This paper extends prior research on the value to society of tornado warnings to examine the value of tornado watches and warning false alarms. False alarms are widely perceived to reduce response to warnings, yet little evidence exists to support the existence of a “false alarm problem.” This paper takes advantage of differences in false alarm ratios across the U.S. to test for a possible false alarm problem in a regression model of tornado casualties using tornadoes from 1986 to 2005. The paper also explores the value to society of tornado watches by testing whether tornadoes occurring within tornado or severe thunderstorm watch areas were less dangerous.