The authors are grateful to Jean-Edouard Colliard (editor) and two anonymous referees for very insightful feedback, as well as to Francesco D’Acunto, Francesco D’Amuri, Laurent Bach, Paul Beaumont, Jean-François Bonnefon, Lisa Bruggen, Giuliano de Rossi, Andreas Hackethal, Michael Haliassos, Ulrich Hege, Fathi Jerfel, Luigi Guiso, Sophie Moinas, Mehrshad Motahari, Peter Pope, Sébastien Pouget, Sofia Ramos, Michael Reher, Alberto Rossi, Charline Uhr, and Wanyi Wang for very useful discussions. The authors also thank Corinne Laboureix, Catherine Leroy, and Fabien Thevenot for insightful explanations on the institutional setting and Sébastien Cadot, Arnaud Delavoet, Anthony Lapeyre, and Mohamed Youssef for help in the data-gathering process. The authors acknowledge funding from the Observatoire de l’epargne européenne; Milo Bianchi acknowledges funding from LTI@Unito and ANR (ANR-17-EURE-0010 grant). The data in this study come from the processing of records and account keeping of AMUNDI ESR employee and pension savings accounts, and they have been analyzed anonymously for scientific purposes. The authors have no material interest that relates to the research described in the paper. The computational reproducibility of the results has been verified by cascad (www.cascad.tech).