Providing Job-Seekers with Labor Market Information

2021-05-03

2021-09-30

Baseline beliefs on labor market parameters.
Impacts on:
- job search behavior: quantity and timing of search effort and the characteristics of jobs applied to.
- belief updating (follow-up survey).
- general labor market outcomes (unemployment duration and hiring: contract types (temporary or long term), geographic and sectoral mobility)
- take-up of public employment services such as vocational training or financial aid for mobility

550,000 job-seekers (registered at Pôle emploi, the French public employment service) are randomized into several treatment or control groups. Control and treated groups are surveyed to collect their beliefs about key parameters of their local labor market. Treated jobseekers are additionally exposed to stylized (charts, text etc...) and personalized (micromarket=profession x local labor market) information at the end of the survey. We separate the treated group into several subgroups that receive different information packages. The information treatments provide distributional moments of wages and tightness and qualitative informationabout mobility along with all pair-wise interactions. The experiment will be implemented from mid-May to end of September 2021.

Using Pole emploi’s rich administrative data and an additional follow-up surveys we analyze the impact of this information on beliefs, job-search behavior and on relevant labor market outcomes.

We first isolate micro-markets to be include in the experiment. These markets will be selected based on the availability of reliable information to create the information packages. We will have 9 treatment arms that correspond to different the information packages and 2 control groups:

1. Wages in the jobseeker's declared sector, locality and experience
2. Labor market tightness for the jobseeker's declared sector, locality and experience
3. 1+2
4. 1 + sectoral mobility information
5. 2 + sectoral mobility information
6. 1 + geographic mobility information
7. 2 + geographic mobility information
8. 1 + sectoral mobility + geographic mobility
9. 2 + sectoral mobility + geographic mobility
10. Information impact control group
11. Survey response control group

Then within the micro-markets in which jobseekers are eligible to receive the survey, we stratify our sample (by micro-market, unemployment duration, education and gender) and randomly assign 550k individual jobseekers within strata with probability 1/11 to groups 1 to 11 to obtain roughly 50,000 jobseekers in each of the 9 treatment arms and 2 control groups. The “information control group” (10) receives a questionnaire but none of the information treatments. The “survey response control group” (11) receives neither the survey nor the information treatment. This group is used to control for possible survey response effects (H0: impact 10 – 11 = 0) using exhaustive administrative data.

We expect to have a response rate of around 10% to the survey with little differential attrition by treatment status. Hence we will have roughly 5,000 respondent in each treatment arm (1-9) and the information control group (10). We thus measure intention to treat effects (ITT) on the entire sample and average treatment effects (ATEs) on the "compliers" of survey respondents.

A primary goal of the survey is to first elicit each jobseeker's beliefs about the key labor market parameters that, as benchmark theory tells us, determine the (expected) returns to job search effort. This allows us to construct a key dimension of heterogeneity through which we will examine treatment effects: Does information provision have different impacts on those who hold accurate versus those that hold inaccurate priors about moments of the wage distribution, job finding probability/competition and opportunities in other sectors and/or localities?

We use the rich administrative data made available by the public employment service (PES) to track job search effort and job finding results. These data also allow us to see the characteristics of the jobs that individuals apply to (sector, initial wage offer, contract type, experience, location and qualification) as well as job finding rates, type of contract and location of eventual job. We can also track data on jobseeker vocational training and financial aid undertaken through the PES.

We will conduct a short follow-up survey in Fall 2021 to re-measure beliefs to provide evidence on the efficacy of the information treatment and potentially collect additional data that we believe might be important following analysis of the initial survey results and treatment impacts.

Finally, we also create an additional “super control group” by randomizing sampled micro-markets before the individual level random assignment. We do this to be able to partial out any significant equilibrium effects of the experiment as a whole. Though this is a "light treatment" and is not expected to cause measurable externalities on non treated/surveyed jobseekers, we nevertheless want to give ourselves the possibility for a robust analysis of aggregate treatment effects. No action will be taken towards this super control group.

Stratified randomization based on individuals characteristics (unemployment duration, education, micro-market, gender). Randomization will be done on a computer in the offices of the French Public Employment Service.

Jobseeker

No

550,000

550,000

50,000