woensdag 18 november 2015

Terrorists, traffic safety and the availability bias

The day after the terrorist attacks in Paris, the Belgian policy held a nation-wide campaign of speed controls. The campaign had been planned and announced well in advance, but was not cancelled following the attacks.
This led some moral outrage that was forcefully aired on the social media: “we are under attack and the Belgian police is checking whether we are complying with speed limits, where are these guys’ priorities?”
Well, the answer is less obvious and thus much more interesting than the author of these considerations may think.
Let us, for a start, assume that this person was not simply self-serving in his moral outrage, and is a law-abiding citizen who always complies with speed limits and who is genuinely worried about our safety.
Is he right that the Belgian police should have cancelled the planned speed controls?
The behavioural economist in me cries: 

“The reality is that, annually, 1.2 million people are killed in road traffic accidents worldwide, an average death rate of 17.3 per 100,000 population. http://healthintelligence.drupalgardens.com/content/road-traffic-death-rates-across-countries-world-2013 On the other hand, less than 33 000 people died from terrorist attacks in 2014 http://www.statista.com/statistics/202871/number-of-fatalities-by-terrorist-attacks-worldwide/ . Thus, this cry of outrage on the social media is a typical example of the so-called availability bias: the number of victims of road traffics may be around 360 times larger than the number of victims of terrorist attacks, but when people die from terrorist attacks, a lot of people die at the same time, so we have the impression that terrorism is a much larger threat than road traffic (especially when it happens in the city where we got engaged)”.


A smug smile runs across my face and I think: “mission accomplished, another beautiful example of deviation from rational thinking identified in the real world.”
Oh, but wait a second: who says that police resources should be allocated on the basis of the numbers of victims? A serious economic analysis also requires us to calculate how many deaths (from road accidents or terrorism) can be prevented by a given investment in police resources.
And this is where we run into trouble.
For traffic death prevention, the analysis is (in principle, if not in practice) quite simple: we “simply” need to know how much value the “speeder” attaches to driving faster than the legal limit, and we know that he will be deterred from speeding if the value of speeding is smaller than the expected value of a ticket for speeding. And this expected value is simply the product of the fine and the probability of getting caught, which is in turn a function of the resources invested in enforcing speed limits. Well, I know there are some details that still need to be developed (does the speeder’s assessment of the detection probability correspond to the objective probability, for instance?), but, in principle, all these issues can be dealt with if we invest the necessary resources in serious empirical research.
But what about terrorism prevention? Do we know how the probability of catching the terrorists before the act is affected by our investment in intelligence and police work? Well, I am not sure we even can. When security on planes increased, terrorists started attacking trains. If we increase security measures on trains, they will start attacking buses. If we increase security measures on buses, they will start attacking supermarkets. In other words, the more we protect specific targets, the more the terrorists will act at random – which is precisely what they have done in Paris. And if we can prevent them from undertaking random attacks in cities, they will move to random attacks in villages.
What is even more annoying with this type of terrorists is that it is not even clear what would deter them even if they were caught with certainty. After all, the only way to stop them is to kill them, and that is precisely what they want. And, because of the availability bias, it is not sure that the effect on public fear of a failed attack  is that much smaller than the effect of a successful attack.

So what does all this imply? Well, I am not so sure anymore. Maybe I should just take this smug smile from my face and start some real hard thinking about the subject. And, maybe, after all, this whole discussion isn’t worth having. After all, this guy on the social media is obviously a self-serving asshole who wants to drive any speed he wants without getting caught, and who is using the terrorists as cheap excuse. Is anyone suggesting that I am suffering from “representativeness” or “stereotyping” bias now?