Background. The increasing prevalence of obese Americans over the last several decades has been well documented. A number of studies have analyzed the relationship of obesity and mortality in community-dwelling elderly persons, but little work has analyzed this issue within the institutionalized elderly population. Methods. In an analysis of the 1996 Medical Expenditures Panel Study, we used logistic regression methods to examine the excess mortality associated with obesity, as defined by body mass index (BMI), over calendar year 1996 for existing and new nursing home residents. Results. Across the total sample of existing and new residents, there was not a statistically significant difference in mortality for "obese" (BMI > 28 kg/m2) nursing home residents (odds ratio [OR] 0.89; 95% confidence interval [CI], 0.67-1.17) compared to the "normal" group, but obesity was associated with significantly less mortality among existing residents (OR 0.75; 95% CI, 0.57-0.98). For "thin" (BMI < 19 kg/m2) nursing home residents, there was significantly higher mortality among both current residents (OR 1.40; 95% CI, 1.11-1.77) and new admissions (OR 1.63;95%CI, 1.17-2.28). For "very obese" (BMI > 35 kg/m2) individuals, there was a significantly higher mortality among new admissions (OR 1.75; 95% CI, 1.10-2.80), but not existing residents (OR 0.67; 95% CI, 0.38-1.15). These effects persisted for "very obese" individuals (BMI > 40 kg/m 2). Conclusions. Very obese nursing home residents experience higher mortality early in their stay, but this association diminishes over time with some evidence suggesting that a higher BMI may be protective among long-stay residents. Copyright 2005 by The Gerontological Society of America.