VoxEU Column Education Health Economics

Sleeping together: Network effects and sleep duration

Sleep is a key determinant of educational attainment among young adults, and carries with it longstanding health implications. This column provides evidence of network effects in adolescent sleeping decisions using a novel econometric approach. Young adults respond to the sleeping behaviour of their peer group, holding constant other observables. This compounding effect suggests a group approach to solving behavioural problems associated with sleep deprivation.

Nearly a third of a person's life is spent in slumber. In the US those with insomnia spend about $1 billion a year on prescription sleep aids, and another $1 billion on over-the-counter sleep medications (Yaniv 2004). The economic costs of sleep disorders in the US in 2004, both direct (expenditure within the health system) and indirect (absenteeism, low productivity, and work-related injuries), was estimated to be $109 billion (Hillman at el. 2006).

Yet sleeping behaviour has received little attention from economists. While sleep is primarily a function of the body's internal biological clock (the circadian rhythm), individual choice plays an important role in determining the timing and duration of sleep. Biddle and Hamermesh (1990) posit a simple economic model that accounts for the endogenous nature of sleep choice, but empirical work on the subject has been very limited.

In particular, there is virtually no evidence on the importance of social interactions in shaping sleeping behaviour. In many circumstances, the decision of agents to exert effort in some activity cannot be adequately explained by their personal characteristics and the intrinsic utility derived from the activity. Rather, its rationale may be found in how peers and others value the activity. There is indeed strong evidence that the behaviour of individual agents is affected by that of their peers.

Individual utility when allocating time to work or leisure may depend on the decisions made by peers. As a consequence, social interactions might be important for understanding the duration of sleep, which is the residual activity. Biddle and Hamermesh (1990) study the demand for sleep in this perspective without social incentives.

Data and estimation issues

We exploit the unique information contained in the National Longitudinal Survey of Adolescent Health (AddHealth) to provide evidence on sleeping patterns among adolescents in the US. Sleeping behaviour during teenage years is of particular interest because of its effect on human capital formation. Research suggests that lack of sleep reduces attendance, increases tardiness, and lowers grades of adolescent students (Eide and Showalter 2012). Furthermore, lack of sleep in youth is correlated with health and behavioural problems such as moodiness, depression, difficulty controlling behaviour, and increased frustration – all of which make learning in school difficult (National Sleep Foundation, Mitru et al. 2002).

The AddHealth data contain unique information on friendship relationships among a representative sample of students from US high school teenagers, together with basic information on individual, family, neighbourhood, and school characteristics (available in the in-school survey). The survey design also includes a questionnaire administered to a random sample of those students collecting information on more sensitive topics (such as health issues, crime, drug use, and sexual behaviour), including time and duration of sleep on week days during the school year (available in the in-home survey). The use of this additional information comes at a cost. The in-home sampling scheme may result in missing observations on the behaviour of friends who were not sampled, and induces measurement error to the endogenous peer effect variable given by the average behaviour of friends. As a result, the existing estimation methods for network models of social interactions (See, for example, Bramoullé et al. 2009, Lee et al. 2010)  are not generally valid with sampled observations on the response variable. In our study, we generalise the nonlinear least squares (NLS) estimator in Wang and Lee (2013) to estimate social network models. The proposed NLS estimator is consistent with sampled observations on the response variable.


Our results show that the sleeping behaviour of friends is important in shaping own sleeping behaviour, holding constant the impact of individual and friend characteristics. Unique information on siblings and their friends allows us to check the robustness of our evidence to unobserved family factors. In terms of magnitude, the effect of peers is non-negligible. In an average group of four people, an additional hour of sleep of each of the friends translates into an increase of about 45 minutes in an individual’s sleeping duration.

Adolescent sleep patterns deserve particular attention because of their potential to affect school performance. Side effects associated with sleep deprivation – inattention, irritability, hyperactivity, and impulse control problems – are likely to show up in school. It is important for educators to screen for sleep problems when concerns exist about a student's attention or behavioural problems. Our analysis suggests that an effective intervention must be measured not only by direct sleep disorder reduction, but also by group interactions.


Biddle, J E and D S Hamermesh, (1990) “Sleep and the allocation of time”, Journal of Political Economy 98: 922.943.

Bramoullé Y, H Djebbari, and B Fortin (2009), “Identification of peer effects through social networks”, Journal of Econometrics 150: 41.55.

Eide, E R and M H Showalter, (2012), “Sleep and student achievement”, Eastern Economic Journal 38: 512.524.

Hillman D, A Murphy, R Antic, and L Pezzullo, (2006), “Economic cost of sleep disorders”, Sleep 29: 299.305.

Lee L F, X Liu, and X Lin, (2010), “Specifcation and estimation of social interaction models with network structures”, The Econometrics Journal 13: 145.176.

Mitru G, D L Millrood, and J H Mateika, (2002), “The impact of sleep on learning and behavior in adolescents”, Teachers College Record 104: 704.726.

Wang, W. and L. F. Lee, (2013), “Estimation of spatial autoregressive models with randomly missing data in the dependent variable”, The Econometrics Journal 16: 73.102.

Yaniv, G (2004), “Insomnia, biological clock, and the bedtime decision: an economic perspective”, Health Economics 13: 1.8.

1,575 Reads