Schneier on Security
A blog covering security and security technology.
« Research in Explosive Detection |
| Election Fraud in Kentucky »
March 23, 2009
Fear and the Availability Heuristic
Psychology Today on fear and the availability heuristic:
We use the availability heuristic to estimate the frequency of specific events. For example, how often are people killed by mass murderers? Because higher frequency events are more likely to occur at any given moment, we also use the availability heuristic to estimate the probability that events will occur. For example, what is the probability that I will be killed by a mass murderer tomorrow?
We are especially reliant upon the availability heuristic when we do not have solid evidence from which to base our estimates. For example, what is the probability that the next plane you fly on will crash? The true probability of any particular plane crashing depends on a huge number of factors, most of which you're not aware of and/or don't have reliable data on. What type of plane is it? What time of day is the flight? What is the weather like? What is the safety history of this particular plane? When was the last time the plane was examined for problems? Who did the examination and how thorough was it? Who is flying the plane? How much sleep did they get last night? How old are they? Are they taking any medications? You get the idea.
The chances are excellent that you do not have access to all or even most of the information needed to make accurate estimates for just about anything. Indeed, you probably have little or no data from which to base your estimate. Well, that's not exactly true. In fact, there is one piece that evidence that you always have access to: your memory. Specifically, how easily can you recall previous incidents of the event in question? The easier time we have recalling prior incidents, the greater probability the event has of occurring -- at least as far as our minds are concerned. In a nutshell, this is the availability heuristic.
Although there are many problems associated with the availability heuristic, perhaps the most concerning one is that it often leads people to lose sight of life's real dangers. Psychologist Gerd Gigerenzer, for example, conducted a fascinating study that showed in the months following September 11, 2001, Americans were less likely to travel by air and more likely to instead travel by car. While it is understandable why Americans would have been fearful of air travel following the incredibly high profile attacks on New York and Washington, the unfortunate result is that Americans died on the highways at alarming rates following 9/11. This is because highway travel is far more dangerous than air travel. More than 40,000 Americans are killed every year on America's roads. Fewer than 1,000 people die in airplane accidents, and even fewer people are killed aboard commercial airlines.
Consider, for example, that the 2009 budget for homeland security (the folks that protect us from terrorists) will likely be about $50 billion. Don't get us wrong, we like the fact that people are trying to prevent terrorism, but even at its absolute worst, terrorists killed about 3,000 Americans in a single year. And less than 100 Americans are killed by terrorists in most years. By contrast, the budget for the National Highway Traffic Safety Administration (the folks who protect us on the road) is about $1 billion, even though more than 40,000 people will die this year on the nation's roads. In terms of dollars spent per fatality, we fund terrorism prevention at about $17,000,000/fatality (i.e., $50 billion/3,000 fatalities) and accident prevention at about $25,000/fatality (i.e., $1 billion/40,000 fatalities).
I've written about this sort of thing here.
Posted on March 23, 2009 at 12:31 PM
• 42 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
Not everyone stoped flying after 9-11 because they though flying was more dangerous, but rather putting up with the crap required before you are allowed to fly.
Perception of control Man.
Also, for 17 million per, I want sharks with laser beams on their heads
It makes no sense to compare funding based on the current levels of fatalities. What needs to be compared is fatalities prevented, fatalities which could be prevented given a larger budget, and excess fatalities which would result from a smaller budget.
I doubt this is true, but it's at least theoretically possible that the current DHS budget is preventing hundreds of thousands of terrorism deaths per year, whereas increasing the NHTSA budget would not significantly reduce road deaths. In that case, the current budget sizes would be best.
I personally think "homeland security" type stuff is highly over-funded, and mundane stuff like road safety is under-funded, but comparing the current death rates doesn't really help make the case.
Agreed with Anonymous. I stopped flying after 9-11 because it's just to much hassle putting up with all the security theater BS. Fear of flying has exactly zero to do with it.
Saying "40,000 people die on the road but 1,000 die on airplanes" is a terrible way to state relative risk. The complex of factors that go into determining risk (which is nicely referenced above) also apply to other methods of travel.
More people also traveled by passenger train after 9/11, which is much *safer* than air travel. It's enormously safer using the dumb comparison above (number of people killed regardless of context or usage), and it is quite a bit safer by more reasonable measures (for example passenger mile). It's even safer when ignoring the effects of crime (in general, air safety statistics ignore hijackings).
Both air and train travel are "safe" and you would have to be irrational to think air travel unacceptably risky. But it's very easy to deride irrational risk evaluation out of one side of one's mouth only to substitute another bunch of sloppy thinking to prove one's point.
Michael Ash is exactly right. we're basically talking about marginal cost per life saved versus average cost per life saved. And the numbers here make sense, more or less:
There is a massive fixed cost associated with running a military - infrastructure, training, force readiness, etc. There is very little fixed cost associated with wunning the NHTSA, which consists largely of safety studies and public awareness campaigns. So it makes sense that, even if we place equal monetary value on a life saved on the highway and a life saved from a terrorist, the AVERAGE cost of saving a single life from a terrorist would be higher.
To play the devil's advocate, auto fatalities have negligible economic impact beyond the people close to the deceased. Money spent on preventing another 9/11 could be seen as spent on avoiding the economic repercussions of such an event.
I harp about this sort of thing a lot as well. I do have to say, though, that 'More than 40,000 Americans are killed every year on America's roads", but "Fewer than 1,000 people die in airplane accidents" only supports the author's reasoning if the precentages fit. For instance, if 40 million people drive on America's roads every day but only 1,000 Americans fly total, then this supports an entirely different conclusion :-)
@ Michael Kirkland: "auto fatalities have negligible economic impact" unless you count the rippling of costs to insurance companies that are passed on to all consumers in the form of premium rate hikes, not to mention the costs of police and emergency response and court processes for either criminal or civil litigation. All this multiplied by the thousands of motorist injuries and deaths every year can add up to quite a bit. The kicker is that these costs reliably continue to mount, whereas Bin Laden only got in one good rib kick before he was chased off.
You also have to factor in the outrage factor that comes from a terrorist action. There are far more negatives inflicted than just the loss of life. Don't underestimate the value of a society that feels secure, even if its a delusional feeling.
I think the costs/benefits of airline security are tough to measure. Up front, I'll disclose it is my belief too much is spent for not enough value. Having said there, I'd like to make the following contradictory considerations...
We all know that a great many people don't fly because of the headaches of security. I still fly and haven't found it to be so unbearable that I won't (certainly, the inconvenience of liquids and taking off my shoes hasn't been such a horror that I'd spend an extra 12 hours in a car to avoid it). There is also a converse group of travelers--the ones where security theater has reassured them that it is safer to fly, the travelers where the overkill has brought their fears more in line with the actual risk, which is small.
I'm not saying it is good to overspend on security, but I am saying the consumer confidence is important as well (as is not needlessly harrassing people).
I also do not think the car vs plane analogy is completely fair. 40 times as many people may die in car crashes, but instinct tells me that most people spend 40 times more time in a car than a plane. However, while the percentages per travel time may not be as off, the fact remains that since more people drive than fly that the need for heightened travel safety would skew towards cars.
Maybe all speed limits should be cut in half to reduce wrecks and the likelihood of fatality in the event of a wreck. (Okay, that part was a joke.)
I'm with Anonymous -- I fly less not because I'm afraid of flying but because it's now a huge pain in the rear.
Regarding flying vs. driving, I think one of the factors is that when flying, you're placing your safety in the hands of others while when driving, you at least have some modicum of control, so even if it is quantitatively more dangerous, one tends to feel safer. Yes, it is entirely a matter of perception.
@Steve: "Regarding flying vs. driving, I think one of the factors is that when flying, you're placing your safety in the hands of others while when driving, you at least have some modicum of control, so even if it is quantitatively more dangerous, one tends to feel safer. Yes, it is entirely a matter of perception."
Fair point. Not to mention that if the car breaks down or there is a colission, you're not five miles above ground in a car.
"More people also traveled by passenger train after 9/11, which is much *safer* than air travel."
Passenger trains are actually roughly as safe as airlines, and while both are extremely safe when compared to something like highway travel, the long-term average has roughly a factor of 3 advantage to airlines. Looking only at the United States, in the period from 1990 to 2006, the airline fatality rate was 0.2 deaths per passenger-mile, whereas the rate on heavy rail in that same period was 0.7.
Trains look safe because they're not very popular. In the period in question, airlines carried 7.8 trillion passenger-miles, but trains only carried about 200 billion, a factor of 40 lower. As such, accordingly fewer people died in train accidents, even though the rate was quite a bit higher.
Isn't all this really just adding window dressing to what actuaries do?
in case you haven't noticed, most people are really dumb. as carlin said, think about how stupid the average person is, then remember that half of them are stupider than that. additionally, i heard a neurology lecture in which the guy said "if it happened more than three months ago, it doesn't matter." i think about this whenever gas prices decline temporarily, then start to rise a few months later, and everyone is shocked! "do you think gas is going to get expensive again?" "gosh, i don't know--there's less of it every year--what do you think?" or when everyone was in a mad rush to buy a house two years ago, as if the market would just go up and up forever. so you can't really expect 90% of the population to EVER comprehend something like the availability heuristic, even if they could say those two words together without stumbling over them.
apart from that, i prefer driving to flying for several reasons:
1. i have a much more active role in a car than a commercial aircraft, and the feeling of control is comforting, whether or not it would save my life.
2. i would much prefer to die instantly in a car crash, than by falling out of the sky, spending a minute or two wondering what it's going to feel like on impact.
3. i can bring toothpaste and hair gel in the car, and even a nail clipper, if i'm feeling reckless.
Availability is probably a reasonable heuristics when dealing with things that happen frequently enough you can expect to have encountered them several times in your life. That the heuristic breaks so badly for the threats relevant to our every day life is partly a testament to how wonderfully safe we've managed to make our environment.
Just a nitpick: if you die in a plane crash, you won't be spending a minute or two falling out of the sky wondering what it's going to feel like on impact. Time of useful consciousness at typical airliner altitudes is about 15 seconds, so without oxygen or pressurization it's lights out real quick. Additionally, most airliner accidents happen with perfectly intact airframes hitting the ground, so again, it's a sudden thing.
That said, I prefer to drive as well, and only fly when the time saved is significant. But it's all about your #3 for me, it's convenience, not safety or even the comforting illusion thereof.
I agree with other posters above, this is highly dubious. What's more, there is a huge difference between motorists and airline passengers - motorists drive themselves! They're not simply at the mercy of barricades and signs and pavement surfaces, they're behind the wheel and about 99.9% responsible for what happens - even if the car breaks down. It's a totally different deal in the air.
Are you possibly trying to relate things that can't be related, or shouldn't be correlated? If shoplifting at the mall jumped 200% the summer I moved into town, should they pay me to move out?
I see a couple of posters think that there is a difference between car accidents and plane accidents because motorist drive themselves. Being in a car, you have less control of the situation than you think.
Here are some things to think about.
Not everyone in a car is behind the wheel. Everyone thinks they drive better than the next guy. Drivers tend to be tired, stressed, distracted, emotional, drunk, or drugged. The structure of the road system causes cars to be more densely packed in cities and highways. Greater density means more possibilities of accidents. Weather conditions also affect car traveling which is compounded by the human condition. Drivers do not have much control when the collisions are from the side or rear.
To sum up, you have no control of the weather conditions of the road, no control of the other drivers on the road, very little control over the passengers in your own vehicle and you may have less control of yourself than you think.
I'd wager that the chances of being killed by another bad driver on the highway through no fault of your own are *still* significantly higher than the chances of being killed in an airliner accident. A good driver will have a lower chance of death than a bad driver, but a good driver can still be easily taken out by a drink, an cell-phone chatterbox, a tractor trailer with bad brakes, and many others. Being in control may make you feel better but it won't get you anywhere near the safety level of an airliner.
"the airline fatality rate was 0.2 deaths per passenger-mile, whereas the rate on heavy rail in that same period was 0.7."
In that case, I think I'll stick to single-passenger planes and keep my trips under 5 miles.
I'm pretty sure I meant to write "per million passenger-miles" there.
I would not rush conclusions by a dollar to dollar comparison.
I mean, one dollar spent for cause A can have a good impact, while the same dollar spent for cause B can have an infinitely smaller impact... then what would you do, don't even try cause B?
e.g. would you shout down research on rare diseases because the same dollar saves less lifes?
Focusing only on best money return would mean in many cases, as in the example, the loss of a lot of opportunities to learn and improve.
Moreover, "results" are not instantaneous, a dollar spent today in x can give results, while is spent in y can give results only in 10-100 years, what would you do? Shut down everithing that doesn't give immediate return (e.g. basic research)?
And again, results does not nicely grow in a linear trend with expeses.
Let me say again, investiment's retur is rarely linear.
If you double the investiment in different things you will rarely double the return, in some cases you will have marginal improvements, in other you will have huge improvements.
Bottom line: don't make the money the absolute meter of measure, expecially when comparing apples and oranges.
I refused to stop flying after 9/11 or change my actions in any way. I contnued using the London Underground system after the 7/7 bombings too.
If you let terrorists change your behaviour as a result of their efforts, then they are winning. I refuse to bow down.
Bruce: "In terms of dollars spent per fatality, we fund terrorism prevention at about $17,000,000/fatality (i.e., $50 billion/3,000 fatalities)"
_We_ don't fund terrorism prevention. It is _them_ who take our money under the false pretenses and wasting it on their own salaries and maintenance of their little bureaucratic empires.
The War On Terror makes sense only as a massive scam.
The real psychology at play there is not "availability heuristic" but rather Stockholm syndrome widespread in the populace which is bossed around by the "servants of the people".
Let's forget the administration "Terrorism Prevention" & "Highway Traffic Safety" for a while, and turn the problem in another way.
Let's say i just know that one administration spend $50 billion for protecting people from a threat, and another administration spend $1 billion for protecting them from another threat.
The first one manage to protect all but ~100 americans/year (on the average), and the second one manage to protect all but 40,000 americans/year.
Maybe without the $50 billion funding allocated to the first administration, the threat it is protecting people from would hurt/kill more that 40,000 people/year. What you think?
ps1 : Please do not associate "first administration" and "Homeland Security", because it could twist your judgment.
ps2: I know i'm playing the devil's advocate, but - according ps1 - tell me where the mistake in my reasoning is
Foster's calculation in the third quoted paragraph is completely muddled. For one thing, he is comparing costs per (very conservatively estimated) death prevented to costs per death not prevented, which makes no sense at all; under this comparison, the more road deaths there are, the more effective NHTSA appears to be!
For another, his estimate of terrorism deaths prevented, whilst superficially very conservative, assumes that no terrorism deaths at all were prevented prior to the formation of the DHS. This is quite unsound as in fact most of DHS's security components are re-badged agencies that existed long before 9/11. It is not conservative, but simply muddled, to assume that they are 100% effective now and 0% effective before they were rebadged. Unfortunately his calculation falls apart if we take this into account: if we admit that any US agencies might have ever prevented any terrorist attacks, whether actively of passively, then it becomes impossible to make an estimate of number of deaths prevented per year.
Additionally, he compares NHSTA's budget to the entire DHS budget, and describes this as the budget for "preventing terrorism". In fact DHS is a mega-department with very broad functions, most of which have nothing to do with prevention of terrorism. Actually due to the way DHS is organised, it is quite difficult to assess the fraction of its budget that is spent on terrorism prevention, but it is probably quite small. For example, consider the U.S. Customs and Border Protection Agency (one fairly large component of DHS, budget about $8.8 billion), which was formed from the former US Customs Service and parts of the INS. It has -- like many DHS agencies -- terrorism prevention as a primary goal in its mission statement. However this sort of airy persiflage has little to do with reality and in fact the vast majority of CBP operations are concerned with illegal narcotics, illegal immigration, and agricultural protection. Looking at their detailed budget proposed for FY09, I can only identify $6 million (0.07%) as specifically for anti-terrorism measures.
I just wanted to add to Bruce's explanation of the availability heuristic that it is probably not just the last incident that is remembered but the number of deaths associated with that type of incident.
In other words, there would need to be 1000 car crashes with an average of 3 people killed in the memory of a person to equal one 9/11 scale disaster. I don't think I have seen or even watched on the news 1000 car crashes in my entire life, let alone 1000 crashes that involved 3 fatalities since 9/11. If I have, I certainly don't remember them which is what is important.
Most car crashes, even those involving fatalities are not reported individually in the media. Often we will only get statistics at the end of a holiday season on how many people died on the roads during that time. Often this is accompanied by a percentage rise or fall from last year's statistics. These sorts of reports lack the emotional impact to stick in your memory.
The fact that there is video footage of 9/11 and that it has been shown and talked about constantly for more than 7 years now means that it is still available in people's minds whereas the few car crashes that made it into people's memories in that time have mostly faded. Only the most recent memories are available.
Regarding the old argument between plane and car crashes I read somewhere (maybe even here, but probably on the Freakonomics blog) that someone ran the numbers and determined that while the per mile deaths for air travel was far far lower than driving, the per minute death rate were remarkably similar.
Now the numbers might be out of date as safety has gotten remarkably better in the last decade, but it goes to show that human intuition shouldn't be so casually thrust aside. Yes if you drive from NY to LA you have a much higher risk of dying than flying in a plane, but the car trip takes 45 hours of driving time and the plane trip takes about 6 hours of flight time. Just off the bat is 7 times more time that something could happen to you in your car.
It's the difference between running through Baghdad and taking a casual stroll and then proclaiming running to be a safer activity on a per mile basis because those who run get shot less.
So why do I bring this up? Because when one is trying to measure fear and security you have to measure the right thing. The common rate metrics like per mile and per 100,000 population often miss the target when trying to describe what constitutes safety.
The problem with the argument that anti-terrorist spending might have saved hundreds of thousands of lives is that there's darn little evidence of that, and what evidence there is (for or against) tends to be classified and therefore unavailable. I can make up scenarios in which any particular government spending saves many lives, and some of them will even be plausible, as long as I don't have to provide evidence.
So, to assume that the current level of anti-terrorist spending is vaguely justified, we need to have some level of trust in the people who have what information that is. People who have instituted invasive security measures that clearly aren't going to help, who have been known to ban fingernail clippers and not glass bottles, who cause massive and vulnerable crowds at checkpoints, who have created a "no-fly" list of those who aren't guilty enough to try but aren't considered innocent enough to have full rights (and there was one report that real suspects might not be on the no-fly list, lest that tip them off to the fact that they're watched).
This is a case where the government, to keep credibility, needs to not only do things right but look like it's doing things right. That isn't happening.
This is government spending and one has to think on a government/political level, not just dollars per fatality.
40,000 deaths on the highways in a year is tragic. Each of those deaths rips to the heart of every family involved. However, few will change their political inclinations over those deaths.
The U.S. government treats fatalities under the watch of the DHS differently because there is more of a chance that they could effect significant change in the entire nation's view of government.
No one has mentioned the reason I drive rather than fly whenever I can -
I don't want the bastards to keep me down.
Even if driving is more dangerous, it still means I am not kowtowing to some Kafkaesque system. Rights and freedoms don't mean squat unless you regularly exercise them. Driving, even for long distances, incurs only a marginal increase in expense and risk. It seems like the obvious route to take for anyone who does not approve of the way the TSA does business.
The availability heuristic breaks down quite a bit -- not because of frequency, but because of scale and control.
Under "natural" conditions of small scale life, it's quite good, since "memory manipulation" is low -- what you remember is mostly actual events that happened to you.
In a mass society, most of your memories are constructed, not just of big things like 9/11, but of every-day little things.
What tastes good? Well, if you go by memory access, the barrage of commercials is much more salient than your actual tasting of food -- you can remember a Cheetohs commercial you've seen 5 times a day since you were two much better than the gustatory sensation.
Just apply generally. Calculating risks and benefits rationally is just too expensive -- rational thought is a late add on that isn't available without significant effort. An approximation by salience, primarily depending on memory, is a tried and true method tested by 100's of millions of years. Unfortunately, mass society breaks the basic premises, and falling back on rational calculation for most of life's demands is simply impossible -- if you did that, you'd be frozen most of the day trying to come to every decision.
The TSA’s budget was $0.00 in 2001. $1,345,000,000 in 2003. The TSA requested $4,810,000,000 in 2004. In 2008, they’re spending $6,814,000,000 and they’re requesting $7,100,000,000 for 2009. Without doing complete research on total expenditures and whatever other functions the TSA has grown to provide, it’s still obvious that we’ve spent over $10,000,000,000 since 2001 *just enforcing airline rules*.
Ten. Billion. Dollars.
How many lives could that have saved if we spent the money on medical research? How much safer would we be, as a nation, if we had spent ten billion dollars on upgrading our air traffic control system? Or upgrading the nation’s highways? The Federal Motor Carrier Safety Administration’s 2007 budget was $455 million dollars, or about 1/15th of the TSA’s 2008 budget. And yet the FMCSA is responsible for reducing crashes, injuries, and fatalities involving large trucks and buses -which average about 1,300 deaths per year just on the employee side (ie, the drivers of the large trucks and buses), to say nothing of their passengers or other cars that might be involved in the accident(s) - quantifiably a problem bigger than airline terrorism by an order of magnitude.
How does any of this make any sense?
(original post here, blog plug: http://padraic2112.wordpress.com/2008/08/04/...
You can tell what the weather is like outside with little effort. With the internet, you can find out the weather and predictions anywhere in the world.
Weather can be a large safety factor for small planes and very large ones. In the last couple of months a medium small plane crashed in inclement weather. Yesterday a fedex cargo plane crashed while landing in wind.
You have absolute control of whether or not you board a plane (or drive) in bad weather. There may be economic expenses attached, but surviving is worth a little inconvenience from time to time.
There is also the psychological issue that most people have resigned into accepting traffic fatalities as unavoidable -- they're on the news every night, they "just happen."
But people also think that terrorism -- or let's limit that to domestic terrorism -- is preventable.
By that reasoning, it makes sense to invest into preventing something that's preventable rather than spending money on the unavoidable.
Start discussing statistics, and you lose people's interest. After all, reducing annual traffic fatalities from 40,000 to 30,000 a year does not make them any less unavoidable.
Here's a thought regarding $17M/fatality for Homeland Security vs $25k/fatality for highway safety.
Risk is typically expressed as (Probability) x (Gravity). In this case, I would assert that calculating whether the amount to spend to mitigate risk also needs a factor in the equation to account for "what we know about what needs to be done."
In the case of highway safety, we have a lot of data about what can and does happen and we pretty much know what works to prevent accidents and minimize the chance that an accident is fatal. With terrorism, we have very little experience with what can go badly. (Lots of speculation about things like "dirty nuclear bombs," but hardly any data.) Also, less information on what really works to prevent terrorist acts, compared with how much we know about highway safety. Consequently, it makes sense to me that we need to throw more money per historical fatality into anti-terrorism due to the uncertainties. Whether a factor of 680X is the right amount can be debated. However, a couple of orders of magnitude to account for the uncertainties does not seem out of line.
I leave it to the reader to apply these concepts to thinking about how much an organization ought to spend to prevent Heartland/RBS-sized data breaches.
@Oliver - exactly right. The problem with the author's logic is that you don't want to think in terms of cost per life lost, but cost per life saved. How many more people would die if the NHTSA was not writing all the reports they do? How many more people would die from airline terrorism if the TSA wasn't there? And the answer is nobody knows. You can guess, but framing it in strict numerical terms is a bit shady without acnowledging that.
The problem with making the comparison between car crashes and terrorism is that the worst case for terrorism is gigantic, thus the calculation for probability x cost is very sensitive. On the other hand, the worst case for car crashes is pretty easy to calculate, and the benefits and costs of safety measures are relatively easy to measure.
I'm not saying the DHS anti-terrorism spending is at all reasonable; I'm simply saying that multiplying a gigantic number by a tiny uncertain number can easily be distorted to support just about any kind of cost/benefit analysis.
Not sure if anybody has mentioned this already but part of the reason why more people die on roads is that more people travel by car than flights. Deaths per hour travelled would therefore be a more reliable statistic.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.