Does Obesity Affect Employment? Evidence from the NCDS

Abstract

Obesity has been associated with inferior labour market outcomes. Although obesity is endogenous to labour market outcomes, few studies have addressed this issue in a convincing manner, however. This paper uses data from the NCDS to estimate the causal effect of obesity on labour market outcomes. The NCDS offers several distinct advantages for our purposes. First, using the rich information on family background, parental inputs, and cognitive abilities, we are able to control for some of the unobserved heterogeneity that previous studies have suffered from. Second, the data records the body size of the mother, which we use as an instrument for own body size. Third, we utilize the panel feature of the NCDS in order to remove some of the time-invariant unobserved heterogeneity, for instance the existence of common or correlated genes determining both obesity and employment.

Simple OLS results show a sizable and significant correlation between obesity and employment at age 42 for females and a less strong but significant correlation for males. Including controls for potentially important factors, such as family background, parental inputs, and cognitive ability does not affect the association between obesity and employment. Moving on to IV-estimation, we note that our IV-strategy hinges on the assumption that the non-genetic factors that determine obesity are not the same as those that determine labour market success. Prior twin- and adoption studies support the assumption that the correlation in body weight between biological relatives is due to genetics, while environmental factors within the household play no role. We make additional checks of this assumption by (1) examining if the strong correlation in body size between biological relatives is affected when controlling for a rich source of environmental factors during childhood and adolescence and (2) whether the correlation in body size between the respondent and his/her mother is different for adopted than for natural children. The results provide suggestive evidence that the body size of a biological relative mainly predicts genetic variation in body size, making it potentially useful as an instrument.

Next, we perform IV-estimation separately by gender on the employment probability as a function of obesity and a wealth of other factors, potentially affecting labour market outcomes. Instrumenting for obesity renders the previously found significant correlation insignificant.

Finally, there still exists the possibility that the same genes that are related to obesity are also directly related to labour market outcomes, thereby posing a threat to our IV-strategy. We therefore utilise the panel feature of the NCDS and conduct analyses on first differences, in order to remove some of the unobserved heterogeneity. Finally, we instrument the change in obesity with the mother's obesity status in levels and, again, find no significant effect of obesity on employment.

Keywords: obesity, body mass index, health, labour, employment, endogeneity

JEL Classification: I12, J71