Dismissals are costly and are often subject to stringent employment protection regulations. This column explores how France’s 2008 introduction of separation by mutual agreement affected separation patterns. Despite fast and wide adoption of such agreements, they only replace a small share of dismissals and substitute more for quits rather than dismissals. Highly conflictual dismissals remain unchanged. The reform failed to improve employer–employee relations during separations and added costs to the unemployment insurance scheme.
Dismissals are costly for firms to implement and have lasting consequences for workers. Two defining features of individual dismissals are that they are typically unilateral decisions by employers and are subject to stringent employment protection regulations (such as advance notice, just-cause requirements, and unemployment-insurance eligibility rules). As a result, in most settings, labour law offers little scope for workers and employers to negotiate a dismissal on mutually agreed-upon separation terms.
There are few systematic studies on how dismissals proceed in practice and how the parties interact. For instance, the literature has investigated firms’ choice to use dismissals rather than downward wage adjustment (Bewley 1999, Davis and Krolikowski 2025, Bertheau et al. 2025) and the decision to separate as such (Jäger et al. 2019, 2023). But the dismissal process largely remains a black box.
In Carry and Schoefer (2024), we study the introduction of a unique way to settle a dismissal amicably at the separation stage. Our context is France, a country known for conflictual labour relations. For instance, 25% of dismissals for personal reasons (which were about 77% of all dismissals, as opposed to collective or economic dismissals) are litigated in labour courts, and about 60% of first-instance rulings are appealed. In part to reduce labour disputes, in 2008 France introduced the possibility for a worker and an employer to strike a flexible and mutually beneficial deal through a rupture conventionnelle, or ‘separation by mutual agreement’ (SMA). In fact, a core motivation of the policy was to make dismissals more amicable by replacing them with SMAs.
The 2008 introduction of ‘separations by mutual agreement’
SMAs allow the employer and the employee to mutually agree to terminate a permanent contract in a flexible framework, where parties can negotiate two parameters: the level of compensation paid to the employee and departure date. The severance pay must be at least equal to the amount paid in case of dismissal, but parties can negotiate additional compensation beyond this minimum. SMAs also preserve the employee’s right to unemployment insurance while precluding recourse to labour courts since both parties consent to the separation.
Crucially, the process is significantly simpler (including a standardised template) and less risky than a dismissal for personal reasons. While dismissals require strict procedures – specifying and formalising a just cause, preliminary meetings, formal notifications, and respecting advance notice periods – SMAs eliminate these obligations and entirely preclude the substantial litigation risk in France.
Have SMAs succeeded in settling dismissals?
Given these advantages, basic economic theory and the policy motivation would predict large take-up of SMAs in separations otherwise leading to a dismissal for personal reasons.To assess whether this theoretical prediction holds, we estimate the share of dismissals that have been replaced by SMAs.
Indeed, following their introduction in 2008, SMAs were quickly and widely adopted. Figure 1 shows that their use climbed quickly to reach a steady 15%–18% share of all open-ended contract separations.
Figure 1 Take-up and diffusion of separations by mutual agreement
Notes: Share of SMAs among all open-ended contract terminations between 2007 and 2022. Open-ended contract terminations include quits, economic, or personal dismissals, and SMAs.
Source: Carry and Schoefer (2024).
However, despite SMAs having become part of the toolkit in the French labour market, the data provide a resounding rejection of the idea that they succeed in replacing dismissals. (Instead, it is plausible that SMAs mostly replace quits, as we discuss later in this column.) Only about 12% of personal dismissals are replaced by SMAs. We reach this estimate using three unique and independent estimation methods. As illustrated in Figure 2, estimates vary between 10.9% and 11.8%.
Figure 2 Conversions of personal dismissals into SMAs: Three estimates
Notes: Share of dismissals in total employment, observed and counterfactual absent SMAs, across three independent estimation methods. The share replaced by SMAs is computed by comparing counterfactual to observed dismissals.
Source: Carry and Schoefer (2024).
The first method compares the total number of personal dismissals before and after the introduction of SMAs. Between 2003–2006 and 2012–2014, personal dismissals (relative to total employment) decreased by 12%.
The second method relies on differences in SMA usage between labour market cells defined by company size, sector, profession, seniority, and worker age. After introducing SMAs, the decrease in dismissals is more pronounced in cells with higher SMA usage. Combining this difference-in-differences estimator with aggregate SMA usage yields an 11.8% conversion share.
The third approach relies on a survey of employees who recently concluded an SMA. When asked what would have happened to their employment relationship if they could not use an SMA, 22% responded they would have been dismissed, which implies a 10.9% conversion share.
Hence, across three different estimation strategies, we find strikingly similar conversion shares.
Which dismissals take up SMAs?
Additionally, we present evidence indicating dismissals converted into SMAs are more cooperative and amicable than typical dismissals before the reform, as shown by pre-reform separation patterns near retirement. Under French labour law, dismissed workers over 50 can receive up to three years of unemployment benefits (versus the usual two), which in practice lets workers use unemployment as a bridge to retirement starting three years before retirement age.
Before the reform, workers and employers could effectively negotiate dismissals to access early retirement, producing excess dismissal rates at the three-years-to-retirement mark (Figure 3). After SMAs were introduced, these excess dismissals declined as a large share of such cooperative separations shifted to SMAs. The implied conversion share at three years before retirement is 37%, well above the overall 12% share.
Figure 3 Cooperative dismissals and SMAs before retirement
Notes: Comparison of dismissal rates before (2004–2007) and after (2011–2014) SMA introduction by distance to statutory retirement age, and distribution of SMAs over 2011–2014 by the same distance.
Source: Carry and Schoefer (2024).
Conversely, the most conflictual dismissals are not converted, as evidenced by the complete lack of decrease in labour court cases challenging personal dismissals (Figure 4).
Figure 4 Number of labour court cases following a dismissal for personal reasons
Notes: Annual number of labour court cases following personal dismissal. Cases are counted in the year the judgement is rendered.
Source: Carry and Schoefer (2024).
Why do SMAs fail to settle dismissals?
To shed light on the low conversion share of SMAs, we conducted a survey of 210 human resource directors who decide (78%) and supervise (95%) dismissal implementation in their companies. The survey allows us to identify the main mechanisms why separations by mutual agreement are not concluded (Figure 5). When we asked these human resource directors why the last personal dismissal they implemented was not converted into an SMA, three mechanisms stood out.
First, hostility between employer and employee and conflictual relationships are the most frequently cited (62%) reason for lack of conversion. This covers situations where employees deliberately create difficulties for the company and, more generally, where one party finds interest in harming the other by imposing costs rather than negotiating.
Second, conversions are also precluded by the usage of dismissals as a disciplinary tool to incentivise other employees (50% of respondents). By firing workers, employers send a dissuasive signal to remaining employees: sanctionable behaviour is followed by dismissal, making such behaviour less attractive.
Finally, asymmetric beliefs about labour court outcomes also contribute to the lack of conversions (49% of respondents). These are cases where either the employer or employee is overly optimistic about their chances of success in a labour court.
Figure 5 Three main sources of conflict in dismissals
Notes: Share of human resource directors reporting that each mechanism contributed to not signing an SMA in the last dismissal they implemented. Listed mechanisms include: (i) hostility or a too-conflictual relationship between employer and employee; (ii) using dismissal as a disciplinary device toward other employees; and (iii) divergent beliefs about the outcome of a potential labour court case (likelihood of going to court, compensation amounts). Bars report the shares answering ‘Yes, a lot’ or ‘Yes, somewhat.’ Reading: 60% report that hostility or a conflictual relationship played a role, of whom 30% ‘Yes, a lot’ and 30% ‘Yes, somewhat.’
Source: Carry and Schoefer (2024).
When asked about a hypothetical scenario absent these three factors – no hostility, no incentive effect for other employees, and no uncertainty about labour court outcomes – responses indicate the conversion share would reach around 67% (compared to 12% currently).
All three of these mechanisms point to forms of conflict in the dismissal process as explaining the inability of parties to minimise joint costs through an SMA.
More evidence on the role of conflict
To provide additional evidence for the key role of conflict, we use worker survey data measuring conflict proxies in their companies and find that limited recourse to SMAs correlates with the existence of conflict (Figure 6). For instance, conversion shares are lower in companies whose employees report that their efforts aren’t valued or that supervisors don’t listen to their opinions. They are also lower where managers indicate employees have contested dismissals in labour courts or frequently use sick leave. These results align with our earlier analyses on the central role of conflict in limiting conversion, this time using a worker-side survey.
Figure 6 Relationship between the conversion share of dismissals into SMAs and employer–employee conflict
Notes: Regressions at the labour-market-segment level (sector, establishment size, occupation, age) linking the average conversion share of dismissals into SMAs to the frequency of conflict situations reported by workers in the REPONSE survey. Bars show estimated coefficients with 95% confidence intervals. Values in parentheses indicate each factor’s average frequency.
Reading: Segments where a larger share of workers report that their effort is not valued exhibit lower conversion shares; a 10-point increase in that share is associated with a 1.1–1.4 point decrease in conversion.
Source: Carry and Schoefer (2024).
What do SMAs do if not replace dismissals? The importance of quits
The low number of conversions of dismissals into SMAs suggests a crucial question: what do the SMAs do? Our findings imply that about 24% of SMAs are converted dismissals, meaning 76% replace either a quit or continued employment. Our study also sheds light on the main alternative margin of substitution: for quits. Indeed, to the extent that employees can receive unemployment insurance otherwise unavailable when resigning, they have an incentive to turn to SMAs rather than quit.
Our evidence suggests that such replacements are substantial. First, the total number of quits (relative to total employment) decreased by about 19% after the reform (Figure 7). Second, a 2012 survey of employees who terminated contracts by SMA indicates nearly 40% would have resigned absent SMAs (versus 22% who would have been dismissed, and 28% who would have remained employed).
Figure 7 Trends in quits and SMAs
Notes: Trends in the share of quits in total employment, and of the combined share of quits and SMAs in total employment, 2003-2006 versus 2012-2014.
Source: Carry and Schoefer (2024).
While the obligation for the employer to pay severance when separating by SMA should prevent replacing quits, the evidence of substantial substitution indicates that employees compensate employers for this cost, particularly by facilitating transition and extending their notice period. Indeed, following an SMA, most employees remain longer than the mandatory month of notice, and this extension is associated with higher severance (each additional day spent in the company is associated with a 2.3% increase in severance relative to monthly salary).
This suggests that SMAs partially replace resignations because the employee is able to negotiate obtaining the SMA in exchange for rendering these services to the employer. More often, such employees occupy a more strategic position in the company and are therefore better placed to negotiate SMAs. Among managers, 19% of separations take the form of an SMA, compared to 11% among workers and employees. This also corresponds to higher-salary workers, increasing the cost to unemployment insurance.
Finally, we show that the introduction of separations by mutual agreement in France only replaced a minority of dismissals. Amicable and less conflictual dismissals are more likely to be converted to SMAs than conflictual dismissals. Instead of converting litigious and conflict-heavy dismissals to amicable separations, the majority of SMAs yield to new unemployment insurance (UI) claimants. At a time when France is discussing a potential change in the taxation of the SMA severance, our paper shows that the reform was largely unsuccessful at improving employer–employee relationships during separations, while generating an additional cost for the UI system.
Source : VOXeu
