A few days ago, a friend said she was worried that her daughter had been kidnapped by zombies because she hadn't texted her as promised. I replied that this was an example of base rate neglect; the chances of a girl forgetting to text her mum are much higher than of her being kidnapped by zombies, even in London.

And then Bryan Caplan says that parents in Virginia should not worry that their sons will be imprisoned for normal teenage behaviour because of the state's stupid laws because the chances of them being so are remote.

As statisticians, Bryan and I are bang right. As humans, however, we're not. It is perfectly reasonable for parents to worry about their children, even if the statistics say otherwise. It's easy to see how such a habit might have evolved. The stone-age mother who feared that every rustling of the trees hid a sabre-toothed tiger and who ran away with her child had children who grew to adulthood and who passed on risk-averse habits. The statistically literate mother who thought it was just the wind was right 99 times in 100, but her child was dead the 100^th and became an evolutionary dead-end.

This erring on the side of caution is not, however, a universal trait. Many safety-obsessed parents voted for a reckless gamble on Brexit. The point broadens. We have different attitudes to risk in different domains: some people who enjoy risky sports avoid financial risk and vice versa; I never go into a bookie's but have many times my salary invested in equities (at least in the winter); and many of us buy insurance but take gambles in other ways.

It's not clear that such differences can be explained by the conventional textbook theory of risk aversion -which, as Matthew Rabin and Richard Thaler have pointed out (pdf), doesn't make much sense anyway.

Instead, I suspect that what happens is that we code gains and losses differently in different ways; this, I think, is the big message of this recent paper (pdf) by Michael Woodford and colleagues.

In thinking about their kids, parents focus heavily upon worst-case outcomes even if these are statistically improbable: I'm not sure this is wholly due to the availability bias Bryan describes but to a more reasonable desire to keep them alive. But in thinking about economic policy choices, those same parents might think: what have we got to lose?

Similarly, I don't bet on horses because the coding doesn't seem right to me: so what if I occasionally win £100 or even £1000? Such sums won't change my life much but similar-sized losses would make me feel guilty (I haven't wholly escaped my Methodist upbringing). Betting on shares, however, codes better for me. This isn't because I believe there's a big equity premium, but because the reasonable chance of being able to retire early appeals whilst the downside - working longer - is just about tolerable.

These differences in coding mean that, contrary to conventional theory, money is not fungible. Our house being burgled is not just different in degree from losing £100 on the gee-gees, but is different in kind. (This inference is strengthened if we throw mental accounting into the mix, as we should).

Maybe attitudes to risk can't be reduced to simple maths but instead depend upon our imaginations, which vary from context to context.

What's more, in the real world it might be tricky to distinguish between perceptions of risk and tastes thereof. Is the mother worrying about her child really mistaken about numerical probabilities, or is she just very risk averse? What looks like a mistake might actually be a useful way of protecting the child.

I guess the point I'm groping towards is that this is one arena in which there is a gulf between conventional economics and real human beings - and perhaps the humans aren't wholly wrong.