The objective of this study was to identify longitudinal predictors of depressive symptoms in autistic children and youth. Participants were youth with a diagnosis of autism who were part of the Province of Ontario Neurodevelopmental Disorders Network longitudinal substudy. Depressive symptoms were assessed using the child behaviour checklist (CBCL) affective problems subscale. Univariate and multivariable logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between clinical and demographic characteristics at baseline (T1) and clinically elevated depressive symptoms (CEDS) approximately 4 years later (T2). The mean age of participants (n = 75) at T1 was 9.8 years (SD = 2.7) and at T2 was 14.1 years (SD = 2.8). A total of 37% and 35% of participants had CEDS at T1 and T2, respectively. Additionally, 24% of participants had CEDS at both T1 and T2. T1 characteristics associated with T2 CEDS were: loneliness (OR = 3.0, 95% CI, 1.1 to 8.8), self-harm (OR = 4.0, 95% CI, 1.1 to 16.9), suicidal ideation (OR = 3.9, 95% CI, 1.0 to 16.5), more social and adaptive skills (OR = 0.3, 95% CI, 0.1 to 0.9), elevated restricted and repetitive behaviours (OR = 3.8, 95% CI, 1.3 to 11.6), psychotropic medication use (OR = 3.0, 95% CI, 1.1 to 8.4), attention-deficient/hyperactivity disorder (OR = 2.8, 95% CI, 1.1 to 7.8), and T1 CEDS (OR = 8.8, 95% CI, 3.1 to 27.0) (uncorrected for multiple comparisons). Associations persisted after adjusting for age and intelligence quotient (IQ) differences. Age, sex, IQ, teasing/bullying on the CBCL, family psychiatric history and family income were not associated with T2 CEDS. Our results highlight both high prevalence and high potential for the persistence of depressive symptoms in autism and emphasize the importance of early support to address loneliness and social participation.