Suboptimal sleep durations and depression frequently cooccur. Short-sleep and long-sleep are commonly thought of as symptoms of depression, but a growing literature suggests that they may be prodromal. While each represents a process of mutual influence, the directionality between them remains unclear. Using polygenic scores (PGS), we investigate the prospective direction involved in suboptimal sleep durations and depression. Male and female participants, aged ≥50, were recruited from the English Longitudinal Study of Ageing (ELSA). PGS for sleep duration, short-sleep, and long-sleep were calculated using summary statistics data from the UK Biobank cohort. Sleep duration, categorised into short-sleep ("≤5 h"), optimal-sleep (">5 to <9 h"), and long-sleep ("≥9 h"), was measured at baseline and across an average 8-year follow-up. Subclinical depression (Centre for Epidemiological Studies Depression Scale [≥4 of 7]) was also ascertained at baseline and across an average 8-year follow-up. One standard deviation increase in PGS for short-sleep was associated with 14% higher odds of depression onset (95% CI = 1.03-1.25, p = 0.008). However, PGS for sleep duration (OR = 0.92, 95% CI = 0.84-1.00, p = 0.053) and long-sleep (OR = 0.97, 95% CI = 0.89-1.06, p = 0.544) were not associated with depression onset during follow-up. During the same period, PGS for depression was not associated with overall sleep duration, short-sleep, or long-sleep. Polygenic predisposition to short-sleep was associated with depression onset over an average 8-year period. However, polygenic predisposition to depression was not associated with overall sleep duration, short-sleep or long-sleep, suggesting different mechanisms underlie the relationship between depression and the subsequent onset of suboptimal sleep durations in older adults.