This paper investigates the relationship between local crime rates and the retirement decisions of older Americans. We do so by linking data from the Health and Retirement Study with measures of local crime patterns taken from the Federal Bureau of Investigation’s Unified Crime Reports. If we condition on crime rates alone, there is either a weakly positive or no relationship between local crime patterns and older men’s propensity to retire early. But unobservable factors associated with early retirement may be correlated with residence in higher-crime rate cities, so next we condition on both the expectation for the crime rate and deviations from average crime levels. We find a positive and statistically significant association between early retirement and expectations for murder rates, and a positive but, on average, imprecisely estimated positive association between early retirement and unexpected increases in crime. The effect of unanticipated increases in crime is greatest, and significant for those in poor health. In this latter group, men are 14 percent more likely to retire early given a standard deviation increase in unexpected murder rates. These findings are consistent with a pattern of more early retirement among those who live in higher crime areas, and earlier retirement among those in poor health when crime levels rise above anticipated levels.