Fear and bad statistics

Freakonomics has an interesting post on our irrational fear of strangers. Many Americans (especially my grandmother) think dying from a random stranger is common. In fact, by far most murders and rapes are done by someone the victim knows.

I recently looked up the statistics for child kidnapping during an argument with friends to find, as Stephen Dubner notes, that its mostly done by family members. In one year only 115 kidnappings, out of over 260,000, are the stereotypical kind of we see on TV.

The problem of trust has been on my mind recently, especially after an interesting discussion on NPR with Eric Weiner, who's new book, The Geography of Bliss, is now on my reading list. Political scientists have been tracking measures of trust in the US and are finding the numbers are decreasing lately.

During the interview, Weiner makes the strange comment that political scientists have a good way of determining trust: asking people "do you trust others?". This question is of course nonsense, and so I don't trust the responses at all. But how do you ask people about trust?

I am a very trusting person. I leave stuff around and have (almost) never been stolen from. I also try to exceed strangers trust of me. But, it isn't so clear cut. I'm writing this in a coffee shop now. If I had to go to the bathroom, I would ask the girl or two older people near me to watch my laptop. I would not ask the single guy that is talking to himself. I rely on heuristics of trust, not blind trust across all groups. If I was asked if I trust people, I would say "very much". If I was asked what percentage of people I trust, I would say 95%. If I was asked if I trust the mentally insane, I would say no.

Lets assume that my number of people I trust is very high, especially for the average American. If the average person trusts 50% of people around them (say of the same age, or ethnicity, or religion), what are they really saying when they respond to "do you trust people?" type questions? Are they thinking about it as a percentage of population, or as an even more vague heuristic? Either one could produce very different results, such as whether they saw a car chase on the news before the interview.