The aim is to determine concurrent validity of a single self-report habitual physical activity (PA) question against accelerometer-based PA and mobility variables, and corresponding changes in self-reported PA and mobility. Cross-sectional and longitudinal data of the "Life-space mobility in old age" (LISPE) cohort and its substudy on PA were utilized. At baseline, 848 community-dwelling, 75- to 90-year-old people living independently in central Finland participated in home-based interviews. One and 2 years later, 816 and 761 of them were reassessed by phone, respectively. Tri-axial accelerometer data over 7 days were collected following the baseline assessments in a subsample of 174. Self-reported habitual PA was assessed based on intensity and duration using a single question with seven response options (range: mostly resting to competitive sports). Mobility variables were as follows: life-space mobility, walking difficulty over 500 m, and short physical performance battery. Statistically significant correlations were found between self-reported habitual PA and mobility (Spearman correlation coefficient Rs = 0.40-0.61) and accelerometer-based PA variables [step counts (Rs = 0.49), time in moderate (Rs = 0.49) and low intensity (Rs = 0.40) PA, and time in sedentary behavior (Rs = -0.28)]. A decline in self-reported habitual PA over time was associated with 5-10p decline in life-space mobility (PA improvement with 0-3p increase) and with developing a higher degree of walking difficulty (in 35-44% of participants). In conclusion, based on these results, the self-report question to assess habitual PA is valid and responsive to change and thus useful for epidemiological research in community-dwelling older people, also in follow-up studies.