A central criticism of early childhood education programs is that their impacts on test scores strongly diminish or disappear shortly after the end of the program. This is called “fadeout”. In this paper I study the precise timing of the fadeout and its possible causes. I first study the period-by-period impact of the programs. Evidence from two experimental datasets shows that (i) there are strong gains on skills for the treatment group only two months after program entry; (ii) there are no additional gains after the first year of the program; and (iii) 83–99% of the fadeout is concentrated in the year of school entry. Fadeout at school entry can be explained by either (i) the loss of previous gains due to program exit or (ii) differential gains from attending subsequent schooling. Using a dataset of a third experimental program that counts with independent variation in early childhood education in two periods, I separately identify the program-exit effect, finding that it causes little or no fadeout in the short term. I conclude that the main explanation for fadeout is that the control group benefits more than the treatment group from the first year of school.