This article estimates the cost effectiveness of tornado shelters using the annual probability of a tornado and new data on fatalities per building struck by a tornado. This approach differs from recent estimates of the cost effectiveness of tornado shelters in Reference 1 that use historical casualties. Historical casualties combine both tornado risk and resident action. If residents of tornado-prone states take greater precautions, observed fatalities might not be much higher than in states with lower risk. Estimation using the tornado probability avoids this potential bias. Despite the very different method used, the estimates are 68 million US dollars in permanent homes and 6.0 million US dollars in mobile homes in Oklahoma using a 3% real discount rate, within about 10% of estimates based on historical fatalities. The findings suggest that shelters provide cost-effective protection for mobile homes in the most tornado-prone states but not for permanent homes.