Tornado-level estimates of socioeconomic and demographic variables

Tornadoes create a threat to human life. Knowing the conditions that make people vulnerable to this threat is vitally important. Yet, socioeconomic and demographic data are not consistently available at the tornado level making it hard to obtain this knowledge. In response to this limitation, here a method to estimate socioeconomic and demographic variables in a consistent manner at the tornado level for historical events is implemented and assessed. The daysmetric method uses data from the 1990 and 2000 Census, as well as the 2010 American Community Survey together with tornado reports over the period 1995--2016. Results show that a typical casualty-producing tornado affects 34 people with an interquartile range between 4 and 198 people. Results also show that the 2 July 1997 Detroit, Michigan tornado with its 90 known injuries likely affected over 101,752 people. Comparisons between estimates using the actual path and a simplified modeled path show strong correspondence (percent errors averaging less than 10%) and estimates compare favorably (correlations exceeding .90) with known demographic numbers from a sample of tornadoes indicating the procedure provides useful information for statistical studies of tornado vulnerability. The code for making the estimates and the resulting numbers are available on GitHub.