Research

Job loss may push workers into homelessness, and homelessness may keep them out of work. I provide the first causal evidence on both directions, quantifying the risk of homelessness upon job loss and how much shelter access mitigates the negative impact of homelessness on labor market trajectories. To do so, I combine the universe of shelter requests with administrative employment and unemployment insurance records in France. To identify each direction, I use a matched difference-in-differences design around economic layoffs and an instrumental variables design based on the leniency of call center agents allocating shelter. Losing a job raises the risk of homelessness by 50 percent. Getting a shelter bed at the first request cuts job loss by 26 percent for those still working and raises employment a year later by 58 percent for the unemployed. This evidence suggests that homelessness is an uninsured risk of unemployment for low-skilled workers. In turn, rapid shelter access can partly mitigate its negative consequences for employment.

We experimentally evaluate an at-scale expansion of public employment services in Addis Ababa, Ethiopia, that provided up-to-date vacancy information through local offices. Benefits were concentrated among relatively lower-skill women (LSW), the most marginalized job seekers before the expansion: offers and wage employment rose, work in household enterprises fell, and earnings increased by 24 percent. Impacts exceeded what the listed vacancies could explain; instead, LSW changed expectations and aspirations. Job match quality also increased. Notably, LSW job offers arrived through friends and family, not through the service directly – evidence that social networks can carry information to those hardest to reach.

Which soft skills both matter for finding a job and can be trained in adulthood? In baseline data on 52,352 French job seekers, we show that personality traits and job-search self-efficacy both predict job finding and earnings, and self-efficacy further predicts finding open-ended instead of fixed-term jobs. A randomized evaluation of a one-to-two-week soft-skills training run by the public employment service then shows that self-efficacy is malleable while traits are not: the program raises self-efficacy by 0.57 standard deviations and leaves the Big Five and locus of control unchanged. Participants shift from fixed-term to open-ended contracts, raising their employment rate in open-ended jobs by 8.6 percentage points over six months, on a base of 9.6 percent. Further analysis suggest that search effort and search criteria are not the main driver, in fact they are not affected, instead the increase in soft skills seems to lead to a higher success per application.