Obesity persistence and duration dependence: evidence from a cohort of US adults (1985-2010).

This study investigates dynamic patterns of obesity persistence and identifies the determinants of obesity-spell exits and re-entries. We utilize longitudinal data from the NLSY79 covering the period 1985-2010. Non-parametric techniques are applied to investigate the relationship between exit from obesity and spell duration. Multivariate discrete hazard models are also estimated, taking into account duration dependence and observed and time-invariant unobserved heterogeneity. In all cases, the probability of exiting obesity is inversely related to the duration of the obesity spell. Without controlling for unobserved heterogeneity, the probability of exit after one wave in obesity is 31.5 per cent; it is reduced to 3.8 per cent after seven or more waves. When time-invariant unobserved heterogeneity is taken into account, the estimated probabilities are slightly larger and broadly similar (36.8 and 10.3, respectively), which suggests that the identified negative duration dependence is not primarily due to composition effects. The obtained results indicate that public health interventions targeting the newly obese may be particularly effective at reducing incidence of long durations of obesity.