Childhood socioeconomic status (SES) predicts executive function (EF), but fundamental aspects of this relation remain unknown: the developmental course of the SES disparity, its continued sensitivity to SES changes during that course, and the features of childhood experience responsible for the SES-EF relation. Regarding course, early disparities would be expected to grow during development if caused by accumulating stressors at a given constant level of SES. Alternatively, they would narrow if schooling partly compensates for the effects of earlier deprivation, allowing lower-SES children to "catch up". The potential for later childhood SES change to affect EF is also unknown. Regarding mediating factors, previous analyses produced mixed answers, possibly due to correlation amongst candidate mediators. We address these issues with measures of SES, working memory and planning, along with multiple candidate mediators, from the NICHD Study of Early Childcare (n = 1009). Early family income-to-needs and maternal education predicted planning by first grade, and income-to-needs predicted working memory performance at 54 months. Effects of early SES remained consistent through middle childhood, indicating that the relation between early indicators of SES and EF emerges in childhood and persists without narrowing or widening across early and middle childhood. Changes in family income-to-needs were associated with significant changes in planning and trend-level changes in working memory. Mediation analyses supported the role of early childhood home characteristics in explaining the association between SES and EF, while early childhood maternal sensitivity was specifically implicated in the association between maternal education and planning. Early emerging and persistent SES-related differences in EF, partially explained by characteristics of the home and family environment, are thus a potential source of socioeconomic disparities in achievement and health across development.