Only teenagers who start school late get the recommended eight hours sleep
Those who had to get up at the crack of dawn to be at school for 7am or 7.30am got 46 minutes less time in bed on average in the morning compared to those who started after 8.30am.
Pupils that started after 8.30am spent more time in bed sleeping between 27 to 57 minutes more in the morning compared to those with an earlier school start.
And a later start did not mean teenagers simply stayed up, the US study of urban teenagers found.
Associate professor of biobehavioural health Orfeu Buxton at Pennsylvania State University said: "Teens starting school at 8:30am or later were the only group with an average time in bed permitting eight hours of sleep, the minimum recommended by expert consensus
"Later school start times were associated with later wake times in our large, diverse sample.
"The presumption is if you let kids start school later they will simply go to sleep later and still not get enough sleep.
"But that's a hypothetical scenario. There wasn't data to back that up."
While those with an earlier start went to bed earlier, they still failed to get at least eight hours sleep.
Only those teenagers with schools that had a start time of 8:30am or later actually got the recommended amount of sleep.
Prof Buxton said one theory is that, despite going to bed earlier than their peers, those with an earlier start may not get enough shut eye because of the anxiety of an early alarm call the following morning.
Sleep deprived teenagers may also be sleeping in at the weekends to catch up on missed sleep leading them to wake up consistently and significantly later than those on school days.
This could misalign their body clocks interfering with having consistent sleep.
The study published in Sleep Health, the Journal of the National Sleep Foundation, followed 413 high school teenagers who completed sleep diaries.
Data collection included daily diary data from a subsample of the parent study, the Fragile Families and Child Wellbeing Study, which follows a longitudinal birth cohort of children born between 1998 and 2000 in 20 United States cities.