Most observational studies can’t separate correlation from causation, but one of my favorite Freakonomics episodes ever is about an observational study that used time zone boundaries as a natural experiment to estimate the causal effect of sleep on earnings.
First, they note prior research that found a negative correlation between sleep and wages attributed to a causal effect in the opposite direction: higher wages raise the opportunity cost of sleep, so high wages cause people to sleep less. Luckily, time zones provide a natural quasi-randomized experiment of sleep. When the sun goes down earlier, people tend to go to bed earlier, but they don’t go to school/work any earlier (or at least not by as much as they go to bed earlier). Within a given time zone, the sun sets earlier for people on the Eastern edge than those on the Western edge, so the Easterners end up getting a little more sleep.
What they find is that if a whole city starts sleeping an additional 1 hour per week, just 8.5 minutes per night, the average worker makes nearly 5% more money per hour. If the workers in their sample worked the same hours regardless of how much they slept, this would work out to an increase of $2,350 per year in trade for an extra 52 hours of sleeping. In effect, these workers would get paid $45 per hour for their extra sleep. The workers in these cities don’t actually work the same hours, though. Their extra hour of sleep trades off with about 45 minutes of work (and 15 minutes of leisure) and their annual income increases by about $1,570.
I’ll include two of the key figures here for your inspection. This first table shows that age, race, gender and education are all the same in cities with early or late sunsets, but that cities with later sunsets sleep 0.8 hours/week less and make $15 per week less.
This second table reports the coefficients from a regression of wages on the features reported above. With high confidence they show that a later sunset causes less sleep and lower earnings.
Some caveats are in order. Because the (quasi-)randomization happens at the level of cities and not individuals, this research still can’t quite say what would happen if an individual started sleeping more. It could be that if an entire city’s workforce is more productive prevailing wages will go up but if an individual worker is more productive they don’t get a raise.
How could you possibly make an entire state go to bed earlier? Well, California is voting on a proposition to enact permanent daylight saving time. If permanent DST were enacted (it still requires approval from the US Congress even if California voters approve it) it would result in the sun going down later for half of the year. Based on this research, that would actually cause California to lose significant productivity, even if we would all have fun with the later sunsets.
Getting paid more to stop working 45 minutes early and go to bed an hour early sounds like a pretty good tradeoff to me. In contrast to the terrible Wall Street advice that “money never sleeps” and the Glengarry Glen Ross advice, “always be closing,” the better advice is that money sleeps and if you want more of it you should always be sleeping.
How to 115 