Abstract Aim To elucidate the association between the occurrence of falls and timed “up and go” (TUG) test score in a dual-task condition among community-dwelling older adults by age group. Methods This longitudinal observation study included 987 community-dwelling older adults at baseline. A TUG test (single-TUG) and a TUG test while counting aloud backward from 100 (dual-TUG) were conducted at baseline. The dual-task cost (DTC) value was computed from these results. Data on fall history were obtained using a self-administered questionnaire at the 1-year follow-up. At follow-up, 322 participants had dropped out and six participants had missing data for falls. The final analysis included 658 individuals (follow-up rate: 658/987, 67%) divided into a young-older adult group (aged 60–74 years) and an old-older adult group (aged 75 years or older). Associations between the occurrence of falls and TUG-related values were analyzed by age group using multivariate logistic regression models. Results For old-older adults, there were significant associations between the occurrence of falls and DTC value (odds ratio OR 0.981, 95% confidence interval CI: 0.963–0.999, p = 0.040) and single-TUG score (OR 1.129, 95% CI: 1.006–1.268, p = 0.039). However, no significant associations were observed for young-older adults. Conclusions Slower single-TUG test score and lower DTC value are associated with the occurrence of falls among old-older adults but not among young-older adults. Dual task assessment is useful for predicting falls in TUG fall assessment for old-older adults.