Is obesity contagious? Social networks vs. environmental factors in the obesity epidemic☆,☆☆
Introduction
The United States has experienced a startling increase in average weight and in obesity over the past few decades (Flegal et al., 2002, Hedley et al., 2004). Though this phenomenon is by now well known and has been widely discussed and debated, there is still little consensus on its causes. One proposed explanation for the increase in obesity is long run technological changes that have impacted food prices as well as the propensity to exercise (Philipson and Posner, 2003, Cutler et al., 2003). Though some observers include genetic variation as a potential explanation for the rise of obesity because of the large estimates of heritability of obesity (Stunkard et al., 1990, Coady et al., 2002), most researchers acknowledge that genetic explanations are unlikely to explain the rapid increase in obesity over a relatively short period of time.
One particularly interesting hypothesis recently explored by Nicholas Christakis and James Fowler (henceforth, CF) in the New England Journal of Medicine is that obesity may spread through “social networks effects.”2 In fact, CF report that their findings suggest that social networks indeed facilitate the spread of obesity.3 This provocative finding was detailed in many media sources, including the front page of the New York Times.4 USA Today coverage indicated that “Obesity is contagious” and “…pick your friends carefully…” (Hellmich, 2007).5 CF suggest some potential mechanisms by which this may occur, including that having obese peers may change a person's tolerance for being obese or may influence weight-related behaviors such as eating habits, smoking, or exercise. Additional mechanisms suggested by CF include infectious causes of obesity or physiological imitation.
However, as is well known in the economics literature, there are alternative hypotheses that also potentially explain the empirical finding that friends’ weight is correlated across time that do not require the presence of social network effects. As CF identify in their study, there are at least three reasons why the weight status of individuals could be clustered within reference groups.6 The first is that individuals could choose their friends based on factors associated with weight or weight trajectories. In economics, this is typically referred to as selection (CF as homophily). Thus, friendship selection could directly lead to the correlation between friends’ weight or weight gain without an individual's weight causally affecting his friend's weight through a social network effect. Second, individuals may adjust behavior because of exposure to common influences. These effects are typically referred to as contextual influences (CF as confounding). For example, the opening of a fast food restaurant, convenience store, gym, etc. near a school could simultaneously affect the weight of all friends in a school's social network. Importantly, the presence of (often unmeasured) shared surroundings can lead to erroneously implicating social network effects in individual outcomes where none exist.7 Finally, individuals may alter their behavior as others in their group change theirs. Economists are now generally labeling this an endogenous social effect (CF as social network effects).
We point to three problems with the CF method. First, CF do not include a sufficiently broad set of contextual effects to account for a range of hypothesized causes of the epidemic. Second, the CF method of controlling for selection is much too narrow in scope. Third, the CF dynamic model as estimated produced coefficients with large degrees of bias (Liu et al., 2006).
Once the first two errors are corrected, evidence for endogenous causes of obesity is thin. We find that the CF results are not robust. In fact, the econometric evidence points strongly to shared environmental factors as the principle operative social mechanism underlying the positive correlation in weight status within reference groups.8 We find this remarkable given the preponderance of contexts in which endogenous effects appear present and the fact that this class of empirical models appears to generate the appearance of effects quite easily (Krauth, 2006). Our findings point to the difficulty in labeling the source of social effects, particularly in contexts with a direct policy reference. The public health implications given endogenous versus contextual drivers of obesity are quite different.
Section snippets
Data
We use the Add Health dataset to examine whether there are social network effects in weight outcomes for a national sample of adolescents who transition into early adulthood.9 Importantly, we have information on friends for approximately 5000 individuals, nearly 2000 of whom are followed over time along with at least one same-sex friend.10
Social network effects vs. shared experience
Central to our discussion is the distinction between endogenous effects, also labeled “induction” or social network effects by CF, and contextual effects. In the case of obesity, one can think of endogenous effects as describing the propensity to become obese because of the direct interaction with another individual. One may decide to eat more (or higher caloric foods) because their friend, spouse, neighbor does so. Because the two individuals are directly connected, they may influence each
CF specification and replication
CF use data on obesity status for an individual (in their terminology, an “Ego”) at a given point in time and estimate its relationship to the obesity status of a friend, spouse or relative (an “Alter”) as well as its relationship to the Ego's age, gender, educational level, and past obesity status. The CF specification uses the BMI of an Ego (i) who lives in community (c) at time (t + 1)14 as a
Conclusion
Our evaluation suggests that the spread of obesity is related to the environment in which individuals live. Though we do not completely rule out the possibility of induction and person-to-person spread of obesity, our results suggest that shared environmental factors can cause the appearance of social network effects. While comparing results across datasets that are quite different in design and focus is usually fraught with difficultly, we were encouraged to be able to closely replicate
Acknowledgements
The authors thank Elizabeth Bradley, Paul Cleary, John Mullahy, David Paltiel, and Jody Sindelar for very helpful comments and Jonathan Morse for research assistance. This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due to Ronald R.
References (18)
Neighborhood effects
Simulation based evidence of peer effects
Journal of Econometrics
(2006)Identification and estimation of econometric models with group interactions, contextual factors and fixed effects
Journal of Econometrics
(2007)- Bramoullé, Y., Djebbari, H., Fortin, B., 2007. Identification of Peer Effects through Social Networks. University of...
- et al.
The spread of obesity in a large social network over 32 years
The New England Journal of Medicine
(2007) - et al.
Genetic variability of adult body mass index: a longitudinal assessment in Framingham families
Obesity
(2002) - Cohen-Cole, Ethan, Fletcher, J.M., 2008. Is Obesity Contagious? Social Networks vs. Environmental Factors in the...
- et al.
Why have Americans become more obese?
Journal of Economic Perspectives
(2003) - Dembner A., 2007. Obesity spreads to friends, study concludes. Boston Globe; July...
Cited by (381)
Peer effects in weight-related behaviours of young people: A systematic literature review
2024, Economics and Human BiologyHigh on Bitcoin: Evidence of emotional contagion in the YouTube crypto influencer space
2023, Journal of Business ResearchGenetic nurture in educational attainment
2023, Economics and Human BiologyPeers, gender, and long-term depression
2022, European Economic Review
- ☆
The views in this paper are solely those of the authors and do not reflect official positions of the Federal Reserve Bank of Boston or the Federal Reserve System.
- ☆☆
Further exchanges between the authors may appear on their home pages. Jason Fletcher http://www.med.yale.edu/eph/faculty/fletcher.html; Ethan Cohen-Cole http://www.bos.frb.org/economic/econbios/cohen-cole.htm; James Fowler http://jhfowler.ucsd.edu/; Nicholas Christakis http://christakis.med.harvard.edu/.
- 1
Tel.: +1 617 973 3294; fax: +1 617 573 5494.