Elevated morbidity and mortality among jobseekers may be partly explained by adiposity, but previous studies of unemployment and body mass index (BMI), which have usually modelled associations as linear, have produced inconsistent results. However, both underweight and obesity are associated with mortality, and both weight loss and weight gain associated with a stressful environment. If unemployment is associated with both underweight and obesity for different subgroups, these associations may previously have masked each other, whilst affecting health through divergent pathways. We investigated whether there is a previously overlooked U-shaped association of unemployment and BMI, which could help explain jobseekers' elevated morbidity and mortality, and identify groups vulnerable to underweight and obesity during unemployment. We used multinomial models to simultaneously investigate associations of unemployment with BMI-defined underweight, overweight, and obesity in 10,737 working-age UK adults from Understanding Society (UKHLS) in 2010–12. Moderating impacts of unemployment duration, demographic factors and smoking were explored. Current jobseekers were more likely to be underweight (Odds ratio (OR): 4.05, 95% confidence interval (CI): 2.12–7.73) and less likely to be overweight (OR: 0.71, CI: 0.55, 0.92) adjusted for gender, age, education, health, smoking and physical activity, while unemployed non-smokers had increased odds of obesity (OR: 1.52, CI: 1.06–2.18). Underweight and overweight associations were more apparent for longer-term jobseekers, men, and jobseekers from lower-income households. We conclude that unemployment is associated with underweight and, in nonsmokers, obesity. Results show the unemployment-adiposity relationship cannot be properly studied assuming unidirectionality of effects, and suggest unemployment may affect health of different groups via divergent adiposity-mediated pathways.