Obesity and Long-Term Earnings

Besides being a major public health issue both in the U.S. and elsewhere, obesity is associated with how individuals fare on the labour market. First, low wages may be a cause of obesity. This might be the case, if e.g. those who earn less cannot afford healthy food. Second, obesity may lead to lower wages, if those who have overweight suffer from employer discrimination, or have lower productivity at work. Third, obesity and low wages may covary due to an unobservable third factor that causes both of them.

A number of recent studies appear to suggest that obese women earn less. For men, the findings are more mixed. However, reflecting the difficulty of dealing with the endogeneity of obesity, many of the prior estimates should probably be interpreted as (partial) correlations (Cawley, 2004, J. of Human Resources). Only a few of the prior studies have been able to control for unobserved genetic differences and family effects. As far as we are aware, none of the prior studies have analyzed the effect of obesity on long-term earnings net of such effects.

We study the long-term earnings effects of obesity by using Finnish data on a large number of identical (monozygotic, MZ) and non-identical (dizygotic, DZ) twin pairs, matched to detailed register-based income data (Finnish Longitudinal Employer-Employee Data, 1990–2009). Our analysis differs from the prior studies in four ways: First, we analyze how obesity in early adulthood affects the long-term earnings during later adulthood. Second, we control for the unobserved genetic differences and family effects by using within-twin pair variation in the data. Third, we examine whether physical activity moderates the relation between obesity and earnings. It might, if there is a difference between being obese and having a high body mass index (BMI). Fourth, the longitudinal dimension of the data allows us to run within-twin pair IV-regressions that use a lagged measure of obesity as an instrument for the within-twin pair differences in the current obesity.

The initial sample of twins contains nearly all same-sex DZ and MZ twins of the Finnish population born before 1958. Our primary earnings measure is calculated over a twenty-year period and it covers the prime working age of the twins. The earnings data are taken directly from tax and other administrative registers, so it does not suffer from recall error, underreporting, or top coding. To ensure that the main dependent variable is measured over the individuals’ prime working age, only twin pairs born 1950–1957 are included to the final estimating sample. The data also include a number of control variables that allow us to condition the within-twin pair analyses on differences in socioeconomic characteristics, education level, health and health-related behavior and working conditions.

We aim at providing new information on the mechanisms that lead to an association between long-term earnings and obesity. This information is useful for policymakers evaluating various public health measures.