Airport Security
Matthew Martin
6/10/2013 11:05:00 AM
Tweetable
The usual counterargument is that you can't do cost benefit analysis, because you can't put a value on a human life. This is all kinds of wrong. First, people put a value on human life all the time when, for example, they decide how much they're willing to pay for a safer car, bicycle helmet, medical prodedures, vaccines, seat belts, airbags, etc. There's even an excellent paper out now extrapolating the implied value of a statistical life from the willingness of fishermen to pay to upgrade their navigation systems for their fishing boats (this is a great paper since fishing is pretty dangerous, navigation systems have a large impact on fatalities, and weather data provides a convincingly exogenous instrument for riskiness). Point is, on a daily basis we make tons of financial decisions that shorten and lengthen our lives, which means that everyone is forced to put a dollar valuation on life.
But even if you accept the notion that we can't possibly know what the right dollar valuation of lives saved is (the benefit side of the cost-benefit equation), that doesn't mean we should ignore the cost-benefit analysis. Resources are scarce, and every x-ray we put in an airport means fewer x-rays in hospitals, every worker in the TSA means fewer paramedics, every dollar spent on airport security is a dollar that could have been spend on cancer screenings or HIV treatments. The government still must prove that more lives will be saved putting those resources into airport security than in the best alternatives.
I agree with Yglesias when he says that if we got rid of all that security, there'd be about as many airplane bombings as there are train bombings--which is to say, there is no measurable benefit to any of these security measures. And to the extent that airport security causes people to drive instead, we are actually increasing fatalities on two fronts: 1) driving instead of flying increases the person's own risk of fatality because driving is less safe, and 2) the increase in road traffic will increase existing drivers' individual likelihoods of a fatal car crash, since traffic volume is the biggest correlate with individual probability of fatal crashes.
So what do we know about the efficacy of airport security? One thing we know is that it basically never catches anyone. All of the airline terror plots since 9/11 have either successfully boarded the plane (ie, the underwear bomber and the shoe bomber), or been stopped prior to reaching the airport by undercover FBI agents (see list here).
Moreover, there isn't even a theoretical reason to suspect that the airport security would be able to identify a bomb. Take for example the full body scanners. What they do is detect liquids, and they do that extremely well. Many explosives are also liquids, so in a lab experiment, you can use the detector to find hidden explosives in a box of junk. But here's the problem: humans are just large sacks of liquids, so the full body detector gives a positive reading for the entire area a human body occupies. If you are smuggling a sack of explosives strapped to your belly, the scanner won't be able to discriminate between the explosives and you--the positive result it would have given anyway can't get any more positive. In actual trials, it was determined that the full body scanner would be unable to identify an explosive liquid most of the time. So naturally, they are now deployed in every airport. The TSA seems to feel that lots of individually ineffective proceedures somehow adds up to an effective system. Go figure.
On the otherhand, suppose that the combination of metal detectors and liquid detectors was successful at identifying all bombs that pass through. That still doesn't mean it is effective to implement those scanners--accurate scanners are meaningless unless you have enough computational power to process all of the scanner data. At present--as far as I can tell--the only computational power the TSA uses consists of low-paid, low-educated on-site TSA agents. Each agent sits infront of three large computer monitors watching the output data from the various scanners scroll by at high speed, expected to be able to spot something suspicious as it blips past the screen. At Dulles airport, the agent was gleefully spining in circles in his swivel chair while the x-rays of my luggage scrolled by on the screen. I could have gotten away with putting just about anything in there. And to put this agent's task into perspective, I find the comparison to medical radiologists to be informative: they recieve upwards of 10 years worth of intensive training, and $315,000 a year in salary to do basically the same thing--spotting anomolies in x-ray and other types of scans. But even if we hired radiologists to watch all the screens, the scanners would still be basically useless, because relative to even rarest cancer tumors, actual terrorist plots are extraordinarily uncommon. And the evidence shows that radiologists typically aren't able to spot unusual, incredibly low-probability anomolies. My point is that to effectively process all of the scanner data, the TSA needs a supercomputer smarter and more powerful than Watson at every airport in the country.
That said, there is one technology that studies show are quite effective at detecting bombs: dogs. Dogs handily outperform all scanners, in-person interviews, lie-detectors, pat-downs, and robots. Trained dogs are the only technology proven to reliably detect concealed explosive materials. And that's not all. Dogs also work fast--when I landed in Dulles they had both a bomb-sniffing beagle as well as metal detectors and full-body scanners--the beagle scanned the crowd in 5 minutes, the scanners took 30 minutes. Besides, the dog is much more aggreable to a crowd of weary travellers than having to strip down for the scanners and then be groped by TSA agents.