Freelancers Advertising RCT
How to cite
Sahni, Navdeep; Waisman, Caio; Haanwinkel, Daniel. 2026. "Freelancers Advertising RCT." AEA RCT Registry. AEARCTR-0019372. https://doi.org/10.1257/rct.19372-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
The study is done in collaboration with a large freelancer platform.
Prior to the experiment, the platform had Premium freelancers: those who pay the platform for services such as ability to reach employers and get more visibility.
The experiment randomized employers into treatment and control group.
Control group got the status quo experience with no sponsored listings.
Treatment group employers were eligible to see sponsored listings featuring Premium freelancers. For any search conducted by a treatment group employer, the platform searched for the most relevant Premium freelancers; if a relevant Premium freelancer was found, the platform placed the most relevant one on top of the listings shown to the employer. At most one sponsored listing was shown in a page.
All other aspects of the platform (including other ways in which Premium freelancers are visible) is same across treatment and control employers.
Intervention
The intervention was promotion of Premium freelancers (FLs). Treatment group employers were eligible to see sponsored listings featuring Premium FLs.
Experimental design
We conducted an employer-level randomized field experiment in partnership with a large online labor platform. The experiment ran from April through July, 2019. Employer accounts were assigned with equal probability to a treatment group or a control group when they entered the experiment during the study period. The realized analysis sample consisted of 464,986 employers: 233,211 assigned to treatment and 231,775 assigned to control.
Employers assigned to the control group experienced the status quo version of the platform, without ads for workers. Employers assigned to the treatment group were eligible to see premium-membership workers displayed as sponsored listings (ads) on worker search and recommendation pages. For each page request, the platform first generated its usual organic list of workers using its existing matching algorithms. It then identified premium-membership workers within that organic list. On the first page of results, the highest organically ranked premium worker was displayed as an ad above the organic listings; on the second page, the second-highest organically ranked premium worker was displayed as an ad, and so forth. The ad added an additional listing rather than replacing an organic listing. A similar procedure was used for worker recommendations shown to employers after they created a job posting.
Treatment assignment determined eligibility for advertising rather than guaranteeing exposure. An ad could be displayed only when the underlying organic candidate set contained a premium-membership worker.
Outcomes are measured at the employer level during the 60 days following assignment. The outcomes include whether and how frequently an employer clicked on worker profiles, invited workers to apply for jobs, and hired workers. These outcomes are measured for all workers and separately according to workers’ premium-membership status. The primary analysis follows an intention-to-treat design, comparing employers according to their randomly assigned treatment status.
Design details
Cited by
No citation data available for this registration yet.