As depression is common in older people and confers a significant risk for dementia, its accurate assessment is essential to monitor and treat the condition. The 15-item Geriatric Depression Scale (GDS-15) is a widely used assessment tool for measuring depression in aged populations, and its psychometric properties have been recently improved using Rasch analysis. However, its temporal reliability and ability to distinguish between dynamic and enduring symptoms of depression have not been examined using the appropriate methodology. Distinguishing enduring aspects of depression helps to estimate the risks of depression and the long-lasting effects of an intervention while identifying and targeting dynamic symptoms may enhance the efficiency of a treatment. Generalizability theory (G-theory) is a suitable method to distinguish between enduring and dynamic symptoms of depression, evaluate the overall reliability of the GDS-15, and identify sources of measurement error. This study applied G-theory to the longitudinal GDS-15 data of 354 participants aged 70 years and older from the Sydney Memory and Ageing Study, collected over ten years with 2-4 years intervals. The GDS-15 demonstrated strong reliability and generalizability of its scores across the sample population and assessment occasions in measuring enduring symptoms of depression (Ga= 0.82, Gr=0.90). In addition, three dynamic symptoms of depression were identified, namely helplessness, hopelessness, and loss of interest in activities, which did not affect the overall strong reliability of the GDS-15. This study demonstrated that the GDS-15 is a reliable measure for assessing enduring symptoms of depression and can be used to evaluate the efficacy of depression treatments and monitor depression levels in older adults. Dynamic symptoms identified by this study are more amendable and can be targeted in the first place to enhance effectiveness of a treatment.