Experimental Design
In the first phase of the project, we conduct a two-sided randomized experiment involving about 1.2 million job-seekers and 40,000 establishments, in 94 commuting zones in France (about one fourth of the French labor market).
On the business-as-usual mode, the LBB website directs each job-seeker toward a list of firms most likely to hire him according to the location and occupation criteria he enters. During the experiment, while the platform remains available to all, we introduce two experimental treatments. First, we randomly select a subset of firms among those short-listed by the LBB algorithm. During four weeks, those “treated” firms are displayed in priority in response to job-seekers’ requests on the website, while the remaining “control” firms are not displayed (or displayed at the bottom of the list if there are too few treated firms satisfying the search criteria). Second, we randomly draw two thirds of the 1.2 million job-seekers to receive two or four emails pushing the LBB service, with specific recommendations toward up to eight of the treated firms.