Does automatic recommendations lead to better matching between job seekers and employers?

How to cite

Barbanchon, Thomas Le; Rathelot, Roland; Hensvik, Lena. 2020. "Does automatic recommendations lead to better matching between job seekers and employers?." AEA RCT Registry. AEARCTR-0003616. https://doi.org/10.1257/rct.3616-1.2000000000000002

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Thomas Le Barbanchon* , Roland Rathelot , Lena Hensvik

* Primary investigator

Registry ID
AEARCTR-0003616
DOI
https://doi.org/10.1257/rct.3616-1.2000000000000002
First registered
Last updated
Status
On going
Trial dates
2019-01-15 – 2022-06-30
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/3616

Abstract

Understanding the search process of job seekers and employers is at the heart of labor economic research and policy. Still our understanding of the nature and consequences of job search is inadequate, mainly due to lack of data. Our project proposes to provide new empirical evidence on the search strategies of both job seekers and of recruiters in the Swedish labor market. This evidence will enhance our understanding of the information asymmetries at the root of search frictions.
We will leverage the opportunity offered by online job boards, which record search activities in detail. We will analyze matched online search activity data with administrative data from unemployment-employment registers. This will enable us to jointly observe search activity and core economic outcomes (wage, job duration) on large samples.
We will evaluate the impact of recommending new matches to job seekers and recruiters, using randomized controlled trials. This will allow us to test for the extent of geographical and skill mismatch in the labor market. We will also learn whether targeted information can be used to improve the matching efficiency.
Our results will guide policy-makers who design job boards and the unemployment insurance system. While online job portals have flourished in the recent years, we still know very little about the impact on unemployment duration and the quality of employment. This project will shed light on the most important information frictions and test improvements. We hope that our research will help to reduce frictional unemployment and to increase productivity through a reduction of mismatch in the labor market.

Intervention

The population of interest are users of Platsbanken, the online job board of Arbetsförmedlingen (Swedish Public Employment Service). The intervention consists in recommending matches to users, using an automatic recommendation system.

Experimental design

We will randomise users into two groups, a treatment and a control group.

Design details

Randomization method
Randomisation done by a computer.
Randomization unit
Individual level. Local labor markets (since October 2021)
Clusters
Per day: around 30,000 vacancies potentially included in the recommendation systems Per day: around 12,000 job-seekers and 600,000 users
Observations
Per day: around 30,000 vacancies potentially included in the recommendation systems Per day: around 12,000 job-seekers and 600,000 users
Arms
Per day: around 15,000 vacancies included in the treatment group and 15,000 in the control group Vacancies remain in their randomization group from one day to another. Per day: around 6,000 job-seekers/300,000 users included in the treatment group and around 6,000 job-seekers/300,000 users in the control group
Primary outcomes
Search activity on Platsbanken, unemployment duration, recruitment duration, employment and wages outcomes.

Cited by

No citation data available for this registration yet.

Keywords