School hours and junk science

I am indifferent about the start time for high school. The children in our household have and will continue to succeed because they and their parents manage their time responsibility under any schedule.

I strongly object, however, to spending my money for social engineering of the start times based on junk science. The proposed change will cost a half million dollars – that’s $250 per high school student and the equivalent of the total annual income of four average Ridgefield households. Of course, only a government bureaucracy would come up with a plan to move exactly the same students from exactly the same addresses to exactly the same schools but at a cost an additional half-million dollars.

Professionally, I teach and advise researchers globally on efficient and effective experimental designs. The research offered to support the proposed time change might well be the case study for one of my favorite examination questions: “Identify ten fatal flaws in the following research study.” Here are a few of those flaws:

1. The Rhode Island study is totally irrelevant since this boarding school provided in loco parentis lights-out enforcement that could likely be the real reason for the improvement.
2. The Edina study made multiple changes at the same time. Not only were the start times changed, but there was also wide community education on sleep deprivation and schools promoted the notion. We have no idea which of the interventions produced any observed outcomes.
3. The Edina study moved the start time by 65 minutes, but Ridgefield proposes 35 minutes, so even if the study were otherwise valid, there is no evidence that 35 minutes would have any effect.
4. The reported improvements are selective: fewer accidents in Lexington, more hot breakfasts in Rhode Island, and fewer visits to the nurse in Edina. Unless they are all measured on the same metrics, it is entirely possible that the selected benefits are the result of random noise.
5. The data showing that teens have more accidents than older people in the category “fall asleep” proves absolutely nothing. Teens are higher in all categories of driver-related accidents.
6. Of all the reported “improvements,” there is nothing about the actual performance in school. Did test scores, for example, improve? The fact that educational performance is not cited, suggests it was not affected.
7. The case for sleep deprivation of teens is based on old national averages. We have no idea what the actual data are for Ridgefield in 2013.
8. The whole argument depends on the hypothesis that teens fall asleep late because their biology requires it. If that is true, then where is the evidence that teens go to bed earlier but simply lie there awake until 11:30? There is none. The critical piece of evidence is missing. Without any evidence, it is more plausible that teens go to bed late because they lack the discipline, incentive, or desire to go to bed timely.
9. The biological studies in combination imply that teens have no choice but to sleep from 11:30 pm to 8:45 am – hardly possible with the proposed 8:00 start time. If one believes the studies, why ignore the full implication?
10. The timing of melatonin release is based strictly on averages. What is the variation? Even if changing the time helped the “average” student, how many will be disadvantaged, and by how much. We have no idea.
11. The claim of different circadian rhythms for teens is suspect. (a) There is no evidence that the study desensitized the subjects to external cues. (b) There is no control or evaluation of entrainment factors. (c) The circadian conclusions fly in the face of two million years of evolutionary natural selection. Teens have been rising before the sun to hunt prey, do the farm chores, or open the family business far longer than they have been going to school. Those who didn’t would be selected out.
12. There is one and only one relevant example cited. Even it were an otherwise valid study, this is equivalent of saying that if a single coin toss were to result in heads then the coin must have two heads.
13.  The Edina study claims that students slept more because they said so. We are to believe this because some other study of sleep reporting in general said the method was accurate. But even if the general study were correct, the responses would likely be biased in a non-neutral setting where specific interventions had been taken and expectations were promoted for more sleep.
14. More generally, we do not have all the data needed to assess sample sizes, sampling variation, and non-response bias.

A shorter version of this piece appeared as a letter to the editors in the Dec. 5 Press.