The Impact of Job Search Requirements and the Salience of Monitoring and Sanctions: Evidence from Integration Agreements for Job Seekers in France

The intervention is implemented through the internal case management software used by caseworkers of the French public employment service (France Travail) during meetings with registered job seekers. During these meetings, caseworkers prepare the Contrat d’Engagement (Engagement Contract), which is then signed by both the caseworker and the job seeker. The contract is an integration agreement that specifies commitments of both the job seeker and the caseworker, including the actions the job seeker agrees to undertake to support labor market integration.

Depending on the treatment arm, caseworkers encounter specific banners in the software prompting them to emphasize different aspects of the contract. These prompts either encourage the caseworker to set a precise individual specific quantitative target for the number of job applications to be submitted by the job seekers, or to explicitly discuss monitoring procedures and potential sanctions for non-compliance. In addition, caseworkers in treated agencies receive a digital toolkit containing guidance and suggesting talking points, accessible through links embedded in the software banners.

2026-03-13

2026-12-16

(1) Leaving unemployment for work
(2) Employment rates
(3) Number of job applications

We use administrative records from France Travail that can be matched with administrative data on employment outcomes. These data allow us to observe, for each individual in our sample, whether they are registered at France Travail, whether they receive unemployment insurance (UI) benefits, and their employment status for up to 18 months after leaving unemployment. Using these administrative records, we will estimate (1) the probability of exiting unemployment for work before unemployment duration t, and (2) the probability of being employed at time t since signing the first integration agreement after registration at France Travail.

The administrative records from France Travail also contain information on job search behavior on the France Travail platform. Our first measure of the number of job applications submitted by the job seeker will be based on the applications sent through the France Travail platform. A second measure will be based on the surveys conducted among job seekers. In these surveys, respondents report the total number of applications sent to employers during the previous four weeks. They are also asked about the number of applications submitted through different channels, including through France Travail.

We will investigate treatment effect heterogeneity with respect to the predicted reemployment rates for the individuals in our sample. The prediction model will be based on a large set of characteristics observed in the administrative data of France Travail, including socio-demographic characteristics (such as education, age and gender), previous employment history, participation in active labor market policy programs, potential benefit duration and UI benefit amounts, and indicators of local labor market conditions. Another dimension of heterogeneity we will examine is job seekers’ locus of control. In the first job seekers survey, conducted around four to six weeks after signing the integration agreement, we measure locus of control using eight items. We plan to conduct around 40,000 interviews in the first survey wave.

We will also explore effect heterogeneity depending on local labor market conditions, the caseworker types and activities, and individual characteristics such as age, gender, education, and eligibility for UI benefits.

(1) Reservation Wage
(2) Search Effort (time)
(3) Earnings
(4) Type of employment contract
(5) Sectoral and occupational mobility

In both job seeker surveys, respondents report their reservation wage and the time they spent on job search during the previous week.

The administrative data contain information on realized labor market outcomes, including earnings, the type of employment contract individuals enter (permanent or temporary), and whether job seekers transition to jobs in different sectors or occupations relative to previous employment.

We observe industry and occupation codes for both previous jobs and newly accepted jobs, which allows us to identify actual sectoral and occupational transitions. In addition, we observe the occupations associated with the vacancies to which job seekers apply. This information allows us to examine the extent to which job seekers apply to positions in occupations that differ from their previous employment.

In addition to the primary and secondary outcomes, we will explore several additional dimensions related to job search behavior and labor market outcomes. These include measures of job search activity and strategies, such as the number of interviews, clicks on job postings, the use of formal versus informal search channels, the willingness to make concessions on job characteristics, and the evolution of the average wage associated with the vacancies to which job seekers apply on the France Travail platform. We will further examine subjective expectations and perceptions, including perceived probabilities of monitoring and sanctions, perceived job finding prospects, and stress experienced during the search process. Additional exploratory outcomes relate to the interaction with the public employment service, including the relationship with the caseworker and the subjective evaluation of the integration agreement and counseling process. Further, we will explore outcomes related to job quality and employment trajectories, such as wages at re-employment, job satisfaction, duration of employment spells, and employer quality (e.g., AKM firm effects).

Using a dedicated caseworker survey, we will also examine how the treatments affect caseworkers' practices - including how they structure the engagement contract, set application targets, and communicate about monitoring and sanctions - as well as their perceptions of the policy.

The experiment is based on a randomization at the agency level. A total of 882 France Travail agencies are randomly assigned with equal probability to one of three treatment arms or to a control group. Caseworkers in treated agencies receive digital toolkits containing guidance and talking points corresponding to their agency’s treatment assignment.

The four groups are:

Treatment Arm 1 — Target Setting: The software displays a banner prompting the caseworker to establish and record a specific quantitative target in the job seeker's action plan, most notably a minimum number of job applications. The banner also contains a link to a toolkit. This toolkit provides guidance on how to define an appropriate target and clarifies what is expected from caseworkers.

Treatment Arm 2 — Monitoring Transparency: The software displays a banner prompting the caseworker to explicitly explain the monitoring procedures and potential sanctions associated with the job seeker's obligations. The banner also contains a link to a toolkit. This toolkit provides suggested talking points on obligations and sanctions.

Treatment Arm 3 — Combined Treatment: Caseworkers receive both the Target Setting and the Monitoring Transparency prompts, along with the corresponding toolkits.

Control Group: No banner or toolkit is provided. Caseworkers follow standard procedures.

This design allows us to separately estimate the causal effects of the two treatments (Target Setting and Monitoring Transparency), and also to detect potential interaction effects between the two treatments.

Our main sample consists of jobseekers who had no previous contact with France Travail in the 6 months before their current registration, and who sign a contract in the first four months after their registration. The overall inflow sample will consist of 9 monthly cohorts of newly registered job seekers. We will measure outcomes for this target group based on administrative data and based on two surveys. The first survey will be conducted around one month after signing the integration agreement. The second survey will be conducted around six months after signing the integration agreement. For both surveys, we target a sample size of around 40.000 completed interviews.

Although outside of our main analysis sample, our treatment will also affect job seekers registered at France Travail who do not meet the sample selection criteria and who sign an integration agreement during the timeframe of our experiment. We will explore effects of the different treatments for this group based on administrative data in a secondary set of analyses.

Not available

For randomization, we employ a recursive matching procedure implemented in R (nbpMatching). All participating agencies are first separated into Mainland France and Overseas territories. The mainland agencies are grouped into blocks of eight based on similarity across three administrative variables observed over the preceding year: the average monthly inflow of new jobseekers eligible for unemployment insurance (UI) benefits, the average employment rate six months after registration for this inflow, and the inflow of jobseekers eligible for welfare benefits (RSA - Revenu de Solidarité Active- beneficiaries). Overseas territory agencies are grouped into blocks of eight based on the inflow of jobseekers eligible for welfare benefits.

Within each block, two agencies are allocated randomly to each treatment arm.

Those agencies which are not assigned to a block due to missing values, or which were left out to reach a multiple of eight, are randomly allocated to the either one of the treatment groups or the control group using the statistical software R.

The experiment is clustered at the agency level.

Yes

882 France Travail employment agencies.

We expect a monthly inflow of around 100.000 job seekers signing a new integration agreement. This will lead to a total number of around 900.000 job seekers.

Target Setting: around 220 agencies with overall around 225.000 job seekers
Monitoring Transparency: around 220 overall agencies with around 225.000 job seekers
Combined: around 220 agencies with overall around 225.000 job seekers
Control: around 220 agencies with overall around 225.000 job seekers

For the power calculation, we assume an inflow sample of 900,000 unemployed job seekers for whom outcomes can be measured using administrative data. In addition, we will conduct two surveys among job seekers and target 40,000 interviews for each wave. We assume an intraclass correlation coefficient of 0.01 at the agency level. We calculate minimum detectable effect (MDE) sizes for comparisons between each treatment arm and the control group. We get a MDE of 1.4 percentage points for the employment rate, which corresponds to a relative effect of 3.5% (assuming an average employment rate of 40% 6 months after unemployment entry). For the number of applications sent – observed in the surveys -- we get a MDE of 0.86, which corresponds to a relative effect of 8.2% (assuming an average number of applications in the last four weeks of 10.5 and a standard deviation of 18).