BACKGROUND: Lengthy separation from migrant parents raises concerns for the wellbeing of 60 million left-behind children (LBC) in rural China. This study aimed to investigate the effect of current and previous parental migration on the psychosocial wellbeing of left-behind children, and the mitigating or exacerbating factors in family care and support. METHODS: Children were recruited from schools by random stratified sampling in rural Zhejiang and Guizhou provinces where out-migration was high. A self-administered questionnaire measured children’s psychosocial wellbeing (Strengths and Difficulties Questionnaire), demographics, household characteristics, and family and social support. Multiple linear regression models examined the effects of parental migration and other factors on psychosocial difficulties. Informed consent was obtained from children and parents, and ethics approval from UCL Research Ethics Committee. FINDINGS: Data from 1930 current, 907 previous, and 701 never LBC were included (mean age 12.36, SD 2.08; 1696 boys, 1874 girls). Adjusted model showed both previous (B=0.75, 95% CI 0.26-1.24, p=0.003) and current (0.57, 0.10-1.04, p=0.017) parental migration was associated with significantly higher psychosocial difficulties score. Age, sex, household wealth, primary caregiver, and presence of siblings did not affect the outcome. Parental divorce (1.00, 0.48-1.54, p=0.0002) and lack of available support (3.03, 2.30-3.75, p<0.0001) demonstrated a strong association with greater total difficulties. While children in Guizhou had much worse (0.92, 0.52-1.33, p<0.0001) psychosocial outcomes than those in Zhejiang, adjusted subgroup analysis showed similar magnitude of between-province disparities, regardless of parental migration status. However, having divorced parents was a greater psychosocial risk factor for current (F=9.88, 95% CI 9.56-10.23, p=0.002) and previous LBC (5.58, 5.40-5.78, p=0.018) than for never LBC (0.03, 0.00-0.05, p=0.85); similarly, social support was more protective in current LBC (59.11, 57.21-61.21, p<0.0001) and previous LBC (16.01, 15.49-16.58, p<0.0001) than in never LBC (9.04, 8.75-9.36, p=0.003) INTERPRETATION: Parental migration is associated with independent, long-lasting adverse outcomes in children. Psychosocial wellbeing of LBC appears to be associated with relationships between nuclear family members and availability of support, rather than socioeconomic status. FUNDING: UK Economic and Social Research Council.