
Cass Sunstein on Worst-case Scenarios 11/19/2007
What this covers
Russ Roberts hosts Cass Sunstein, a law professor, to discuss how societies and individuals should handle low-probability events with catastrophic potential. Drawing on Sunstein's book Worst-Case Scenarios, the conversation examines why humans fail at probabilistic reasoning about risk and what coherent policy actually requires. The core tension is this: the precautionary principle, which seems to counsel caution about worst-case scenarios, is logically incoherent—any response to a risk itself carries costs and probabilities, so the principle simultaneously forbids action, inaction, and everything in between.
The discussion traces why different risks capture public and political attention so unequally. Probability neglect, the availability heuristic, and the power of identifiable villains or salient events (like 9/11) cause people to collapse nuanced risk into a binary safe-or-unsafe judgment. This explains why terrorism received hundreds of billions in post-9/11 spending while climate change was treated as unaffordable—even though the latter poses far larger aggregate risks. Sunstein argues that good policy requires separating two distinct ideas: whether future lives matter equally to present ones (they do), and whether future dollars should be discounted (they should, since societies will likely be wealthier). He also contends that irreversible losses deserve weight, but that aggressive responses themselves create new risks and costs—including civil liberties costs in counterterrorism and poverty effects in climate regulation—that a coherent approach must weigh alongside the harms it seeks to prevent.
Sunstein argues that humans cope poorly with low-probability/high-cost risks by collapsing them into a binary safe/unsafe judgment, and that good policy requires probabilistic thinking, attention to the costs and risks of responses themselves, and separating intergenerational moral neutrality from monetary discounting.
- People use heuristics (availability, probability neglect, outrage) that distort worst-case risk assessment, as seen in differing reactions to terrorism vs climate change.
- Responses to risks carry their own risks and costs, so the precautionary principle is incoherent because it bans both action and inaction.
- Discounting future lives (intergenerational neutrality) must be distinguished from discounting future money, since the future will likely be richer.
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
One must separate two distinct ideas: 'intergenerational neutrality'—that a life 100 years from now is worth no less than a life today, so we should be equally concerned about future and present people—from the discounting of money, where a dollar today is genuinely worth more than a dollar in 2100; failing to discount money while honoring life-neutrality could itself harm future people by impoverishing the present and forgoing prosperity that yields human goods.
“we should be as concerned about future people as present people supposing they're going to be people ... we have to separate the idea of intergenerational neutrality which is good from the idea of treating a dollar today as worth no more than a dollar in 2100”
The precautionary principle in its strong versions (placing the burden on a proponent to prove an activity is safe) is incoherent and paralyzing because it forbids the very steps it requires: banning nuclear power on precautionary grounds itself creates probabilistic risks (including climate change), so the principle simultaneously compels and prohibits action, inaction, and everything in between.
“the problem with it is that it forbids the very steps that it requires so that the things that the precautionary principle bans are banned by the band”
A single salient event (like 9/11) can have a massive, possibly disproportionate effect on intuitive cost-benefit analysis of worst-case scenarios; terrorism had such an event while climate change lacked one (Hurricane Katrina failed to serve as climate's 9/11), and this difference largely explains the asymmetric policy responses.
“the existence of an available salient event can have a big effect on people's responses to worst-case scenarios we had that for terrorism we didn't have that for climate change and that made all the difference”
Testing for a rare disease with an imperfect test can be harmful: a 95%-accurate test for a condition with one-in-a-thousand prevalence (e.g., lupus) yields only about a 2% chance of actually having the disease given a positive result, so when no treatment exists, such testing mainly creates false alarm and is a disastrous thing to do.
“even though the test is 95% accurate ... getting a positive result the test means your odds of having the disease is 2%”
Outrage—particularly the presence of an identifiable human face responsible for a risk—is a powerful magnifier of attention to worst-case scenarios; terrorism has a face (Osama bin Laden/Al Qaeda) that magnifies concern, while climate change has no villainous face and is therefore diminished in salience.
“if there's a person who is responsible for the creation of a risk that can be a tremendous magnifier on people's attention to the worst case scenario”
When an emotionally gripping outcome is at stake, people exhibit 'probability neglect'—they focus on how terrible the outcome would be while ignoring its likelihood; terrorist attacks have this emotional grip while gradual, hard-to-visualize climate change does not.
“when there's a vent that has it gets about people's emotions really going then they often neglect probability altogether and they just think how terrible the outcome would be”
Some social risks are exploited by self-interested 'bootleggers and Baptists' coalitions where politicians look green or protective while funneling resources to favored special interests—e.g., the corn lobby benefiting from ethanol policies, or European farmers manufacturing GMO-food fears as a protectionist tool to scare consumers away from American products.
“politicians exploit our hope romance etc and by looking green they funnel resources to particularly favored special interests”
After 9/11, almost as many people died from foregoing air travel in favor of driving as would be expected, because statistical analysis even shortly after 9/11 showed driving was at least not safer than flying, yet people remained scared of flying while comfortable driving.
“after the 9/11 attack almost as many people died from foregoing air travel in favor of driving”
Aggressive responses to catastrophic risks themselves impose costs and create new risks: aggressive greenhouse-gas regulation produces certain adverse effects (higher energy prices harming the poor) and a remote but nonzero risk of international instability from economic distress, while anti-terrorism responses create civil-liberties risks and can increase other terrorist risks.
“if we did something very aggressive to control greenhouse gases that would produce a certainty of some adverse effects and with some probability very serious risks”
Humans have an evolutionarily rooted propensity to classify situations as simply 'safe' or 'unsafe' (an on/off 'danger sign'), which is unhelpful for good decision-making because real risks come in gradations rather than as a binary switch.
“there's a human propensity which probably has evolutionary roots easier to think it's okay we don't have to worry or to have read danger sign on in your head which completely takes over”
Collective-action problems lead people to believe their own small contribution will matter or spur others even when it will not, which underlies confused individual and subnational efforts on climate change.
“it's like with many collective action problems that people think that their own contribution will do something or spur something even if it won't”
People (including trained statisticians and law students) struggle to reason differently about a 10% probability versus a 30% probability of harm, because the mind does not naturally treat probabilities along a continuum, especially after an event has occurred or when reasoning anticipatorily about personal stakes.
“to think that you you know think very differently about a case involving 10% probability from a case involving 30% of the probability of some harm the mind doesn't naturally work that way”
The availability heuristic causes people to assess probabilities not like statisticians but by how easily they can recall an example; this explains why people wrongly believe homicides exceed suicides and accidents exceed diseases as causes of death, when the reverse is true, because homicides and accidents are more salient/available.
“people typically believe that more people die from homicides and from suicides though the opposite is true more people actually die from suicides and homicides and the reason is that homicides are very salient”
The tendency to respond to a salient tragedy by removing its cause—such as requiring baseball coaches to wear helmets after a one-in-a-million freak fatality, or laws and precautions during the shark-attack scare—is a remarkable human phenomenon that is usually hard to justify on all-things-considered grounds, especially given possible unintended consequences like reduced caution from a false sense of safety.
“it's a remarkable human phenomenon that when these kind of events happen we have such a natural visceral instinct to remove the cause”
U.S. cities and states (notably California) adopting costly emissions reductions that deliver essentially no climate benefit is explained by a combination of confusion (not seeing that the problem requires broad international participation), hope (that local action may trigger a cascade spurring international action), and a perceived moral obligation toward vulnerable nations like India and African countries that face severe harm even from moderate (~2.5°C) warming.
“I think this is a combination of confusion hope and a perceived moral obligation”
The right way to handle intergenerational tradeoffs is to ask what future generations would want: sometimes they would want us to take precautions, but often they would say 'grow, and we will come up with ways to cope,' since being wealthier enables coping—so we should neither control all pollution nor none of it.
“I think the right way to think about it is what would they want and the answer is sometimes they'd want us to take precautions a lot of times they'd say no grow and we'll come up you will come up people will come up along the way with ways to cope”
'Worst-case entrepreneurs' fan the flames of fear by exploiting public ignorance and making low-probability risks more salient; they attacked nuclear power successfully by invoking Three Mile Island (which was not actually a disaster) and Chernobyl, giving nuclear power an undeserved 'bum rap' that has harmed U.S. energy development.
“the attack on nuclear power was driven partly by worst-case entrepreneurs who invoked Three Mile Island and Chernobyl”
Government risk regulation legitimately differs from individual decision-making: because governments operate population-wide, they can reasonably worry about one-in-a-hundred-thousand risks or less, whereas for most individuals most of the time such precautions make no sense because the associated anxiety is itself a cost (and an adverse health effect) that exceeds the precaution's benefit.
“governments because they're dealing on a population wide basis there probably are definitely legitimately concerned with one in a hundred thousand risk or less but for most of us most of the time it probably doesn't make ... any sense”
Irreversibility is where concern for future generations genuinely bites: irreversible losses (e.g., the Grand Canyon or a starlit sky) that deprive future generations of options with no good substitutes warrant special weight—but an irreversible loss should not be treated as infinitely valuable, and many feared harms (like skin cancer) will likely be cured within a century.
“the issue that ... I think is relevant where the future generation thing does bite is irreversibility obviously if we destroy the Grand Canyon ... there aren't good substitutes for that”
The U.S. under Reagan aggressively attacked ozone depletion and was more aggressive than Europe, with Washington Republicans pushing for action, because the Council of Economic Advisers showed that phasing out ozone-depleting chemicals was a tremendous bargain (~$21 billion cost) yielding far larger benefits in reduced cataracts and skin cancer—even unilateral U.S. action would have been justified.
“the Montreal Protocol it turned out cost the US about 21 billion dollars significant but not massive ... in terms of cataracts and skin cancer this is a tremendous bargain for the United States”
A feeling of personal control (such as being at the wheel while driving) can diminish perceived risk even when statistically unjustified, which partly explains why people fear flying more than driving.
“a feeling of personal control can be diminished feelings of risk even in cases in which statistically that's just not sensible”
The Kyoto Protocol was rejected on a bipartisan basis (the Senate voted 95-0 against, no Democratic senator supported ratification) because it would have cost the U.S. upwards of $300+ billion (bearing 50-80% of worldwide costs) while delivering near-zero monetized U.S. benefits—since the developing world was excluded and the existing stock of greenhouse gases unaffected—reducing anticipated 2100 warming by only about 0.03°C.
“the Kyoto Protocol would have cost ... about upwards of three hundred billion dollars ... the United States as a first approximation would have gotten zero in terms of monetized benefits”
Imposing large costs on the private sector causes deaths because people become poorer and less employed; data suggest roughly a $7 million expenditure causes at least one statistical death, so a $40 million regulatory expenditure (not unusual) probabilistically causes five or six or more deaths, which itself violates the precautionary principle.
“there's data suggesting a seven million dollar expenditure will cause at least one death so a 40 million dollar expenditure ... causes five or six deaths”
The differing American reactions to terrorism (hundreds of billions spent post-9/11) and climate change (similar costs viewed as unaffordable) can be partly explained as rational self-interested cost-benefit analysis: climate remedies were perceived to mostly benefit foreigners in the distant future at high domestic cost, while terrorism precautions were seen as cheap relative to high perceived domestic benefits.
“Americans have thought not unreasonably that certain steps to protect against climate change would mostly benefit people in other countries in the distant future and cost us a fair bet”
Cheney's '1% doctrine'—treating a 1% chance of a catastrophe like a nuclear terrorist attack as if it were a near-certainty—is flawed because, even in the domain of catastrophe, a 1% chance is fundamentally different from a 100% chance and should not be treated identically.
“they're they're wrong in thinking that a 1% chances is anything like a hundred percent chance even in the domain of catastrophe”
Even several years after 9/11, Americans across the country estimated something like a 20-30% probability that they or someone they knew would be killed in a terrorist attack within the next year—a wildly inflated personal threat assessment that is hard to justify on available information, undermining the 'rational updating' explanation.
“people thought there was a very high probability that they or someone they knew would be killed in the next year in a terrorist attack something like twenty or thirty percent”
Because human history suggests the future will be much richer than the present, it makes no sense for the relatively impoverished present to transfer resources to the far wealthier future; sacrificing economic growth now also hurts the future because long-term growth produces citizens who live healthier, longer, and better lives.
“if human history is any guide the future will be much richer than the present and it makes no sense to say that the relatively impoverished present should transfer its resources to the far wealthier future”
There is no such thing as a truly 'safe' drug, treatment, surgery, or device; for both individuals and the FDA there are only gradations of risk, and when stakes are high one should evaluate both the magnitude of the danger and its probability rather than relying on a safe/unsafe label.
“for an individual doctor or patient or for the Food and Drug Administration they're gradations of risk not as if there's an on/off switch”
Worst-case entrepreneurs invoking the risks of DDT (e.g., via Rachel Carson's Silent Spring) helped drive an international consensus to ban it, which—while justified in the U.S.—created a worse-case scenario of its own in the developing world, where DDT is the most effective and cheapest way to prevent malaria, leading to large numbers of human deaths; risks frequently exist on both sides of social situations.
“in the development world DDT is most effective and cheapest way of preventing malaria and the worst case entrepreneurs invoking the risks associated with DDT actually breathe a worse case of their own which is large numbers of human deaths”
Once a country becomes dependent on a technology like nuclear power, it is very hard to reverse course, as in France, where an early decision with a powerful trusted government and little intense opposition entrenched nuclear power—analogous to Finland's unresponsiveness to cell-phone/cancer fears because Nokia and cell-phone use are central to its economy.
“once the country was nuclear power very hard to change it for various reasons one is that the country's dependent on it”
When probabilities cannot be assigned with confidence (a wide band of uncertainty), the standard expected-value approach of multiplying outcomes by probabilities breaks down, so decision-makers may need simplifying rules of thumb instead.
“if you have a wide band we don't know really the probability is very well then you're going to struggle and you might need simplifying rules”
The political process acts as an aggregator of individual preferences, and politicians respond to voter voices, so the U.S. banned ozone-depleting chemicals (a good deal for Americans) but stayed indifferent to costly global-warming action—whereas in poor countries ruled by tyranny the voice of the people matters far less, possibly explaining differential international responses to risks like DDT bans.
“in the United States politicians respond pretty much to the voices of the voters ... we pretty much get what we want”
We already transfer enormous resources into the future—not merely through bequests but through the immense time and effort spent on our children and loved ones—a fact often ignored in discounting debates.
“we transfer tons of resources into the future already not not in through the estate ... but just through the fact that the amount of time and effort we spend on our children and loved ones is is immense”
Driving more rather than flying can now be a rational choice because of the added time costs and inconvenience of flying (security lines), even if driving is statistically more dangerous.
“I think it is sensible to drive a little bit more than in the past because of the added time costs of flying”
Mammals generally operate in a binary fight-or-rest mode—'I can go to sleep or I have to stay awake'—which mirrors the human safe/unsafe instinct.
“I like to think of it either as I can go to sleep or I have to stay awake look at mammals generally that's how they are”
Entrepreneurs who claim they only need '1% of the market' or '1% of the book market' frame it as a cautious goal, but achieving even 1% of a large market is actually a remarkably ambitious claim.
“if I can just get 1% of the market ... you think well that's only 1% ... but of course it's really rather remarkably ambitious claim”