The Direct and Indirect Effects of Employment Decisions on Body Weight

This paper examines the direct and indirect effects of employment decisions on body weight. While the existing literature suggests that there is at least an associative relationship between employment decisions and body weight, the mechanism of that effect is not well understood. 

I develop and estimate a dynamic model of occupational choice; allocation of time to work, exercise, home production and leisure; and allocation of money to grocery food; restaurant food; and inedible goods.  An occupation is defined by the intensity of physical demands, social demands, and contextual stress.  Both the requirements/hours of the chosen occupation and the individual’s allocation of time and money are permitted to affect body weight.  The model captures three mechanisms by which employment decisions affect body weight.  The model captures the contemporaneous direct effect of employment decisions on body weight, i.e. burning more calories by working more hours at a more physically demanding job.  Second, by permitting the individual’s job requirements to affect the demand for exercise and food, the model captures whether the requirements of the chosen job lead to substitution of sedentary leisure for exercise or restaurant meals for food consumed at home.  Finally, the model captures how work related weight gain this period affects future health investment behaviors (e.g. exercise).  The model is estimated using Geocoded PSID data from 1999-2009; the O*NET (Occupational Information Network) on job requirements, Zip Code Business Patterns, and food price data from ACCRA.  As the set of occupations are defined over a continuous space, I estimate the individual’s choice of occupation using a continuous logit model from the spatial choice literature. This continuous logit is estimated jointly with a system of simultaneous equations for the money/time allocation decisions and a flexible error structure that allows for correlation in the permanent and time-varying heterogeneity that affects the outcomes in the model. Preliminary results indicate that a 1 standard deviation increase in job stress leads to a 20% decrease in exercise frequency and subsequent 0.08 annual increase in BMI.  Working ten additional hours per week leads to 10% substitution of restaurant meals for food consumed at home.  This, in turn, results in an annual increase of 0.05 BMI points.  While these single year effects are small, these results translate to a 10 pound weight gain over 10 years for a male of average height and weight.  Controlling for selection and unobserved heterogeneity, I find no significant direct effects of job requirements on body weight.