•Even in a highly resourced early childhood program, staff turnover was 30% annually.•Survival analysis allowed for quantifying the timing of turnover.•Predictors of staying were a positive work environment and teaching infants/toddlers.•Teaching staff with more years of education were more likely to leave.•Some predictors of staying were different for teachers compared to assistants/aides. This study of a large sample of classroom teaching staff at 23 early childhood schools across the U.S. serving children birth to age 5 used survival analysis to investigate both the timing of staff turnover and the characteristics associated with turnover. The data were collected from over 2,000 teachers, assistant teachers and aides between 2007 and 2019. Survival analysis allowed for investigation of both the timing of staff turnover and the time-varying and time-invariant characteristics associated with turnover. Over time, staff who were more positive about their work environment, had more years of teaching experience, taught in infant-toddler classrooms, or identified as Black were more likely to stay, whereas staff who had more years of education were more likely to leave. The probability of leaving was greatest in the first 2 to 3 years of employment. Some predictors of retention differed between lead teachers and assistants/aides, for example, the risk of leaving was greater for teachers with depressive symptoms but not for assistants/aides. While the results point to some areas where school leaders can effect practice changes to improve retention, the findings point to multiple areas where systemic and policy changes may be needed to reduce turnover.