Eating disorders are an important and growing health concern, and Bulimia Nervosa (BN) accounts for the largest fraction of eating disorders. BN is characterized by recurrent episodes of binge-eating followed by compensatory purging, and is especially serious given its increasingly compulsive nature. However, remarkably little is known about its addictive nature and the factors determining the incidence of BN. We use a unique panel data set to examine whether bulimic behavior satisfies the economic definition of addiction. Further, we examine whether BN is a rational addiction in the sense of Becker and Murphy (1988), where both the lead and lag of bulimic behaviors should help explain current behavior after controlling for observed and unobserved heterogeneity. We find there is substantial persistence in BN over time due to state dependence and that this result is very robust to different model specifications and identifying assumptions. Thus BN satisfies the economic definition of addictive behavior. Our results have important implications for public policy. They suggest that: i) treatment for BN should receive the same insurance coverage as treatment of other addictive behaviors such as alcoholism and drug addiction; ii) preventive educational programs that facilitate the detection of BN at an early age should be coupled with more intense (rehabilitation) treatment for individuals exhibiting addictive behaviors. Moreover, bulimic behavior also appears to be a rational addiction, although this result is more tentative as the data (and econometric theory) do not permit similar robustness checks. Finally, we find that bulimic behavior is decreasing in income and parent's education; moreover when race plays a role, African Americans are more likely to exhibit bulimic behavior. These results stand in stark contrast to the popular conceptions of who is most likely to experience BN, and we argue that this is a result of differences in diagnosis of BN across racial and income classes.