How Does Time Use Affect the Likelihood of Becoming Obese?

We undertake a careful examination of the role of time use in obesity using the Eating and Health module of the American Time Use Survey (ATUS).  The ATUS provides a detailed look at how people spend their time over a 24-hour period.  But body mass is affected by behavior over a longer period of months or years.  Given the considerable day-to-day variation in time use, the time diaries can be viewed as measuring long-run time use with error.  Furthermore, behavior and obesity might be jointly determined, or obesity might influence behavior.

The mismatch between the time period of interest and the time period observed in the data has been underappreciated in previous research. This mismatch introduces bias even if researchers are uninterested in causal effects. For example, previous work has regressed BMI on time use variables in an attempt to estimate the association of time use on a typical day with BMI.  What they actually estimated was the association of time use yesterday with BMI today.

The good news is that both the endogeneity of time use and the measurement error due to the mismatch between time periods can be addressed using instrumental variables. The bad news is that finding valid instruments that predict long-run time use is extremely difficult. 

In the absence of traditional instruments, we address these issues using an approach developed by Lewbel (2011) that exploits heteroskedasticity in mismeasured or endogenous explanatory variables to construct instrumental variables. This estimator replaces traditional exclusion restrictions with assumptions about the covariance of certain variables with the error terms. Fortunately, these covariance assumptions can be tested using familiar first-stage F-statistics and tests of overidentifying restrictions.

The resulting estimates suggest that time spent exercising (defined as physically active leisure) reduces body mass and the probability of being obese for women; however, exercise does not lower the body mass of men, possibly due to changes in muscle mass. On the other hand, time spent walking or biking that is not leisure (e.g., commuting or walking a dog) reduces the body mass of both men and women. We also find some evidence that time spent in market work increases body mass for both genders, and time spent in “secondary eating” (grazing) results in lower body mass for women.