Can algorithmic predictions and online advice help to reduce unemployment?
Short permalink to this article: https://bit.ly/2S5PlXa
Yagan Hazard (master PPD)
Since the early 2000s, the Internet has created among economists and policy makers some hope for an increase in job search efficiency (1). Many thought the Internet would enable a much larger number of vacancies and job seekers to “meet” each other, given the low cost of both posting a vacancy on the web and searching for a job on online platforms that would centralize those job offers. Yet the first attempts to evaluate the effect of online job search and online platforms tempered the enthusiasm, as no significant effect were found (2). With the increasing capacity of online platforms to concentrate and classify more and more efficiently job vacancies, the original hope for an online-augmented labor market was revived in the past decade, and growing number of public placement services have been investing in innovative ways to provide job search assistance. In France, Pôle emploi has its own start-up incubator; in the US, Ohio announced recently a partnership with LinkedIn to help its job seekers. However, as mentioned above, there is a lack of experimental studies that provide robust insights on the effectiveness of such policies. Recent research suggests that providing tailored online advice about job search could broaden the set of jobs considered by job seekers, and ultimately increase their chances to get job interviews (3). This would be true in particular for individuals who are searching narrowly (in terms of the set of occupations considered) in the first place.
The evaluation of La Bonne Boîte carried out by Yagan Hazard provides evidence on the efficiency of online job search assistance. La Bonne Boîte is a start-up that uses administrative data about past hiring behavior and some prediction algorithm to forecast hiring at the firm level. This information is then made available on a website to help job seekers to search more efficiently. In his study, Yagan Hazard takes advantage of a randomized control trial (RCT) conducted by Pôle emploi in 2015. It encouraged (through e-mails) 75 000 randomly selected (“treated”) job seekers to use La Bonne Boîte: the comparison between those individuals and 75 000 other random (“control”) job seekers allows to estimate the causal effect of this encouragement. On average, no statistically significant effect on the job finding rate is detected. However, the study reveals the existence of some heterogeneity. Indeed, in “tight” labor markets — i.e. labor markets on which the ratio of vacancies to job seekers is high —individuals encouraged to use La Bonne Boîte have a 0.5 percentage point higher chance to find a job after 6 months compared to those who are not encouraged. This is consistent with La Bonne Boîte acting as a vacancy sorter: it would allow job seekers to focus their search effort on firms with the highest probability of hiring in the end. If it is the case, then intuitively its effect should be larger when there are numerous firms to consider in the first place — and that would explain a larger effect on tight labor markets. Lastly, another heterogeneity analysis suggests that individuals with the worst employment prospects benefit more from this online job search assistance than others. More precisely, in the population with below median probability to find a job (in the absence of treatment), individuals who were encouraged to use La Bonne Boîte have a 3 percentage point higher job finding rate after 6 months compared to the one of individuals who are not encouraged. This result suggests that online job search assistance might be a useful tool to fight long-term unemployment, an important policy objective given the high costs associated with it.
***
References
(1) Autor, D.H. (2001). Wiring the Labor Market. Journal of Economic Perspectives.
(2) Khun P., Mansour H. (2014). Is Internet Job Search Still Ineffective? The Economic Journal.
(3) Belot M., Kircher P., Muller P. (2017). Providing Advice to Jobseekers at Low Cost: An Experimental Study of Online Advice. The Review of Economic Studies.
Master’s thesis title: “Online Platforms and the Labour Market: Learning (with Machines) from an Experiment in France”
Under the direction of: Luc Behaghel
Available at: https://dumas.ccsd.cnrs.fr/dumas-02407555
Photo credit: Raw Pixel (Shutterstock)