Sit f (ω̄it )
e e e
(1 − δ + rtK ) + W e (I.26)
(1 − ςt )Cit + ςt Kit+1 = Kit
1 − Rt Sit g(ω̄it )
Solvent entrepreneurs who exit the economy at the end of the period will consume all their net worth. Thus,
the entrepreneurial consumption, Cit
e
, is given by:
Rt Sit f (ω̄it ) ςt
e e
(1 − δK + rtK ) + W e − Ke (I.27)
Cit = Kit
1 − Rt Sit g(ω̄it ) (1 − ςt ) it+1
P ws
= (1 − ςt ) t Yitws f (ω̄it ) (I.28)
Pt
Solvent entrepreneurs who do not exit the economy at the end of period t will keep accumulating net worth
using their realized return. Consequently, from equation (I.24) the evolution of the entrepreneurial real net
worth is given by:
Ptws ws
Xit+1 = W e + ςt Y f (ω̄it ) (I.29)
Pt it
40 3. MODEL
Thus, ςt is interpreted as a shock to entrepreneurs' net worth. It evolves according to:
ςt = ς¯ςt (I.30)
iid
log(ςt ) = ρς log(ςt−1 ) + uςt , ρδ ∈ (0, 1) where uςt ∼ N (0, σς2 )
where ς¯ is the steady state value of the shock to entrepreneurs' net worth, ςt .
Firms optimality conditions
Optimality conditions for a rm i are delivered by maximizing equation (I.22) subject to equations (I.12)
and (I.26):
Ptws ws
(Kt ) ωmt Y = rtK Rt St (I.31)
Pt K
1 − ςt λt+1 St+1 f (ω̄t+1 )
(Kte ) 1 = βEt K
(1 − δK + rt+1 ) (I.32)
ςt λt 1 − St+1 g(ω̄t+1 )
Zt γ
(Vt ) = Rt St (I.33)
Pt p(θt )
Zt P ws (1 − α)Ytws
(nft ) = ωmt t − Rt St Wt Ht (I.34)
Pt Pt nft
λt+1 Zt+1
+(1 − δ)βEt
λt Pt+1
where Zt is the Lagrange multiplier associated to the employment laws of motion constraint.
In order to keep the representative- rm context, we assume that the threshold value of the entrepreneurial
productivity, ω̄t , is the same for all wholesale rms. At the equilibrium, this is an important assumption
since all rms will have the same nancial mark-up, St . Consequently, according to this assumption, the
subscript i is dropped21 .
Combining equations (I.33) and (I.34) delivers the job creation condition under nancial frictions in the
credit market:
γ P ws (1 − α)Ytws λt+1 γ
Rt St = ωmt t f
− Rt St Wt Ht + (1 − δ)βEt Rt+1 St+1 (I.35)
p(θt ) Pt nt λt p(θt+1 )
The evolution of credit market conditions changes the opportunity cost for resources used to create new
jobs through the role played by St . Thus, it alters the dynamics of job vacancies. Credit spreads are a key
element to understand the cyclical behavior of job creation and the general dynamics of labor markets.
Indeed, from the rst-order conditions obtained, asymmetric information in the credit market generates
ine ciencies in all markets, the wholesale-good market, the capital market and the labor market. On the
21 Paustian (2004) shows that thanks to linear monitoring technology and constant returns to scale in production, the
heterogeneity in rms productivity does not require to keep track of the distribution of net worth of entrepreneurs.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 41
one hand, marginal productivity of capital and labor are higher than their corresponding real marginal
cost. And the nal real price of the wholesale good is augmented by a nancial mark-up, St > 122 , used to
overcome the agency problem between entrepreneurs and banks.
The higher are monitoring cost and break-even entrepreneur-speci c productivity, the higher is this nancial
mark-up. Indeed, if monitoring costs are high, the bank bears a greater burden when it has to monitor the
rms that declare themselves bankrupt, all other things being equal. Therefore, when reaching the optimal
contract with a wholesale rm, knowing these high monitoring costs, the bank sets a high margin to cover
these monitoring costs. And as a result, the rm passes on this nancial mark-up on its own sales prices
which will then increase.
Similarly, if break-even entrepreneur-speci c productivity is high, the bank knows it will be di cult for
wholesale rms to reach this threshold. The risk of default is high. Therefore, when signing the optimal
contract with a wholesale rm, the bank sets a larger margin to compensate this high risk. And as previ-
ously, the wholesale rm passes on this nancial mark-up on its own sales prices which will then increase.
As a consequence, credit market conditions matter because through the nancial mark-up, they aect
rms' marginal costs and they are transmitted to the rest of the economy through the selling price's mark-up.
Banks have a margin behavior, that will pass through the rest of the economy by the wholesale-good price.
This framework breaks down the Modigliani-Miller theorem and makes the rms' external borrowing costs
higher than internal funds opportunity costs. Indeed, rms must borrow at a premium over the risk-free
rate.
Figure 5 shows the evolution of the risk premium as a convex increasing function of the entrepreneurial pro-
ductivity and management e ciency threshold, ω̄ . This result is quite familiar in the nancial accelerator
literature. A higher default probability of rms induces a higher cost of lending for banks and consequently,
a higher loan spread.
On the other hand, the cost of a new hire, Zt , depends also on the nancial contract conditions. Besides
the unit cost stemming from hiring, γ , and the average duration of vacancies, 1/p(θt ), the total cost of
recruitment is augmented by the same nancial mark-up, St . For any positive monitoring cost, nancial
frictions increase the average cost of lling a vacancy. Financial contract conditions aect the labor market
e ciency through the total vacancy posting cost, that becomes an endogenous variable. This relation is
represented by gure 6, where the real posting cost is also an increasing and convex function of ω̄ and its
slope raises with monitoring costs, µ. For a higher default likelihood (higher ω̄ ), banks charge a higher
risk premium, ∆, so that entrepreneurs obtain their credit at a higher lending rate, Rl , which makes their
external funds more expensive and reduces their willingness to open vacancies.
An increase of µ shifts the real vacancy posting cost upward. For a xed level of ω̄ , the real cost of a new
hire raises with monitoring costs. As these costs are expressed in terms of currency, and not in terms of
22 See appendix A for the proof.
42 3. MODEL
R is k pr e mium,∆ t
Z/P
ω̄
ω̄
Figure 6: Vacancy posting cost as a function of ω̄ for dierent
Figure 5: The risk premium as a function of ω̄ values of monitoring costs: µ = 0.15 (solid line), µ = 0.2 (dotted
line) and µ = 0.25 (dashed line)
physical goods, they do not generate a loss of resources through a destruction of goods, which could have
been used for consumption. However, they generate an additional cost taken into account by banks when
agreeing on an appropriate interest rate on loans. Fluctuations in monitoring costs and bankruptcy rates
will have an impact on welfare only indirectly, through their implications on the mark-up pricing.
3.4 Wage and hours bargaining
Real hourly wages and hours worked per worker are determined through a period-by-period Nash bar-
gaining23 . This bargaining takes place after the nancial contract determination. So entrepreneurs and
households integrate in their bargaining the existence of a nancial mark-up set up by banks. The Nash
bargaining consists of maximizing the net surplus of the employment relationship for a representative rm,
(Zt − JtV ), and a representative household, (Wtn − WtU ), depending on workers and entrepreneurs Nash
bargaining powers respectively η and 1 − η .
Bellman equations
The household's surplus of a new employment relationship in terms of current consumption at period t
is given by:
Ht1+τ
λt+1
Wtn − WtU = Wt Ht − − b + βEt n
(1 − δ)(1 − q(δt+1 ))(Wt+1 U
− Wt+1 ) (I.36)
(1 + τ )λt λt
The entrepreneur's surplus of a new employment relationship in terms of current consumption at period
23 Whether workers are newly hired or not, they see their wages and hours bargained period-by-period.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 43
t is given by:
Ptws (1 − α)Ytws λt+1
Zt − JtV = ωmt f
− Rt St Wt Ht + (1 − δ)βEt V
(Zt+1 − Jt+1 ) (I.37)
Pt nt λt
Since there is a free-entry condition on the labor market, at the equilibrium the value to the entrepreneur
of an unemployed worker, JtV , is such that JtV = 0 is satis ed. Indeed, if JtV > 0, the entrepreneur has an
incentive to post vacancies as the value of a vacant job is positive. As the number of vacancies increases
relative to the number of unemployed workers, the probability to ll a vacancy job, p(θt ), decreases. In-
deed, the labor market tightness, θt , increases and p(θt ) decreases since ∂p(θt )/∂θt < 0. That reduces the
incentive to post vacancies and diminishes the value of JtV until it equals zero.
Nash bargaining, wage and hours setting
The chosen real hourly wage and the chosen hours of work per worker are the ones that maximize24 :
max (Wtn − WtU )η (Zt − JtV )(1−η) (I.38)
Wt ,Ht
Therefore, the rst-order necessary condition for the Nash wage bargaining solution is given by:
η
Zt = Wtn − WtU (I.39)
1−η
The following Nash real hourly wage is thus obtained:
" # " #
η Ptws (1 − α)Ytws λt+1 γ (1 − η) Ht1+τ
W t Ht = ωmt + (1 − δ)βEt Rt+1 St+1 + b+ (I.40)
Rt St Pt nt λt p(θt+1 ) Rt St (1 + τ )λt
The Nash real hourly wage splits by de nition the joint surplus of the employment relationship between an
entrepreneur and a worker according to the bargaining power of worker, η , associated here with the nancial
mark-up, St . Workers obtain a fraction η of the rm's revenues, as well as a fraction η of the rm's expected
saving on total hiring costs, including the nancial mark-up St+1 , depending on the probability that the
match will not be destroyed exogenously. If the matching is broken, workers and entrepreneurs have to look
for another partner in the next period, which is costly for entrepreneurs, especially if nancial frictions are
high. This cost is thus incorporated in the wage. And workers are compensated for a fraction 1 − η for the
disutility they suer from supplying hours of work and for the foregone unemployment bene ts.
The rst-order necessary condition for the Nash hours per worker bargaining solution is then given by:
Ptws ws Hτ
ωmt Yt = t+1 (I.41)
Pt λt
24 Recall that matches do not depend on any idiosyncractic component.
44 3. MODEL
3.5 Intermediate and nal-good rms
Final-good production
Final-good rms proceed in a perfectly competitive market and are owned by households. They purchase
a continuum of dierentiated intermediate goods indexed by j ∈ [0, 1] and aggregate theses varieties to
produce Yt units of nal goods. They have no other costs or inputs, except the one to buy to intermediate-
good rms the dierent varieties. Final goods are produced using a standard constant return to scale
technology given by:
"Z
# −1
1 −1
Yt = Yjt
dj (I.42)
0
where > 1 is a parameter governing the degree of monopolistic competition in the intermediate good sector
(or the elasticity of substitution between intermediate goods).
Each competitive nal-good rms choose their own input demand functions for each variety of interme-
diate goods, Yjt , so as to maximize their nominal pro ts de ned as:
Z 1
Pt Yt − Pjt Yjt dj
0
where Pt is the consumer's price index that corresponds to the aggregate price level. The solution to the
maximization problem25 yields the following demand function for the intermediate good of variety j :
" #−
Pjt
Yjt = Yt (I.43)
Pt
So, the demand for each variety of retail goods is a downward sloping demand curve, which gives to the
retail rms some pricing power as we will see below.
Then, as we are in a competitive setting, the zero-pro t condition applies at the equilibrium for nal-good
rms and it yields:
Z 1
Pt Yt = Pjt Yjt dj (I.44)
j=0
From this condition, the aggregate price level or the consumer's price index can be easily derived by plugging
25 Final-good rms maximize their expected stream of pro ts, which is equivalent to maximize their pro ts period-by-period
since they purchase intermediate goods at the same frequency.
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 45
the demand function (I.43) into the zero-pro t condition (I.44), which gives:
1
Z 1 1−
Pt = 1−
Pjt dj (I.45)
j=0
Intermediate-good production
There is a continuum of monopolistically competitive retailers. These retailers are owned by house-
holds26 . They buy homogeneous wholesale goods at the price Ptws . They dierentiate costessly each unit
of these goods into a unit of retail goods, Yjt . These rms are assumed to have no other inputs or costs
than the homogeneous good. Following Yun (1996), Christiano et al. (2005) and Trigari (2009), a price
stickiness for these rms is formulated in the spirit of Calvo (1983). Every period, only a random fraction
(1−%) ∈ [0, 1) of rms is able to fully re-optimize their nominal prices. The hazard rate, %, is constant across
rms and time. The remaining fraction of rms does not re-optimize their prices and following Christiano
et al. (2016), they keep their prices unchanged. So the price set by a retailer j , Pjt , corresponds to:
Pjt−1 with probability %
Pjt = (I.46)
P ∗ with probability 1 − %
t
where Pt∗ is the optimal price set by the fraction % of retailers who are able to re-optimize their prices at
time t. Note that Pt∗ does not depend on j because all rms that can re-optimize their prices at time t
choose the same price as they are assumed to be symmetric. Then, for rms not able to re-optimize their
prices, there is no price indexation to replicate the observation that many prices can remain unchanged over
time (Eichenbaum et al. (2011) and Klenow and Malin (2010)). So, the price index (I.45) is thus given by:
hZ 1 1
i 1− h 1− 1
1− i 1−
Pt = 1−
Pjt dj = (1 − %) Pt∗ + % Pt−1 (I.47)
0
The price index is a CES aggregate of all retail goods prices in the economy. The sum in equation (I.47)
can be transformed into a convex combination of two prices because retail rms of each type are assumed
to be respectively symmetric. And a costless price regulation mechanism is assumed, which guarantees that
a consumer pays the same price whatever the rm at which he realizes his purchases27 . So, the price index
corresponds to a weighted average price of the fraction % of rms who can re-optimize their prices after the
aggregate shock, At , and the fraction (1 − %) of rms who can not.
Let's now determine the optimal price Pt∗ . Retail rms that can re-optimize their price, maximize the
expected discounted value of their pro ts given the demand for the good they produce, since rms expect
to keep this price for more than the current period. They take into account that the price may be xed for
26 So retailers have the same stochastic discount factor as households.
27 The matching of consumers and rms is here ignored.
46 3. MODEL
many periods. If the expected probability of price stickiness is high, rms able to re-optimize their price
at the period t will be relatively more concerned about the future when they make their current pricing
decisions.
Thus, the retail rm's problem is stated as:
∞
" 1− − ws #
X
s λt+s Pjt Pjt Pt+s
max Et (%β) Yt+s − Yt+s
Pjt
s=0
λt Pt+s Pt+s Pt+s
Note that % is integrated in the discount rate because there is a probability %s that the price chosen is still
applied in s periods of time.
The rst-order condition of the problem is given by:
∞
X λt+s (−1) ws ∗
(%β)s Pt (− − 1) + (1 − )Pt∗ (−) = 0
Et Pt+s Yt+s Pt+s
s=0
λt
The optimal price, Pt∗ , sets by retail rms who are able to re-optimize their prices is:
P∞ s λt+s ws −1
Et s=0 (%β) λt Pt+s Pt+s Yt+s
Pt∗ = (I.48)
− 1 Et ∞ s λt+s P −1 Y
P
s=0 (%β) t+s t+s
λt
The size of the mark-up depends on the time-varying elasticity of substitution between retail goods, . So
exible-price retail rms set their price such that it equals the present discounted value of marginal costs.
The optimal price is a mark-up over a weighted average of future marginal costs. Note that if there is no
price-stickiness, % = 0, the monopoly standard mark-up formula is obtained:
Pt∗ = P ws , where >1 (I.49)
− 1 t+s −1
3.6 Monetary and scal policy
Monetary policy
The monetary policy is decided and carried out by the central bank following an interest rate Taylor-type
rule28 . The nominal interest rate is set depending on deviations in output, in ation and nominal interest
rate from their steady-state levels:
!ρr " !ρπ !ρY #1−ρr
Rt Rt−1 πt Yt
= (I.50)
R̄ R̄ π̄ Ȳ
where ρR is the degree of interest rate smoothing, ρY and ρπ are respectively the response coe cients to
output and in ation variables. Variables without a time subscript are steady state values. Rt is the interest
28 The same kind of Taylor rule is used by Krause et al. (2008) and Trigari (2009).
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 47
rate on deposits decided and set at time t that pay o in period t + 1.
Fiscal policy
Unemployment bene ts, (1−nt )b, are provided by the government. He nances also other real exogenous
spending, Gt , that are not used for any productive object. Government outlays are nanced through nominal
lump-sum taxes by households, Tt . The ow government budget constraint is:
Tt
Gt + (1 − nt )b = (I.51)
Pt
3.7 Equilibrium
A dynamic equilibrium consists of state-contingent endogenous processes {Ct , Cte , nt , Ht , Ut , Vt , Kt+1 ,
e
Kt+1 h
, Kt+1 , Ytws , rtK , Wt , Pt , Tt , Rt , St , ω̄t } that satisfy the following conditions: the vacancy creation con-
dition (I.35), the Nash wage rule (I.40), the hours per worker rule (I.41), the household's Euler equation for
deposits (I.6), the household's surplus of having a new employed worker (I.8), the household's Euler equation
for capital (I.7), the entrepreneur's equation for consumption (I.27), the entrepreneur's Euler equation for
capital (I.32), the bankruptcy threshold condition (2.5), the price index (I.47), the aggregate law of motion
for employment (I.11), the aggregate capital market clearing Kt = Kth + Kte , the entrepreneur's leverage
ratio (I.19), the rm's capital law of motion (I.26), the aggregate resource constraint Ct + It = Ytws , the
interest rate Taylor-type rule (I.50) and the government budget constraint (I.51). Stochastic processes are
taken as given {s0t , σωt , ςt }∞
t=0 .
4 Quantitative exercise
The quantitative exercise is conducted through numerical approximation of the model by linearization
around the steady state. We use Matlab to perform the model simulation. We relied on Matlab codes from
dierent sources, especially those of Cristiano Cantore and Paul Levine, obtained during a Summer School
"The Science and Art of DSGE Modelling" at the University of Surrey in 2015.
4.1 Calibration
Table 1 lists our parameters setting. Parameter values are chosen to be consistent with those standard
in literature and to match some empirical facts observed in data for the United-States between 1960-Q1
and 2007-Q4.
Preference and technology parameters are standard in the literature. The household discount factor, β ,
48 4. QUANTITATIVE EXERCISE
is set to 0.99 implying an annual real interest rate of 4%. The inverse of the elasticity of worked hours to
real hourly wage (or the Frish elasticity), τ , is equal to 10. This value is chosen by Trigari (2009), that has
in her paper extensive and intensive margins as we do. The external habit persistence parameter is xed to
0.5 as in Gertler et al. (2008). The capital share in production function, α, is 0.36. The capital depreciation
rate, δK , is xed to 0.025 corresponding to an average annual depreciation rate of 10 per cent. This value
is based on calculation of Kydland and Prescott (1982) using US time series data.
The labor market parameters are also standard in the recent literature. Elasticity of matches to un-
employment, ρ, is set to 0.4 as in Blanchard and Diamond (1989), Merz (1995), Andolfatto (1996) and
Mortensen and Nagypal (2007)29 . The value, 0.4, is given to the bargaining power, η , to deliver Hosios
e ciency. The vacancy posting cost, γ , is calibrated to 0.14, and the destruction rate parameter is set to
0.08. It is compatible with those used in the literature which range from 0.07 in Merz (1995) to 0.15 in
Andolfatto (1996). The unemployment bene ts, b, is equal to 0.71 as in Petrosky-Nadeau (2014).
For the intermediate-good sector, the monopolistic mark-up or the elasticity of substitution across in-
puts, , is xed to 11 to have a conventional price-mark-up on marginal costs at 10% as in Walsh (2005)
and Trigari (2009).
The steady state entrepreneurial survival rate is equal to 0.9728 derived from the default rate data from
Moosdy's US speculative-grade. The gross external nance premium, s1 , is set to 4.97. The long-run mean
of idiosyncratic productivity and management e ciency, ωm , is normalized to 1.
Finally, the Calvo stickiness of prices, %, is set to 0.66 as in Gertler et al. (2008). And for the Taylor rule,
conventional value are also taken. The interest rate smoothing coe cient, ρR , is set to 0.75. Coe cients
for the responses of interest rate to in ation, ρπ and to the output gap as in are xed respectively to 1.7
and 0.125 as in Gertler et al. (2008).
29 This value is almost within the range of values of 0.5 to 0.7 reported by Petrongolo and Pissarides (2001) in their survey
of the literature on the estimation of matching function. Others values are used such as 0.72 in Shimer (2004) and Petrosky-
Nadeau (2014) or 0.5 in Gertler et al. (2008).
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 49
Table 1: Parameters values for quantitative analysis
Parameter Value Description
Preferences
β 0.99 Households' quarterly discount factor
τ 10 Inverse of the elasticity of worked hours
to real wage
h 0.5 External habit persistence
Technology
α 0.36 Capital's share in production function
δK 0.025 Capital depreciation rate
Labor markets
ρ 0.4 Elasticity of matches to unemployment
γ 0.14 Unit cost of job vacancies
b 0.71 Unemployment bene ts
η 0.4 Nash bargaining power of workers
δ 0.08 Quarterly job separation rate
Entrepreneurs and nancial markets
s1 4.973 Gross external nance premium
s¯0 7.3678 Steady-state value of monitoring cost shock
ς¯ 0.9728 Steady-state value of net worth shock
ωm 1 Productivity and management e ciency mean
Intermediate-good markets
% 0.66 Calvo price stickiness
Exogenous process
ρ s0 0.95 Monitoring cost shock persistence
ρω 0.97 Idiosyncratic volatility shock persistence
ρς 0.5 Net worth shock persistence
ρR 0.75 Interest rate smoothing coe cient
ρπ 1.7 Response to in ation
ρY 0.125 Response to output
50 4. QUANTITATIVE EXERCISE
4.2 Results
We choose to consider three types of shocks: net-worth, monitoring cost and idiosyncratic volatility
shocks to investigate the role played by nancial shocks on labor market outcomes. Shocks originating from
the nancial sphere were indeed considered of prime importance during the Great Recession, as being at the
source of damages on the real sphere of economies, especially on labor markets : higher credit imperfections
may be the source of a slowdown of the economy, and not its consequence. The three shocks considered are
interesting to consider as they are representative of dierent aspects of the nancial frictions environment:
the riskiness of the debtor, its degree of dependence on external nancing and the cost the lender has to
bear to tackle asymmetric information. It should be noted that, with the exception of the idiosyncratic
volatility shock, public decision-makers can in uence rms' net worth accumulation (for example through
taxation, which is however not taken into account in this model), as well as monitoring costs (through
banking regulations for example).
Other shocks could have been considered, especially those originating from labor markets, as job sep-
aration rate or bargaining shocks. Shocks originating from the labor market could be at the source of
uctuations in the nancial sphere. For example, lower rms bargaining power or higher unemployment in
the economy may induce banks to increase their nancial mark-up as they may expect an increase in the
default risk of rms: a kind of "unemployment accelerator" as called by Blanco and Navarro (2016) could be
investigated. However, it is not the purpose of our paper (and of my dissertation) to consider this causality
link, from labor markets to nancial markets despite the interest of the question. It could however pave the
way for future research.
Net worth shock
We assume a one-% expansionary shock to the entrepreneurs net worth (the net worth of entrepreneurs
is 1% higher). According to the model, it corresponds to a positive increase in ςt . The aggregate real net
worth is increasing as the probability of dying for each entrepreneur is decreasing. A positive net worth
shock induces that all entrepreneurs will carry on to the next period t + 1 a higher net worth. Recall
that surviving solvent entrepreneurs carry their pro ts as a part of their net worth. And entrepreneurs,
based on their net worth amount and their expected production and returns, borrow funds from nancial
intermediaries to nance their capital rental costs, wage payments and hiring costs. So the net worth is a
mean for entrepreneurs to lower their use of nancial intermediaries' loans.
Figure 7 shows impulse response functions to a positive % net worth shock. Firstly, an increase in the
aggregate real net worth has a positive eect on the nancial sphere. By a positive eect, we mean that
it decreases the nancial mark-up, highlighting a nancial accelerator mechanism. The higher is the level
of the aggregate real net worth, the lower is the nancial mark-up charged by banks to entrepreneurs as
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 51
found by Zanetti and Mumtaz (2011). As a consequence of the real aggregate net worth increase and of the
nancial mark-up decrease, the default rate is decreasing. This decrease induces an increase in the gross
share of returns going to the rm according to the theoretical model.
These results on the nancial sphere impact the real sphere of the economy, especially the labor market.
−4 Output −5 Consumption −5 Investment
x 10 x 10 x 10
2 6 6
1 4 4
0 2 2
−1 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5
Employment: intensive margin −4
Employment: extensive margin −5 Unemplyment
x 10 x 10 x 10
5 1 0
−2
0 0.5
−4
−5 0 −6
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5 Vacancies −3 Job finding rate −4
x 10 x 10 x 10 Labor market tightness
1 1 1.5
1
0.5 0.5
0.5
0 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Real wage −3 Inflation −3
Nominal interest rate
x 10 x 10
0.2 0 0
−1 −1
0
−2 −2
−0.2 −3 −3
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5 Real Net worth −4Financial mark−up −3 Default rate
x 10 x 10 x 10
10 5 1
5 0 0
0 −5 −1
−5 −10 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Figure 7: IRF to positive networth shock
An increase in the aggregate real net worth of entrepreneurs leads to a decrease in the unemployment rate,
following a negative hump-shaped, thanks to an easing in their nancing costs. Indeed, entrepreneurs are
now able to nance by themselves a higher part of their production and vacancies costs, and for the part
nanced by loans, the nancial mark-up charged by banks is reduced. So the aggregate real wholesale rm's
52 4. QUANTITATIVE EXERCISE
marginal costs are reduced according to the model. This reduction induces entrepreneurs to post more
vacancies, and the unemployment rate decreases as a consequence as found in the estimation of Zanetti and
Mumtaz (2011). The job nding rate for workers increases and the labor market tightness from the rm
point of view increases.
More precisely, a substitution eect between hours worked per employee and the amount of employed work-
ers appears for the labor market in this calibration exercise as in the estimation of Christiano et al. (2011).
This phenomenon highlights the interaction between intensive and extensive margins on labor markets. As
a consequence of the easing of nancing costs, entrepreneurs are willing to hire more workers, but they
decrease their demand for hours worked per employee. This fact is known to be observed in data. Extensive
margin is indeed more reactive than the intensive margins when the economy is hit by a shock.
Furthermore, the link from the nancial market to the labor market goes also through wages. An easing
in nancial constraints increases the labor market tightness from the rm point of view. And the reduction
in aggregate real marginal costs is passed through prices according to our model, leading to a lower in ation
rate as observed in the calibration exercise. These two facts contribute to an increase of the real wage in
the economy. However, these results are dierent from the ones of Zanetti and Mumtaz (2011). The reason
may be that they integrate only the capital stock as the production cost to pay in advance of production.
Obviously, the previous results on the nancial sphere have also an impact on goods market. An increase
in the net worth of entrepreneurs generates a positive hump-shaped in consumption, through according to
the model, the higher wages, more vacancies posted and less unemployment. The net worth positive shock
induces also a positive hump-shaped, larger than the one for consumption, in investment. This increase
in investment can be explained by the model as the consequence of more employed workers and higher
wages for households that are able to invest more, and through a higher demand for capital expressed by
entrepreneurs, as the net worth increases and nancial mark-up decreases (Zanetti and Mumtaz (2011)).
The conjunction of higher consumption and higher investment leads to an increase in output, after a rst
decrease, possibly due to higher wages and higher vacancy posting costs.
So the economy is better o, at least in terms of consumption and employment, when all rms are able at
the aggregate level to self- nance a higher part of their production and vacancies costs. Financial frictions
are indeed less at stake as a consequence, through a decrease in the amount of loans demanded by rms
and a decrease in the level of marginal costs, realized by a lower nancial mark-up. The most important
impact appears to be on the labor market, where vacancies increase and unemployment rate decreases as
expected according to the theoretical model.
Monitoring cost shock
A positive monitoring cost shock is realized in the economy through a decrease in the monitoring cost, µt .
More precisely, the recovery rate de ned in the model following Livdan et al. (2009) and Petrosky-Nadeau
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 53
(2014) increases (so the monitoring cost decreases), by an increase of s0t interpreted as a monitoring cost
shock. Recall that in the model a monitoring cost appears because wholesale-good production is subject to
an idiosyncratic shock observed privately by entrepreneurs and not by nancial intermediaries. Thus, banks
have to pay a monitoring cost for bankrupt entrepreneurs to check the real output produced.
Figure 8 shows impulse responses to a positive monitoring cost shock. As for the net worth shock, a
positive monitoring cost shock has an impact on the nancial sphere. A decrease of the monitoring cost
leads to a reduction of the default rate. It was strongly expected as in the model, the monitoring cost
spending is a direct synonym of bankruptcy. The lower are the monitoring costs, the lower is the default
rate. Then, the positive monitoring cost shock leads to a reduction of the nancial mark-up charged by
banks to entrepreneurs. It could be easily explained through the model by the expression of the nancial
markup St that depends negatively on the recovery rate. A positive monitoring cost shock means that the
recovery rate for banks increases. The proportion of the real amount recovered by banks is higher. So banks
do not need to charge, everything else equal, the same amount of nancial mark-up to entrepreneurs.
For the real part of the economy, following the model, the agency problem between banks and en-
trepreneurs alters the real hiring cost and the marginal cost of production for wholesale rms. We observe
in this quantitative exercise that it is true: a positive monitoring cost shock induces an increase of va-
cancies and a decrease of unemployment following a negative hump-shaped with a peak at one year and
half. Petrosky-Nadeau (2014) nds in the same spirit following a negative monitoring cost shock, called
credit shock, a positive hump-shaped for unemployment with a peak just over a year after the shock, that is
considered as comparable to the research of Jermann and Quadrini (2012), but with a speci c insight into
the labor market. So, lower marginal costs, thanks to lower nancial mark-up, induce rms to post more
vacancies and to hire more. The labor market tightness from the rm point of view increases and the job
nding rate for households increases as a consequence as in Petrosky-Nadeau (2014).
The substitution eect between hours worked per employee and number of employed workers does not
appear here after a positive monitoring cost shock. On the contrary, a complementary eect is observed:
employed workers are more numerous, and each of them are working more. That is a rst reason why we
observe an immediate positive impact on the output compared to the previous case.
Concerning the output, in the model, we assume that the monitoring cost spending, synonym of bankruptcy
is spent in terms of currency, and not in terms of physical goods. Bankruptcy has no direct impact on the
real output. However, the simulation shows a clear indirect impact on output. Indeed, monitoring costs
are additional costs taken into account by banks when agreeing on an appropriate interest rate on loans. A
reduction in monitoring costs and bankruptcy rates has an impact on welfare indirectly, through their im-
plications on the mark-up pricing. The calibration illustrates this phenomenon: the decrease in monitoring
costs reduce real marginal costs of rms, reducing the level of in ation in the economy, as well as the level
of nominal interest rate. It generates an increase in consumption and investment that in turns increase the
output level.
54 4. QUANTITATIVE EXERCISE
−5 Output −4 −6
x 10 xEntrepreneurial
10 Consumption x 10 Real Net worth
10 1 3
5 2
0.5
0 1
−5 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5
Employment: intensive margin −5
Employment: extensive margin −5 Unemplyment
x 10 x 10 x 10
3 3 0
2 2 −1
1 1 −2
0 0 −3
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−6 Vacancies −4 Job finding rate −5
x 10 x 10 x 10Labor market tightness
4 4 4
2 2 2
0 0 0
−2 −2 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Real wage −3 Inflation −4
Nominal interest rate
x 10 x 10
0 1 5
0 0
−0.1
−1 −5
−0.2 −2 −10
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Monitoring cost −3Financial mark−up −3 Default rate
x 10 x 10
0 0 0
−0.1 −0.5 −2
−0.2 −1 −4
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Figure 8: IRF to positive monitoring cost shock
Finally, the aggregate real net worth increases as a consequence of the reduction of the monitoring cost
and of marginal costs. Solvent dying, and not dying, entrepreneurs have indeed a larger share of the net
output to consume, and to save to the next period.
Thus, a positive monitoring cost shock pushes down marginal costs and prices, as well as hiring costs
by a nancial mark-up depending on the level of monitoring costs. A reduction of monitoring costs in
an economy has a strong impact on the vacancies and employment levels as expected through the model.
Finally, we obtain as observed on gure 1 a positive correlation between unemployment and monitoring
costs (approximated in the gure by the risk premium, Baa-Aaa spread, in the economy).
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 55
Idiosyncratic volatility shock
An idiosyncratic volatility shock is de ned as a variation in the variance of the idiosyncratic shock
concerning entrepreneur's productivity30 . According to the model, this variance re ects the entrepreneurs
riskiness for banks. As a consequence, a positive idiosyncratic volatility shock means that banks are facing a
higher risk when they decide to lend to entrepreneurs. Recall that banks do not observe rms idiosyncratic
shocks, but they only know the cumulative and density functions of these idiosyncratic shocks.
Figure 9 shows impulse responses to a positive idiosyncratic volatility shock. As expected, after an in-
crease in the level of rms idiosyncratic volatility, banks charge a higher nancial mark-up to protect them
against asymmetric information. The default rate increases in quite huge proportion due to the increase of
idiosyncratic volatility in the economy, and due to the increase in the nancial mark-up that increases rms
real marginal costs.
On the real side, a higher idiosyncratic volatility decreases employment, output, consumption and
investment. As a consequence, labor market tightness from the workers point of view increases. Indeed,
rms are posting fewer vacancies. The reason is linked to the increase of the nancial mark-up. After an
increase in the level of idiosyncratic volatility, according to the model, the increase of the nancial mark-up
is passing through real marginal costs and prices in the economy. Firms are induced to post fewer vacancies,
to hire fewer workers because of this increase in nancing costs. Furthermore, in ation increases as well as
the nominal interest rate. Intensive and extensive margins appear here to move in the same direction as
after a monitoring cost shock. Hours worked per employee decrease also. So, fewer workers are employed
in the economy and each employed workers are working less. It explains why consumption and investment
are going down. Households have a less amount of resources to invest and the demand for capital by rms
decreases due to the increase in the nancial mark-up.
Thus, the economy reaches a negative position after an increase in the idiosyncratic volatility of whole-
sale rms. Facing higher risks of default, banks need, to overcome the agency problem they face with
entrepreneurs, to increase the nancial mark-up that they charge to them. The real marginal nancing
costs of entrepreneurs increase. It reduces as a consequence the value of a new hire and leads to lower va-
cancies. The unemployment goes down. Furthermore, the higher marginal costs are passed through prices
to the rest of the economy, leading to higher in ation, less consumption and less output in a new-Keynesian
perspective. Finally, we obtain as observed on gure 1 a positive correlation between unemployment and
entrepreneurs riskiness for banks (approximated in the gure by the default rate in the economy).
30 We took inspiration from nancial economics literature (Fu (2009) for example who uses GARCH models to estimate
expected idiosyncratic volatility) to name this shock.
56 5. CONCLUSION
−4 Output −5 Consumption −5 Investment
x 10 x 10 x 10
0 0 5
−0.5 −2 0
−1 −4 −5
−1.5 −6 −10
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5
Employment: intensive margin −4
Employment: extensive margin −4 Unemplyment
x 10 x 10 x 10
5 0 1.5
0 −0.5 1
−5 −1 0.5
−10 −1.5 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
−5 Vacancies −3 Job finding rate −4
x 10 x 10 x 10Labor market tightness
0 0 0
−0.5 −0.5
−0.5
−1 −1
−1.5 −1 −1.5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Real wage −3 Inflation −3
Nominal interest rate
x 10 x 10
0.5 6 4
4
0 2
2
−0.5 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Monitoring cost −3Financial mark−up −3 Default rate
x 10 x 10
−0.005 1.5 6
−0.01 1 4
−0.015 0.5 2
−0.02 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Figure 9: IRF to positive idiosyncratic volatility shock
5 Conclusion
We construct a new-Keynesien DSGE model integrating sticky prices, nancial frictions on the credit
market and asymmetric information on the labor market. We nd that credit market frictions may be the
source of lower posting vacancies and higher unemployment levels. Financial frictions push up wholesale
rms' marginal costs as well as hiring costs by a nancial mark-up charged by nancial intermediaries. This
nancial mark-up is then transferred by these rms on prices. Thus, it aects their hiring behavior, as
well as wage and employment levels in the economy. Then, we calibrate the model and we consider three
CHAPTER I. CREDIT IMPERFECTIONS, LABOR MARKET FRICTIONS AND
UNEMPLOYMENT: A DSGE APPROACH 57
shocks: a net worth shock, a monitoring cost shock and an idiosyncratic volatility shock. By moving three
dimensions involving dierent degrees of nancial frictions, we observe through the quantitative exercise a
clear impact from the nancial market shocks to the labor market. Increasing asymmetric information in
nancial markets (through higher monitoring costs, higher idiosyncratic volatility or lower entrepreneurial
net worth) leads clearly to higher unemployment in the economy through dierent channels. These channels
converge all to the role of the nancial mark-up that is charged by banks to overcome agency problems. This
nancial mark-up is passed through the rest of the economy by higher marginal costs and higher in ation.
That in turn reduces the levels of vacancies posting, employment, wages and consumption, and nally the
level of output. The evolution of credit market conditions changes the opportunity cost for resources used
to create new jobs. Thus, it alters the dynamics of job vacancies and unemployment. For the three shocks,
the unemployment jumps up or goes down to its highest level in the rst period and then slowly converges
back to its steady state level. The model could therefore also be able to generate the persistent evolution
of unemployment following a shock on credit markets. This model could then be estimate with Bayesian
methods to determine if it is able to to faithfully replicate unemployment, wages and vacancies data for the
United-States.
58 5. CONCLUSION
Appendices
59
61
A Proof
In this appendix, we prove that St ≥ 1, ∀ ω̄t ∈ [0, ∞), where St = {1 − µt [Γ(ω̄t ) + ω̄t h(ω̄t )f (ω̄t )]}−1 .
Using the assumption that ω̄t h(ω̄t ) is increasing in ω̄t and taking derivatives, we obtain:
µt d(ω̄t h(ω̄t )) 1
St0 = > 0.
1 − Φ(ω̄t ) dω̄t St2
Given this result, St is an increasing function of ω̄t .
Now taking limits of Γ(ω̄t ) and ω̄t h(ω̄t )f (ω̄t ) at the lower bound of ω̄ , we get:
lim Γ(ω̄t ) = 0, lim ω̄t h(ω̄t )f(ω̄t ) = 0
ω̄→0 ω̄→0
Thus, lim St = 1.
ω̄→0
Combining the previous results, St ≥ 1, ∀ ω̄ ∈ [0, ∞).
62 A. PROOF
Chapter II
Credit Constraints and Labor Market:
the role of Wage Bargaining Regimes
1 Introduction
Labor market institutions of each country are diverse and are sources of potential rigidity leading to a
multiplicity of equilibrium (Nickell and Layard (1999)). This chapter takes into account a speci c labor
market institutional form that are the wage bargaining schemes. Two bargaining schemes are considered,
the e cient bargaining (EB hereafter) and the right-to-manage bargaining (RTM hereafter). I compare
them in a labor search model with credit constraints on the credit market. The objective is to study under
each type of bargaining regimes, the impact on unemployment, wages and hours worked per employee of
labor and credit markets frictions. Many papers analyzing the eect of asymmetric information on labor
and credit markets assume that wages and hours worked per employee are determined through an e cient
bargaining between rms and workers. However in actual practice, rms have often the right-to-manage
workers. Hours per workers are unilaterally chosen by rms rather than determined by a bargaining between
rms and workers. So the question raised in this chapter is: in a world with credit frictions, how dierent
wage bargaining regimes aect labor market outcomes?
Borrowing constraints are sources of higher unemployment when the bargaining is called 'e cient'1 . By
modifying the bargaining regime, where hours worked are chosen unilaterally by rms, I obtain dierent
impacts of nancial and labor markets frictions on wages, hours worked per workers and unemployment
through the existence of a speci c mark-up that increases the degree of markets ine ciencies.
This chapter merges two strands of the literature. First, it is related to the literature dealing with the
1 The bargaining is not entirely 'e cient' anymore as soon as borrowing constraints are integrated as it will be showed in
this chapter. However I use as in Trigari (2006) the term 'e cient' bargaining to identify a speci c regime of bargaining, where
workers and rms bargain over wages and hours worked per worker.
63
64 1. INTRODUCTION
impact of nancial frictions on macroeconomic performances2 , including labor market outcomes. In my
upcoming model, nancial frictions are modeled following Kiyotaki and Moore (1997), and more precisely
Jermann and Quadrini (2012), where rms borrowing is limited by a collateral constraint. Jermann and
Quadrini (2012) investigate the consequences of the use of debt or equity nancing by capitalists on the dy-
namic of real and nancial variables. Credit shocks here directly aect the borrowing capacity of economic
agents.
Second, this chapter introduces labor market frictions à la Mortensen and Pissarides (1994), associated with
the limited borrowing capacities. Christiano et al. (2016), Garin (2015), Monacelli et al. (2011) and Zanetti
and Mumtaz (2011) introduce also these kinds of frictions in their models. They analyze the impact of -
nancial shocks on labor markets. However they do not distinguish between intensive and extensive margins.
And they are both using an e cient bargaining framework on labor markets.
Third, two bargaining regimes are considering in this chapter following the paper of Trigari (2006), that
introduced the distinction between 'e cient' bargaining and 'right-to-manage' bargaining.
My model is a partial equilibrium model that features search-and-matching frictions in the labor market,
enforcement constraints in the credit market and bilateral wage bargaining à la Mortensen and Pissarides
(1994). I introduce credit frictions through a collateral constraint that is recognized to be a powerful tool
to specify credit crunches. Then, I introduce a second layer of frictions in the labor market: the model
integrates Non-Walrasian labor market features, namely matching frictions generating unemployment in
equilibrium. I allow for variable hours per worker such that labor input can be adjusted along two margins,
the extensive margin (employment) and the intensive one (hours per worker). Finally, two types of bar-
gaining are considered, the Nash bargaining over both wages and hours per worker, the so-called 'e cient
bargaining' (EB), and the so-called 'right-to-manage bargaining' (RTM) where rms adjust hours unilater-
ally3 .
Before production takes place, a rm needs an intra-period loan to be able to pay in advance the work-
ing capital4 . However, due to asymmetric information, there is a risk of rm's default for the lender. To
protect himself against this risk, the lender imposes an enforcement constraint to the rm. Then, knowing
this enforcement constraint, the rm posts vacancies in the labor market at a unit vacancy posting cost,
where unemployed workers are searching for jobs. Matching occurs. Wages are bargained either through an
e cient bargaining or through a right-to-manage one. Hours per worker are determined jointly in an EB
set-up, whereas they are determined unilaterally by the rm in a RTM set-up. In that case, it is said that
wages are allocational for hours. With the intra-period credit, the rm has to pay vacancy posting costs
and production costs (capital and wage bill). The production of the rm occurs. At the end of the period,
2 A more developed literature review about this literature is developed in the chapter I of this dissertation.
3 About the terminology, I follow Trigari (2006).
4 See Garin (2015), Jermann and Quadrini (2012) for papers assuming a payment in advance of production of working
capital.
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 65
production is sold to households and capitalists and the intra-period loan is reimbursed.
I nd that the two bargaining regimes modify the extent to which nancial frictions interact with labor
market variables, meaning wages, hours worked per employee and unemployment. In EB and RTM set-ups,
nancial frictions reduce the relative bargaining power of the rm compared to a case without nancial
frictions. An ine ciency gap appears, the so-called nancial mark-up. Workers are able to extract a bigger
rent from the surplus. But, in a RTM set-up, nancial frictions reduce the relative bargaining power of
the rm to a lesser extent as the rm has a higher degree of freedom in the bargaining. Another source of
ine ciency appears, the so-called wage mark-up. So with a RTM regime, the impact of nancial frictions
exists but it is mitigated by the fact that the rm is able to modify the level of hours worked by each worker.
Thus, a rm compensates the existence of nancial frictions by reducing the level of hours demanded for
each worker, that increases aggregate markets ine ciencies.
Thus in both cases, a higher level of collateral constraints leads to an increase in the worker's bargaining
power. They ask for a bigger rent, but their capacities to extract a bigger part of the surplus depend on the
bargaining regime: a RTM regime appears to restore partly the bargaining power of rms by giving them
a higher degree of freedom in the bargaining process that is not internalized by workers. I identify to that
purpose two ine ciency gaps compared to a case without any friction, a nancial and a wage ine ciency
gap, the last one being present only under a RTM regime. Firms use intensive margins to alleviate nancial
frictions under a RTM bargaining regime, that triggers higher markets ine ciencies.
This chapter is organized as follow. Section 2 describes the model integrating labor and credit frictions
with two possible bargaining regimes (EB and RTM). Section 3 analyses the implication of both models in
terms of ine ciencies generated by the presence of labor and credit market frictions, under RTM and EB
regimes. Section 4 concludes.
2 Model
2.1 Model overview
The model is an in nite horizon partial equilibrium model and is in discrete time. The economy fea-
tures three sets of agents: households and capitalists that own and manage rms. Agents interact in three
dierent markets: labor, credit and good markets.
Each household works, consumes and saves through a one-period riskless discount bond used to nance
inter-temporal loans for rms. Household members are either employed or unemployed workers. Each cap-
italist consumes, owns and manages rms. They do not borrow or save in the form of bonds as it will be
66 2. MODEL
done on their behalf by the rms they own. They only earn dividend incomes. Firms own the capital stock
of the economy, post vacancies, recruit workers and produce goods. Labor is hired in a frictional labor
market modeled as in Mortensen and Pissarides (1994).
So as to nance its activities, each rm has access to two sources of nancing: debt and equity. For debt,
there are an inter-period loan and an intra-period loan. The inter-period loan, nanced with the one-period
riskless discount bonds, is required for the lender to have an interest to keep the rm active in case of default
of the rm on the intra-temporal loan5 . This intra-period loan aims to nance in advance of production
the production bill as in Jermann and Quadrini (2012), the dividends, as well as the cost of job vacancy
posting since the rm does not possess its own cash. However, due to uncertainty related to recovering this
loan, intra-period loan issuance is collateralized by the capital stock of the rm. So the idea behind this
kind of nancial frictions is that rm uses its capital stock as a collateral in order to nance their working
capital requirements. After the sale of the production, the intra-period loan is reimbursed without default
as it is always enforced (default is thus not possible at the equilibrium in this kind of model), and there is
no interest.
On the labor market, the search and matching process for jobs is costly. A matching technology brings
together unemployed workers to vacancies. Job creation occurs when a rm and a searching worker meet
and agree to match at a bargained wage, which depends on the parties' bargaining power. This bargaining
power is dierent under the dierent bargaining regimes (RTM or EB regimes). The match continues until
the parties exogenously end the employment relationship at the beginning of each period.
The model builds on the work of Garin (2015), and the modeling of the labor market with an e cient
or a right-to-manage bargaining follows the article of Trigari (2006).
Figure 10: Timing of events
5 As developed in appendix A.
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 67
Each period is decomposed in four steps. First, technological and credit shocks are realized6 and a
fraction of last period matches are destroyed exogenously. Second, an intra-period loan is contracted by
the rm due to a cash- ow mismatch to nance its total working capital. The borrowing capacity of the
rm is constrained by the amount of capital accumulated before as there exists a risk of rm's default. The
rm decides, knowing this constraint, how much it wants to post vacancies, hire and invest. Third, after a
matching process on the labor market, the rm bargains with the worker over the hourly wage, and over the
hours worked in an EB regime. In a RTM regime, hours worked per employee are determined unilaterally
by the rm, depending on the wage bargained. Then, the rm pays in advance the working capital. Fourth,
the rm produces and reimburses the intra-period loan (as well as the inter-period loan from period t − 1).
2.2 Labor market
The labor market is modeled using a standard Mortensen and Pissarides (1994) search and matching
framework. At each period t, rms post vacancies, Vt , and unemployed workers, Ut , are searching for
jobs. Each rm and each unemployed worker meet randomly. The matching of workers and rms is es-
tablished by a standard matching function, Mt , measuring the aggregate ow of hires in a unit period :
M (Ut , Vt ) = νUtσ Vt1−σ , where ν > 0 is a scale parameter re ecting the e ciency of the matching process
and σ ∈ (0, 1) is the relative weight the pool of unemployed workers receives in this matching process. As
standard, the number of new hires increases with the number of unemployed, as well as with the number of
vacancies.
Assumption 1. Transitions between in and out the labor force is not considered. There is full participation
of household members at all times, that are either employed or searching for a job.
Assumption 2. Job-to-job transition is not considered. Only unemployed workers can look for a job. And
an employed worker can be employed in at most one job.
The labor market tightness from the rm point of view is de ned as the ratio between vacancy jobs
and unemployed workers searching for a job: θt := Ut .
Vt
The probability that a vacancy gets lled during
the matching process of time t, the job lling rate, is: qt := Mt
Vt = νθt−σ . And the probability for an
unemployed worker to nd a job during the matching process of period t, the job nding rate, is de ned
as: st := Mt
Ut = νθt1−σ , such that st = θt qt and ∂qt /∂θt < 0, ∂st /∂θt > 0. The higher is the labor market
tightness from the rm point of view, the higher is the probability for an unemployed worker to nd a job
and the lower the probability for a rm to ll a job. It represents the congestion externalities of new vacan-
6 I introduce here two shocks that will not be used in this chapter for any empirical or quantitative exercise. However, as
for future research it would be interesting to develop a general stochastic dynamic equilibrium model, I choose to set these two
shocks.
68 2. MODEL
cies or of new unemployed workers for respectively other rms and other unemployed workers searching for
employment. Both unemployed workers and rms are assumed to take qt and st as given.
The number of employed workers at the beginning of period t is denoted by Nt−1 . Employment rela-
tionships may be severed for exogenous reasons at the beginning of any given period, before matching of
period t takes place7 .
Assumption 3. Workers whose the match has been exogenously destroyed at the beginning of period t
participate immediately to the matching process of period t, as the match destruction takes place before
the matching process within the period. There is no need for them to wait until the next period.
Assumption 4. New hires of period t start working immediately once they are hired as the matching
process takes place before production within the period. Unmatched unemployed workers remain jobless.
I denote by ρ the exogenous destruction probability, so that employment evolves according to the fol-
lowing law of motion:
Nt = (1 − ρ)Nt−1 + M (Ut , Vt ) (II.1)
According to the assumptions made, employment at time t, Nt , i.e. the number of workers who will
participate to the production of period t, is given by the number of total matches from t − 1 that survived
to the next period, (1 − ρ)Nt−1 , plus the newly hires of period t, M (Ut , Vt ).
Finally, the number of unemployed household members searching for a job in period t is Ut = 1−(1−ρ)Nt−1 8 .
As the labor force is normalized to one, the number of unemployed members at the end of period t or after
the matching process of period t is given by ut = 1 − Nt .
2.3 Households
There is a representative household seen as a large family composed of a continuum of members rep-
resented by the unit interval. Each period, each household member can be either employed by a rm or
unemployed according to the process described in the previous section.
As in Merz (1995) and Andolfatto (1996), there is a full risk sharing of consumption in order to avoid
distributional issues due to heterogeneity in incomes among household members. The household pools the
income of all its members such that a perfect consumption is fully insured for all members9 . So Ch,t denotes
the consumption enjoyed by a given household member as well as overall household consumption. Each
7 As separation is exogenous, I omit the fact that a rm that nds a worker prior to period t decide to produce or not. The
"no production" point is used as a threat point in the wage bargaining process. Implicitly, at equilibrium, the bargaining set
will be always non-empty.
8 This unemployment rate is evaluated after the exogenous matches destruction occurs at the beginning of period t.
9 Unemployed members are better o than the employed ones since they have the same consumption but they enjoy all the
leisure. It is a strong assumption, however commonly used in the literature.
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 69
household member has the following separable utility function:
vh (Ch,t ) − g(Ht ) (II.2)
Ht1+φ
where vh (Ch,t ) = log(Ch,t ) and g(Ht ) = ϕ 1+φ , ϕ > 0, φ > 0. Ht is the measure of hours worked by each
employed household members at time t, ϕ accounts for the weight on hours in labor disutility and φ is the
inverse of the intertemporal elasticity of labor substitution.
The representative household has the following expected lifetime utility function:
∞
X
E0 βht [vh (Ch,t ) − G(Ht , Nh,t )] (II.3)
t=0
where βh is the household's intertemporal discount factor, Nh,t is the number of employed household mem-
bers at time t and G(Ht , Nh,t ) is the sum of disutilities from labor supply of employed members at time
t, de ned as G(Ht , Nh,t ) = Nh,t g(Ht ). It is thus equal to the household's disutility from supplying labor
hours.
In the representative household, Nh,t household members are employed and each of them earns a total wage
Wt Ht . 1 − Nh,t household members are unemployed as the labor force is normalized to one and each of them
earns unemployment bene ts, bt . Unemployment bene ts are de ned as bt = bw Wt Ht , where bw ∈ (0, 1)
corresponds to the ratio of unemployment bene ts or period t to the wage bill of period t. Finally, the
household is holding one-period discounted bonds, Bt . These revenues are used by the household to nance
its expenditures. The household consumes, Ch,t , contributes to unemployment bene ts trough lump-sum
taxes, Tt , and purchases in period t the riskless bonds that pay o in period t + 1, Bt+1 .
Conditional on {Ht }∞
t=0 and taking as given the set of prices {Wt , Rt }t=0 , the representative household
∞
chooses sequences of consumption {Ch,t }∞
t=0 , the number of employed workers {Nh,t }t=0 and discounted
∞
t=0 to maximize its expected discounted utility (II.3) under the following budget constraint:
bonds {Bt+1 }∞
Bt+1
Wt Ht Nh,t + (1 − Nh,t )bt + Bt ≥ Ch,t + + Tt (II.4)
Rt
The household does not choose hours of work per worker. These hours are determined at the labor market
equilibrium with search and matching frictions, through decentralized bargaining between a representative
rm (described in section 2.5) and a representative household (e cient bargaining, EB) or unilaterally by
the rm (right-to-manage, RTM). The wage, Wt , is in both cases set through decentralized Nash bargaining
between representatives rm and household.
Finally, the household takes the job nding rate as given, hence perceiving employment evolving according
70 2. MODEL
to:
Nh,t = (1 − ρ)Nh,t−1 + M (Ut , Vt ) = (1 − ρ)Nh,t−1 + st Uh,t (II.5)
The problem of the representative household is written recursively. The vector of individual states is
ωh,t = {Bt , Nh,t−1 }. These variables are also called control variables, meaning that they are the ones that
have been chosen by the household. The vector of aggregate states is Θt = {Kt , Nt−1 ; At−1 , ξt−1 }. Kt
denotes the capital stock of the economy for period t decided in t − 1 by rms. At−1 is the aggregate
technological shock and ξt−1 is the credit shock, that will be de ned in the subsequent section.
The vector of aggregate states is a set of variables that in uences the agent's return within the period but by
assumption these variables are outside of the agent's control within the period t. These state variables evolve
over time in a way that can be in uenced by the control variables. Given a current aggregate state and
current individual states, the aggregate states vector for the subsequent period is determined. The aggregate
state vector completely summarizes all of the information from the past that is needed to make a forward-
looking decision. The idea is to replace the attempt to locate equilibrium sequences of contingent functions
with the search for time-invariant equilibrium decision rules. These decision rules specify current actions
as a function of a limited number of state variables which fully summarize the eects of past decisions and
current information. Knowledge of these state variables provides the economic agents with a full description
of the economy's current state. Their actions, together with the realization of the exogenous uncertainty
determine the values of the state variables in the next sequential time period.
The value function for the representative household, Ωt (ωh,t ; Θt ), is de ned as:
Ωt (ωh,t ; Θt ) = max vh (Ch,t ) − G(Ht , Nh,t ) + βh Et Ωt+1 (ωh,t+1 ; Θt+1 )
Ch,t ,Bt+1 ,Nh,t
Bt+1 (II.6)
subject to Wt Nh,t Ht + (1 − Nh,t )bt + Bt ≥ Ch,t + + Tt
Rt
Nh,t = (1 − ρ)Nh,t−1 + M (Ut , Vt )
which yields the two standard e ciency conditions, relative to consumption and discounted bonds:
0
λh,t = vh (Ch,t ) (II.7)
λh,t
= βh Et λh,t+1 (II.8)
Rt
where λh,t is the Lagrange multiplier of the household' budget constraint.
These conditions lead to the usual consumption Euler equation that describes the consumption decision:
0
1 v (Ch,t+1 )
= βh Et h 0 (II.9)
Rt vh (Ch,t )
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 71
The last e ciency condition relative to employment is useful to determine because it gives a measure of the
surplus value of a marginal worker for the household:
g(Ht )
Sth = Wt Ht − bt − + (1 − ρ)Et Λht+1|t (1 − st+1 )St+1
h
(II.10)
λh,t
where Sth is the ratio between the Lagrange multipliers relative respectively to the law of motion for Nh,t
and to the budget constraint, λh,t . Sth represents the marginal contribution of a newly created job to the
household in terms of utility, that is given by the wage net of the disutility of work and unemployment
bene ts that are foregone (the outside option), plus the expected discounted surplus from continuing the
matching in t + 1. 1 − ρ(1 − st+1 ) employed household members continue to match with a rm in the next
period and st+1 unemployed members nd a job in the next period10 .
Finally, the discounted intertemporal marginal rate of substitution de ned in equation (II.9) is equal at the
0
vh (Ch,t+1 )
equilibrium to the household's stochastic discount factor Λht+1|t ≡ βh 0
vh (Ch,t )
.
2.4 Capitalists
Capitalists have a lifetime utility vc (Cc,t ). They are the only ones to have access to the ownership of
rms in respective individual shares zt 11 . dt are the aggregate dividends paid by all rms. Capitalists'
optimization problem is written recursively as:
Γt (zt ; Θt ) = max vc (Cc,t ) + βc Et Γt+1 (zt+1 ; Θt+1 )
zt+1 ,Cc,t
(II.11)
subject to zt dt = Cc,t + zt+1
where Γt (zt ; Θt ) is the value function for capitalists which depends on the aggregate state Θt and on the
individual state, the share of rms owned by each capitalist, zt .
βc is the capitalist discount factor. I assume that capitalists are more impatient than households, so that
βh ≥ βc 12 . Furthermore, as standard in the literature, I assume that capitalists do not borrow or save in
the form of bonds because it will be done on their behalf by the rms they own as described in section (2.5).
The rst-order conditions of the optimization problem give the typical Euler equation:
0 0
vc (Cc,t ) = βc Et vc (Cc,t+1 )dt+1 (II.12)
Capitalists are homogeneous agents and they earn only dividend incomes from rms. Therefore, in equi-
10 The net matching value of a marginal worker for the household can be also expressed as Sth ≡ Wth − Uth where Wth =
t+1 )Ut+1 and Ut = bt + Et Λt+1|t st+1 Wt+1 + (1 − st+1 )Ut+1 .
g(Ht )
+ Et Λh h + ρ(1 − s h h h h h
Wt Ht − λh,t t+1|t
(1 − ρ(1 − st+1 ))Wt+1
11 Workers save only in the form of bonds. The market is segmented.
12 This condition allows to insure that at the steady state the rm borrows from the household and that the borrowing
constraint faced by the rm is always binding at the steady state. See appendix B for more details.
72 2. MODEL
librium zt = zt+1 = 1 (all the representative capitalists have indeed a fraction of all the representative
rms) and thus Cc,t = dt . It follows that the representative capitalist discounts future dividends by
0
vc (dt+1 )
Λct+1|t = βc Et vc0 (dt )
. Since the capitalist owns the rm, this will also be the discount factor used by
the representative rm.
2.5 Firms
Firms produce a homogeneous good, Yt , following a decreasing return to scale function13 , using capital,
Kt , and labor, Nc,t Ht :
Yt = At Ktα (Nc,t Ht )1−α (II.13)
where At is an aggregate technological shock. As there is no idiosyncratic shocks, all rms choose the same
allocations in equilibrium. So all rms are identical and symmetry among them is assumed ex-ante.
The representative rm is owning the capital stock of the economy. Consistent with the standard timing
convention, the capital stock used in t is selected at time t − 1. Therefore it is predetermined at time t.
Instead the labor force can be changed in period t, at the extensive margin as well as at the intensive margin.
The law of motion for capital accumulation is:
Kt+1 = (1 − δ)Kt + It (II.14)
where δ is the capital depreciation rate and It is the investment.
The objective of the rm is to maximize the sum of its expected discounted dividends:
∞
X
max E0 Λct+1|t dt (II.15)
dt
t=0
where Λct+1|t is the stochastic discount factor of the capitalist as it owns the rm.
The rm faces the following budget constraint:
at+1
Yt + = dt + Wt Ht Nc,t + It + at + κVt (II.16)
Rt
where κ is unit vacancy posting costs, at is the amount of the matured inter-temporal debt and at+1 /Rt is
the value of the new debt.
The rm's budget constraint equals the dierence between revenues from production, Yt , plus the new
at+1
inter-temporal liabilities, Rt , net of equity payout, dt , labor wages, Wt Ht Nc,t , investment, It , vacancies
posting costs, κVt and the reimbursement of inter-temporal liabilities, at .
13 The same production function is de ned by Faccini et al. (2011).
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 73
Credit market
The rm begins each period with inter-temporal liabilities at . The payments to workers, capitalists,
expenses related with capital accumulation, with creating a vacancy and current debt net of the new issue
is assumed to be made before the production occurs. To be able to do so, the rm contracts an intra-period
loan, lt , to cover the cash ow mismatch during the period. The intra-period loan is written as:
at+1
lt = dt + Wt Ht Nc,t + κVt + It + at − (II.17)
Rt
This loan has to be entirely repaid at the end of the period t after the production and sale take place. As
it is repaid within the period, there is no interest. Note from equation (II.16) that lt = Yt , meaning that
the intra-period loan equals the total expected revenue.
As in Kiyotaki and Moore (1997), nancial frictions arise due to a costly contract enforcement relative
to the intra-period loan. The rm is subject to collateral requirements because it can default on its reim-
bursement obligations. Indeed, the lender can not force the rm to repay its debt unless this debt is secured
by collateral assets. The collateral constraint nally limits the amount that is borrowed as a function of the
rm's nancial wealth.
In my framework, the repayment obligation of the rm may not exceed a fraction ξt of the value of the
capital stock accumulated by the rm. ξt measures the degree of credit frictions. It is interpreted as a
credit shock, or more precisely as a collateral shock, since it aects the rm's capacity to borrow from the
lender14 . As in Jermann and Quadrini (2012), this shock is assumed to be stationary and evolves according
to15
iid
log(ξt ) = ρξ log(ξt−1 ) + uξt , ρξ ∈ (0, 1) where uξt ∼ N (0, σξ2 )
A negative shock implies a deterioration in credit market conditions.
Assumption 1. Following a default, production can not be seized by the lender.
Assumption 2. Following a default and before next period's capital incorporation, the lender con scates
the rm and sells each unit of capital at ξt qK,t .
Assumption 3. Following a default, the lender has no bargaining power in the debt bargaining and it does
not value the stock of workers within the rm16 .
14 The nancial behavior of the lender is not described here. It is done in appendix A where the borrowing constraint is
derived.
15 This shock is here speci ed in the perspective of future research where a general dynamic stochastic equilibrium could be
developed and estimated.
16 It could be interesting to modify this assumption in future research so as to take into account a potential "unemployment
accelerator" as in Blanco and Navarro (2016) where workers value directly aect the rms' nancial conditions.
74 2. MODEL
Formally with these assumptions, the collateral constraint faced by the rm is given by17 :
at+1
lt ≤ ξt qK,t Kt − (II.18)
Rt
The expected value of the recoverable capital net of inter-temporal debt must be at least higher than the
intra-period loan needed to produce. In this constraint, the lender uses the shadow value of the capital,
given by the marginal Tobin's Q, qK,t , to assess the collateral value and set the borrowing limit accordingly.
This constraint represents a rm's borrowing capacity restriction in the credit market. The rm is limited
by the collateral constraint when it wants to invest, hire workers or pay dividends to capitalists, which is a
function of rm's beginning of period capital stock.
From now on, I distinguish the rm's decisions depending on the bargaining regime that is considered
on the labor market. Indeed, the choice variables are dierent for the rm in the dierent regimes. In EB,
the rm does not choose the hours worked per workers, whereas under RTM, it is the case.
Firm's optimization decisions with a right-to-manage bargaining
The rm chooses dividends {dt }∞
t=0 , new inter-temporal debt {at+1 }t=0 , investment {It }t=0 , hours
∞ ∞
worked per employee {Ht }∞
t=0 and vacancies {Vt }t=0 so as to maximize the discounted sum of dividends
∞
(II.15) subject to the budget constraint (II.16), the collateral constraint (II.18), laws of motion of capital,
Kt+1 (II.14) and of employment, Nc,t .
The problem of the representative rm is written recursively. The vector of individual states is ωc,t =
{at , Nc,t−1 , Ht−1 , Kt }. ∆t (ωc,t ; Θt ) is the value function for the rm under a right-to-manage bargaining,
that is de ned as:
∆t (ωc,t ; Θt ) = max dt + βc Et ∆t+1 (ωc,t+1 ; Θt+1 )
dt ,Kt+1 ,It ,at+1 ,Nc,t ,Ht ,Vt
at+1
subject to At Ktα (Nc,t Ht )1−α + = dt + Wt Ht Nc,t + It + at + κVt
Rt
at+1
lt + ≤ ξt qK,t Kt
Rt
Nc,t = (1 − ρ)Nc,t−1 + qt Vt
Kt+1 = (1 − δ)Kt + It
17 The derivation of this constraint is given in appendix A.
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 75
The rst-order conditions are respectively:
(dt ) µc,t = 1 − µb,t (II.19)
(at+1 ) 1
Rt
1
= Et Λct+1|t ( 1−µ b,t
) (II.20)
(Kt+1 ) µk,t = Et Λct+1|t [µb,t+1 ξt+1 qK,t+1 ] (II.21)
α−1
+Et Λct+1|t (1 − µb,t+1 )At+1 αKt+1 (Nc,t+1 Ht+1 )α + µk,t+1 (1 − δ)
(It ) µk,t = µb,t + µc,t (II.22)
(Nc,t ) Stc = (1 − µb,t )(At (1 − α)Ktα Ht (Nc,t Ht )−α ) − Wt Ht + (1 − ρ)Et Λct+1|t St+1
c
(II.23)
(Vt ) Stc = κ
qt (II.24)
(Ht ) Wt = (1 − µb,t )(1 − α)At Ktα (Nc,t Ht )−α (II.25)
where µc,t , µb,t and µk,t are respectively the Lagrange multipliers associated to the budget constraint, the
borrowing constraint and the law of motion of capital.
Stc is the Lagrange multiplier on the law of motion of employment and measures the net matching value
of a marginal worker for the rm. The surplus for the rm from having a new worker corresponds to the
pro t generated by a worker at each period, given by (1−µb,t )At Ktα Ht (Nc,t Ht )−α −Wt Ht and this matching
continues to next period with a probability (1 − ρ). The pro t generated by an individual worker integrates
here the shadow value of the use of external nancing, µb,t .
The expected productivity of capital and labor are augmented by a wedge that depends on the tightness
of the enforcement constraint. A tighter credit constraint increases the cost of having an extra unit of labor
and capital and reduces as a consequence their demands.
Then, equation (II.24) implies that the unit cost of lling a vacancy, κ, times the average vacancy
duration, qt must equal the surplus value of employment for the rm. With the equation (II.23), I obtain
the standard hiring condition:
κ κ
= (1 − µb,t )(At (1 − α)Ktα Ht (Nc,t Ht )−α ) − Wt Ht + (1 − ρ)Et Λct+1|t (II.26)
qt qt+1
A rm posts vacancies until the cost of posting and maintaining a vacancy equals the bene t of having a
new employment match. The bene t for the rm of hiring a new worker is his shadow value integrating
the tightness of the credit constraint, plus the vacancy posting costs saved if the employment relationship
continues.
Hours per worker are determined here through the equation (II.25) as the rm has the right-to-manage
hours for a given wage that will be determined in the next section. The rm unilaterally chooses the hours
of work for a given bargained wage to maximize the shadow value of a new worker, Stc . I nd that the wage
76 2. MODEL
is equal to the marginal product of an hour worked by a worker augmented by a wedge that depends on
the tightness of the collateral constraint. After rearranging the equation, I obtain the hours demand under
right-to-manage bargaining conditional to WtRM :
! α1
−α
(1 − µb,t )(1 − α)At Ktα Nc,t
HtRM = (II.27)
WtRM
Firm's optimization decisions with an e cient bargaining
With an e cient bargaining, the only dierence in the rm's optimization decisions is the absence of
the equation (II.25) relative to hours, as hours worked per workers are determined through Nash bargaining
between workers and rms. So the rm chooses dividends {dt }∞
t=0 , number of riskless bonds {at+1 }t=0 ,
∞
investment {It }∞
t=0 and vacancies {Vt }t=0 so as to maximize the discounted sum of future dividends (II.15)
∞
subject to the budget constraint (II.16), the enforcement constraint (II.18), the law of motion of capital,
Kt+1 (II.14) and the law of motion of employment, Nc,t .
The problem of the representative rm is written recursively. The vector of individual states is here
ωc,t = {at , Nc,t , It }. I obtain for the rst-order conditions the same equations: (II.19)-(II.24).
2.6 Wage contracts
Once a worker is matched with a vacancy, the representative rm and the representative worker bargain
through a generalized Nash bargaining, so that the outcome of the bargaining process maximizes the Nash
product:
max (Sth )η (Stc )1−η (II.28)
where the rst term in brackets is the household surplus value from an employment relationship de ned
in equation (II.10) and the second term in brackets is the rm surplus from an employment relationship
de ned in equation (II.23). η re ects the relative bargaining power of workers.
The e cient bargaining concerns the wage and working hours per workers, whereas the right-to-manage
bargaining covers only the wage, hours worked having been implicitly de ned by the equation (II.25).
The right-to-manage bargaining
A rm and a worker bargain over the wage through the generalized Nash bargaining (II.28), as the
rm unilaterally chooses the optimal hours of work determined by equation (II.25) for a given bargained
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 77
wage. In practice, the rm is able to adjust the total hours worked by changing the worked hours per
workers. Indeed, the e cient bargaining solution derived below where every adjustment in working hours
is associated with a renegotiation of wages seems out of step with reality. A large proportion of wages in
Europe for example are covered by collective bargaining agreements at the sectoral level, leaving the rm
with the optimal decision on the demand for hours only (part time or full time jobs for example). As a
consequence, the right-to-manage model seems to characterize the institutional setup in Europe better than
the e cient bargaining model.
There are two stages in the right-to-manage bargaining in each period. First, wages are determined by
Nash bargaining between a worker and a rm so as to maximize their joint surplus. Second, expected this
wage, the rm chooses the level of hours per worker so as to maximize the matching value for the rm,
Stc , as derived in equation (II.25). So before the rm sets the hours worked per employee, the rm and the
worker choose the wage so as to maximize the Nash product, taking as given the eect of wages on hours.
The wage chosen by the match satis es the following optimal condition:
ηδth Stc = (1 − η)δtc Sth (II.29)
where δth and δtc denote respectively the marginal surplus from an increase in the wage to the worker and
to the rm. Thus, the wage is set to equate the proportional marginal surplus to each party, weighted by
each party's bargaining power. Given Wt , the marginal contribution of wages to the value of a match to
the worker is:
∂Sth ∂Ht
δth = = Ht + (Wt − mrst ) (II.30)
∂Wt ∂Wt
The marginal contribution of wages to the value of a new worker to the rm is:
∂Stc ∂Ht
δtc = − = Ht + (Wt − (1 − µb,t )mpht ) (II.31)
∂Wt ∂Wt
Hφ
where mrst = ϕ λh,t
t
and mpht = (1 − α)2 NYt H
t
t
, corresponding respectively to the marginal rate of substi-
tution for the worker and to the marginal product of hours worked per worker for the rm.
Finally, I derive from equation (II.29) the wage equation for the right-to-managed bargaining regime:
κ
WtRM α −α
= χt (1 − µb,t )At (1 − α)Kt (Nt Ht ) + (1 − ρ)Et Λt|t+1 c
(II.32)
qt+1 Ht
bt mrst h χ t+1 κ
+(1 − χt ) + − (1 − ρ)Et Λt|t+1 (1 − st+1 )
Ht 1+φ 1 − χt+1 qt+1 Ht
78 2. MODEL
where
ηδth
χt = (II.33)
ηδth + (1 − η)δtc
The e cient bargaining
A rm and a worker determine jointly the wage, Wt , and the hours worked, Ht , so as to maximize the
Nash product (II.28). The outcome can be seen as privately e cient as a consequence and is equivalent to
a model where hours are chosen to maximize the joint surplus of a match, while the wage is set to split the
surplus according to the parameter η , the worker bargaining power.
Wage Bargaining. The wage for the e cient bargaining satis es the following optimal condition:
ηStc = (1 − η)Sth (II.34)
Here, δth = Ht and δtc = Ht , so that these terms disappeared. It is the consequence of the fact that hours
worked per worker and the wage are now determined jointly (hours do not depend here on the wage).
Substituting the expressions for the matching values, Sth and Stc yields:
bt mrst
WtEB = η (1 − µb,t )At (1 − α)Ktα (Nt Ht )−α + (1 − η) (II.35)
+
Ht 1+φ
κ κ
+η(1 − ρ) Et Λct|t+1 − Et Λht|t+1 (1 − st+1 )
qt+1 Ht qt+1 Ht
Condition (II.35) splits the wage into costs and bene ts of an employment match according to the bargaining
power η . The wage compensates the worker up to a fraction η of its marginal productivity and the saving
of hiring costs from subsequent periods, and up to a fraction 1 − η of the endogenous outside option of a
worker that is disutility of labor and the foregone unemployment bene t. The labor marginal productivity
is reduced by the tightness of the collateral constraint, µb,t .
Hours Bargaining. Hours worked per worker are chosen according to the following optimal condition derived
from the Nash product (II.28):
(1 − µb,t )mpht = mrst (II.36)
The marginal rate of substitution equals the marginal product of hours worked per worker, times a wedge
that depends on the tightness of the collateral constraint. Any change in hours will be associated with a
change in the wage so that any renegotiation of hours and wages will yield an agreement on the contract
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 79
curve. Equation (II.36) corresponds to:
ψHtφ
(1 − µb,t )(1 − α)2 At Ktα (Nt Ht )−α = (II.37)
λh,t
which gives nally for the hours worked per worker under e cient bargaining:
1
(1 − µb,t )(1 − α)At Ktα Nt−α φ+α
HtEB = λh,t (II.38)
ψ
2.7 Market clearing
The good market clearing, including posting vacancy costs, is given by:
Ct + It + κVt = Yt (II.39)
where Ct = Cc,t + Ch,t .
The credit market clearing is given by:
at+1 = Bt+1 (II.40)
I abstract from modeling government spending for simplicity, and assume that all unemployment bene ts
are nanced by lump-sum taxes:
bt (1 − Nt ) = Tt (II.41)
The recursive equilibrium is de ned by the following equilibrium conditions: the household's Euler
equation (II.9), the household's surplus of having a new employed worker (II.10), the capitalist's Euler
equation (II.12), the rm's Euler equation for capital (II.21)-(II.22), for dividends and bonds (II.19)-(II.20),
the rm's capital law of motion (II.14), the vacancy creation condition (II.26), the aggregate law of motion
for employment (II.1), the credit market clearing constraint (II.40), the aggregate resource constraint (II.39),
the government budget constraint (II.41), and for the EB regime, the EB wage rule (II.35) as well as the
EB hours per worker rule (II.38), or for the RTM regime, the RTM wage rule (II.32) as well as the RTM
hours per worker rule (II.25). The stochastic process are taken as given {ξt }∞
t=0 .
3. SOURCES OF AGGREGATE INEFFICIENCIES UNDER RIGHT-TO-MANAGE AND EFFICIENT
80 BARGAINING REGIMES
3 Sources of aggregate ine ciencies under right-to-manage and ef-
cient bargaining regimes
As stipulated by Gali et al. (2007), market frictions can trigger ine cient uctuations in the allocation
of resources. I use their measure of aggregate ine ciency to examine under both wage bargaining regimes,
the impact of credit and labor market frictions on the economy. This measure is called 'ine ciency gap'
compared to the standard equilibrium conditions. It corresponds to the wedge between the marginal product
of hours worked per worker and the the marginal rate of substitution between consumption and leisure.
Indeed, if the marginal product of hours is equal to the marginal rate of substitution, hours per worker are
at their e cient level. Hence, deviations from zero represent an ine cient allocation of hours worked per
worker.
Let de ne the 'ine ciency gap' as:
mpht
gapt = (II.42)
mrst
As explained by Gali et al. (2007), the gap variable represents the distance between the perfectly competitive
hours supply and hours demand curves, evaluated at the current level of hours. I decompose this ine ciency
gap between two mark-ups : a wage and a nancial mark-up.
I de ne the nancial mark-up as the ratio between the marginal product of hours worked per worker
and the wage, ωf,t = mpht
Wt and the wage mark-up as the dierence between the wage and the marginal rate
of substitution, ωw,t = mrst .
Wt
Basically, the wage mark-up corresponds to a measure of search and matching
frictions, whereas the nancial mark-up corresponds to a measure of nancial frictions. So frictions are
here associated to variations in the aggregate level of macroeconomic e ciency. These two kind of frictions
are supposedly going to divert from the equilibrium standard conditions, the hours of work per worker, as
well as equilibrium wages. The question is thus to determine if these mark-ups exist and are signi cantly
dierent under both bargaining regimes, and to what extent.
Proposition. The ine ciency gap under EB and RTM can be decomposed as:
1+φ
(1 − µb,t )−1 under RTM
mpht 1 − χt
≡ (ωf,t ωw,t ) = (II.43)
mrst
(1 − µb,t )−1 under EB
where nancial and wage mark-ups are respectively:
(1 − µb,t )−1 under RTM
ωf,t = (II.44)
Stc −(1−ρ)Et Λct+1|t St+1
c
(1 − µb,t )−1 (1 − α)−1 1 + under EB
Wt Ht
CHAPTER II. CREDIT CONSTRAINTS AND LABOR MARKET: THE ROLE OF WAGE
BARGAINING REGIMES 81
1+φ
under RTM
ωw,t = 1 − χt (II.45)
c c c −1
(1 − α) 1 + St −(1−ρ)Et Λt+1|t St+1
under EB
Wt Ht
In the e cient bargaining regime, the ine ciency gap comes from only nancial frictions measuring
ine ciency in credit markets, (1−µb,t ). Indeed, a nancial mark-up exists related to asymmetric information
on credit markets that increases the cost of a new hiring for rms. The higher is the shadow value of the
use of external nancing, the higher is the ine ciency gap under e cient bargaining.
In the right-to-manage bargaining regime, the ine ciency gap includes in addition to the nancial mark-up,
an endogenous wage mark-up regarding the ine ciency in labor and credit markets. This wage mark-up
depends on the eective bargaining power of workers, that is itself function of nancial frictions. So in a
right-to-manage bargaining framework, the rm puts up a wage mark-up depending on the level of borrowing
constraints they face. Financial frictions interact here clearly with labor markets frictions.
Under an e cient bargaining regime, rms are not able to modify the hours worked per employee as it
is determined through the bargaining with workers. Thus it is not possible for them to use this degree of
freedom to tackle the existence of nancial frictions.
Under a right-to-manage bargaining regime, rms are able to adjust unilaterally the number of hours worked
per worker (the intensive margin). They are able so to compensate the existence of nancial frictions by
using the intensive margin as a tool to alleviate extra-costs due to asymmetric information on credit markets.
Indeed, under RTM, a nancial mark-up still exists, but another mark-up appears that is the wage mark-up
linked itself to nancial frictions through the eective bargaining power of workers. Finally, hours worked
per employee under RTM are lower than under EB and standard equilibrium.
Figure 11 delivers a synthetic way to see how labor and nancial frictions do not have the same impact in
terms of labor market allocations depending on the bargaining regime that takes place:
Figure 11: Sources of ine ciencies depending on bargaining regimes
graph.png
In both bargaining regimes, a nancial mark-up exists due to asymmetric information between the lender
and the rm. Under a right-to-manage regime, an other mark-up exists, a wage mark-up due to the way
hours per worker are determined. It leads to lower hours worked per worker compared to e cient bargaining
or standard equilibrium conditions.
82 4. CONCLUSION
4 Conclusion
The impact of credit frictions on unemployment, wages and hours worked per employee is not the same
depending on the way hours and wages are bargained. With an EB regime, the wage splits the surplus
of a match on the labor market according to the rm's bargaining power that depends negatively on the
level of collateral constraints. So, credit frictions increase the bargaining power of workers: they extract a
higher rent from the bargaining relatively to a framework without credit frictions. With a RTM regime, the
impact of nancial frictions exists but it is mitigated by the fact that the rm is able to modify the level
of hours worked by each worker. Thus, a rm compensates the existence of nancial frictions by reducing
the level of hours demanded for each worker. So in both cases, a higher level of collateral constraints leads
to an increase in the worker's bargaining power. They ask for a bigger rent, but their capacities to extract
a bigger part of the surplus depend on the bargaining regime: a RTM regime appears to restore partly the
bargaining power of rms by giving them a higher degree of freedom in the bargaining process that is not
internalized by workers. I identify to that purpose two ine ciency gaps compared to a case without any
friction, a nancial and a wage ine ciency gap, the last one being present only under a RTM regime. Firms
use intensive margins to alleviate nancial frictions. This mechanism has been potentially being used in
particular countries, as Germany after the Great Recession by the use of part-time jobs.
Note that the model developed in this chapter is a partial equilibrium model. It may be relevant to develop
in future research a stochastic general equilibrium model in order to be able to implement a calibration, or
even an estimation of this enriched model to see for example if a credit shock has a dierentiated impact
on labor markets outcomes depending on the way hours and wages are bargained.
Appendices
83
85
A Borrowing constraint computation
Following Jermann and Quadrini (2012), the rm has the possibility to default after the production is
realized but before the intra-period loan is reimbursed. After the selling of its production, the rm's total
at+1
liabilities are the intra-period loans and the new debt acquired: lt + Rt .
So the rm holds liquidity lt in period t. I assume that the rm is able to divert the liquidity holdings,
lt but not the capital stock, Kt . So, the lender can only recover the capital stock with a probability ξt ,
which is stochastic. With a probability (1 − ξt ), the lender recovers nothing. This probability, ξt , can be
interpreted as resulting from the decision of regulators in terms of default regulation or as the probability
of nding a buyer for the capital stock of the rm or as an uncertainty on the liquidation value of the capital.
Let's now derive the collateral constraint. Recall the assumptions made in this chapter II:
Assumption 1. Following a default, production can not be seized by the lender.
Assumption 2. Following a default and before next period's capital is incorporated, the lender con scates
the rm and sells each unit of capital at ξt qK,t .
Assumption 3. Following a default, the lender has no bargaining power in the debt bargaining and they
do not value the stock of workers within the rm.
In case of default, the lender and the rm can decide to bargain a payment after the liquidation value
of the capital stock is realized.
First, if the liquidation value of the capital stock is zero, the lender decides to keep the rm as he is waiting
for the reimbursement of the intertemporal loan, at+1 . So the rm keeps the liquidity, lt . The total value
of default for the rm when the liquidation value is zero is:
lt + Et Λct+1|t ∆t+1 (46)
where Et Λct+1|t ∆t+1 is the expected present value of the rm if the rm continues to operate.
Now, if the liquidation value is ξt qK,t Kt , the rm bargains over a payment Pt to deter the lender to close
the rm. The net surplus to the rm of avoiding the liquidation is:
lt + Et Λct+1|t ∆t+1 − Pt (47)
And for the the lender, the net surplus from reaching an agreement with the rm is:
at+1
Pt + − ξt qK,t Kt (48)
Rt
86 B. BINDING BORROWING CONSTRAINT
If the rm has all the bargaining power, the threat value is obtained by the lender, either Pt = ξt qK,t Kt+1 −
at+1
Rt , so as to ensure that the rm will not be closed.
The expected total net surplus of reaching an agreement between the rm and the lender is therefore
equal to:
at+1
lt + Et Λct+1|t ∆t+1 − Pt = lt + Et Λct+1|t ∆t+1 + − ξt qK,t Kt (49)
Rt
Incentive compatibility requires that the expected surplus of defaulting for the rm does not exceed the
value of not defaulting, Et Λct+1|t ∆t+1 :
at+1
Et Λct+1|t ∆t+1 ≥ lt + Et Λct+1|t ∆t+1 + − ξt qK,t Kt (50)
Rt
at+1
lt + ≤ ξt qK,t Kt (51)
Rt
B Binding borrowing constraint
I follow Garin (2015) to show that the credit constraint is binding in the steady state.
The Euler equation for the household is:
0
1 v (Ch,t+1 )
= βh h 0 (52)
Rt vh (Ch,t )
The Euler equation of the rm is:
1 1
= Et Λct+1|t (53)
Rt 1 − µb,t
where µb,t is the Lagrange multiplier on the credit constraint. Combining these two equations at the steady
state, it follows that for all t:
βh − βc
µb = (54)
βh
Thus, as I assume that households are more patient than capitalists (βh > βc ), the borrowing constraint is
binding in the steady state, µb > 0.
87
C Proof
Financial mark-up under e cient bargaining
The standard hiring condition (II.23) can be expressed as:
Stc −(1−ρ)Et Λct+1|t St+1
c
Nt +Wt Ht Nt
(1 − µb,t ) = (1−α)Yt (55)
Stc −(1−ρ)Et Λct+1|t St+1
c
Nt Wt Ht
= (1−α)Yt Wt Ht + Wt Ht Nt
(1−α)Yt (56)
Stc −(1−ρ)Et Λct+1|t St+1
c
= Wt Ht Nt
(1−α)Yt 1+ Wt Ht (57)
Stc −(1−ρ)Et Λct+1|t St+1
c
= Wt
(1−α)mpht 1+ Wt Ht (58)
which gives:
Stc −(1−ρ)Et Λct+1|t St+1
c
1+ Wt Ht
EB
ωf,t = (59)
(1 − α)(1 − µb,t )
Financial mark-up under right-to-manage bargaining
In addition to the previous hiring condition, the e ciency condition at the intensive margin (II.25) is given
by:
Wt = (1 − µb,t )mpht (60)
Therefore, ωf,t
RT M
= (1−µb,t ) .
1
Wage mark-up under e cient bargaining
The bargaining between a rm and a worker over hours per worker yields the e ciency condition (II.36) for
the intensive margin:
(1 − µb,t )mpht = mrst (61)
It implies that the aggregate ine ciency gap, ωtEB , under e cient bargaining is equal to (1 − µb,t )−1 .
I know that the nancial mark-up, ωf,t
EB
, under e cient bargaining is equal to
c −(1−ρ)E Λc
St c
t t+1|t St+1
1+ Wt Ht
(1−α)(1−µb,t ) .
So, as ωtEB = ωf,t ωw,t ,
EB EB
!−1
Stc − (1 − ρ)Et Λct+1|t St+1
c
EB
ωw,t = (1 − α) 1 + (62)
W t Ht
88 C. PROOF
Wage mark-up under right-to-manage bargaining
Under RTM, equation (II.36) is not valid any more. I use equation (II.32) that linked marginal rate of
substitution to right-to-manage wages:
Wt = RT M
ωw,t mrst (63)
1 − χt
= mrst (64)
1+φ
It follows:
1+φ
RT M
ωw,t = (65)
1 − χt
Chapter III
Do Corporate Credit Conditions alter
Labor Market Dynamics? A SVAR
analysis in a Transatlantic Perspective
1 Introduction
The Great Recession is characterized by a sharp disruption in credit markets together with a deteriora-
tion in labor markets in a lot of industrialized countries. The relationship between credit market conditions
and labor market dynamics is usually seen as being positive. Better credit market conditions, as a higher
level of credit in one economy, is associated with better labor market outcomes as rms have a better access
to external nancing. They are less nancially constrained and are thus able to post more vacancies and
hire more workers. Theoretical papers (Garin (2015), Iliopulos et al. (2014), Monacelli et al. (2011), Perri
and Quadrini (2018), Petrosky-Nadeau (2014), Zanetti et al. (2015) or the chapter I of this dissertation) are
nding this positive causal relation for the United-States.
However, after the Great Recession, dierences on labor markets outcomes have been observed among coun-
tries (Sala et al. (2012)). Furthermore, the positive relation between credit market conditions and labor
market dynamics has not been empirically confronted to many countries other than the United-States1 .
Countries may dier deservedly in the way their respective labor markets react after a sharp fall in credit
conditions especially depending on their institutional and legal environment. To ll this gap, this chapter
studies labor markets dynamics in two dierent countries that are Germany and the United-States after
credit shocks by using structural VARs.
1 In their theoretical paper, Christiano et al. (2011) are using Swedish data and nd the same result as for the United-States.
89
90 1. INTRODUCTION
To motivate my empirical analyses about potential discrepancies between these two countries on their
labor markets dynamics after credit shocks, I consider gures 12 and 13. They depict rates of unemploy-
ment, vacancies and the growth rate of non- nancial corporations credit between 1952:Q2 and 2016:Q2 for
the United-States and 1991:Q1 and 2016:Q2 for Germany2 . For the United-States, a negative correlation
(−0.4) between unemployment and corporations credit is found, together with a positive correlation (0.42)
between vacancies and corporations credit. The higher is the credit for corporations in the United-States,
the lower is the unemployment rate and the higher are job vacancies (and vice versa as I observe here
correlations and not causality links). For Germany, relationships among variables are weaker and intriguing
for unemployment. A slightly positive correlation between unemployment and corporations credit (0.0071)
is observed, as well as for vacancies and corporations credit (0.1612).
Germany is a country worth to consider due to data availability and due to dierences in labor market
structures compared to the United-States especially concerning the governance of labor market institutions.
In Germany labor unions keep an active role in the bargaining process with rms compared to the United-
States. In Germany, in 2017 18% of workers were members of labor unions according to the OECD. In
the United-States, in 2017 only 10% of workers were members of unions according to the Bureau of Labor
Statistics. Furthermore, some empirical papers notice potential dierences between these countries with
regard to labor market and output dynamics to credit and technological shocks (Bachmann and Balleer
(2010), Belke and Osowski (2017) and Sala et al. (2012) among others). Bachmann and Balleer (2010)
compare the eects of technological shocks on labor market dynamics in Germany and in the United-States.
They nd signi cant dierences between these two countries concerning job nding and separation rates'
responses to technological shocks. Belke and Osowski (2017) nd a positive response of German output
to an increase in the Euro area uncertainty whereas for the United-States the response is negative. Sala
et al. (2012) investigate labor market outcomes dierences across various countries after the Great Reces-
sion. They show that Germany follows dierent dynamics compared to other countries and especially to the
United-States. From an empirical point of view, they show that unemployment in Germany tends to decline
since 2005 ( gure 13) with a relatively small increase during the Great Recession, at least compared to the
United-States. Unemployment is even lower than before the nancial crisis. The authors claim that these
countries may have been hit by dierent types of shocks and dierently by the so-called risk-premium shock.
These three empirical papers suggest that dierences seem to exist between Germany and the United-States
in the light of their economy' responses to credit or technological shocks. In this chapter, I try to pursue
the investigation by studying how their respective labor markets react to these shocks.
The following empirical investigation aims to determine whether non- nancial corporations credit con-
ditions may alter or not labor market dynamics and to what extent, in countries having dierent larbor
2 See appendix A for a detailed description of data.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 91
market institutional environments. Thus, this chapter asks the following questions: Are there dierences
in labor market dynamics to technological and credit shocks between Germany and the United-States (US)?
How dierences in labor market dynamics to credit shocks could be explained? As there is already some
literature about the impact of technological shocks on these economies, I use them as a natural benchmark
against which I assess the impact of credit shocks.
Figure 12: Unemployment, job vacancies and non- nancial corporations credit growth between 1952-Q1 and 2016-Q1 for the United-
States
2.5
1
3
3
Non−financial corporations credit
Non−financial corporations credit
2
2
2
.5
Unemployment
Vacancies
1
1
0
0
1.5
0
−1
−1
1
−.5
−2
−2
1952 1959 1966 1973 1980 1987 1994 2001 2008 2015 1952 1959 1966 1973 1980 1987 1994 2001 2008 2015
Quarters Quarters
Unemployment Non−financial corporations credit Vacancies Non−financial corporations credit
Figure 13: Unemployment, job vacancies and non- nancial corporations credit growth between 1991-Q1 and 2016-Q1 for Germany
3
1
4
4
Non−financial corporations credit
Non−financial corporations credit
2.5
.5
2
2
Unemployment
Vacancies
0
0
2
0
−2
−2
−.5
1.5
−4
−4
−6
−1
−6
1
1991 1995 1999 2003 2007 2011 2015 1991 1995 1999 2003 2007 2011 2015
Quarters Quarters
Unemployment Non−financial corporations credits Vacancies Non−financial corporations credit
Firstly, I nd that a positive one-standard deviation shock to technology in both countries leads to a
decrease in unemployment and an increase in vacancies. This result is consistent with theoretical models
integrating search and matching frictions as well as nancial frictions on labor and credit markets. Second,
a positive one-standard deviation shock to the level of corporations credits in the United-States leads in
the short-run to a statistically signi cant decrease in unemployment and an increase in vacancies, as well
as in output. It is consistent with ndings of theoretical papers (as the ones of Garin (2015), Iliopulos et al.
(2014), Monacelli et al. (2011), Perri and Quadrini (2018), Petrosky-Nadeau (2014), Zanetti et al. (2015)
and results of the chapter I of this dissertation). Better credit conditions allow corporations to take on
external debt more easily, either through a lower risk premium or through the relief of collateral constraints,
to nance the posting of vacancies and the hiring of new employees. That in turn induces an increase
92 1. INTRODUCTION
in production. Conversely in Germany, results are ambiguous: a one-standard deviation shock to level of
corporations credits leads in the short-run to a signi cant increase in unemployment and to insigni cant
impacts on vacancies and output. In this particular country, an increase in the level of non- nancial corpo-
rations credits does not necessarily lead to better conditions on labor markets.
To understand what may happen, I investigate two potential explanation: a 'Schumpeterian' eect or
a 'search for conciliation' eect. Through a Schumpeterian eect, German rms may substitute after a
positive credit shock labor for capital, hence explaining why unemployment increases after a positive credit
shock. This explanation, although attractive, is not con rmed by empirical results. Another explanation
has its roots in the particular functioning of the German labor market institutions. German rms have
a tradition of great negotiations with labor unions. When corporations credit is high, labor unions are
less demanding to German corporations. Laid-o workers could potentially more easily nd an other job
when credit is abundant in the economy. Conversely, when corporations credit is restricted in the economy,
labor unions know it will be di cult for a laid-o worker to nd a new job. Labor unions and corporations
reach agreements in order to keep jobs, as part-time jobs for example. A phenomenon of 'labor retention'
emerges. This retention of labor will then come to an end when the level of credit increases. The excess
payroll constituted during low credit levels periods will then be reduced, resulting in a counter-intuitive
result of rising unemployment during a period of high credit in Germany. I tend to validate this explanation
using data of rates of in ow and out ow of unemployment in Germany.
Finally, the so-called 'search for conciliation' explanation is reinforced by a result given by robustness
checks of the benchmark analysis. I nd non-linear impacts in the way the German labor market reacts to
credit shocks depending on whether the state of nature is normal or in crisis. In times of crisis, I observe
the 'search for conciliation' mechanism just exposed. In normal times, this mechanism no longer appears
and the German labor market returns to a functioning more consistent with theoretical models, where a
positive credit shock leads to a decrease in unemployment. Non-linear potential eects pave the way for
further theoretical and empirical researches.
Results are delivered by structural vector auto-regressive (SVAR) models for the United-States and Ger-
many. Restrictions are imposed on responses of variables on each others based on an underlying VAR model
with the objective to interpret as causality links the contemporaneous relationships among variables (Stock
and Watson (2001)). I identify technological and credit shocks by using short-term restrictions relying on
economic theory. I use a shared denominator from various theoretical frameworks (Garin (2015), Iliopulos
et al. (2014), Jermann and Quadrini (2012), Kiyotaki and Moore (1997), Monacelli et al. (2011), Perri and
Quadrini (2018), Petrosky-Nadeau (2014), Zanetti et al. (2015)) that assume that external nancing is in-
evitable prior to production for rms to pay in advance wages as well as vacancies posting costs, reminiscent
of the so-called 'wage-fund doctrine'. Hence the level of credits assigned in one economy is seen in those
models as critical ex-ante to enable rms' hiring and determine the level of employment and output in the
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 93
economy. I specify more precisely a positive credit shock as an increase in the level of credit in the economy,
that can be due to a decrease in the risk premium (making the credit less expensive for rms and thus
more available) or an increase in the fraction of collateral that rms are allowed to borrow (lenders are less
demanding). In both mechanisms used to model nancial frictions in the literature ( nancial accelerator
or collateral constraints), the same assumption is made about the need for rms to borrow before paying
employees or hiring new ones, and then producing. Assuming rms have to borrow before producing and
observing their productivity level enables to create asymmetric information between lenders and borrowers.
So I use this timing assumption to identify shocks in SVAR models and to confront the theoretical results
to empirical data by considering two countries that have dierent labor market institutional environments.
The remaining of this chapter is organized as follows. Section 2 presents data, empirical models and
identi cation choices. Section 3 displays estimated structural impulse responses and variances decomposi-
tion. Section 4 provides robustness analyses. Sections 5 and 6 explore the likely explanations of the results
previously obtained by considering succinctly a Schumpeterian eect and a so-called 'search for conciliation'
eect. Finally, section 7 concludes.
2 Empirical investigation
Firstly, I justify why I use of a short-term SVAR model after presenting the basis of the VAR analysis.
Second, I present the data and the SVAR models used. Third, I discuss my identi cation strategy based on
short-term restrictions.
2.1 VAR and SVAR methodology
Vector auto-regressive (VAR) models are a multi-variate way to model time series introduced by Sims
(1980). These models explain the evolution of each variable with its own past realizations and past re-
alizations of other variables included in the VAR. The general dynamic of the system is estimated. The
originality compared to standard regression models is to not distinguish between endogenous and exogenous
variables (even if it is also possible in VAR models to specify exogenous variables). A VAR model can be
used to describe and summarize time series, to compute forecasts and mainly to understand the eect of a
shock over time on variables, to understand how they interact and to understand the contribution of each
shock to the behavior of the dierent variables.
94 2. EMPIRICAL INVESTIGATION
A VAR(p) process can be de ned as:
Xt = δ + Σpi=1 Bi Xt−i + et (III.1)
0 0
where E(et ) = 0, E(et et ) = Ω and E(et es ) = 0 ∀t 6= s
where p is the lag length, Xt is a (n × 1) vector of stationary endogenous variables, δ is a (n × 1) vector of
constants, Bi is a (n × n) matrix of coe cients, et is the (n × 1) vector of exogenous shocks to each variable
and Ω the variance-covariance matrix of these error terms. The correlation of the error terms re ects the
contemporaneous relation between endogenous variables. Here, each variable in Xt is expressed as a linear
function of its own past realizations and past realizations of all other variables. The VAR process is thus
de ned in its reduced-form.
By de ning a VAR process through its reduced-form, the error terms (et ) can not be interpreted as struc-
tural shocks. Indeed, if the endogenous variables are contemporaneously correlated with each others, then
error terms will also be correlated. Reduced-form VAR models have parameters that are not explainable:
nothing can be said about the structure of the described economy. To make it possible, some economists
since Sims et al. (1986) are looking at orthogonal shocks with an economic sense. Shocks are identi ed
by setting reasonable identifying restrictions based on economic theory. These models are called structural
vector auto-regressive models (SVAR).
In a SVAR model, restrictions are set on the responses of variables on each others based on the un-
derlying VAR model. These restrictions are established according to conventional economic reasoning and
theory instead of using an automatic procedure lacking of economic sense. Structural parameters are hence
obtained out of the estimated reduced-form model by orthogonalizing the shocks in accordance with eco-
nomic reasoning. The objective is, of course, to be able to interpret impulse response functions and variance
decomposition in terms of causality and not only in terms of correlation. A precise number of restrictions
is required to make it possible as developed in section 2.3.
Various restrictions have been introduced in the literature: zero short-run restrictions as in Sims et al.
(1986), zero long-run restrictions (or Blanchard and Quah (1989) restrictions), both of them as in Gali
(1992) and sigh restrictions as in Uhlig (2005). The variety of possible restrictions implies that there are as
many structural models as restrictions exist. That is why the justi cation of restrictions used is particularly
important.
Long-term restrictions are set on the long-run accumulated eects of shocks. They have been introduced
by Blanchard and Quah (1989). They chose to put restrictions on the long-term dynamic of the system to
study the impacts of demand and supply shocks. They assumed that demand shocks have no in uence in
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 95
the long-run whereas supply shocks have.
Sign restrictions are set on the expected signs eects of shocks. An example of sign restrictions proposed by
Uhlig (2005) consists of a contractionary monetary policy shock that should raise the federal funds rate and
lower prices. The eect of this monetary policy shock on the output is contrariwise unrestricted in terms of
sign as it corresponds to the research question Uhlig (2005) wanted to answer.
Finally, short-term restrictions are set on the contemporaneous variance-covariances between shocks. I use
these restrictions in this chapter.
I de ne SVAR models with short-run restrictions to identify technological and credit shocks. The identi-
cation of both shocks is relying on the assumption made for the credit variable. Indeed, the identi cation
of credit shocks has been a lot discussed in the literature. I choose to develop structural VAR models with
short-term restrictions and credit and technological shocks for three reasons.
First, the use of a structural model appears to be fundamental as the traditional Cholesky identi cation
of nancial variables does not rely on economic theory3 . Financial variables are usually placed after slow-
moving macroeconomic variables only because they are supposed to move faster. However, there is a priori
no reason to consider that it is true for the level of credit in one economy. So I choose to rely on economic
theory (and to use structural VARs instead of recursive or unrestricted VAR) inferring a speci c order for
variables used as a lens to look at the data.
Then, in the literature, papers are considering either credit shocks as I do (Helbling et al. (2011)), credit
spread shocks (Gilchrist et al. (2009), Gilchrist and Zakraj²ek (2012), Meeks (2012)) or uncertainty shocks
(Belke and Osowski (2017), Bloom (2009), Bloom et al. (2012), Caggiano et al. (2014), Popescu and Smets
(2010) among others) to model ' nancial' shocks. I choose to consider the so-called 'credit' shock as it
allows an identi cation scheme compatible with a large range of theoretical papers. These papers model
nancial frictions in dierent ways ( nancial accelerator or collateral constraints) but they all assume that
external nancing is needed for rms prior to hiring and producing (Garin (2015), Iliopulos et al. (2014),
Jermann and Quadrini (2012), Kiyotaki and Moore (1997), Monacelli et al. (2011), Perri and Quadrini
(2018), Petrosky-Nadeau (2014), Zanetti et al. (2015)). So I de ne the credit shock as a shock to the level
of corporations credit in one economy without considering if the shock comes from a collateral constraint
channel or from a credit spread channel. My objective being to develop an identi cation strategy fully
supported by various theoretical frameworks, I use a type of shock on which I am able to apply the same
assumption, which is that rms need external nancing before hiring and producing.
Finally, using zero short-term restrictions appears to be the most appropriate and agnostic choice for
my empirical analyses. I do not want to impose sign-restrictions as it would imply to specify reactions
of macroeconomic variables to credit shocks. However, it is not possible ex-ante to infer particular signs
3 The Cholesky identi cation using the recursive ordering of variables has been however a lot used as in Bassett et al.
(2014) without any theoretical support. I propose to go a step further by looking at a structural model relying on theoretical
assumptions.
96 2. EMPIRICAL INVESTIGATION
of causality between unemployment, vacancies and non- nancial corporations credit as gure 13 suggests.
Moreover, it is precisely the explicit purpose of this chapter to determine these signs of causality between
those variables for the two countries that are Germany and the United States. It would therefore make no
sense to restrict the expected signs of eects of technological and credit shocks on unemployment, vacancies
and output. Lastly, I do not use long-term restrictions as I am not able to determine which shock has
more long- than short-term eects on variables included in the VAR. It would be possible to consider that
credit shocks have short-term eects compared to technological shocks with long-term eects. However, it
is known in particular since the Great Recession that less external nancing can have long-term impacts
through hysteresis eects (Ball (2014)) by reducing investment and hiring capacities for rms. Furthermore,
we are not able for now to determine if credit shocks are demand or supply shocks.
2.2 Data and SVAR de nition
I investigate impacts of technological and credit shocks on labor market variables in the United-States
and Germany via structural vector auto-regressive models (SVAR). The economic dynamic of a given country
is assumed to be represented by a dynamic system whose structural equations are:
A0 Xt = α + Σpi=1 Ai Xt−i + t (III.2)
0
where E(t ) = 0 and E(t t ) = Σ
where Xt is a (n × 1) vector of stationary endogenous variables, A0 is an invertible (n × n) matrix describing
contemporaneous relations among endogenous variables, α is a vector of constants, p is the lag length, Ai
is a (n × n) matrix of coe cients and t is the vector of exogenous uncorrelated or orthogonal structural
shocks to each variable. Σ is the matrix of variance-covariance of these structural shocks.
The vector of endogenous variables is de ned as:
∆ct
vt
Xt =
ut
∆yt
In the benchmark speci cation, following Ravn and Simonelli (2007) endogenous variables included in the
VAR for each country are the rst-dierence of the log of non- nancial corporations credit (∆ct ), the rst-
dierence of the log of real output (∆yt ), logs of unemployment rate (ut ) and job vacancy rate (vt ). I use
quarterly data from 1952:1 to 2016:2 for the United-States, and for Germany quarterly data from 1991:1
to 2016:2. The time period dierence is rst due to data availability and second to avoid the structural
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 97
break due to reuni cation in Germany4 . The non- nancial corporations credit data are obtained from the
BIS total credit statistics. The credit is provided by domestic banks, all other sectors of each economy and
non-residents. Non- nancial corporations are both private-owned and public-owned. In terms of nancial
instruments, credit covers loans and debt securities. Precise de nitions and sources of all variables are given
in appendix A.
As a preliminary check, I test if series are stationary. To estimate a VAR model properly, data need
to be stationary. I use for this purpose the augmented Dickey-Fuller (ADF) test, a unit-root test. Results
of these tests are in appendix C for variables expressed either in levels or in rst-dierence. I include a
number of lagged dierences following the Schwarz and Hannan-Quinn information criteria to eliminate
auto-correlation in the error terms of the Dickey-Fuller regression. When data show a clear trend, I use
the trend option. As for all data, their means appear to be greater than zero, I use the constant option.
Philips-Perron tests lead to the same conclusions as ADF tests. All series of variables used in the VAR are
shown to be stationary as tests do not detect the presence of unit roots5 .
Compared to an unrestricted VAR (equation (III.1)), an additional matrix A0 of parameters appears in
equation (III.2). This matrix multiplies the vector of contemporaneous endogenous variables, Xt , allowing
contemporaneous relationships between variables.
From the structural equation (III.2), I derive the following VAR model of order p:
Xt = δ + Σpi=1 Bi Xt−i + et (III.3)
0 0
where E(et ) = 0, E(et et ) = Ω and E(et es ) = 0 ∀t 6= s
and δ = A−1 −1 −1
0 α is a vector of constants, Bi = A0 Ai and et = A0 t .
So matrix A0 represents the contemporaneous relationships between reduced-form, et , and structural shocks,
t :
t = A0 e t (III.4)
Although estimating the equations of an unrestricted VAR does not imply identi cation assumptions, inter-
preting impulse response functions or variance decomposition requires identifying restrictions. Indeed, the
residuals from the reduced-form VAR (et ) have to be transformed into structural residuals (t ) such that
they can be interpreted. So I de ne now the short-term restrictions based on theoretical assumptions to be
able to interpret the empirical results obtained.
4 As
a robustness check, I restrict the time period analysis for the United-States from 1991:1 to 2016:2.
5 As
tests for unemployment and vacancies for Germany are less signi cant, I test the results in a robustness check for the
inclusion of unemployment and vacancies in rst-dierence.
98 2. EMPIRICAL INVESTIGATION
2.3 Identi cation of shocks
The conditional distribution of Xt is completely characterized by a reduced-form VAR (equation (III.1)).
However, the structural form of the model has n2 additional parameters coming from the matrix A0 com-
pared to the reduced-form. An identi cation problem appears. Indeed disturbances et from the reduced-form
VAR are typically correlated. They represent the unpredictable shocks in endogenous variables of Xt . They
could be correlated with one another because of contemporaneous causal relationships among the variables
included in the VAR estimation. It is as a consequence not possible to interpret et as structural shocks.
Given values of the reduced-form parameters δ, Bi , Ω, it is not possible to uniquely solve the system for the
structural parameters α, A0 , Ai , Σ. To go from the reduced-form to the structural one (and so to be able to
interpret the results), n2 identifying restrictions have to be imposed, that is 16 restrictions as n = 4 here.
Firstly, it is now common to consider Σ as being a diagonal matrix as structural shocks are by de nition
uncorrelated (if structural shocks were not independent, it would say that unexplained causal relationships
between them remain) and to normalize the main diagonal of A0 to 1 as it has been done by Sims (1980). It
ensures that each variable of Xt is allocated to its own structural equation and allows to give an economic
n(n−1)
interpretation to each shock. With this 10 restrictions of normalization, I have to determine 2 other
restrictions to identify technological and credit shocks, that is 6 restrictions. I choose to use short-term
restrictions relying on theoretical assumptions6 .
These restrictions consist of assumptions on the dynamic contemporaneous relationships between shocks
and variables and are thus imposed on matrix A0 . I interpret credit shocks as shocks to the level of non-
nancial corporations credit in one economy. To build the short-term restrictions, I rely on a fundamental
assumption of various theoretical papers: credit has to be delivered ex-ante to rms, before posting vacan-
cies, hiring and producing as they have to pay in advance vacancy posting costs and wages. Despite various
modeling assumptions, this assumption is common to a set of theoretical papers exhibiting asymmetric
information between lenders and borrowers (Garin (2015), Iliopulos et al. (2014), Jermann and Quadrini
(2012), Kiyotaki and Moore (1997), Monacelli et al. (2011), Perri and Quadrini (2018), Petrosky-Nadeau
(2014), Zanetti et al. (2015)). This assumption is mandatory to obtain credit frictions and nancial con-
straints. Indeed, nancial frictions appear between lenders and borrowers because a lender is not sure to
be reimbursed as the production and the sale occur only after credits are delivered and used. To protect
himself against a risk of default, the lender extracts either an external nance premium (Bernanke and
Gertler (1989), Petrosky-Nadeau (2014)) or asks for a collateral constraint (Garin (2015), Kiyotaki and
Moore (1997), Zanetti et al. (2015).). The higher are external nance premium or collateral constraints,
the higher are nancial frictions and the lower is the level of credits in the economy. As external nancing
6 In appendix B, I show how the identi cation is properly obtained thanks to zero short-term restrictions.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 99
is needed prior to production7 , a credit shock may have a direct and contemporaneous impact on the level
of employment, job vacancies and thus on real output8 . So an exogenous shock to the level of credit has an
immediate impact on labor market variables and output as the credit is used forthwith to nance wages and
vacancy posting costs. Thus, in the identi cation scheme I assume that credit is predetermined and does
not react to other shocks within the same quarter. It depends only on its own history and on lagged values
of other variables. Note that the previous cited papers are using a version of the 'wage-fund doctrine' to
build their theoretical models. The wage-fund doctrine implies that working capital have to be deducted
from a xed amount of funds available to employers at each period, here corresponding to the credit level.
The wage-fund corresponds to wages rms have to pay before the production is sold. I extend this idea by
including also the vacancy posting costs.
Then, to be consistent with this wage-fund doctrine, I assume that the output responds contemporaneously
to innovations to credit, unemployment and vacancies. So an exogenous shock to the technology impacts
other variables with one-period lag9 .
Finally, I assume that job vacancies react contemporaneously to innovations in credit only and unemploy-
ment responds contemporaneously to innovations in credit and vacancies shocks10 .
To conclude, I identify a shock on credit markets by imposing restrictions that appear to be reasonably
shared by various theoretical models11 . Caggiano et al. (2014) or Nodari (2014) for examples choose also
to put rst their ' nancial shock', that is in their paper an uncertainty shock (VIX index and FRPU shock
respectively) even if they do not use a theoretical support to justify it.
Thus, I propose as short-term restrictions consistent with the previous assumptions:
1. No restrictions are imposed on the eect of credit shocks on real output, vacancies and unemployment
to not exclude short-term eects from credit shocks on the real economy. All the other variables start
to respond in the rst quarter. So the non- nancial corporations credit variable is ordered as rst
variable.
Credit conditions evolve exogenously, technological and other shocks can not aect contemporaneously
the credit. Therefore, acy = acu = acv = 0, where aij are coe cients of matrix A0 .
2. Technological shocks aect other variables after one quarter. The growth rate is ordered as the last
variable. So auy = avy = 0.
7 In these models, rms inevitably resort to credit: they can not fully nance themselves. For this reason it is also technically
assumed either that entrepreneurs/ rms die at a constant rate at the end of each period or that entrepreneurs/ rms have a
lower discount factor than consumers ( rms are less impatient).
8 The assumption made in some theoretical papers to be sure that credit shock will impact the output immediately is that
new hired workers are productive within the same period as it is assumed in Blanchard and Galí (2010), Garin (2015), Iliopulos
et al. (2014), Zanetti et al. (2015)
9 In robustness checks, I test for dierent ordering of variables with notably credit ordered last.
10 In robustness checks, I test also for dierent order of variables.
11 In the spirit of Uhlig (2005), even if I am not working with sign restrictions, I try to be minimalist in the sense that
restrictions are based on a straightforward idea that before producing, rms need a given amount of external nancing to
nance this production.
100 3. RESULTS
3. The residual shocks to unemployment and vacancies are identi ed using a non-structural triangular
restriction that requires that avu = 012 .
With this identi cation scheme, I de ne for the benchmark models the matrix A0 as follow:
1 0 0 0
avc 1 0 0
A0 =
auc auv 1 0
ayc ayv ayu 1
where the remaining aij coe cients are the ones being estimated. I do not impose more restrictions that
required by the just identi cation necessity13 .
3 Results
SVAR benchmark models are estimated using an optimal lag order of p = 2 for both countries. Appendix
C delivers results of selection criteria ( nal prediction error (FPE), Akaike's information criterion (AIC),
Schwarz's Bayesian information criterion (SBIC), and the Hannan and Quinn information criterion(HQ))
that lead to choose this lag length14 . After each estimation, I test for the stability of the obtained SVAR
model by looking at moduli of the eigenvalues being or not less than one. All lie inside the unit circle, so
estimated models are stable.
I document various sets of results, impulse responses functions, cumulative impulse responses and the vari-
ance decomposition conditional on the technological and credit shocks for the United-States and Germany.
3.1 Credit shock
Structural impulse response functions (SIRF). Impulse responses illustrate dynamic responses over
time of all variables to a structural shock. Impulses are normalized to a one-standard deviation in the
underlying variable. Responses of endogenous variables to this deviation are presented in percentage points
(pct). Figure 14 presents the dynamic adjustment process of unemployment, vacancies, credit and real
output to a positive credit shock for respectively the United-States (US) and Germany.
Responses of US vacancies and employment are statistically signi cant and suggest an increase in the
number of vacant jobs and in the number of people employed after an increase in non- nancial corporations
credit. This result is consistent with theoretical papers of Garin (2015), Petrosky-Nadeau (2014) and Zanetti
12 An other way to obtain a just-identi ed system is to assume for residual shocks that auv = 0. Results based on this
alternative identi cation scheme are similar as shown in robustness analysis.
13 Sims (1980) considers in his famous critic that over-identi cations are entirely unrealistic.
14 For the United-States, a lag order of 6 could be also selected. In robustness analysis, I discuss the results obtained.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 101
US Germany
0
.03
Unemployment
.02
−.02
.01
−.04
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.02
.04
.01
0
.02
Vacancies
−.01
0 −.02
−.03
−.02 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .2
.2 .1
Output
.1 0
0 −.1
−.1 −.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
1.5
.6
1
Credit
.4
.5
.2
0
0
−.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 14: Structural impulse responses to credit shock. Benchmark model.
Notes: The gure reports structural impulse responses to an unanticipated credit shock. Columns report the SIRFs for the
United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90%
level. The horizontal axis identi es quarters.
102 3. RESULTS
et al. (2015). A shock increasing the level of credits15 positively aects the real economy as corporations are
able to use more credit to nance job vacancies and employment (the marginal value of an open vacancy
is higher). It leads to an increase in production and hence the real gross domestic product increases with
an immediate peak at nearly 0.2 pct that dies out very quickly. Garin (2015) and Zanetti et al. (2015) nd
also in the empirical section of their papers that a credit shock has no persistent dynamics on output in the
United-States as well as Gilchrist and Zakraj²ek (2012) and Meeks (2012) in their respective papers.
Unemployment rate decreases signi cantly and persistently16 until almost three years with a peak at 0.03
pct after one year and a half in response to a rise in the non- nancial corporations credit. It then adjusts
gradually back to their pre-shock level. I nd as in Caggiano et al. (2014) (who are using an uncertainty
shock) a hump-shaped path for unemployment after a credit shock in the United-States. The impact on
vacancies is shorter at two years reaching a peak at 0.025 pct after ve quarters.
So a positive credit shock in the United-States generates an economic upturn with a positive eect on out-
put statistically signi cant and short-lived, and positive eects on employment and vacancies statistically
signi cant and quite persistent.
For Germany, a positive non- nancial corporations credit shock leads to ambiguous responses of unem-
ployment and vacancies. The positive response of unemployment is signi cant for almost two years and
is quite persistent. A peak is reached at 0.015 pct after six months. The impact on vacancies is a lot of
more ambiguous as the one-standard error band includes zero, suggesting to not overstate the impact of
credit shock on vacancies in Germany. The impulse response for vacancies is within [−0.03; 0.015] interval
around zero at any point during the rst ve years after the credit shock. So the reaction of vacancies is
either positive or negative after such a shock. While it could be consistent with theoretical models results
of increasing vacancies after a positive credit shock, the data does not seem to give full weight to this result.
Finally, the impact of a positive credit shock on real output in Germany is also ambiguous. A one-standard
deviation credit shock in Germany may leave the output unchanged or may drive it up to 0.05 or down to
0.03.
These results raise questions: expansionary credit shocks do not necessarily have expansionary eects on
vacancies and output in Germany unlike the United-States. Firstly, I nd that a credit shock has partic-
ularly 'non-real' impacts in Germany compared to the United-States as impacts on vacancies and output
of a credit shock appear to be insigni cant. Second, an increase in the level of non- nancial corporations
credit leads to a slight increase of unemployment and not to a decrease as it could be expected according
to theoretical models (as the one developed in chapter I), and how as it is observed in the United-States.
The idea that something dierent is going on for Germany in face of credit shocks is actually not entirely
15 According to nancial frictions literature, corporations borrow more either because the risk premium is reduced (Petrosky-
Nadeau (2014)) or because they can borrow a larger fraction of their collateral (Garin (2015), Zanetti et al. (2015)).
16 I nd in the chapter I of this dissertation in section (4.2) a persistent evolution of unemployment following a shock on
credit markets in the United-States.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 103
new. Belke and Osowski (2017) nd a slight signi cant positive impact on German output of an increase in
the Euro area uncertainty and a signi cant negative impact on the US output. Sala et al. (2012) suggest
in their paper that Germany has been less directly hit by nancial shocks than the United-States. In next
sections, I investigate ways to explain these dierences.
Cumulative impulse responses. I plot the cumulative eects of a permanent shock to non- nancial
corporations credit17 .
For Germany, a permanent shock to credit has a signi cant positive impact on unemployment (0.3 pct)
and an ambiguous impact on vacancies and real output as expected. For the United-States, a permanent
positive credit shock of one-standard deviation triggers unemployment to decrease by 0.4 pct, vacancies to
increase by 0.2 pct and output to increase by 0.4 pct.
3.2 Technological shock
Structural impulse response functions. Figure 15 shows dynamic responses to technological shocks of
unemployment, vacancies, real output and credit for respectively the United-States and Germany.
For the United-States, a positive technological shock lowers unemployment for more than three years.
The eect is the largest after one year where the shock reduces unemployment by 0.023 pct. This shock
increases vacancies for two years and half with a peak at one year where the shock increases vacancies by
0.02 pct. Eects hereafter die out. Ravn and Simonelli (2007) nd the same impacts for the United-States
after a permanent technological shock that increases vacancies and employment with large and marked
hump-shaped responses. For the response of non- nancial corporations credit, technology shock triggers a
signi cant increase during two years consistent with the idea that credit is pro-cyclical in the United-States.
For Germany, results are similar even if standard error bands are larger. After a technological shock,
unemployment decreases for four years with a peak at two years and half by 0.015 pct. Nordmeier and
Weber (2013b), Popescu and Smets (2010) nd also for Germany that the unemployment rate goes down
after a positive technological shock. Vacancies increase for three years after the shock with a peak at one
year and half at 0.022 pct. Finally, the impact on non- nancial corporations credit is ambiguous as the
standard error band includes zero.
17 Figures are found in appendix D.
104 3. RESULTS
US Germany
0 0
−.01
Unemployment
−.01
−.02
−.02
−.03
−.03
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.03
.04
.02 .03
Vacancies
.02
.01
.01
−.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
.6
.6
.4
.4
Output
.2 .2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.15
.2
.1
.1
Credit
.05
0
0 −.1
−.2
−.05
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 15: Structural impulse responses to technological shock. Benchmark model.
Notes: The gure reports structural impulse responses to an unanticipated technological shock. Columns report the SIRFs
for the United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals
at 90% level. The horizontal axis identi es quarters.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 105
These results for both countries are consistent with standard theoretical models integrating search and
matching frictions as well as nancial frictions where working capital needs to be paid in advance of pro-
duction (as the ones of Garin (2015), Petrosky-Nadeau (2014) and Zanetti et al. (2015)). In those models,
a positive technological shock increases the expected surplus from a match such that rms decide to post
more vacancies and unemployment decreases as a consequence of the increase in labor market tightness from
the rm point of view.
Cumulative impulse responses. For Germany, a permanent positive shock of one-standard deviation
to real output causes unemployment to be about 0.6 pct lower and vacancies to be about 0.6 pct higher.
For the United-States, the unemployment rate is about 0.2 pct lower and vacancies 0.15 pct higher. This
magni cation comes from two eects. First, shocks to technology tend to persist for periods after the shock,
so unemployment decreases more and vacancies increase more as a result. Second, a positive shock to tech-
nology increases employment, which nally feeds back positively on real output.
3.3 Forecast errors variance decomposition
The variance decomposition of the forecast errors establishes the relevance of shocks for variations in
endogenous variables at a given horizon. Even if I nd insigni cant results for the impact of credit shocks on
labor markets variables for Germany, it does not mean that this shock could not be important for variables
dynamics (Helbling et al. (2011)). Table 2 highlights the contribution of technological and credit shocks in
explaining short-run uctuations in labor markets variables for Germany and the United-States. Note that
variance decomposition is certainly based on orthogonal shocks but I do not obtain 100% as it exists other
potential unidenti ed shocks completing the rest of the variance18 .
At very short forecast horizons, only small fractions of forecast error variance decomposition of unem-
ployment and vacancies are accounted for by credit and technological shocks. Beyond the six years horizon,
both shocks contribute relatively more to unemployment and vacancies volatility.
More precisely in the shortrun, the volatility of unemployment for both countries is more dominated by
credit shocks, especially for the United-States. At a six quarters horizon, credit shock accounts for 6% of
the forecast error variance of unemployment in Germany, whereas the technological shock account for 3%.
For the United-States, percentages are respectively 8% compared to 4.6%. Thus, it suggests that credit
shocks play a key role in employment uctuations, even more than technological shock.
Then, at a six quarters horizon, credit shock explains 0.4% of the volatility in vacancies in Germany, con-
versely to the United-States where it is equal to 7%. For the United-States, the credit shock plays a stronger
18 As monetary policy or government spending shocks that are found by Ravn and Simonelli (2007) to contribute to unem-
ployment and vacancies volatility in the United-States. However, it is not the purpose of this chapter to investigate the impact
of such shocks.
106 4. ROBUSTNESS ANALYSIS
role in explaining vacancies uctuations than technological shock. Credit shock explains 7% of the variation
in vacancies, whereas the technological shock counts for 3.5%. The inverse is observed for Germany: credit
shock does not matter a lot for vacancies uctuations.
Table 2: Forecast error variance decomposition of labor market variables - Germany and United-States technological and credit shocks
(percentage)
Germany Unemployment Vacancies
Forecast horizon Techn. shock Credit shock Techn. shock Credit shock
1 0 0.018 0 0
6 0.033 0.061 0.043 0.004
12 0.047 0.058 0.057 0.012
18 0.05 0.062 0.059 0.017
20 0.051 0.063 0.059 0.018
United-States Unemployment Vacancies
Forecast horizon Techn. shock Credit shock Techn. shock Credit shock
1 0 0.029 0 0.02
6 0.046 0.085 0.035 0.07
12 0.059 0.121 0.038 0.089
18 0.06 0.129 0.038 0.088
20 0.06 0.129 0.038 0.087
4 Robustness analysis
This section assesses the sensitivity of previous results along various dimensions to con rm that results
do not depend on a given speci cation. First, I proceed by modifying the lag length for the United-States (6
lags). Identifying assumptions are after inspected. Then, I address some data issues by considering unem-
ployment and vacancies in rst-dierence for Germany and non- nancial corporations credit as a percentage
of the GDP. I modify also the data time period notably by taking into account the Great Recession. This
point is of particular importance as this will infer on a potential explanation to the previous results. Fi-
nally, I introduce an extra-variable in the benchmark model. Robustness checks are available in appendix E.
Lag length. I re-estimate the benchmark model with a higher lag of p = 6 for the United-States as
suggested by three selection criteria (Appendix C). Indeed, using short lag length can induce spurious sig-
ni cance of parameters (unexplained information could be left in error terms). The results for a higher lag
length are quantitatively and qualitatively identical. Dierences concern impacts of the technological shock
on unemployment and vacancies that are larger and longer-lasting. The drop in unemployment reaches
−0.03 at 6 quarters (versus −0.023 at 4 quarters in the benchmark model) and vacancies increased by
0.023 to 6 quarters (compared with 0.02 to 4 quarters). Responses are however less signi cant: that is not
surprising as a higher number of parameters is estimated.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 107
Identifying assumptions. First, an other way to obtain a just-identi ed system is to assume for residual
shocks that auv = 0. Job vacancies react contemporaneously to innovations in credit and unemployment,
whereas unemployment responds contemporaneously to innovations in credit only. Results are consistent
for both countries, except for the impact of a technological shock on credit for the United-States where the
positive result is slightly signi cant.
Then, I set the non- nancial corporations credit variable at the end in the recursive order and real output
rst. Credit is not pre-determined anymore and all other shocks can impact this variable. On the other
side, credit shock and other shocks can not aect the level of the output within the same period. Results
are similar in terms of signs. I lose in terms of signi cance and magnitude for the impact of the credit shock
for both countries, but the impact of the technological shock is of greater signi cance and magnitude.
Data de nitions. The ADF test can reject the null hypothesis of non stationarity only at the 5% level
for the unemployment and vacancies in Germany. I check the results by including these variables in rst-
dierences. Results are the same in terms of signs. However, I lose the result signi cance also for unem-
ployment. It reinforces the idea that the impact of credit shocks on labor markets variables in Germany
is highly ambiguous. A contractionary credit shock does not have necessarily a negative impact on labor
markets variables.
Then, I use for the credit variable, the non- nancial corporations credit as a percentage of the GDP. This
measure facilitates data comparability. However, it necessarily modi es the identi cation scheme. Credit
expressed as a percentage of the GDP has to be ordered last and the output rst. Indeed, a shock to the
output has an immediate impact on the credit to GDP ratio by construction. Results are robust to this
change, except for the impact of credit shock on unemployment that is also not signi cant anymore for
Germany. Furthermore, impacts of technological shock in both countries are stronger on labor markets
variables. And I obtain a signi cant decrease of credit after a technological shock in both countries.
Extra-variable introduction. I test the robustness of my results to the introduction of an extra-variable,
consumption19 . Four additional identifying restrictions have to be set to identify the ve dimensional SVAR
models. I assume that consumption is made at the end of each quarter, after the production takes place,
that appears to be an uncontroversial assumption. So I set the following restrictions:
1. Consumption is aecting contemporaneously by all shocks.
2. Consumption impacts other variables with one period lag. acC = ayC = auC = avC = 0
19 Data de nition and source are found in appendix A
108 4. ROBUSTNESS ANALYSIS
The matrix A0 becomes:
1 0 0 0 0
avc 1 0 0 0
A0 =
auc auv 1 0 0
ayc ayv ayu 1 0
aCc aCv aCu aCy 1
SVAR models are estimated using an optimal lag order of p = 2 for both countries20 . Results are robust
to the introduction of consumption. I note that after technological shocks in both countries, consumption
increases (0.3 pct for the United-States and 0.22 pct for Germany). Ravn and Simonelli (2007) nd also
a positive impact on consumption of technological shocks in the United-States. After credit shocks, con-
sumption increases also in both countries (up to 0.15 pct in both countries). These empirical results are
in contradiction with some mechanisms of theoretical models with nancial and labor markets frictions.
Indeed, as in Garin (2015), after a positive credit shock consumption should decrease as a consequence of
the increase in interest rates, increasing the incentives to save. Finally, impacts of technological and credit
shocks on unemployment, vacancies and output are robust to the introduction of an extra-variable.
Sub-sample analysis. I restrict the data period for the United-States so as to use the same data period
as Germany (1991.1 to 2016.2). I estimate the SVAR model with an optimal lag order of p = 4 according
to selection criteria. The results obtained are consistent with the benchmark model where the date period
is 1952.2 to 2016.2. Responses of US vacancies and employment to credit shock are statistically signi cant.
The number of vacant jobs increases reaching a peak at 0.02 pct after 3 quarters (compared with 0.021
pct at 4 quarters in the benchmark model) and the number of people unemployed decreases with a peak at
−0.018 pct at 5 quarters (compared to −0.03 at 4 quarters in the benchmark model). I note that the impact
of the credit shock is shorter (5 quarters for the impact on unemployment and vacancies, compared to 12
and 8 quarters respectively in the benchmark model). Output increases after the shock with an immediate
peak at 0.15 pct that dies out very quickly as in the benchmark model. Finally, the positive technological
shock lowers unemployment with a peak at −0.03 pct at 8 quarters and increases vacancies reaching a peak
at 0.03 pct at 8 quarters. The major dierence here compared to the benchmark model is that the impact
of technological shocks on unemployment and vacancies is longer lasting (15 and 12 quarters for the impact
on unemployment and vacancies, compared to 14 and 10 quarters in the benchmark model).
20 For the United-States, a lag order of p = 4 could be also chosen. Results are not modi ed taking into account this lag
order.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 109
Do labor markets react in the same way in times of crisis and in normal times?
I extend the sub-sample analysis by considering a question of particular interest in Germany: Do labor
markets react in the same way in times of crisis and in normal times? It may be argued that credit shock
is fundamental in crisis periods but not or less otherwise (Caggiano et al. (2014)). Especially in Germany,
bargaining of wages and working times is done at branch level, and even at rms level. Thus, in a period
of crisis after a negative credit shock, rms and labor unions could bargain to put in place for example
measures of partial unemployment, to resort to internal exibility and limit the increase in unemployment.
In the United-States, external exibility is more widely used either in normal or in crisis periods as labor
unions are weaker in this country. For example, in Germany in 2017 18% of workers are members of labor
unions according to the OECD. In the United-States, in 2017 only 10% of workers were members of unions
according to the Bureau of Labor Statistics. Hence, in normal times in Germany, a credit shock may have
standard 'new-Keynesian' features whereas in crisis period, a credit shock may have dierent impacts on
labor markets variables. In the United-States, 'new-Keynesian' features would be observed in crisis and
normal times as labor unions are quite weaker in this country to play a key role in bargaining processes.
One drawback of VAR exercises is that they are based on a linear structure. In reduced-form VAR
models, coe cients as well as the variance-covariance matrix of error-terms are assumed to be constant over
time. It does not take into account the state of the nature ( nancial crisis period versus normal times). As
a very rst attempt to discuss this argument, I consider how results depend on crisis periods as opposed
to normal times. I investigate here the sub-sample stability of the preceding results21 . I restrict the data
period until the beginning of the crisis, at the end of 200722 . The aim is to avoid the introduction of speci c
eects due to the Great Recession and to see how much this speci c period played a key role.
Responses for the United-States do not change, except for the response of credit to a technological shock
that is not signi cant anymore. For Germany, responses of unemployment and vacancies to technological
shocks are less signi cant and shorter-lasting. Furthermore, responses of unemployment and vacancies to a
credit shock change notably. I observe a decrease of unemployment (at −0.01 pct with a peak at 7 quarters)
and an increase of vacancies (reaching a peak at 0.03 pct at 5 quarters). It illustrates the potentiality of
non-linearity. As a consequence, it appears that in crisis period German rms adopt a dierent manage-
ment of their payroll compared to normal periods. The dierent labor market institutional environment in
Germany compared to the United States could be an explanation, the so-called 'search for conciliation'. I
propose to investigate deeper this explanation in section (6).
21 For futureresearch, more sophisticated tools could be used as Markov-Switching VAR or Threshold VAR.
22 Accordingto NBER recessions date. The Great Recession begins according to OECD recessions date in February 2008 in
Germany. I decide to restrict the data for Germany also at the end of 2007.
110 4. ROBUSTNESS ANALYSIS
US Germany
.01
−.01 0
Unemployment
−.02
−.01
−.03
−.02
−.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.04 .06
.04
.02
Vacancies
.02
−.02 −.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .3
.2 .2
Output
.1 .1
0 0
−.1 −.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8 2
.6
1
Credit
.4
0
.2
0
−1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 16: Structural impulse responses to credit shock. Data period to 2007-Q4.
Notes: the gure reports structural impulse responses to an unanticipated credit shock. Columns report the SIRFs for the
United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90%
level. The horizontal axis identi es quarters.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 111
US Germany
.005
0
Unemployment
−.01
−.005
−.02
−.01
−.03 −.015
0 2 4 6 8 10 12 14 16 18 20 0 5 10 15 20
.03
.02
.02
.01
Vacancies
.01
−.01 −.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.6
.6
.4
.4
Output
.2
.2
−.2
0 5 10 15 20 0 2 4 6 8 10 12 14 16 18 20
.2
.1
0
.05
Credit
0 −.2
−.05 −.4
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 17: Structural impulse responses to technological shock. Data period until 2007-Q4.
Notes: the gure reports structural impulse responses to an unanticipated technological shock. Columns report the SIRFs for
the United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at
90% level. The horizontal axis identi es quarters.
5. ARE CREDIT SHOCKS FOR GERMANY GENERATING SCHUMPETERIAN CREATIVE
112 DESTRUCTION EFFECTS?
5 Are credit shocks for Germany generating Schumpeterian cre-
ative destruction eects?
In Germany, according to the benchmark model, credit shocks have no robust impact on vacancies and
output and have a positive impact on unemployment. Schumpeterian features of credit shocks in Germany
is one potential explanation for this result. German rms may substitute capital for labor. A positive credit
shock would therefore have a positive impact on capital investment, as well as on the level of unemployment.
Workers are replaced by capital.
To test this assumption, I add to the benchmark model the investment variable23 . I develop this new model
for Germany but also for the United-States to see if I get the result I expect for this country. For the United-
States, I expect that a positive credit shock increase capital investment. There would be no substitution
mechanism as strong as in Germany (factors of production would therefore be more complementary at the
aggregate level).
Four additional identifying restrictions have to be set to identify the ve dimensional SVAR model. Following
my previous assumptions, investment is partly nanced by external credit. I make the following restrictions:
1. Investment impacts contemporaneously all variables except credit, aci = 0.
2. Investment is aecting contemporaneously by only credit shock. Therefore, aiy = aiu = aiv = 0.
The matrix A0 becomes:
1 0 0 0 0
aic 1 0 0 0
A0 = avc avi 1 0 0
auc aui auv 1 0
ayc ayi ayv ayu 1
Figures 18 and 19 show the impulse responses of credit and technological shocks for SVAR models with
investment, of lags 2 for both countries according to the standard selection criteria24 .
For both countries, a positive technological shock impacts positively investment for 2 quarters with a
peak at 0.6 pct for Germany and for 6 quarters with a peak at 0.3 pct for the United-States. This result is
consistent with the ndings of Ravn and Simonelli (2007) for the United-States. Then, the eect dies out.
For the United-States, a positive credit shock impacts positively also investment for 3 quarters with a peak
at 0.22 pct. However I do not nd a signi cant impact of credit shock on investment for Germany. As a
23 See appendix A for data de nition and source.
24 I test the results for other identifying restrictions
(investment impacts contemporaneously output, and is aecting con-
temporaneously by credit, unemployment and vacancies shocks ; investment is aecting contemporaneously by all variables).
Results are not modi ed for all these orderings choices.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 113
US Germany
.4 .4
.2
Investment .2
−.2
−.4
−.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.03
−.02
Unemployment
.02
−.04
.01
−.06 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.02
.04
0
.02
Vacancies
−.02
0
−.02 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .2
.1
.2
Output
.1
−.1
−.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
1.5
.6
1
Credit
.4
.5
.2
0
0
−.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 18: Structural impulse responses to credit shock. Investment added.
Notes: The gure reports structural impulse responses to an unanticipated credit shock. Columns report the SIRFs for the
United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90%
level. The horizontal axis identi es quarters.
5. ARE CREDIT SHOCKS FOR GERMANY GENERATING SCHUMPETERIAN CREATIVE
114 DESTRUCTION EFFECTS?
US Germany
1
.4
.5
.2
Investment
0 0
−.2 −.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
0
0
−.01
Unemployment
−.01
−.02
−.02
−.03
−.04 −.03
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.04 .04
.02 .02
Vacancies
0 0
−.02 −.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.6 .6
.4
.4
Output
.2
.2
−.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.15 .4
.1 .2
Credit
.05 0
0 −.2
−.05 −.4
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 19: Structural impulse responses to technological shock. Investment added.
Notes: The gure reports structural impulse responses to an unanticipated technological shock. Columns report the SIRFs
for the United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals
at 90% level. The horizontal axis identi es quarters.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 115
consequence, I reject the assumption of Schumpeterian creative destruction eects of credit shock in Ger-
many. Credit shock in Germany have no signi cant impact on the real sphere according to these results.
As I do not nd Schumpeterian eects of credit shocks in Germany, I choose to investigate deeply the
unemployment dynamics aroused by a credit shock in Germany. Trough which channels an increase in credit
in Germany could lead to a potential increase in unemployment?
6 What drive the unemployment dynamics in the United-States
and in Germany?
To study the unemployment dynamics in the United-States and in Germany, I develop a new SVAR
model with data of unemployment in ow and out ow rates from Elsby et al. (2010)25 for the United-States
and from Hartung et al. (2016)26 for Germany. I use quarterly data from 1952:Q1 to 2013:Q2 for the
United-States and for Germany quarterly data from 1991:Q1 to 2016:Q4. I choose to use data from Elsby
et al. (2010) because these data are on quarterly frequencies (that is not the case in Elsby et al. (2013)
where data are found for Germany and the United-States but only on a annual basis).
The impacts on unemployment of credit shocks observed before may be the result of variation in the
rates at which workers ow into unemployment, variation in the rate at which unemployed workers exit
unemployment or a combination of the two. I use unemployment in ow and out ow rates because of the
following stock- ow relationship:
u̇t = st (1 − ut ) − ft ut (III.5)
where ut is the unemployment rate in period t, and ft and st are respectively the out ow and in ow rates
in period t. The out ow rate, f , corresponds to the number of hires divided by the number of unemployed
workers. The in ow rate, s, corresponds to the number of layos divided by the number of employed
workers. I do not take into account ows in and out of the labor force and provide according to Nordmeier
and Weber (2013a) the 'pure response of the unemployment rate that arises from the worker reallocation
process within the labor force'. Furthermore, I do not use the terms 'separation' and 'job nding' rates
because as well explained by Elsby et al. (2010), separations from employers do no necessarily lead to a
ow into unemployment, as unemployed workers may either nd a new job or ow to non participation
population.
25 They compute these hazard rates according to the method of Shimer (2007). The method is based on the use of time series
for the number employed, unemployment and unemployed less than ve weeks.
26 I thank Professor Jung for giving me access to these data.
6. WHAT DRIVE THE UNEMPLOYMENT DYNAMICS IN THE UNITED-STATES AND IN
116 GERMANY?
Accordingly, the steady state unemployment rate is given by:
s̄
ū = (III.6)
s̄ + f¯
As a consequence, the steady state unemployment rate will change if either f or s or both change.
I assume according to the chapter I of this dissertation that out ows from unemployment are realized
after credits are delivered. And in ows into unemployment are realized after the production takes place27 .
So I set the following short-term restrictions:
1. Out ows from unemployment (f ) impact all other variables except the credit. af y = af s = 0.
2. In ows into unemployment (s) impact other variables with one period lag.
The matrix A0 becomes:
1 0 0 0
af c 1 0 0
A0 =
ayc ayf 1 0
asc asf asy 1
SVAR models are estimated using an optimal lag order of p = 2 for both countries according to the standard
selection criteria.
According to gure 20, in the United-States, a technological shock has a positive impact on out ows
from unemployment, reaching a peak at 0.028 pct at one quarter. In ows into unemployment decrease
with an immediate drop to −0.024 pct. Thus, after a technological shock, unemployment decreases in the
United-States ( gure 15) because more unemployed nd a job and less employed workers loose their jobs.
And I found a positive impact of technological shocks on corporations credit as in section (3.2).
From gure 21, I observe a signi cant impact of credit shock on out ows from unemployment and a
barely non signi cant impact on in ows into unemployment. As a consequence, in the United-States, credit
shock impacts negatively the unemployment (as observed on gure 14) mainly through the out ows rate:
after a positive credit shock, more unemployed are nding a job leading to a decrease in unemployment. I
found also positive impacts of credit shocks on output.
27 I test the results with dierent ordering speci cations (namely by setting the in ow rate at the end and just after the
out ow rate). Results do not change.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 117
US Germany
.03
.015
.02
.01
Outflows
.01 .005
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.01
0
−.01
Inflows
−.01
−.02
−.02
−.03
−.03 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
.8
.6 .6
Output
.4 .4
.2 .2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2 .2
.15
0
Credit
.1
−.2
.05
0 −.4
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 20: Structural impulse responses to technological shock. Unemployment dynamics.
Notes: The gure reports structural impulse responses to an unanticipated technological shock. Columns report the SIRFs
for the United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals
at 90% level. The horizontal axis identi es quarters.
6. WHAT DRIVE THE UNEMPLOYMENT DYNAMICS IN THE UNITED-STATES AND IN
118 GERMANY?
US Germany
.02 .02
.01
.01
Outflows
−.01
−.01 −.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.01
.03
.005 .02
Inflows
.01
−.005
−.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .2
.2
.1
Output
.1
−.1
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8 2
.6 1.5
Credit
1
.4
.5
.2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 21: Structural impulse responses to credit shock. Unemployment dynamics.
Notes: The gure reports structural impulse responses to an unanticipated credit shock. Columns report the SIRFs for the
United-Sates (US) and Germany. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90%
level. The horizontal axis identi es quarters.
In Germany, a positive technological shock leads to an increase in the out ow rate and a decrease in
the in ow rate into unemployment. That is why I found previously on gure 15 that unemployment rate
goes down after a positive technological shock in Germany. The response of the out ow rate is signi cant
for 4 years, reaching a peak at 0.01 pct at 3 quarters. The response of the in ow rate is signi cant for 14
quarters, reaching a peak at −0.024 at 2 quarters.
CHAPTER III. DO CORPORATE CREDIT CONDITIONS ALTER LABOR MARKET DYNAMICS?
A SVAR ANALYSIS IN A TRANSATLANTIC PERSPECTIVE 119
Nordmeier and Weber (2013b) using long- and short-run restrictions, found in the same vein that job
nding rate increases after a positive technological shock, whereas the separation rate decreases (with an
insigni cant result for the separation rate). In our case, both margins appear to react to technological
shocks in Germany and not only the out ow rate.
Finally, in ow rate into unemployment in Germany goes up after a positive credit shock and out ow rate
goes down. As a result, as I observed on gure 14, unemployment increases after a positive credit shock in
Germany. Here, the response of out ow rate is borderline signi cant and should not be overstated. Hence, in
Germany, the increase in unemployment after a positive credit shock appears to be mainly due to an increase
in the in ow rate into unemployment. This result may illustrate a kind of 'search for conciliation' argument
playing for German rms. More credit in the economy could be seen as a sign of good health of the economy
(more projects are carrying on by rms as they have access to more external nancing for example). Labor
unions are considering that workers will nd easily new jobs in this context. So rms and labor unions allow
themselves to dismiss, leading nally at the aggregate level to an increase in unemployment. Conversely,
less credit in the economy is a bad news for the employment perspectives. Labor unions tend to bargain
with rms so as to make them keeping workers with part-time jobs for example. As a result, unemployment
goes down for a while.
German rms have a tradition of great bargaining with labor unions. When corporations credit is high,
labor unions are nally less demanding to German corporations. Laid-o workers could potentially more
easily nd an other job when credit is abundant in the economy. Conversely, when corporations credit is
restricted in the economy, labor unions know it will be di cult for a laid-o worker to nd a new job. A
phenomenon of 'labor retention' emerges. This retention of labor will then come to an end when the level
of credit increases. The excess payroll constituted during low credit levels periods will then be reduced,
resulting in a counter-intuitive result of rising unemployment during a period of high credit in Germany.
The 'search for conciliation' argument could explain why in period of crisis especially (as shown in section
3), the German labor market dynamics may be counter-intuitive. It would be linked to the labor market
institutional environment.
7 Conclusion
This chapter investigates the labor market dynamics in Germany and in the United-States after tech-
nological and credit shocks. Higher levels of non- nancial corporates credit have various impacts on labor
markets depending on countries. In the United-States, a positive credit shock has a positive impact on the
labor market, through a decrease of unemployment and an increase in vacancies. This result is consistent
with theoretical models developed in recent years. In Germany, a positive credit shock has a negative im-
pact on the labor market, through an increase of unemployment. The impacts on vacancies and output are
ambiguous: I nd no evidence of a robust eect of credit shock on these variables. So theoretical frameworks
120 7. CONCLUSION
modeling credit and labor markets frictions should not overestimate the role of credit shocks on labor mar-
kets dynamics, without taking into account the institutional speci city of each country as for the German
case. A common view widespread today is to consider that more credit in one economy will be the source of
better labor markets conditions. My empirical result suggests that this view can be challenged and discussed
as an increase in the level of credit in an economy is not necessarily the source of a better situation on the
employment front. In Germany, a 'search for conciliation' mechanism seems to appear. German rms tend
to adjust their wage bill when credit conditions are favorable, and thus they will separate from the excess
workers at that time. This argument is reinforced by the fact that I found non-linearity eects: the German
labor market reacts dierently in normal periods compared to periods of crisis. More investigation about
these non-linear eects would be interesting to be able then to improve theoretical models by integrating
regimes switching for example.
Appendices
121
123
A Data de nitions and sources
Table 3: De nitions and source of data - Germany
Variable De nition Source
Population Index of total population 2010:2=1 OECD
Seasonally adjusted (s.a)
Output First dierence of logarithm of real gross Eurostat
domestic product. Billions of chained 2010
euros divided by population, s.a
Unemployment Logarithm of unemployment rate, s.a OECD
Vacancies Logarithm of index of total un lled job OECD
vacancies divided by labor force, s.a
Price level Gross domestic product, Implicit price Eurostat
de ator 2010=100
Corporations credit First dierence of logarithm of total non- nancial BIS credit statistics
corporations credit divided by price level
and population, s.a
Consumption First dierence of logarithm of private nal OECD
consumption expenditure divided by price level
and population, s.a
Investment First dierence of logarithm of gross xed capital OECD
formation divided by price level and population, s.a
Out ow rate Number of hires on number of unemployed Hartung et al. (2016)
In ow rate Number of bene t claims on number of Hartung et al. (2016)
employed workers
124 B. IDENTIFICATION OF SHORT-TERM SVAR MODELS
Table 4: De nitions and source of data - United-States
Variable De nition Source
Population Index of civilian non institutional BLS
population, n.s.a
Output First dierence of the logarithm of real gross BEA (A191RX1)
domestic product. Billions of chained 2009
dollars divided by population, s.a
Unemployment Logarithm of civilian unemployment rate, s.a BLS (LNS14000000)
Vacancies Logarithm of composite help-wanted index Barnichon (2010)
Price level Gross domestic product, Implicit price BEA (A191RD3)
de ator 2009=100
Corporations credit First dierence of logarithm of total non- nancial BIS credit statistics
corporations credit divided by price level
and population, s.a
Consumption First dierence of logarithm of personal consumption BEA
divided by price level and population, s.a (PCEC, DPCERC1)
Investment First dierence of logarithm of xed non-residential BEA (A008RC)
private investment expenditure divided by price level
and population, s.a
Out ow rate Number of hires on number of unemployed. Elsby et al. (2010)
Computed according to the method of Shimer (2007)
In ow rate Number of layos on number of employed workers. Elsby et al. (2010)
Computed according to the method of Shimer (2007)
B Identi cation of short-term SVAR models
The structural VAR model is de ned as:
0
A0 Xt = α + Σpi=1 Ai Xt−i + t E(t ) = 0, E(t t ) = Σ (7)
Note that Σ is a diagonal matrix as structural shocks are by de nition uncorrelated (if structural shocks
were not independent, it would say that unexplained causal relationships between them remain).
The reduced-form SVAR model is obtained by multiplying the equation (7) by A−1
0 :
p
A−1 −1 −1 −1
0 A0 Xt = A0 α + Σi=1 A0 Ai Xt−i + A0 t (8)
which gives:
0 0
Xt = δ + Σpi=1 Bi Xt−i + et E(et ) = 0, E(et et ) = Ω, E(et es ) = 0 ∀t 6= s (9)
where δ = A−1 −1 −1
0 α, Bi = A0 Ai and et = A0 t .
If A−1 −1
0 is known, we are able to deduce δ , Bi and et . The problem it thus to determine how to obtain A0 .
The identi cation problem for n = 4 is presented as follow:
125
0
E(et et ) = A−1 −1 0
0 Σ(A0 ) = Ω
−1 −10
σ σ12 σ13 σ14 a a12 a13 a14 σ 0 0 0 a a12 a13 a14
11 11 11 11
σ12 σ22 σ23 σ24
a21 a22 a23 0
a24 σ22 0 0 a21 a22 a23 a24
=
σ13 σ23 σ33 σ34 a31 a32 a33 a34 0 0 σ33 0 a31 a32 a33 a34
σ14 σ24 σ34 σ44 a41 a42 a43 a44 0 0 0 σ44 a41 a42 a43 a44
There are 16 unknowns for the matrix A0 and 4 unknowns for the matrix Σ but only 10 equations as the
variance-covariance matrix, Ω, is symmetric. The system is not identi ed.
Based on structural assumptions developed in section (2.3), assume A0 is lower triangular such that:
−1
1 0 0 0
a21 1 0 0
−1
A0 =
a31 a32 1 0
a41 a42 a43 1
Now there are 10 equations with 10 unknowns as 10 elements of A0 are set to precise values, either 0 or 1.
The system is just identi ed.
126 C. SVAR MODELS SPECIFICATIONS
C SVAR models speci cations
Table 5: SVAR lag order selection by selection criteria for the United-States
Maximum lag length LR FPE AIC SBIC HQ
2 2 2 2 2 2
4 4 4 4 2 2
6 6 6 6 2 2
10 10 6 6 2 2
Notes: LR = Likelihood ratio test statistics, FPE = Final prediction error,
AIC = Akaike information criterion, SBIC = Schwarz bayesian information
criterion, HQ = Hannan-Quinn information criterion.
Table 6: SVAR lag order selection by selection criteria for Germany
Maximum lag length LR FPE AIC SBIC HQ
2 2 2 2 2 2
4 2 2 2 2 2
6 5 2 2 2 2
10 10 2 2 2 2
Notes: LR = Likelihood ratio test statistics, FPE = Final prediction error,
AIC = Akaike information criterion, SBIC = Schwarz bayesian information
criterion, HQ = Hannan-Quinn information criterion.
Table 7: Augmented Dickey-Fuller (ADF) tests for the United-Sates
Level First dierence
Model speci cation Test statistic Model speci cation Test statistic
Credit t, c, p= 3 -2.967 c, p = 2 -4.473***
Output t, c, p = 3 -2.218 c, p = 2 -8.563***
Unemployment c, p = 3 -3.893***
Vacancies c, p = 3 -4.038***
Consumption t, c, p = 3 -2.067 c, p = 2 -8.234***
Investment t, c, p = 3 -2.316 c, p = 2 -8.157***
Notes: The ADF regressions cover a number of lags (p) according to the Schwarz and Hannan-Quinn information criteria.
Regressions may include a trend (t) and/or a constant (c). ***, ** and * indicate signi cance at the 1%, 5% and 10%
levels.
127
Table 8: Augmented Dickey-Fuller (ADF) tests for Germany
Level First dierence
Model speci cation Test statistic Model speci cation Test statistic
Credit t, c, p = 2 -1.252 c, p = 2 -5.521***
Output t, c, p = 2 -2.807 c, p = 2 -4.704***
Unemployment c, p = 2 -3.211**
Vacancies t, c, p = 2 -3.393**
Consumption t, c, p = 2 -0.358 c, p = 2 -5.338***
Investment t, c, p = 3 -0.569 c, p = 2 -5.203***
Notes: The ADF regressions cover a number of lags (p) according to the Schwarz and Hannan-Quinn information criteria.
Regressions may include a trend (t) and/or a constant (c). ***, ** and * indicate signi cance at the 1%, 5% and 10%
levels.
128 D. CUMULATIVE IMPULSE RESPONSES
D Cumulative impulse responses
US Germany
0 .6
−.2 .4
Unemployment
−.4 .2
−.6 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2
.4
.3 0
Vacancies
.2
−.2
.1
−.4
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.4
.6
.2
.4
Output
.2
−.2
0
0 2 4 6 8 10 12 14 16 18 20 0 5 10 15 20
2
2.5
2
Credit
1.5
1.5
.5
0 1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 22: Cumulative impulse responses to credit shock
Notes: The gure reports cumulative impulse responses to an unanticipated permanent credit shock. The column report the
CIRFs for the United-Sates and Germany. The solid lines denote the median CIRFs. The dot lines identify con dence
intervals at 90% level. The horizontal axis identi es quarters.
129
US Germany
0 0
−.1
Unemployment
−.5
−.2
−.3 −1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2
Vacancies
.5
.1
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
1.5
.9
.8
1
Output
.7
.5
.6
.5
0
0 2 4 6 8 10 12 14 16 18 20 0 5 10 15 20
.8 1.5
.6 1
Credit
.4 .5
.2 0
0 −.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 23: Cumulative impulse responses to technological shock
Notes: The gure reports cumulative impulse responses to an unanticipated permanent technological shock. The column
report the CIRFs for the United-Sates and Germany. The solid lines denote the median CIRFs. The dot lines identify
con dence intervals at 90% level. The horizontal axis identi es quarters.
130 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
E Robustness analysis - Impulse responses
Figure 24: Structural impulse responses to credit and technological shocks for the United-States. Lag = 6.
Credit Shock Technological Shock
0
0
Unemployment
−.02
−.02
−.04 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.04 .04
.02 .02
Vacancies
0 0
−.02 −.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.6
.2
.4
.1
Output
.2
0
0
−.1 −.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2
.6
.4 .1
Credit
.2
0
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit and technological shocks. The column report
the IRFs for the United-Sates. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90%
level. The horizontal axis identi es quarters.
131
Figure 25: Structural impulse responses to credit shock. Unemployment ordered second.
US Germany
.03
0
Unemployment
.02
−.02
.01
−.04
−.06 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.02
.04
0
Vacancies
.02
−.02
0
−.02 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .2
.2 .1
Output
.1 0
0 −.1
−.1 −.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
1.5
.6
1
Credit
.4
.5
.2
0
0
−.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit shock. I rank unemployment second. The solid
lines denote the median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal axis identi es
quarters.
132 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
Figure 26: Structural impulse responses to technological shock. Unemployment ordered second.
US Germany
0
0
−.01
Unemployment
−.01
−.02
−.02
−.03
−.04 −.03
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.04
.03
.03
.02
Vacancies
.02
.01
.01
−.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
.6
.6
.4
.4
Output
.2 .2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.15
.2
.1
.1
Credit
.05
0
0 −.1
−.2
−.05
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated technological shock. I rank unemployment second.
The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal axis
identi es quarters.
133
Figure 27: Structural impulse responses to credit shock. Output ordered rst.
US Germany
0
.02
−.01
Unemployment
.01
−.02
−.03
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.03
.01
.02
0
Vacancies
.01
−.01
−.02
−.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.15 .2
.1 .1
Output
.05 0
0 −.1
−.05 −.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
1.5
.6
.4
Credit
.5
.2
0 −.5
0 5 10 15 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit shock. I order the output rst and the credit
last. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal axis
identi es quarters.
134 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
Figure 28: Structural impulse responses to technological shock. Output ordered rst.
US Germany
0
−.02
−.02
Unemployment
−.04
−.06 −.04
−.08
−.06
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.1
.09
.06
.05
Vacancies
.03
−.03 −.05
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
.8
.6
.6
Output
.4
.4
.2 .2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .3
.2
.2
.1
Credit
.1
0
−.1
0
−.2
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated technological shock. I order the output rst and the
credit last. The solid lines denote the median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal
axis identi es quarters.
135
Figure 29: Structural impulse responses to credit and technological shocks for Germany. Unemployment and vacancies expressed in
rst-dierence.
Credit Shock Technological Shock
0
.006
Unemployment
.004 −.002
.002
−.004
−.006
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.5
1.2
0
.8
Vacancies
.6
−.5 .4
.2
0
−1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2 .8
.6
.1
Output
.4
.2
−.1
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.1
1.5
0
Credit
−.1
.5
−.2
−.3
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit and technological shocks for Germany.
Vacancies and unemployment are expressed in rst-dierence. The solid lines denote the median SIRFs. The dot lines
identify con dence intervals at 90% level. The horizontal axis identi es quarters.
136 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
Figure 30: Structural impulse responses to credit shock. Credit to output ratio.
US Germany
0 .02
−.01
.01
Unemployment
−.02
−.03
−.04 −.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.02
.03
.01
.02
Vacancies
0
.01
0 −.01
−.01 −.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2 .2
.1
.1
Output
0
0
−.1
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
1.5
.6
.4
Credit
.5
.2
0 −.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit shock. Credit is expressed as the credit to
output ratio. I order output rst and credit last. The solid lines denote the median SIRFs. The dot lines identify con dence
intervals at 90% level. The horizontal axis identi es quarters.
137
Figure 31: Structural impulse responses to technological shock. Credit to output ratio.
US Germany
0
0
−.02
−.02
Unemployment
−.04
−.04
−.06
−.06
−.08
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.1 .1
.05 .05
Vacancies
0 0
−.05 −.05
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8 .8
.6 .6
Output
.4 .4
.2 .2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.05 .2
0
0
Credit
−.05
−.2
−.1
−.4
−.15
−.6
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit shock. Credit is expressed as the credit to
output ratio. I order output rst and credit last. The solid lines denote the median SIRFs. The dot lines identify con dence
intervals at 90% level. The horizontal axis identi es quarters.
138 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
Figure 32: Structural impulse responses to credit shock. Consumption added.
US Germany
0
.03
Unemployment
.02
−.02
.01
−.04
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.02
.04
.01
0
.02
Vacancies
−.01
0 −.02
−.03
−.02 −.04
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.3 .2
.2 .1
Output
.1 0
0 −.1
−.1 −.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
1.5
.6
1
Credit
.4
.5
.2
0
0
−.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2 .3
.15
.2
Consumption
.1
.1
.05
0
−.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports SIRF to unanticipated credit shock. I add consumption ordered last. The solid lines denote the
median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal axis identi es quarters.
139
Figure 33: Structural impulse responses to technological shock. Consumption added.
US Germany
0 0
−.01
Unemployment
−.01
−.02
−.02
−.03
−.03
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.03 .04
.03
.02
Vacancies
.02
.01
.01
−.01
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.8
.6
.6
.4
Output
.4
.2
.2
0
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.1
.2
.1
.05
Credit
0
−.1
−.2
−.05
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.4 .4
.3
.2
Consumption
.2
0
.1
0
−.2
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports SIRF to unanticipated technological shock. I add consumption ordered last. The solid lines denote
the median SIRFs. The dot lines identify con dence intervals at 90% level. The horizontal axis identi es quarters.
140 E. ROBUSTNESS ANALYSIS - IMPULSE RESPONSES
Figure 34: Structural impulse responses to credit and technological shocks for the United-States. 1991.1 to 2016.2 data period.
Credit Shock Technological Shock
.02
.02
0
Unemployment
−.02
−.02
−.04
−.04 −.06
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.04
.04
.02
.02
Vacancies
0
0
−.02
−.02
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.2
.4
.1
Output
.2
−.1 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
.6
.2
.4
.1
Credit
.2
0
0
−.2 −.1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Notes: The gure reports structural impulse responses to unanticipated credit and technological shocks for the United-States.
I set the data time period from 1991.1 to 2016.2. The solid lines denote the median SIRFs. The dot lines identify con dence
intervals at 90% level. The horizontal axis identi es quarters.
General conclusion
This dissertation addressed the interactions between labor and credit markets in an imperfect informa-
tion framework, taking into account the institutional environment surrounding the labor markets.
Chapter I investigated the way nancial frictions interact with frictional labor markets, through which
transmission channels. It showed that nancial frictions interact with labor markets through a marginal
costs channel. Asymmetric information implied a nancial mark-up charged by nancial intermediaries to
rms. It induces less posting vacancies by these rms, less employment in the economy and a higher level of
prices. A calibration exercise is carried out to investigate the impact of a net worth shock, a monitoring cost
shock and an idiosyncratic volatility shock on macroeconomic variables, such as vacancies, unemployment
rate and real wages. Dierent channels of propagation from the nancial sphere of the economy to the
labor market are investigated and appear to be consistent with the theoretical model. However, there are
some limitations to this empirical exercise. Each shock has not been clearly separately identi ed so as to
determine their speci c contributions to the real economy dynamics. Furthermore, a Bayesian estimation
could be an empirical work worth to consider so as to improve the empirical analysis and to see if the model
is able to replicate data behaviors in the United-States.
Chapter II compares two bargaining regimes between workers and rms, the e cient bargaining and
the right-to-manage bargaining in a labor search model with credit constraints on the credit market. It
showed that in an imperfect information framework, dierent wage bargaining regimes aect dierently
labor market outcomes. In an e cient bargaining regime, nancial frictions reduce the relative bargaining
power of the rm. An ine ciency gap appears, the so-called nancial mark-up. In a right-to-manage bar-
gaining regime, nancial frictions reduce the relative bargaining power of rms but to a lesser extent as the
rm has a higher degree of freedom in the bargaining process. Another source of ine ciency appears, the
so-called wage mark-up, that will be in uenced by nancial frictions levels. Depending on the bargaining
regime, rms are able to alleviate more or less easily the cost linked to the presence of nancial frictions.
A right-to-manage bargaining gives them a higher degree of freedom for their choices. However, the nal
allocation of resources in the economy is going even worse. This chapter is based on a partial equilibrium
model. It may be relevant to develop in future research a stochastic general equilibrium model in order to
be able to implement a calibration, or even an estimation of this enriched model.
Chapter III provided an empirical analysis taking into account two dierent countries, that are Germany
141
142 General conclusion
and the United-States. It studied potential dierences on the way their credit markets interact with labor
markets. It showed that higher levels of non- nancial corporations credit have various impacts on labor
markets depending on countries. In the United-States, a positive credit shock has a positive impact on the
labor market, through a decrease of unemployment and an increase in vacancies. In Germany, a positive
credit shock has a negative impact on the labor market, through an increase of unemployment. The impacts
on vacancies and output are ambiguous: no evidence of a robust eect of credit shock on these variables
is found. In Germany, a 'search for conciliation' mechanism seems to appear. German rms tend to keep
their labor force when credit conditions are not good, because of the role played by labor unions in this
country. This argument is reinforced by the fact that the German labor market reacts dierently in normal
periods compared to periods of crisis. The non-linear impact of credit shocks in Germany could be deeply
investigated with non-linear models in futures researches.
Through this dissertation, I studied in an imperfect information environment, labor and credit markets
interactions, depending on the institutional environment surrounding the labor market. Labor market in-
stitutions are often referred as a source of rigidity that impede the e cient adjustment of labor markets.
This dissertation could also bring to think that institutions in the credit market can be a source of potential
rigidity. Thus, it seems relevant to develop future theoretical models that allow to analyse market interac-
tions based on dierent institutional framework across markets, and not on a one-to-one market as I have
done in this dissertation. A more general comprehensive framework would be an interesting perspective to
be able to propose structural reforms which would not have unwanted perverse eects due to the particular
institutional characteristics on the dierent markets.
Bibliography
Amable, B., Demmou, L., and Gatti, D. (2007). Employment performance and institutions: New answers
to an old question.
Andolfatto, D. (1996). Business cycles and labor-market search. American Economic Review, 86(1):11232.
Bachmann, R. and Balleer, A. (2010). What drives labor market dynamics in the us and germany?
Ball, L. M. (2014). Long-term damage from the great recession in oecd countries. Technical report, National
Bureau of Economic Research.
Barnichon, R. (2010). Building a composite help-wanted index. Economics Letters, 109(3):175178.
Bassett, W., Chosak, M. B., Driscoll, J., and Zakraj²ek, E. (2014). Changes in bank lending standards and
the macroeconomy. Journal of Monetary Economics, 62:2340.
Belke, A. and Osowski, T. (2017). International eects of euro area versus us policy uncertainty: A favar
approach. Ruhr Economic Papers, 689.
Bernanke, B. and Gertler, M. (1989). Agency costs, net worth, and business uctuations. American
Economic Review, 79(1):1431.
Bernanke, B. and Gertler, M. (1995). Inside the black box: The credit channel of monetary policy. The
Journal of Economic Perspectives, 9(4):2748.
Bernanke, B., Gertler, M., and Gilchrist, S. (1999). The nancial accelerator in a quantitative business
cycle framework. Handbook of macroeconomics, 1:13411393.
Blanchard, O. and Diamond, P. (1989). The Beveridge Curve. Brookings Papers on Economic Activity,
20(1):176.
Blanchard, O. and Galí, J. (2010). Labor markets and monetary policy: A new keynesian model with
unemployment. American Economic Journal: Macroeconomics, 2(2):130.
Blanchard, O. and Quah, D. (1989). The dynamic eects of aggregate demand and supply disturbances.
American Economic Review, 79(4):655673.
Blanco, J. A. and Navarro, G. (2016). The unemployment accelerator.
Bloom, N. (2009). The impact of uncertainty shocks. Econometrica, 77(3):623685.
Bloom, N., Floetotto, M., Jaimovich, N., Saporta-Eksten, I., and Terry, S. (2012). Really uncertain business
cycles. Technical report, National Bureau of Economic Research.
Caggiano, G., Castelnuovo, E., and Groshenny, N. (2014). Uncertainty shocks and unemployment dynamics
in us recessions. Journal of Monetary Economics, 67:7892.
Calvo, G. A. (1983). Staggered prices in a utility-maximizing framework. Journal of Monetary Economics,
12(3):383398.
Campolmi, A. and Faia, E. (2011). Labor market institutions and in ation volatility in the euro area.
Journal of Economic Dynamics and Control, 35(5):793812.
Carlstrom, C. T. and Fuerst, T. S. (1997). Agency Costs, Net Worth, and Business Fluctuations: A
Computable General Equilibrium Analysis. American Economic Review, 87(5):893910.
143
144 BIBLIOGRAPHY
Carlstrom, C. T. and Fuerst, T. S. (2001). Monetary shocks, agency costs, and business cycles. Carnegie-
Rochester Conference Series on Public Policy, 54(1):127.
Checchi, D. and García-Peñalosa, C. (2008). Labour market institutions and income inequality. Economic
Policy, 23(56):602649.
Christiano, L. J., Eichenbaum, M., and Evans, C. L. (2005). Nominal rigidities and the dynamic eects of
a shock to monetary policy. Journal of Political Economy, 113(1):145.
Christiano, L. J., Eichenbaum, M. S., and Trabandt, M. (2016). Unemployment and business cycles.
Econometrica, 84(4):15231569.
Christiano, L. J., Trabandt, M., and Walentin, K. (2011). Introducing nancial frictions and unemployment
into a small open economy model. Journal of Economic Dynamics and Control, 35(12):19992041.
Christoel, K., Kuester, K., and Linzert, T. (2009). The role of labor markets for euro area monetary policy.
European Economic Review, 53(8):908936.
Eichenbaum, M., Jaimovich, N., and Rebelo, S. (2011). Reference prices, costs, and nominal rigidities.
American Economic Review, 101(1):23462.
Elsby, M. W., Hobijn, B., and ahin, A. (2010). The labor market in the great recession. Technical report,
National Bureau of Economic Research.
Elsby, M. W., Hobijn, B., and ahin, A. (2013). Unemployment dynamics in the oecd. Review of Economics
and Statistics, 95(2):530548.
Faccini, R., Millard, S., and Zanetti, F. (2011). Wage rigidities in an estimated dsge model of the uk labour
market.
Fiore, F. D. and Tristani, O. (2013). Optimal monetary policy in a model of the credit channel. The
Economic Journal, 123(571):906931.
Fu, F. (2009). Idiosyncratic risk and the cross-section of expected stock returns. Journal of nancial
Economics, 91(1):2437.
Gale, D. and Hellwig, M. (1985). Incentive-compatible debt contracts: The one-period problem. The Review
of Economic Studies, 52(4):647663.
Gali, J. (1992). How well does the is-lm model t postwar us data? The Quarterly Journal of Economics,
107(2):709738.
Gali, J., Gertler, M., and Lopez-Salido, J. D. (2007). Markups, gaps, and the welfare costs of business
uctuations. The review of economics and statistics, 89(1):4459.
Galí, J., Smets, F., and Wouters, R. (2012). Unemployment in an estimated new keynesian model. NBER
Macroeconomics Annual, 26(1):329360.
Garin, J. (2015). Borrowing constraints, collateral uctuations, and the labor market. Journal of Economic
Dynamics and Control, 57:112130.
Gertler, M., Kiyotaki, N., et al. (2010). Financial intermediation and credit policy in business cycle analysis.
Handbook of monetary economics, 3(3):547599.
Gertler, M., Sala, L., and Trigari, A. (2008). An estimated monetary DSGE model with unemployment and
staggered nominal wage bargaining. Journal of Money, Credit and Banking, 40(8):17131764.
Gertler, M. and Trigari, A. (2009). Unemployment uctuations with staggered nash wage bargaining.
Journal of political economy, 117(1):3886.
Gilchrist, S., Yankov, V., and Zakraj²ek, E. (2009). Credit market shocks and economic uctuations:
Evidence from corporate bond and stock markets. Journal of Monetary Economics, 56(4):471493.
Gilchrist, S. and Zakraj²ek, E. (2012). Credit spreads and business cycle uctuations. American Economic
Review, 102(4):16921720.
BIBLIOGRAPHY 145
Hartung, B., Jung, P., and Kuhn, M. (2016). Etiopathology of europe's sick man: Worker ows in germany,
1959-2016. Technical report, IZA Discussion Paper No. 10341.
Helbling, T., Huidrom, R., Kose, M. A., and Otrok, C. (2011). Do credit shocks matter? a global perspective.
European Economic Review, 55(3):340353.
Iliopulos, E., Langot, F., and Sopraseuth, T. (2014). Welfare cost of uctuations: when labor market search
interacts with nancial frictions.
Jermann, U. and Quadrini, V. (2012). Macroeconomic eects of nancial shocks. The American Economic
Review, 102(2):11861186.
Kiyotaki, N. and Moore, J. (1997). Credit Cycles. Journal of Political Economy, 105(2):21148.
Klenow, P. J. and Malin, B. A. (2010). Microeconomic evidence on price-setting. In Handbook of monetary
economics, volume 3, pages 231284.
Krause, M. U., Lopez-Salido, D., and Lubik, T. A. (2008). In ation dynamics with search frictions: A
structural econometric analysis. Journal of Monetary Economics, 55(5):892916.
Krause, M. U. and Lubik, T. A. (2007). The (ir)relevance of real wage rigidity in the new-Keynesian model
with search frictions. Journal of Monetary Economics, 54(3):706727.
Kydland, F. E. and Prescott, E. C. (1982). Time to build and aggregate uctuations. Econometrica,
50(6):134570.
Lechthaler, W., Merkl, C., and Snower, D. J. (2010). Monetary persistence and the labor market: A new
perspective. Journal of Economic Dynamics and Control, 34(5):968983.
Livdan, D., Sapriza, H., and Zhang, L. (2009). Financially constrained stock returns. The Journal of
Finance, 64(4):18271862.
Meeks, R. (2012). Do credit market shocks drive output uctuations? evidence from corporate spreads and
defaults. Journal of Economic Dynamics and Control, 36(4):568584.
Merz, M. (1995). Search in the labor market and the real business cycle. Journal of Monetary Economics,
36(2):269300.
Monacelli, T., Quadrini, V., and Trigari, A. (2011). Financial markets and unemployment. Technical report,
National Bureau of Economic Research.
Mortensen, D. and Nagypal, E. (2007). More on Unemployment and Vacancy Fluctuations. Review of
Economic Dynamics, 10(3):327347.
Mortensen, D. T. and Pissarides, C. A. (1994). Job creation and job destruction in the theory of unemploy-
ment. The Review of Economic Studies, 61(3):397415.
Nickell, S. and Layard, R. (1999). Labor market institutions and economic performance. Handbook of labor
economics, 3:30293084.
Nodari, G. (2014). Financial regulation policy uncertainty and credit spreads in the us. Journal of Macroe-
conomics, 41:122132.
Nordmeier, D. and Weber, E. (2013a). Conditional patterns of unemployment dynamics in germany.
Nordmeier, D. and Weber, E. (2013b). Patterns of unemployment dynamics in germany. Technical report,
IAB Discussion Paper.
North, D. C. (1994). Economic performance through time. The American economic review, 84(3):359368.
Paustian, M. (2004). Welfare Eects of Monetary Policy Rules in a Model with Nominal Rigidities and
Credit Market Frictions. Number 597.
Perri, F. and Quadrini, V. (2018). International recessions. American Economic Review, 108(4-5):93584.
Petrongolo, B. and Pissarides, C. A. (2001). Looking into the black box: A survey of the matching function.
Journal of Economic literature, pages 390431.
146 BIBLIOGRAPHY
Petrosky-Nadeau, N. (2014). Credit, vacancies and unemployment uctuations. Review of Economic Dy-
namics, 17(2):191205.
Popescu, A. and Smets, F. R. (2010). Uncertainty, risk-taking, and the business cycle in Germany. CESifo
Economic Studies, 56(4):596626.
Ravn, M. O. and Simonelli, S. (2007). Labor market dynamics and the business cycle: Structural evidence
for the united states. Scandinavian Journal of Economics, 109(4):743777.
Sala, L., Söderström, U., and Trigari, A. (2012). Structural and cyclical forces in the labor market during
the great recession: cross-country evidence. Technical report, National Bureau of Economic Research.
Shimer, R. (2004). The consequences of rigid wages in search models. Journal of the European Economic
Association, 2(2-3):469479.
Sims, C. A. (1980). Macroeconomics and reality. Econometrica: Journal of the Econometric Society, pages
148.
Sims, C. A. et al. (1986). Are forecasting models usable for policy analysis? Quarterly Review, pages 216.
Stiglitz, J. E. and Weiss, A. (1981). Credit rationing in markets with imperfect information. The American
economic review, pages 393410.
Stock, J. H. and Watson, M. W. (2001). Vector autoregressions. Journal of Economic Perspectives,
15(4):101115.
Thomas, C. and Zanetti, F. (2009). Labor market reform and price stability: An application to the euro
area. Journal of Monetary Economics, 56(6):885899.
Townsend, R. M. (1979). Optimal contracts and competitive markets with costly state veri cation. Journal
of Economic theory, 21(2):265293.
Trigari, A. (2006). The role of search frictions and bargaining for in ation dynamics. (304).
Trigari, A. (2009). Equilibrium unemployment, job ows, and in ation dynamics. Journal of Money, Credit
and Banking, 41(1):133.
Uhlig, H. (2005). What are the eects of monetary policy on output? results from an agnostic identi cation
procedure. Journal of Monetary Economics, 52(2):381419.
Walsh, C. E. (2005). Labor market search, sticky prices, and interest rate policies. Review of economic
Dynamics, 8(4):829849.
Yun, T. (1996). Nominal price rigidity, money supply endogeneity, and business cycles. Journal of Monetary
Economics, 37(2-3):345370.
Zanetti, F. et al. (2015). Financial shocks and labor market uctuations. University of Oxford Economics
Working Papers, 746.
Zanetti, F. and Mumtaz, H. (2011). The eect of labor and nancial frictions on aggregate uctuations.
Macroeconomic Dynamics, pages 129.
Title: Financial frictions and labor market
Abstract: Unemployment rates in developed economies are now signi cantly dierent. Labor
market institutions are also diverse and multifaceted. These institutions could be considered as
allowing to increase or to reduce the levels of unemployment. However empirically, there is no
direct and unambiguous link between unemployment rates and institutions in the labor market.
By considering more precisely the way in which rms decide on their payroll, we realize that we
omit, by making this simple correlation link, an essential variable that determines the hiring and
ring behavior of rms, namely the funding variable. The external nancing capacity of rms
may determine the labor demand, conditional on the institutions in the labor market. Thus,
the problem is not whether institutions in the labor market condition its relative performance
but rather whether the couple of institutions in labor and credit markets determines this
performance. A rm is certainly constrained by a greater or lesser exibility in the labor market,
but its computations are part of a broader perspective, which is whether or not it has access
to the funding it needs. The importance of nancial frictions in the credit market determines
the level of the external nancing constraint for rms. This could then have an impact on their
hiring plans and job levels in economies depending on the prevailing labor market institutions.
Financial frictions should therefore in uence the main labor market macroeconomic variables,
namely unemployment, wage level and the number of vacancies, conditional on existing labor
market institutions.
Keywords: labor market, credit market, imperfection information, institutions, frictions,
unemployment, wages.
Titre : Frictions nancières et marché du travail
Résumé : Les niveaux des taux de chômage des économies développés sont aujourd'hui sensiblement
diérents. Les institutions du marché du travail sont elles aussi diverses et protéiformes selon les pays.
Ces institutions pourraient être considérées comme permettant d'accroître ou de réduire les niveaux de
chômage. Or empiriquement, on ne trouve pas de lien direct et univoque entre les taux de chômage et
les institutions sur le marché du travail. Si nous considérons plus précisément la façon dont les rmes
décident de leur masse salariale, on réalise que l'on omet en faisant ce simple lien de corrélation, une
variable essentielle qui détermine les comportements d'embauche et de licenciement des entreprises, à
savoir la variable du nancement. La capacité de nancement externe des rmes pourrait déterminer,
ou non, la demande de travail, conditionnellement aux institutions sur le marché du travail. Ainsi, le
problème ne serait pas de savoir si les institutions sur le marché du travail conditionnent sa performance
relative mais plutôt de savoir si le couple d'institutions sur le marché du travail et le marché du crédit
détermine ces performances. Une entreprise est certes contrainte par la plus ou moindre grande exibilité
existante sur le marché du travail, mais ses calculs s'inscrivent dans une perspective plus large, qui est de
savoir si elle a accès ou non aux nancements dont elle a besoin. L'importance des frictions nancières
sur le marché du crédit détermine le niveau de la contrainte de nancement externe pour les rmes.
Cela pourrait alors avoir un impact sur leurs projets d'embauche et sur les niveaux d'emplois dans
les économies. Les niveaux de frictions nancières devraient donc in uencer le niveau des principales
variables macroéconomiques relatives au marché du travail, que sont le chômage, le niveau du salaire et
le nombre de postes vacants, conditionnellement aux institutions existantes sur le marché du travail.
Mots clés : marché du travail, marché du crédit, information imparfaite, institutions, frictions,
chômage, salaires.
Vous aimerez peut-être aussi
- Strikes Labor Conflicts and Firms Performances inPas encore d'évaluationStrikes Labor Conflicts and Firms Performances in217 pages
- Croissance Économique et Développement FinancierPas encore d'évaluationCroissance Économique et Développement Financier436 pages
- Ciblage d'Inflation: Efficacité et PerformancePas encore d'évaluationCiblage d'Inflation: Efficacité et Performance378 pages
- Tâche 4 - Le Marché - Offre Et DemandePas encore d'évaluationTâche 4 - Le Marché - Offre Et Demande10 pages
- Economie Générale ECS 1 INPHB (Enregistré Automatiquement)100% (2)Economie Générale ECS 1 INPHB (Enregistré Automatiquement)262 pages
- DAVOINE (2007) - La Qualité de L'emploiPas encore d'évaluationDAVOINE (2007) - La Qualité de L'emploi583 pages
- Cours d'Économie Générale - ECES 1Pas encore d'évaluationCours d'Économie Générale - ECES 1292 pages
- Localisation économique et dépendance géographiquePas encore d'évaluationLocalisation économique et dépendance géographique159 pages
- Régimes et Changements Structurels en ÉconomiePas encore d'évaluationRégimes et Changements Structurels en Économie402 pages
- ECO 202 Cours de Macroéconomie À L'universitéPas encore d'évaluationECO 202 Cours de Macroéconomie À L'université113 pages
- Essais sur la politique monétaire DSGEPas encore d'évaluationEssais sur la politique monétaire DSGE183 pages
- Méthodes et modèles en macroéconomiePas encore d'évaluationMéthodes et modèles en macroéconomie11 pages
- Problèmes économiques contemporainsPas encore d'évaluationProblèmes économiques contemporains85 pages
- Modèle macroéconomique : fondements théoriquesPas encore d'évaluationModèle macroéconomique : fondements théoriques20 pages
- Marchés numériques du travail et TICPas encore d'évaluationMarchés numériques du travail et TIC385 pages
- Normes Fondamentales Du Travail Et Developpement ePas encore d'évaluationNormes Fondamentales Du Travail Et Developpement e234 pages
- Inanciarisation Répartition Des Revenus Et Croissance en RancePas encore d'évaluationInanciarisation Répartition Des Revenus Et Croissance en Rance389 pages
- Nouvelle représentation du marché du travailPas encore d'évaluationNouvelle représentation du marché du travail32 pages
- Relation Économiques InternationalesPas encore d'évaluationRelation Économiques Internationales9 pages
- Modélisation mathématique pour le market making et la liquidité financièrePas encore d'évaluationModélisation mathématique pour le market making et la liquidité financière245 pages
- Économie de marché : principes et enjeuxPas encore d'évaluationÉconomie de marché : principes et enjeux30 pages
- Microéconomie et Applications 2023Pas encore d'évaluationMicroéconomie et Applications 2023317 pages
- Concepts clés de macroéconomie avancéePas encore d'évaluationConcepts clés de macroéconomie avancée29 pages
- Principes de la macroéconomie néoclassiquePas encore d'évaluationPrincipes de la macroéconomie néoclassique39 pages
- Mémoire de Master Recherche de TDNPas encore d'évaluationMémoire de Master Recherche de TDN170 pages
- Variabilité Taux de Change MadagascarPas encore d'évaluationVariabilité Taux de Change Madagascar368 pages
- Concepts & Théories Économiques - Desjardins100% (1)Concepts & Théories Économiques - Desjardins220 pages
- Programme d'Économie en CPGE TechnologiquePas encore d'évaluationProgramme d'Économie en CPGE Technologique7 pages
- Économie CPGE : Concepts et ModèlesPas encore d'évaluationÉconomie CPGE : Concepts et Modèles8 pages
- Financement des entreprises innovantesPas encore d'évaluationFinancement des entreprises innovantes217 pages
- Prise en Compte Des PCD Dans Les BIP Des Communes Avril 2025Pas encore d'évaluationPrise en Compte Des PCD Dans Les BIP Des Communes Avril 202516 pages
- Dépenses Publiques en Infrastructures au MaliPas encore d'évaluationDépenses Publiques en Infrastructures au Mali78 pages
- Comprendre les Styles d'ApprentissagePas encore d'évaluationComprendre les Styles d'Apprentissage3 pages
- Évaluation pédagogique : types et fonctionsPas encore d'évaluationÉvaluation pédagogique : types et fonctions2 pages
- Objectif Mémoire Au Lycée Et À Luniversité - Re - Trouvez Le Goût de Travailler Avec Plaisir Et Effica PDF100% (3)Objectif Mémoire Au Lycée Et À Luniversité - Re - Trouvez Le Goût de Travailler Avec Plaisir Et Effica PDF184 pages
- Vas Interface Edition Documents SigfnePas encore d'évaluationVas Interface Edition Documents Sigfne1 page
- Règlement et Comportement au CollègePas encore d'évaluationRèglement et Comportement au Collège29 pages
- Cours de Seminaire de Recherche ScientiquePas encore d'évaluationCours de Seminaire de Recherche Scientique30 pages
- Intercompréhension en langues romanesPas encore d'évaluationIntercompréhension en langues romanes14 pages
- Formulaire D'évaluation D'entretienPas encore d'évaluationFormulaire D'évaluation D'entretien2 pages
- Fiche de Renseignement de L'école Elhadj IbrahimPas encore d'évaluationFiche de Renseignement de L'école Elhadj Ibrahim2 pages
- Appel À L'école de L'habiter Montréal 2026 251126 164209Pas encore d'évaluationAppel À L'école de L'habiter Montréal 2026 251126 16420910 pages
- تقرير الأعمال يناير 25م فارغ تربوي FrançaisPas encore d'évaluationتقرير الأعمال يناير 25م فارغ تربوي Français3 pages
- Recus de Préinscription en Licence D'excelence: Informations PersonnellesPas encore d'évaluationRecus de Préinscription en Licence D'excelence: Informations Personnelles1 page
- 2 Programmes Éducatifs Primaire CP2 Septembre 2025 Unique v1Pas encore d'évaluation2 Programmes Éducatifs Primaire CP2 Septembre 2025 Unique v1313 pages
- 2022-2!22!972 - Le Quotidien D'oranPas encore d'évaluation2022-2!22!972 - Le Quotidien D'oran24 pages
- 002 Communiqué Portant Ouverture de Deux Concours Directs - 20240717 - 0001Pas encore d'évaluation002 Communiqué Portant Ouverture de Deux Concours Directs - 20240717 - 00012 pages
