Time use and productivity: The wage returns to sleep
BibTeX
@article{gibson2014time,
title={Time use and productivity: The wage returns to sleep},
author={Gibson, Matthew and Shrader, Jeffrey},
year={2014}
}
Abstract
While economists have long been interested in effects of health and human capital on productivity, less attention has been paid to the influence of time use. We investigate the productivity effects of the single largest use of time–sleep. Because sleep influences performance on memory and focus intensive tasks, it plausibly affects economic outcomes. We identify the effect of sleep on wages by exploiting the relationship between sunset time and sleep duration. Using a large, nationally representative set of time use diaries from the United States, we provide the first causal estimates of the impact of sleep on wages: a one-hour increase in long-run average sleep increases wages by 16%, equivalent to more than one year of schooling. We also document the nonlinearity of the sleep-wage relationship. Our results highlight the economic importance of sleep and pose potentially fruitful questions about the effects of time use on labor market outcomes. (JEL No. J22, J24, J31)
Notes and Excerpts
Using a plausible functional form assumption inan IV setting, we demonstrate that wage-optimizing sleep is approximately 9 hours pernight in the United States. This is higher than the average sleep amount reported in thedata—8.3 hours—and it is much higher than the 7 to 8 hours per night that the medicalliterature generally considers to result in lowest total mortality (Cappuccio et al., 2010)
Biddle Hamermesh 1990 might be interesting to take a look at. SUpposed to be simple model of time use between work, leisure, and sleep. Might be a good teaching example.
Combine time and expenditures to get units of leisure.
Sleep enters linearly into wage and also utility.
Consumption not in budget constraint. They just assume leisure has a cost.
Calculating sunset relies on four inputs: the date, latitude, longitude, and time zone. We can approximate three of these four components using annual average sunset time, solar declination (the angle of the sun relative to the equator) on the diary date, and the interaction of the two.
In a vacuum, we might expect residents of the eastern city to rise earlier. Hamermesh et al. (2008), however, show that work scheduling is not sensitive to solar time, and workers must wake up in time for a coordinated start—this is one reason for the widespread use of morning alarm clocks.
Indeed, state and local governments may petition the Department of Transportation to switch time zones, which has resulted in a long-run westward movement of boundaries (USNO, 2014). This movement means that the precise location of the boundary is endogenous and research designs based on comparing nearby communities on opposite sides of the boundary will be biased. Note, however, that the westward movement of boundaries is the opposite of what we expect if counties are choosing their time zone based on sleep-driven productivity considerations.
To assign locations to individuals in ATUS, we began by merging the ATUS datawith the corresponding CPS data (the match rate was 100%). For a given individual,the CPS data often contain location at the county level. This variable is censoredfor individuals living in counties with fewer than 100,000 residents. When county isavailable, we assign the county centroid as an individual’s location. We have countylocation for approximately 44% of ATUS observations and 42% of workers. For re-maining individuals, ATUS contains location at the state level. We assign the 2010population-weighted state centroid (computed by the Census) as the location for theseindividuals.
This is exactly the prediction of a sorting model like Roback (1982). With perfectworker and firm mobility, the gains from a productive location-specific amenity accrueto owners of land, the fixed factor. Such a model predicts that locations with earlierlocal sunset times will have higher rents and house prices, even without worker sortingon ability. Using county-level Census data from 2010, Table VI provides evidence thatthis is indeed so.