When is safety a good idea?

Faced with that title, the experienced reader will immediately say 'it depends what you mean by safety'. Let's start with a seemingly innocuous proposition: 'if you can make something safer at no or little cost, you should'. One potential counterexample here is bicycle helmets. While the evidence is by no means definitive, it does suggest that making people wear helmets makes cycling less safe. What seems to happen is that people feel safer when wearing a helmet and hence take more risks. They are not that much safer, so the extra risks they take introduce more risk than the helmets remove. It may well be the case that helmets make bike accidents involving banging the head safer, but they make accidents in general more likely, it seems. In short requiring their use involves risk transformation rather than risk reduction.

Another good example is safety glass. Some forms of safety glass (as I have recently found out) don't shatter: a sheet is in fact composed of two pieces of glass bonded either side of a transparent but tough plastic film, so even if the individual glass sheets break, they remain bonded to the plastic. That's all very well if you want to avoid glass shards flying all over the place in the event of a breakage. But it does mean when you want to remove the glass you can't just cover the area with some material to pick up the shards then smash it: you actually have to cut the sheet out of the frame. This is a much more dangerous job than removing ordinary glass as a number of small cuts on my fingers testify. Very minor injuries aside, though, what is interesting is that something that was created as a safety feature actually turns out to increase risk in certain situations. A good question therefore in designing any kind of safety mechanism - in mechanisms, electronics, software, finance or regulation - is:

When might the behaviour we think is unsafe actually be useful? And what will happen then if the safety mechanism prevents it?

One might even argue that collateral is functioning that way at the moment. Clearly in ordinary conditions, calling collateral reduces risk. But if the failure to post collateral actually pushes your counterparty into default, as happened to Carlyle Capital Corp, it might not be such a good idea after all. And, as JPMorganChase recently pointed out, the banking system is currently facing a systemic margin call. Hmmm....