Body mass index and employment status: A new look

Highlights

• Obesity has a heterogeneous impact on various employment outcomes.

• Obesity increases the probability of not being able to work due to disability.

• The findings were supported by instrumental variable models.

Abstract

Earlier literature has usually modelled the impact of obesity on employment status as a binary choice (employed, yes/no). I provide new evidence on the impact of obesity on employment status by treating the dependent variable as a as a multinomial choice variable. Using data from a representative English survey, with measured height and weight on parents and children, I define employment status as one of four: working; looking for paid work; permanently not working due to disability; and, looking after home or family. I use a multinomial logit model controlling for a set of covariates. I also run instrumental variable models, instrumenting for Body Mass Index (BMI) based on genetic variation in weight. I find that BMI and obesity significantly increase the probability of “not working due to disability”. The results for the other employment outcomes are less clear. My findings also indicate that BMI affects employment through its effect on health. Factors other than health may be less important in explaining the impact of BMI/obesity on employment.

Introduction

Obesity has been argued to have an impact on labour market outcomes for the following reasons:

• Obesity is a condition that reduces ability to work through its impact on health, and is a central explanation for the increase in disability among those under the age of 50 in the U.S. (Lakdawalla et al., 2004, Kinge and Morris, 2010). It does so by being a risk factor for a wide number of deliberating diseases, including coronary heart disease, type II diabetes, hypertension and stroke (NHLBI Obesity Education Initiative, 1998).

• Obesity has an impact on certain characteristics that might reduce performance in the labour market, such as lower self-esteem, lower reservation wages and/or higher discount rates (Komlos et al., 2004, Offer, 2001).

• There may be discrimination against the obese due to their physical attributes (Agerström and Rooth, 2011, Finkelstein et al., 2007, Pingitore et al., 1994, Popovich et al., 1997, Rooth, 2009). As discussed by Morris (2007) there are be three reasons for this. First, prejudice by employers, reflecting their distaste for obese workers and the psychological costs incurred when dealing with them (Mclean and Moon, 1980). Second, there may be stereotyping by employers, arising from a belief that the obese are less productive (Everett, 1990). Third, discrimination may arise through uncertainty, or a lack of knowledge about the productivity of obese workers (Pagan and Davila, 1997).

Several studies looked at obesity and labour market outcomes including a special issue in this journal (Kelly, 2014) and they generally found poorer labour market outcomes in women, while the results for men were more mixed (Burkhauser and Cawley, 2008, Cawley, 2000, Cawley, 2004, Norton and Han, 2008, Sabia and Rees, 2012, Renna and Thakur, 2010, Averett and Korenman, 1996, Shimokawa, 2008, Han et al., 2009, Greve, 2008, Johansson et al., 2009, Sarlio-Lähteenkorva and Lahelma, 1999, Bozoyan and Wolbring, 2011, Averett et al., 2012, Asgeirsdottir, 2011, Garcia and Quintana-Domeque, 2007, Brunello and D’Hombres, 2007, Atella et al., 2008, Villar and Quintana-Domeque, 2009, Baum and Ford, 2004). However, it is intrinsically difficult to establish the impact of obesity on employment status due to classic endogeneity issues¹. Hence, earlier literature has used econometric techniques like instrumental variables, time-fixed effects and sibling-fixed effects. A common approach, which is also used in this study, has been to rely on genetic variation in weight as an instrument for BMI.

A notable literature has also used experimental designs, by e.g. sending out weight-manipulation photos or showing videos of applicants in different obesity categories, to study discrimination against the obese (Agerström and Rooth, 2011, Finkelstein et al., 2007, Pingitore et al., 1994, Popovich et al., 1997, Rooth, 2009). The obese were less likely to be called for interviews and seen as capable for the job, than the normal weight.

The aim of this paper is to examine the relationship between BMI and employment status. The likelihood of employment is particularly interesting, as holding a job is a prerequisite for general labour market performance like wages and on-job performance. Two studies have done this using UK data and IV-regression methods. Morris (2007) investigated the impact of obesity on employment (measured as a binary variable) and found a negative effect of obesity on employment in both men and women. Lindeboom et al. (2010) studied the impact of obesity on employment (measured as a binary variable) and found a negative impact of obesity on employment in both men and women. However, the IV-regression models, which used biological relatives BMI as instruments, did not support the findings.

One might organise employment status as a multinomial choice variable, where the effect of the obesity variable is allowed to differ for various employment outcomes (e.g. disabled, retired and employed) (Renna and Thakur, 2010, Terza, 2002). However, a very limited literature has used this approach to study obesity and employment outcomes. Such approaches can provide more complex results, which can answer why and how obesity affects employment status. This study makes three contributions to the literature. First, it estimates the impact of obesity on employment status as a multinomial dependent variable covering a more representative sample of the population than Renna and Thakur (2010) who use data from the US restricted to persons aged 50 or more. Second, it uses instrumental variables, based on genetic variation in weight, as a robustness control. Third, covering a period from 1998 to 2013, the data is more recent than earlier British studies. This is important as obesity rates have increased and the composition of the obese population has changed.