To investigate the differences in self-efficacy (SE) for walking tasks between older patients with knee osteoarthritis (OA) and older adults without knee OA. A cross-sectional design was employed. Older patients with radiographic knee OA and community-dwelling older adults without knee OA as controls were enrolled in the study. SE for the walking task was assessed using the modified gait efficacy scale (mGES). A Wilcoxon rank-sum test was used to compare the mGES between the groups of participants. A Tobit regression model was used to estimate the difference in mGES. The presence of radiographic knee OA was used as an independent variable. Sex (women), age, and body mass index were used as potential confounding variables in the model. After exclusion, 78 participants (n=40 with knee OA, n=38 controls) were included. The mGES was lower in patients with knee OA than in controls. In the Tobit regression model adjusted for confounding factors, mGES in patients with knee OA was estimated to be 26.8 (95% confidence interval [CI]: 15.8-37.8) points lower than in controls. This study demonstrated that mGES was lower in older patients with knee OA than in older adults without knee OA.