Employment protection schemes are notoriously hard to reform. Many governments have proposed ambitious liberalisation agendas, only to water them down after being faced with unrelenting opposition. Given the political costs, one may presume that such reforms are based on ample sound evidence. Unfortunately, the state of research is not as conclusive as expected.
Standard economic models predict that employment protection legislation reduces job destruction and job creation, thus suggesting that it has an ambiguous effect on unemployment (Bentolila and Bertola 1990, Mortensen and Pissarides 2011). The negative impact on job creation arises from the anticipation of separation costs, which reduce expected profits. Although this mechanism is clear in theory, it has very little empirical support (Scarpetta 2014).
The anticipation effect in France
In a recent paper (Cahuc et al. 2019), we show that this anticipation effect does play a key role in reducing job creation but also in increasing job destruction, thus inducing significant employment losses. To establish this finding, we use a discontinuity in the relationship between seniority and severance payments in cases of unfair dismissal. Until September 2017, according the French labour code, employers had to pay at least six months' salary to their employees whose seniority exceeded two years in case of unfair dismissal. Figure 1 shows that this regulation entailed a hike in severance payments. Figure 2 shows that this hike induced a significant rise in the job separation rate just before the two-year threshold, followed by a drop of even greater magnitude.
Figure 1 Average compensation in case of appeal for wrongful contract termination (in monthly wages) and seniority (in quarters)
Note: The graph represents a bin scatter which groups the variable on the horizontal axis into equal-sized bins, computes the mean of the variables on the horizontal and vertical axes within each bin, and creates a scatterplot of these data points using a sample of 27,936 French Appeal Courts rulings over 2003-2012. The compensation for wrongful termination reported on the vertical axis is the average compensation for all litigations for contract breach, including the cases where the employee gets no compensation (39.9% of litigations).
Figure 2 The relationship between seniority and quarterly hazard rate from employment to unemployment
Note: Individuals working in the private sector, aged 15 to 54, with at most high school degree. Subsidised jobs are excluded.
To understand how employment protection influences the dynamics of the labour market, we build a search-and-matching model which accurately reproduces the shape of the separation rate and use the discontinuity at two years to identify its structural parameters. Estimating the model allows us to quantify the hike in layoff costs at the two-year threshold, and to separate the procedural from the severance components. We find that total layoff costs increase with seniority and are about four times larger than the expected severance payments of laid-off workers at two years of seniority (see Figure 3). Procedural costs are therefore very large due to the extreme complexity of rules and procedures defined by the French labour code.
Figure 3 The relationship between estimated layoff costs, severance payments, and job seniority
Note: Layoff costs, F, measured in share of the monthly minimum wage, are estimated from the model. Severance payments, Fe, are computed from the data.
Once these costs have been quantified, we run counterfactual experiments to evaluate their impact on job tenure, labour market flows, and unemployment. Although we focus on the French employment legislation, our approach could be applied to a broader set of countries because discontinuities in severance packages after a seniority threshold are fairly common, as shown by the OECD job protection database.
In our empirical context, job protection reduces expected job duration for low-skilled workers but increases it for high-skilled workers, as shown by Table 1 which reports the actual expected job durations when the actual employment legislation is in force and in the counterfactual situation without dismissal costs, simulated from the model.
The negative impact of job protection on job duration is driven by the anticipation effect: as firms foresee that they will have to pay higher costs in the future in the event of a separation, they find it optimal to anticipate the separation decision. The anticipation effect, which raises the destruction of jobs with short tenure, is counterbalanced by the well-known labour hoarding effect which lowers the separation rate for jobs with long tenure. The anticipation effect dominates for low-skilled workers because they usually land jobs with small initial surpluses. The presence of the minimum wage amplifies this effect because it prevents downwards wage adjustments that could dampen the impact of severance payments on profitability.
These findings rely on an empirically relevant property of the model consistent with the strongly decreasing profile of separation rates for jobs with short tenure. The fragility of low-seniority jobs, mostly overlooked by the literature on employment protection so far, implies that the anticipation effect is empirically important, especially in the presence of wage floors. In our empirical context, this suggests that job protection does not achieve its goal since it makes job retention less likely. However, in the absence of wage floors, or when the surplus of starting jobs is sufficiently high, the anticipation effect can be dominated by the labour hoarding effect. This is why we find that job protection raises the job duration of skilled workers, although it decreases that of unskilled workers.
Table 1 Labour market performance with the actual EPL in force until September 2017 and in the counterfactual situation without dismissal costs
Notes: Unskilled workers are individuals aged 15-54 with at most a high school degree. Skilled workers are individuals aged 15-54 with at least a high school degree. “Duration” stands for the expected jobs duration at their starting date. "Actual EPL" stands for the layoff costs estimated in the benchmark case corresponding to the French EPL. "No EPL" is the case without dismissal costs.
We also take into account the impact that job protection has on job creation when we simulate its effect on unemployment. We find that job protection significantly raises unemployment for skilled and unskilled workers. We estimate that removing all dismissal costs would reduce the unemployment rate by 6.4 percentage points for unskilled workers and by 3.3 percentage points for skilled workers. This result is driven to a large extent by the fragility of low-tenured jobs, which implies that job protection raises the destruction of these jobs and significantly lowers their rate of creation. The unemployment impact of job protection is larger for the unskilled because the surplus of their jobs is lower and the minimum wage is more likely to be binding, but it is also strong and significant for skilled workers.
All in all, our results suggest that employment protection does not work as intended by its proponents since it lowers expected job duration for low-skilled workers, who are the main targets of the legislation. More protection leads to less security because firms anticipate their separation decisions in order to avoid paying more firing costs in the future. This mechanism implies that job protection is strongly detrimental to employment in the French context.
Bentolila, S and G Bertola (1990), “Firing Costs and Labour Demand: How Bad is Eurosclerosis?”, The Review of Economic Studies 57(3): 381-402.
Cahuc, P, F Malherbet and J Prat (2019), “The Detrimental Effect of Job Protection on Employment: Evidence from France”, CEPR Discussion Paper 13767.
Mortensen, D and C Pissarides (2011), Job Matching, Wage Dispersion, and Unemployment, edited by K Tatsiramos and K Zimmermann, Oxford University Press.
Scarpetta, S (2014), “Employment Protection”, IZA World of Labor 12.