Background The Euro-D depression scale consists of symptom clusters that may be differentially related to demographic and cognitive characteristics in older adults. This hypothesis needs further investigation and the role of measurement bias on substantive conclusions remains to be established. Method The study sample comprised 10,405 community-dwelling older adults from six Latin American countries. We applied a Multiple Indicators Multiple Causes (MIMIC) model for a concurrent investigation of measurement bias and of the association between Euro-D symptom clusters and background variables. Results The factorial validity of Euro-D, with a two-dimensional structure - affective suffering and motivation disturbance, was consistently supported in all countries. Although complete measurement invariance could not be assumed across countries, measurement bias was minor. Both Euro-D factors were unrelated to age, but related to gender, as well as to impairment in memory and verbal fluency. Gender differences were larger for affective suffering than for motivation disturbance, whereas differences in verbal fluency impairment were more strongly related to motivation disturbance. Limitations Our analytic strategies could only examine invariance at the level of indicator thresholds. The generalisability of current findings needs to be examined in clinical populations. A wider set of cognitive tests is needed. We did not examine the compositional factors that could have accounted for the variation in Euro-D scores across countries, as this was beyond the aims of the paper. Conclusion The current study adds evidence for the construct validity of Euro-D and for the possible differential association of depression symptom-clusters with gender and verbal fluency in older adults. An understanding of the heterogeneity of late-life depression may carry clinical implications for the diagnosis and treatment of depression in old age.