Abstract Background Time-varying phenotypes have been studied less frequently in the context of genome-wide analyses across ethnicities, particularly for mood disorders. This study uses genome-wide association studies of depressive symptoms in a longitudinal framework and across multiple ethnicities to find common variants for depressive symptoms. Ethnicity-specific GWAS for depressive symptoms were conducted using three approaches: a baseline measure, longitudinal measures averaged over time, and a repeated measures analysis. We then used meta-analysis to jointly analyze the results across ethnicities within the Multi-ethnic Study of Atherosclerosis (MESA, n???=???6,335), and then within ethnicity, across MESA and a sample from the Health and Retirement Study African- and European-Americans (HRS, n???=???10,163). Methods This study uses genome-wide association studies of depressive symptoms in a longitudinal framework and across multiple ethnicities to find common variants for depressive symptoms. Ethnicity-specific GWAS for depressive symptoms were conducted using three approaches: a baseline measure, longitudinal measures averaged over time, and a repeated measures analysis. We then used meta-analysis to jointly analyze the results across ethnicities within the Multi-ethnic Study of Atherosclerosis (MESA, n = 6,335), and then within ethnicity, across MESA and a sample from the Health and Retirement Study African- and European-Americans (HRS, n = 10,163). Results Several novel variants were identified at the genome-wide suggestive level (5??10???8?????p-value?????????5??10???6) in each ethnicity for each approach to analyzing depressive symptoms. The repeated measures analyses resulted in typically smaller p-values and an increase in the number of single-nucleotide polymorphisms (SNP) reaching genome-wide suggestive level. Conclusions For phenotypes that vary over time, the detection of genetic predictors may be enhanced by repeated measures analyses.