The general population prevalence and incidence of late-life agoraphobia was estimated and its clinical characteristics and risk factors described using data from the French ESPRIT study. One thousand nine hundred and sixty-eight persons aged 65 and above were randomly recruited from the electoral rolls of the district of Montpellier. Prevalent and incident agoraphobia diagnosed by a standardized psychiatric examination and validated by a clinical panel was assessed at base-line and over 4-year follow-up. The one-month prevalence of agoraphobia was estimated at 10.4% of whom 10.9% reported having the first-episode at age 65 or over. During the 4-year follow-up 11.2% of participants without agoraphobia at base line were classified as cases giving an incident rate of 32 per 1000 person-years. These 132 incident late-onset cases were associated with higher incident rates of anxiety disorders and suicidal ideation. Only two incident cases had past or concurrent panic attacks, which was not significantly different from non-cases. The principal base-line risk factors for incident cases derived from a multivariate model incorporating all significant risk factors were younger age of onset (OR = 0.94; 95% CI 0.90–0.99, P = 0.02), poorer visuospatial memory performance (OR = 1.60; 95% CI 1.02–2.49, P = 0.04), severe depression (OR = 2.62; 95% CI 1.34–5.10, P = 0.005) and trait anxiety (OR = 1.73; 95% CI 1.03–2.90, P = 0.04). No significant association was found with cardiac pathologies. We conclude that agoraphobia has high prevalence in the elderly and unlike younger cases, late-onset cases are not more common in women, and are not associated with panic attacks, suggesting a late-life subtype. Severe depression, trait anxiety and poor visuospatial memory are the principal risk factors for late-onset agoraphobia.