Séminaire COURNOT – Sylvie Blasco (Université du Maine, Le Mans)
De 14:00 à 15:30
Détails de l'événement :
« The effect of automation risk on mental health » écrit avec Julie Rochut (Cnav-URV) et Bénédicte Rouland (Université de Nantes, Lemna-Tepp).
Over the past decades, new technologies have profoundly changed the labour market and working conditions. Recent developments of artificial intelligence make a wider range of workers possibly exposed to automation and renew the debates about the “end of work”. The economic literature has extensively studied the consequences of automation on employment and labour demand, but there is still little evidence about the health consequences the risk of automation. The health effect of automation risk is theoretically undetermined. In this paper, we aim to empirically evaluate how the risk of automation impacts workers’ working conditions, well-being and mental health. We also aim to test the empirical relevance of theoretical mechanisms, including job insecurity and job intensity. To do so, we use the 2013 and 2016 French Working Conditions Surveys (Drees-Dares) that provide detailed information about working conditions, labour market history and health status for about 27,000 individuals, representative of the French working age population. We take various indicators of mental health, including depression and anxiety, as well as a more global measure of well-being, the WHO-5 score in a dichotomized version. As for the measure of automation risk, we adopt a task approach and classify as exposed to the risk of automation workers who have no interactions on the job, perform repetitive tasks, have to follow strict instructions on how to perform the required tasks and are closely monitored. As workers are not randomly exposed to working conditions nor randomly allocated to tasks and jobs, and as automation risk is correlated to other factors that may also affect health, we implement propensity score matching. This method is particularly well suited here, since the richness of our data helps us satisfying the Conditional Independent Assumption (CIA). We find that workers exposed to the risk of automation have a higher probability to declare anxiety or depression, and to have a low self-assessed well-being. Results are heterogenous with respect to some workers’ characteristics, including gender and age.