Unemployment in France is a long-running and deep-rooted problem (Baudchon 2015). François Mitterrand came to power in 1981 on promises to reduce unemployment, only to see it soar into the double digits (Sachs and Wyplosz 1986). And in the past 40 years, France’s unemployment rate has compared unfavourably to the US and other economic peers. Beyond France, several other European countries experience high levels of unemployment, particularly in Southern Europe, including Italy, Spain, Portugal, and Greece (Figure 1).
Figure 1 Unemployment, total (percentage of the total labour force) in France, Germany, Italy, and the US
Source: International Labour Organization, ILOSTAT database via World Bank.
Unemployment brings with it economic, social, and psychological costs, leading France and other countries to try out new programmes to reduce it.
A large literature tests ways to address labour market forces that keep people out of work. For example, Blundell et al. (2004) use variation in eligibility criteria based on area and age to show that a large programme in the UK providing extensive job assistance and wage subsidies to employers raised transitions to employment by about 5 percentage points. Recent meta-analyses of subsidised employment, training, and job-search assistance programmes find positive, if uneven, effects (Card et al. 2010, 2017).
However, such labour market policies can be expensive. Given the evidence that frictions and inefficiencies in the labour market account for an important share of overall unemployment even when jobs are available (e.g. Mortensen 1970, Cooper and Kuhn 2020), some recent experiments have attempted to home in on the informational and behavioural challenges jobseekers face, and test low-cost ways of alleviating them.
Belot et al. (2019) experiment with automating tailored advice to jobseekers. Working with 300 jobseekers in the UK, the researchers evaluate an experimental platform recommending alternative occupations to users, and giving them information on the situation of the labour market. The results show that the intervention broadens the set of jobs that unemployed workers consider and increases their number of job interviews.
Altmann et al. (2018) focus on providing information more broadly. In a large-scale experiment in Germany, the research team sent recently unemployed people brochures about the nonpecuniary consequences of unemployment, and find modest effects on employment and earnings, which are concentrated among individuals at risk of being unemployed for an extended period.
Abel et al. (2019) focus on closing the gap between jobseekers’ intentions and behaviour. They conducted a field experiment with 1,100 unemployed youths in South Africa and find that jobseekers who fill out a plan template subsequently increase the breadth and efficiency of their search, leading to a substantial increase in the number of job offers and employment.
While these studies had encouraging effects, the programmes they tested were designed for research purposes, and the fact that researchers were involved in their creation may have enabled a high quality of programme content.
A new study by six of us takes a different approach by combining personalised recommendations on where to search with more general advice on how to search (Ben Dhia et al. 2022). We evaluate at scale a programme reaching many thousands of job seekers designed by a nonprofit in collaboration with the French government employment agency.
In contrast to the aforementioned experiments, the intervention had only limited effects on job seekers’ search strategies and no effect on reemployment outcomes in the short or medium run.
Evaluating a private–public partnership
In 2016, the French employment agency, Pôle emploi, collaborated with a private nonprofit technology firm that aimed to use algorithms to provide tailored advice to jobseekers. The website, Bob Emploi (https://www.bob-emploi.fr/), does not host job listings, but rather provides jobseekers with personalised, data-driven advice on sectors and locations to target in their search, offers step-by-step planning assistance, and provides general tips such as on how to behave during a job interview. Bob Emploi also attempts to assist jobseekers in overcoming the psychological costs of looking for work by engaging them with a user-friendly interface, sending them regular reminders and uplifting messages, and encouraging them to engage in activities outside their job search that might brighten their outlook.
Our experiment entailed generating exogenous variation in take-up of Bob Emploi among over 200,000 individuals who had been registered with the French public employment agency for less than a year, and who attended an informational session that included an introduction to the platform.
We conducted follow-up surveys and use Pôle emploi data to follow subjects’ employment trajectories for 18 months after the intervention. To the best of our knowledge, this is the first at-scale experimental evidence of the impact of an actual private website dedicated to job-search assistance.
The results are sobering. Bob Emploi’s effects were null or nonsignificant on the following outcomes:
- Weekly hours spent looking for a job
- Jobseekers’ search scope (in sector or geographical terms) or number of non-Pôle emploi websites used
- Number of applications, and the likelihood of sending unsolicited applications
- Correcting the jobseekers’ underestimation of the time it would take to find employment
- Self-reported wellbeing or participation in activities.
Bob Emploi did have some modest positive effects on a few desired outcomes: jobseekers in the treatment group were slightly more likely to use their personal and professional network and to follow up with recruiting firms after sending their application (as recommended by Bob Emploi), and they were somewhat more likely to report feeling supported in their search for a job.
However, most importantly, there was no effect on any employment outcome, measured based on administrative data.
Furthermore, we can rule out one possible interpretation of the disappointing effects on employment: that Bob Emploi led people to decrease their use of existing public job-search assistance services. Instead, the number of Pôle emploi websites jobseekers used to look for jobs increased, and their participation in training programmes and other programmes organised by Pôle emploi did not decrease.
Future directions for online platforms
Due to the large scale of the experiment, our null effects on employment are precise. It appears that, at best, Bob Emploi can serve as a complement to public job-search assistance services.
The absence of substantial effect is an important finding, as Bob Emploi was widely lauded at launch1 and is now freely accessible across the entire country. Moreover, it is representative of a growing number of private websites offering job-search assistance services to unemployed people in many countries.
This leads us to reflect on the differences between Bob Emploi and the interventions we list above that showed positive results – and what other types of employment-assistance efforts governments might consider.
On the behavioural front, the automated messaging from Bob Emploi did not appear to have the motivating force of human job counselling (e.g. Behaghel et al. 2014). It may be that adding automated features that have proved successful in smaller experiments to close the intention–behaviour gap would also have some effect at scale. In any case, future interventions should be rooted in an in-depth understanding of jobseekers’ behaviour and their needs, to increase the odds of successfully addressing them.
On the informational front, our results do not imply that attempts to improve the functioning of the labour market by leveraging rich administrative data and AI algorithms are bound to fail. Instead, one possible interpretation of our findings is that recommending new locations and sectors to jobseekers is not sufficient. Online platforms which provide more precise and customized data, and connect job seekers with actual job listings matching their skills may be more impactful. It would be interesting to see such platforms evaluated.
Abel, M, R Burger, E Carranza, and P Piraino (2019), “Bridging the Intention-Behavior Gap? The Effect of Plan-Making Prompts on Job Search and Employment”, American Economic Journal: Applied Economics 11(2): 284–301
Altmann, S, A Falk, S Jäger, and F Zimmermann (2018), “Learning About Job Search: A Field Experiment with Job Seekers in Germany”, Journal of Public Economics 64:33–49.
Baudchon, H (2015), “France: Unemployment, a deep-rooted problem”, BNP Paribas Conjecture.
Behaghel, L, B Crépon, and M Gurgand (2014), “Private and public provision of counseling to job seekers: Evidence from a large controlled experiment”, American Economic Journal: Applied Economics 6(4): 142–74.
Belot, M, P Kircher, and P Muller (2019), “Providing Advice to Jobseekers at Low Cost: An Experimental Study on Online Advice”, The Review of Economic Studies 86 (4): 1411–1447.
Ben Dhia, A, B Crépon, E Mbih, L Paul-Delvaux, B Picard, and V Pons (2022), “Can a website bring unemployment down? Experimental evidence from France”, NBER Working Paper 29914.
Blundell, R., M Dias, C Meghir, and J Van Reenen, J. (2004), “Evaluating the employment impact of a mandatory job search program”, Journal of the European Economic Association 2(4): 569–606.
Card, D, J Kluve, and A Weber (2010), “Active labour market policy evaluations: A meta‐ analysis”, The Economic Journal 120(548): F452-F477.
Card, D, J Kluve, and A Weber (2018), “What Works? A Meta Analysis of Recent Active Labor Market Program Evaluations”, Journal of the European Economic Association 16(3): 894–931.
Cooper, M, and P Kuhn (2020), “Behavioral Job Search”, in Handbook of Labor, Human Resources and Population Economics, 1–22. Cham: Springer International Publishing.
Mortensen, D T (1970), "A Theory of Wage and Employment Dynamics", Microeconomic Foundations of Employment and Inflation Theory: 167–211.
Sachs, J and C Wyplosz (1986), “The economic consequences of President Mitterrand”, Economic Policy 1(2): 261–306.
1 See press coverage here and here.