I have sleep on the mind this week — and not just because I lost an hour of it this weekend to daylight saving time.
People in the developed world are sleeping less than they used to. In the US, one 2015 study found the age-adjusted mean sleep duration was 7.18 hours, down from 7.4 hours in 1985.
That’s why the premise of a new study in Nature Human Behavior is so intriguing. Published this week, the study argues that getting less sleep makes us less prosocial — less inclined to vote, to donate money to charity, to sign petitions.
I was intrigued by the study, but I was also suspicious. I’ve written about the replication crisis in psychology research, and this study sounded like the kind that might fail to pan out: a broad and striking claim, several different metrics of interest, even the fact that it’s in a prestigious journal (some analyses have found that the top journals publish more questionable studies than journals more specific to a field).
Also, the effect sizes are pretty large — often a sign that the results won’t hold up to more scrutiny. And there’s a long history of dubious studies claiming outsize effects on voting, caused by everything from television to menstruation.
So I spent some time digging into this sleep research. And, as it happens, it’s pretty solid.
A few lines of evidence suggest sleep matters — a lot
The researchers — political scientists John Holbein and Jerome Schafer and economist David Dickinson — start by observing that survey respondents who report less sleep are less likely to vote. That could easily be confounded by dozens of things — most obviously, people presumably sleep less because they’re busier, which gets in the way of voting but also most other activities.
The researchers try to control for this, but “controlling for” confounders doesn’t work as well as you’d expect, and the reported effect size is suspiciously large. While one sign of poor statistical methodology in a study is that all the findings are right beneath the threshold for statistical significance (that is, the effects the study found count as statistically significant, but only barely), the exact opposite problem — where studies claim astoundingly large effects from small changes — is also a sign of statistical malpractice.
I think of this as the “one weird trick” rule of psychology studies — if they claim that wearing the color red or watching a scary movie or ovulating explains 10 percent or more of the variation in how people vote, eat, or behave, then there’s probably something wrong there. Subconscious effects just shouldn’t be that large — and if they are, aren’t there 10 other equally random ones influencing the results?
To handle that, the researchers turn to a methodology that employs discontinuities in how well Americans sleep near time-zone boundaries. Regression discontinuity analysis can find causal relationships by analyzing how a variable — in this case, likelihood of voting — varies across a boundary where there’s a cutoff in how much people sleep.
It turns out that Americans living “near the immediate eastern side of the US time-zone boundaries have been shown to achieve a significant 20–25 minutes less sleep per night (on average) than those living near the immediate western side of the same time-zone boundary.” That’s a terrifying, startling statistic — does an arbitrary choice of time really affect us that profoundly?
One source for the finding is a different regression discontinuity analysis, published in 2017, which explains:
In counties lying on the eastern (right) side of a time zone boundary, sunset time occurs an hour later than in nearby counties on the opposite side of the boundary. More generally the onset of daylight is delayed by an hour. ... Because of the delayed onset of daylight and the biological link between environmental light and the production of melatonin throughout the day, individuals on the late sunset side of a time zone boundary will tend to go to bed at a later time.
(They speculate that primetime television airing times could be related too.)
While people go to bed later when there’s a later sunset and later sunrise, work and school hours don’t tend to accommodate this. That study found that the disparity amounts to 19 minutes difference in how much people sleep. (There was a bigger difference for people who had to start work before 7 am or drop their children off at school before 8 am; there was much less difference for anyone with a flexible schedule.)
That’s concerning in its own right, but it’s really useful for sleep research — it suggests that we can study the effects of sleep by taking advantage of the data from the quasi-experiment we’ve created with our time zones.
That’s how the new study examines whether the correlation between sleep deprivation and voting holds up. Their regression discontinuity analysis finds that, yep, the sleep-deprived are much less likely to vote.
Finally, the authors conducted a randomized controlled experiment, asking online survey respondents to complete an attitude survey either during the day or in the middle of the night. Again, the sleepy group was less likely to intend to vote, sign a petition, or donate money.
There’s one more obvious objection to the study. Maybe tired people just ... do less of everything, making it a bit misleading to highlight specifically that they do less prosocial activity. The paper considers that too, using the regression discontinuity approach to determine that people who sleep less have more leisure time and spend as much time relaxing, reading for fun, or doing home repairs and work than people who get more sleep.
There’s a variant of that criticism that I think should be taken more seriously: Can we be sure that what makes voting, petition signing, and donating money difficult when you’re sleep-deprived is specifically that they’re prosocial activities, rather than just that they require making significant decisions, which we’re less likely to do when we’re tired? The study doesn’t answer that.
This research fits into some previous work on the effects of exhaustion — I’m thinking particularly of the research into how tired people are less likely to make healthy lifestyle choices. There could be a common cause there. When we’re tired, it’s harder to make ourselves do things that we think are important but don’t find intrinsically motivating. So we’re less altruistic, and also less good to our future selves (which, at least for me, feels motivated by the same impulse as altruism).
Does this study have substantial takeaways? I think it has a couple. First, if you live near a time zone change, move to the western side — or at least take a job with a later start time — if you possibly have a choice. Research suggests it’ll significantly affect your sleep.
For the rest of us, before you make donation decisions, or before you try to do something difficult and altruistically motivated, it might be worth prioritizing a good night’s rest. If daylight saving hit you as hard as it hit me, this week is probably a bad time to consider how much you feel like budgeting for charity.
Sign up for the Future Perfect newsletter. Twice a week, you’ll get a roundup of ideas and solutions for tackling our biggest challenges: improving public health, decreasing human and animal suffering, easing catastrophic risks, and — to put it simply — getting better at doing good.