Observational evidence suggests that there is a possible bidirectional association between obesity and sleep. An important limitation of observational analysis is the ability to infer causation and overcome reverse causality. Mendelian randomization (MR) is a technique, which seeks to overcome this hurdle by using genetic variants identified from the Genome Wide Association Study (GWAS) literature as instruments for the exposure of interest and test whether there is a causal association with the outcome of interest. Here we implement a 2-stage MR to examine whether body mass index (BMI) is causally associated with sleep duration using pooled data, including individual-level data from the English Longitudinal Study of Ageing and Understanding Society. A fixed-effects meta-analysis of ELSA and USoc confirms the existence of a negative relationship, such that for every SD increase in BMI (5.08 kg/m2) we observed an SD decrease in sleep duration of 5.86 minutes. We then performed comprehensive causal analyses and thus, report on the largest Mendelian randomization study of BMI and self-reported sleep duration to date, in which we analysed data from 142,209 individuals using a combination of individual-level data and summary GWAS data.