To study the effect of the 2009 hiring credit on employment in eligible firms, I use a balanced monthly panel of firms between 2007 and 2012. It includes all firms that had between 6 and 14 FTE employees over the first eleven months of 2008, month by month from January 2007 to December 2012. The idea is to compare the evolution of the dependent variables in eligible and ineligible firms relatively to the timing of the implementation of the hiring credit. This approach is similar to that of Cahuc et al. (2019) but the main differences are that the definition of the eligibility andineligibility groups is fixed over time, which allows me to measure the effect of the hiring credit over a longer period of time, and that I try to have a precision at an infra-annual level (month by month). This is the only way to understand whether the effect is due to a reduction of labor cost, a displacement effect or an hiring subsidy. This has also the advantage to guard against the possible reclassification bias in Cahuc et al. (2019). Eligible firms in 2010, 2011 and 2012, control periods, are negatively selected compared to eligible firms in 2009, the treatment period. In a double-difference framework, it may lead to overestimate the impact of the hiring credit.
The identification hypothesis of my approach is that the evolution of employment in ineligible firms in 2008 is a good approximation of what would have happened in eligible firms in 2008 if the credit had not existed (a good counterfactual). This is known as the parallel trend hypothesis. In theory, this hypothesis is untestable, but it is often customary to check that the evolution of the dependent variable in the treatment and control groups is similar before treatment. Figure 1.6 shows that this is not the case for employment as both groups of firms seem to diverge before even the implementation of the credit.
Furthermore, as noted in Cahuc et al. (2019), there is no theoretical reason why employment patterns would remain the same in the absence of hiring credit for firms of different size. Suppose a production function F(L) with labor cost w. The optimal level of employment is such that F0(L) = w. In this case, dL = L: where dL is the change in employment following a 1% decrease in labor cost and epsilon is the elasticity of labor demand. This means that even in the absence of a hiring credit, a common shock also affecting the labor costs of the two groups of firms will have a different effect on employment. For the same reason, looking at the infra-annual evolution of employment can be risky if seasonal cycles affect groups of firms of different sizes differently.
To get to the heart of the matter, I follow the suggestions of Bilinski et Hatfield (2018). That is, I compare the double difference estimator with that of similar models where the existence of different trends between the treatment and control groups is possible and where, for example, seasonal differences are allowed. I estimate three different models with increasing complexity where Yit is the dependent variable for firm i in month t, knowing that t = 0 corresponds to November 2008, i.e. the month preceding the application of the 2009 hiring credit. Ti is a binary variable equal to 1 if firm i was eligible for the credit at the end of November 2008. t and i are temporal and individual fixed effects.
Yi;t = + :[Ti 1ft 2 [1; 16]g] + t + i + ei;t (0.1).
Yi;t = + :[Ti 1ft 2 [1; 16]g] + :[t Ti] + t + i + ei;t (0.2).
Institutional context before 2008
Before 2008, indefinite term labor contracts can be terminated in France either on employers’ initiative (dismissals) or on employees’ initiative (quits), there is no third option. Furthermore, the costs of the procedure are mainly borne by the contracting party who initiate the procedure. Employees who choose to quit lose their eligibility to receive a severance package as well as their eligibility to receive unemployment benefits 5. Employers who decide to dismiss employees have to justify their decisions and run the risk of being sued for unfair dismissal. This risk is often painted as one reason for the sclerosis of the French labor market.
Dismissals can be justified by economic reasons. In such a case, the employer has to prove the seriousness of its economic problems and has to pay severance payments. In case of collective dismissals for economic reasons, the employer has also to justify the choice of who is dismissed and who is not. French labor laws ask employers to dismiss lower seniority workers first, as well as workers with lower family responsibility (see article 1233-5 of French labor laws).
Dismissals can also be justified by non-economic reasons (dismissals for cause), most notably when employers consider that employees are guilty of misconduct. There are three levels of misconduct, namely simple, serious or very serious misconduct 6. Employers have to pay severance payments, except in case of serious or very serious misconduct (article L.1234-9 of French Labor law). The vast majority of litigations follow dismissals for cause 7. Between 1996 and 2003, about 25% of these non-economic terminations have been challenged in French courts (Fraisse et al. (2015)).
The 2008 reform
In June 2008, the French government introduced a third type of labor contract termination, called rupture conventionnelle (hereafter, termination by agreement). When an employer and an employee opt for such a termination, the liability is shared and the consent is mutual.
This was the result of years of negotiation between trade-union and employer representatives. Employer representatives were arguing for the introduction of mutual consent termination since the beginning of the 2000’s and their key demand was that this type of termination would be litigation risk free. Trade-union representatives were responsive but their key demand was that this new termination would give workers eligibility to the unemployment insurance unlike quits. After a call from the government to social partners in 2007 to reflect on how to modernize the French labor market, employer and trade unions representatives agreed on a set of measures in January 2008, including the introduction of terminations by agreement, that led to the June 2008 law.
For employees, terminations by agreement bring several advantages compared to quits. After a termination by agreement, employees remain eligible to receive a severance package at least as important as the one they receive in case of an employer-initiated termination. They also remain eligible to receive unemployment benefits. To the best of our knowledge, France is the first country who introduced a procedure of termination by mutual agreement which does not entail, for employees, the loss of eligibility to receive unemployment benefits and severance packages. For employers, the main advantage of terminations by agreement over dismissals is that terminations by agreement need not be justified. Termination by agreements do not exempt employers from giving layoff notices or paying severance package, but save them from having to explain why they wish to terminate the labor contract 8. This alone reduces dramatically the risk of subsequent litigation 9 and, consequently, the termination costs expected by employers, especially in periods where terminations cannot be motivated by clear economic problems.
Potential effects of the reform : a conceptual framework
Compared to dismissals, terminations by agreements represent an option which is less risky for employers and less stigmatizing for employees. Hence, we can hypothesize that the 2008 reform induced the substitution of terminations by agreements for some dismissals. In fact we can expect such substitutions to be even more likely for dismissals for cause, since they represent by far the greatest risk of litigation and the most stigmatizing terminations for employees (Gibbons et Katz (1991), Okatenko (2010)).
Compared to quits, terminations by agreements represent an option which is much less costly for employees, but not for employers. Hence, we can hypothesize that the 2008 reform had much weaker substitution effects on quits than on dismissals for cause 10.
Eventually, terminations by agreement may in some cases represent an improvement over no termination at all, for both employers and employees. Before 2008, no termination at all means that dismissal would be too costly for the employer while quitting would be too costly for the employee. But, it does not rule out that some workers would prefer to be on unemployment rather than with their current employer : they choose to stay with their current employer because the only possible ways to leave their employer involve either stigmatization costs (dismissal) or the loss of eligibility to receive unemployment benefits (quit). If the number of such would-be movers is significant and if terminations by agreement are perceived by employers as less risky and costly than dismissals, the 2008 reform may induce a rise in overall separation rate, i.e., a rise in terminations by agreement signed by people who would have stayed with their employer before the reform.
In Appendix 2.B, we develop a simple labor demand model that makes more precise how the introduction of terminations by agreement may affect firms’ hiring and termination decisions. Assuming that terminations by agreement are actually less risky and costly than dismissals, the model shows that the introduction of terminations by agreement may or may not entail a rise in overall termination rates depending on the number of would-be movers and on how the magnitude of adverse labor demand shock (denoted ) compares to exogenous outflows of workers (denoted S).
In a nutshell, when is larger than S, the difference S represents the downward adjustment that the firm would like to perform when it is hit by an adverse shock. In practice, the firm performs this downward adjustment only if labor adjustment costs are not too high. Hence, if the adjustment costs associated with terminations by agreement are sufficiently low compared to the adjustment costs associated with layoffs and if there exists a sufficiently large number of would-be movers, it may become possible for firms to make the S adjustment after the reform (using termination by agreements) whereas no adjustment would have been seen pre-reform (because of layoff costs). In the remainder of this paper, we will build on an administrative establishment-level dataset with exhaustive quarterly information on flow of workers to test these different assumptions and to explore the consequences of the 2008 change in employment doctrine.
Table of contents :
0.1 Contribution and outline of the dissertation
0.2 Chapter 1 : Subsidizing Adjustment Or Wages ? Evidence From A Hiring Credit In France
0.3 Chapter 2 : Termination of Employment Contracts by Mutual Consent and Labor Market Fluidity
0.4 Chapter 3 : Labor Market Concentration and Stayers’ Wages : Evidence from France
0.5 Chapter 4 : Rival Guests or Defiant Hosts ? The Economic impact of Hosting Refugee
1 Subsidizing Adjustments Or Wages ? Evidence From a Hiring Credit In France
1.3.3 Monitoring file of Pôle Emploi
1.4 Descriptive statistics
1.5 Econometric analysis
1.6 Robustness tests
1.6.1 Selection in the sample
1.6.2 Placebo tests
1.6.3 Randomization inference
1.A Measure of eligibility criterion
1.B Use of the hiring credit
1.C Selection in the EMMO sample
2 Termination of Employment Contracts by Mutual Consent and Labor Market Fluidity
2.2 Institutional context
2.2.1 Institutional context before 2008
2.2.2 The 2008 reform
2.3 Potential effects of the reform : a conceptual framework
2.4.1 Terminations by agreement and establishments’ survival
2.5 Terminations by agreement and establishments’ exit flows : a graphical analysis .
2.5.1 Terminations by agreement as a substitute for other forms of terminations .
2.5.2 Terminations by agreement and overall separation rates
2.6 Regressions analysis
2.6.1 Regression results
2.6.2 An augmented specification
2.6.3 Interaction between the timing of the reform and the one of the crisis
2.7 Termination by agreement and worker mobility
2.A Figures and Tables
2.B Conceptual Framework
2.B.1 Technology and adjustment costs
2.B.2 First-order conditions and state variables
2.B.3 Pre-reform optimal strategies
2.B.4 After the reform
2.C A “stacked” difference-in-difference approach
3 Labor Market Concentration and Stayers’ Wages : Evidence from France
3.2 Literature review
3.2.1 Modern Monopsony
3.2.2 New classical Monopsony
3.3 Empirical specification
3.3.1 Labor market concentration
3.3.2 Labor market concentration and wages
3.5 Main results
3.6 Secondary results
3.6.1 Selection into the stayer status
3.A Not for publication or for online publication only
4 Rival Guests or Defiant Hosts ? The Local Economic Impact of Hosting Refugees
4.2 Context and Data
4.2.1 Humanitarian migrants and housing centers in France
4.3.1 Econometric model
4.3.2 Identification Hypothesis
4.4.1 The mobility response of natives
4.4.2 Economic consequences of native avoidance
4.5 Estimating the aggregate welfare cost of refugee center openings
4.A Alternative channels
4.A.1 Local Labor market
4.B Refugees and housing centers
4.B.1 Humanitarian migrants in France
4.B.2 Refugee centers openings
4.B.3 Refugee centers in the press
4.C Definition of population
4.E Spillover effect
4.F Main types of crime and misdemeanour
4.G Another amenity shock : day-care centers closures
4.H Other matching algorithms
4.I Local tax rates
4.J Identification problem