The instinct to "do something" if it "helps at all" is indicative of the difficulty of matching up solutions to problems in a quantitative way. The general public has no way to understand the difference between a policy that reduces a risk by 50% vs a policy that reduces a risk by 5%. Assigning those probabilities is speculative in the first place, and all intermediate probabilities (ones that are not "very low" or "very high") tend to be treated in functionally similar ways.
Anything in this intermediate region ends up getting classified as either "it never works" or "it always works" based on anecdotal proof/disproof, depending on personal biases.
In financial risk measurement, the inverse problem (choose a probability, and let that determine your policy) is omnipresent. In the 1990s, JP Morgan started using quantiles of profit&loss distributions as risk measure: you choose a "confidence level", say 95%, and a risk horizon, say one day, and ask your model for the amount of money that you are going to lose with less than 5% probability over one day.
All this may work just fine (with parameters as the above, the model can be back-tested against reality). The trouble starts when you start setting aside capital based on the absolute risk figures. What if you want to be "safer"? Keep the model, and amp up the confidence level (to 99.9%, say) and the risk horizon (to one year, say). Much of the turmoil in financial markets is due to people not understanding that this is a bad idea.
"Common sense" rhetoric is much too popular with politicians. Making laws is not a matter of common sense, and it's hard to predict what impact they will have. That goes to points 1 and 3, 2 as well I suppose.
The instinct to "do something" if it "helps at all" is indicative of the difficulty of matching up solutions to problems in a quantitative way. The general public has no way to understand the difference between a policy that reduces a risk by 50% vs a policy that reduces a risk by 5%. Assigning those probabilities is speculative in the first place, and all intermediate probabilities (ones that are not "very low" or "very high") tend to be treated in functionally similar ways.
Anything in this intermediate region ends up getting classified as either "it never works" or "it always works" based on anecdotal proof/disproof, depending on personal biases.
In financial risk measurement, the inverse problem (choose a probability, and let that determine your policy) is omnipresent. In the 1990s, JP Morgan started using quantiles of profit&loss distributions as risk measure: you choose a "confidence level", say 95%, and a risk horizon, say one day, and ask your model for the amount of money that you are going to lose with less than 5% probability over one day.
All this may work just fine (with parameters as the above, the model can be back-tested against reality). The trouble starts when you start setting aside capital based on the absolute risk figures. What if you want to be "safer"? Keep the model, and amp up the confidence level (to 99.9%, say) and the risk horizon (to one year, say). Much of the turmoil in financial markets is due to people not understanding that this is a bad idea.
It's "Christ or chaos." Looking on at the mess that was once the greatest civilization on earth, this binary rings true, it seems to me.
"Mental illness" is a secular world's all-encompassing catchphrase for what used to be called "evil" by our religious ancestors.
"Common sense" rhetoric is much too popular with politicians. Making laws is not a matter of common sense, and it's hard to predict what impact they will have. That goes to points 1 and 3, 2 as well I suppose.