It’s also critical to understand the datasource. If we’re computing average by crash data, I’m sure I'm not average, but also “know” more people who have died than I am aware of

If their data was from personal surveys, my answer might be 2. The real number is likely higher, and the average is not predictive