Can a Website Bring Unemployment Down? Experimental Evidence from France

We evaluate the impact of "Bob Emploi,'' an online job-search assistance website established to reduce unemployment in France. Bob Emploi was launched in 2016 by Bayes Impact, a private non-profit organization. In designing the website, Bayes Impact benefited from a close collaboration with the French public employment agency, Pôle emploi, which records systematic information on each job seeker's periods of unemployment. Bob Emploi set out to analyze the rich data provided by Pôle emploi with state-of-the-art statistical analysis methods. This innovative partnership sets Bob Emploi apart from other private websites and increases its potential for impact. Bob Emploi endeavors to assist and motivate job seekers by providing personalized and data-informed advice on sectors and locations to target; offering step-by-step planning assistance; sending regular reminders and encouragement messages; and providing general tips, such as how to behave during a job interview.

2017-04-01

2017-12-01

Employment outcomes, including: experiencing any employment episode following the intervention, obtaining a stable job (a contract of more than six months), total number of days in unemployment, and total amount of unemployment bene fits. Up to 18 months after the intervention.

Job-search effort, scope, and expectations, including: job seekers' reemployment expectations and how realistic they proved to be; job search effort (number of hours respondents declare spending on their job search, sending at least one unsolicited application per week, and number of online applications sent through Pôle emploi's job board); and scope of the search (respondents who report applying to jobs located outside the municipality where they live; and number of applications sent by job seekers through Pôle emploi for jobs outside their municipality, outside their target sector, and below their reservation wage).

Job-search methods, including: using one's personal and professional networks when looking for a job; adapting one's CV and cover letters to job listings; following up with recruiting firms; number of Pôle emploi websites and of private websites used by job seekers; and use of public job-search assistance offline services (Pôle emploi workshops and meeting with Pôle emploi caseworkers).

Well-being and life balance: self-declared overall well-being; feeling motivated and supported during one's job search; and participation in non-job-related activities.

Our experiment relied on an encouragement design. Job seekers in the treatment group were invited to an informational session at their local agency. Such sessions are frequently organized by Pôle emploi caseworkers to present free tools and programs to job seekers. In the present case, each meeting was hosted by one or two Pôle emploi caseworkers and lasted approximately 90 minutes. It began by introducing the challenges associated with a job search and the value of widening one’s search methods, notably through personal and professional networking. The caseworkers then briefly presented the "Emploi Store," the Pôle emploi platform that lists job-search services, including Bob Emploi. The session then turned to the presentation of Bob Emploi using screenshots and, sometimes, a real-time simulation. Participating local agencies organized these information meetings at a rate of approximately one per week from April 20 to July 31, 2017. Job seekers in the treatment group were invited to these sessions based on the month in which they were drawn. Specifically, individuals from the first wave, drawn on April 1, were invited to attend meetings taking place between late April and late May. Individuals from the second wave, drawn on May 1, and third wave, drawn on June 1, were invited to attend meetings taking place between late May and late June, and between late June and late July, respectively.

To maximize take-up, we completed the informational sessions by sending three emails encouraging all job seekers in the treatment group to create an account on Bob Emploi. The emails were sent on July 7 and 26 (for individuals drawn in the first two months and in the third month, respectively), September 28, and November 13. The exact text of the emails differed slightly for recipients who had attended an informational session and those who had not.

Randomization done in office by a computer

The randomization was done at the level of the individual job seeker. There was a unique level of randomization.

No

212,277 individuals.

212,277 individuals.

92,752 individuals in the control group, and 119,525 individuals in the treatment group.