I was reading through Kip Viscusi’s Economics of Regulation and Antitrust last night and came across an interesting graph showing how our perceptions of risk don’t correlate well with reality—especially risk of dying from various causes.

I can’t find the graph online to post here, so a verbal breakdown will have to do.

• The public’s off-the-cuff estimated yearly deaths from low-probability events, like botulism tornadoes and floods, overshoot actual reported deaths from such events.
• The public’s off-the-cuff estimated yearly deaths from higher-probability events, like cancer, diabetes and stroke, undershoot actual reported deaths from such events.

At first thought, this made sense to me. Deaths by cancer and heart disease aren’t well-publicized, while death tolls from natural disasters are media favorites. I’d only expect people to overestimate the risk of dying from well-publicized events. But that raises the question of why accidents and natural disasters are publicized more often than deaths by disease.

Turns out Viscusi was one step ahead of me. Just after the graph, he points out that events are more feared by the public that more often result in relatively early deaths (i.e. natural disasters) than those resulting in death at older ages (i.e. heart disease). In other words, a longer estimated length of life lost is associated with higher perceived risk of dying.

Additionally, Viscusi cites another study showing that respondents’ answers to perceived risk are influenced by their age—young people over-estimate risk of dying from a tornado or flood and underestimate risk of dying from “old people” things, like heart disease. Vice versa for older people, I assume.

In other words, “world-revolves-around-me” attitudes extend even to our general estimates regarding risk of dying and give us inaccurate perceptions about risk. Since our perceptions of risk affect our preferences and behavior, perhaps causing us to spend extra for unnecessary safety features or not enough for better ones, being selfish can be harmful to your health!

But it’s not totally irrational to overestimate risk of dying from lower-probability events, like accidents and natural disasters. Viscusi writes:

Even if we are equipped with this knowledge of how people err in their risk beliefs, it is sometimes difficult to tell whether we are overreacting to seemingly small probabilities. In 2002 a sniper in Washington, D.C., created extraordinary public fears that led football teams to cancel outdoor practices and thousands to change their daily routines. The daily risks were seemingly small—surely under a one-in-a-million chance of being killed, based on the sniper’s past record. But what if one day the sniper decided to kill dozens of people, not just one? Then the odds would become much more threatening. The change that one would be shot by the sniper consequently was not well understood, since we have a small sample of observations that might not reflect accurately the extent of the risk.

I was excited to see this comment, as I was living in DC in 2002 and remember having soccer games and practices cancelled because of the sniper. It wasn’t so much about the odds, but about the unpredictability of the sniper—like Viscusi said, how could we have known he wouldn’t one day shoot dozens of people?

I think the same goes for natural disasters. They can be totally unpredictable and result in unexpected death. It’s entirely possible that tornadoes will kill 500 people tomorrow in Oklahoma, or that an earthquake will destroy large swaths of San Francisco. These events are unlikely, of course, but for no other reason that those things haven’t happened often in the past. We have no other reason than this. Diseases, on the other hand, progress slowly. We can take measures to dramatically reduce our likelihood of getting one. We can predict with more certainty whose susceptible and why.