It has been argued that the temperature increase caused by anthropogenic climate change will produce a significant increase in violent crime. Support for that prediction is often based on statistical analyses of seasonal temperature and crime data cycles across days, months and quarters, and sometimes on large geographic areas. Within year temperature changes are very large, however, compared to the thirty-year temperature increases employed to measure climate change. In addition, because temperature trends associated with climate change vary geographically, analyses should employ small geographic units where temperature changes are measured over yearly intervals, and for long periods of time. To address these conditions, this study examined the long-term temperature-crime association for homicides in New York and London for the years 1895 through 2015. Consistent with previous studies examining seasonal weather and crime patterns, we found a positive correlation between annual homicide rates and temperature, but only at the bivariate level. This relationship became statistically insignificant in both New York and London when gross domestic product is controlled. Moreover, the bivariate relationship between temperature and homicide is statistically insignificant when correcting for non-stationarity. Thus, it does not appear that climate change has led to higher rates of homicide in New York and London over the long term. These non-findings are important as they suggest that studies of climate change and violence might do well to consider alternative mechanisms that mediate the relationship between climate change and violence.