Fear of Flying After Terror Attacks: What the Road Deaths Show

The figure most people have heard is 1,500. After the attacks of 11 September 2001, the story goes, Americans stopped flying, drove instead, and about 1,500 more of them died on the roads in the following year than would otherwise have done. It is one of the most quoted findings in the psychology of risk, and it deserves to be, because the mechanism behind it is real.

It is also one estimate among five. The published figures for the same months run from about 350 to 2,300, depending on the baseline, the time window and the statistical model. One reanalysis found that most of the extra deaths happened on local roads rather than the interstate highways a long drive instead of a flight would use. And when Madrid was bombed in 2004, Spaniards avoided the trains but did not die on the roads at all.

This is what the research actually found about “dread risk” — the fear of rare events that kill many people at once — what it costs, and why the cost depends much more on the roads, bicycles and trains around you than on how frightened you are.

What a dread risk is

Risk researchers use “dread” for hazards that kill many people at one moment, feel uncontrollable and arrive without warning. A plane crash is the classic case. A year of ordinary road deaths is the opposite: it kills far more people, but one or two at a time, in places no camera is pointed at.

The two do not register the same way. In a reanalysis of a classic risk-perception dataset, Robin Gregory and Robert Mendelsohn found that people’s ratings of how risky something was tracked its expected death toll, but their ratings of dread did not. Dread is its own dimension. It moves behaviour without moving in step with the numbers.

Gerd Gigerenzer, a psychologist at the Max Planck Institute for Human Development, put the consequence in one sentence in 2004: if people avoid a dread risk, the alternative they choose may be more dangerous than the thing they are avoiding. After September 2001, the alternative to flying was driving. Per mile travelled, American roads kill far more people than American airlines.

Five estimates of the same second strike

Extra US road deaths attributed to people driving instead of flying after 11 September 2001. Most figures are the authors’ rounded estimates.

Gigerenzer, 2004 (Oct to Dec 2001)350

Sivak and Flannagan, 2004 (Oct to Dec 2001)1,018

Blalock and colleagues, working paper (total)1,200

Gigerenzer, 2006 (12 months)1,500

Blalock and colleagues, 2009 (upper bound)2,300

Gigerenzer, Psychological Science 2004 (as summarised by Sivak and Flannagan) and Risk Analysis 2006; Sivak and Flannagan, Transportation Research Part F, 2004; Blalock, Kadiyali and Simon, Applied Economics, 2009, and a circulated working-paper version. Grey marks the unpublished version.

The five estimates, and why they differ

Gigerenzer’s first paper looked only at October to December 2001. It compared monthly road deaths with a baseline from earlier years and concluded that the extra deaths exceeded the number of people killed aboard the four hijacked aircraft. Michael Sivak and Michael Flannagan at the University of Michigan then argued that the baseline should come from the trend in road deaths between 2000 and 2001 for January to August. Using that trend, they put the excess for the same three months at 1,018, roughly three times Gigerenzer’s figure.

Gigerenzer’s 2006 paper widened the window to twelve months. Air travel fell, interstate highway travel rose, and in every month for a year the number of fatal highway crashes exceeded the baseline of previous years. His estimate for the whole period was about 1,500.

The economists Garrick Blalock, Vrinda Kadiyali and Daniel Simon took a regression approach, controlling for time trends, weather and road conditions. Their published paper attributes 327 driving deaths a month to the response in late 2001 and says as many as 2,300 deaths may be attributable to the attacks. A circulated working-paper version of the same study put the figure at 242 a month and about 1,200 in total. The difference is a reminder that model choices move these numbers by hundreds of lives.

None of the five is wrong in the sense of being careless. They answer slightly different questions over different windows. What they share is the direction: every analysis finds more road deaths, not fewer.

The part that does not fit the simple story

The tidy version of the finding is a modal shift: someone who would have flown from Chicago to Denver drives the same trip on the interstate instead, and the extra miles on a fast road produce the extra deaths.

Sivak and Flannagan checked where the deaths actually happened. The major part of the increase was on local roads, not interstates. Driver deaths did not rise out of proportion to passenger deaths. And pedestrians and cyclists carried a disproportionate share of the increase.

That pattern is hard to explain as people swapping a flight for a drive. It looks more like a general change in how people behaved on the roads in the months after the attacks, of which substituting the car for the plane was only one part. Sivak and Flannagan themselves say their results argue against a simple modal shift to the same destinations.

Wolfgang Gaissmaier and Gigerenzer returned to the question in 2012 and found that the rise in deaths varied widely by region. The weaker predictor was proximity to New York, where stress reactions after the attacks had been measured as highest. The stronger predictor was driving opportunity: the number of highway miles, and the number of registered cars per inhabitant. Fear started the process. The environment decided how many people it killed.

Four attacks, four different second strikes

What people avoided after each attack, and what the roads showed afterwards.

Attack	What people avoided	What happened on the roads
New York and Washington, 2001	Flying	More fatal crashes each month for a year; estimates from about 350 to 2,300 deaths
Israel, 2001 to 2002	Not measured directly	A lull in minor crashes, then a 35% spike in fatal crashes on the third day after an attack
Madrid, 2004	Trains	No rise in car travel or in fatal accidents
London, 2005	The Underground (about 10% fewer journeys over 20 weeks)	214 extra cyclist casualties over six months; no detectable rise in car crashes

Gigerenzer 2006; Sivak and Flannagan 2004; Blalock and colleagues 2009; Stecklov and Goldstein, PNAS 2004; López-Rousseau, Psychological Science 2005; von Winterfeldt and Prager 2010; Ayton, Murray and Hampton, Judgment and Decision Making 2019.

Madrid and London

If fear of a dread risk automatically produced road deaths, the 2004 Madrid train bombings should have produced them too. Alejandro López-Rousseau analysed train travel, highway traffic and fatal highway accidents in Spain in the months after 11 March 2004. Spaniards did avoid the trains. But highway traffic did not rise and neither did fatal accidents. He offered a sociopolitical interpretation; the plainer point is that avoiding one risk does not force you into another unless an alternative is sitting there.

London was different again. Peter Ayton, Samantha Murray and James Hampton looked at the aftermath of the 7 July 2005 bombings on the Underground and a bus. Tube journeys fell and cycling rose. Over the next six months there were 214 additional bicyclist road casualties, a 15.4% increase. They found no detectable increase in car accidents.

In London, the alternative to the Underground was the bicycle, so that is where the second strike landed. In Spain, trains were avoided without a comparable shift onto the roads. In the United States, a country built around the car, the alternative was driving.

Israel shows a shorter rhythm. Guy Stecklov and Joshua Goldstein analysed traffic flows, accidents and attacks from January 2001 to June 2002 and found a temporary lull in light accidents after an attack, followed by a 35% spike in fatal accidents three days later. They read it as a population-level behavioural response, measured in the traffic record rather than in surveys.

London after 7 July 2005

Six months after the bombings, compared with what would otherwise have been expected.

+214additional bicyclist road casualties, a 15.4% increase

0detectable increase in car accidents

Ayton, Murray and Hampton, Judgment and Decision Making, 2019. The authors stress that the effect depended on an alternative mode being available.

Fear is not the whole explanation

It would be easy to turn this into a lesson about irrational people. The research does not quite support that.

Surveys taken during a sustained campaign of bus bombings in Jerusalem and Haifa found that people knew perfectly well that road crashes were more dangerous than terror attacks. Wafa Elias, Gila Albert and Yoram Shiftan reported that fear and perceived risk still shaped whether people used buses, and that women reported more fear and changed their travel more. Knowing the numbers did not switch the feeling off.

A simulation study by Nicolai Bodemer, Azzurra Ruggeri and Mirta Galesic argued that reacting more strongly to a single catastrophe than to the same number of deaths spread over time can be ecologically rational. In their models, a dread event does more lasting damage to a population than a continuous risk with the same toll, particularly when it kills young people. That is a modelling argument, not an observed result, but it is a serious one.

And the data from Israel are not uniform either. A conference paper by M. H. Cale and Elit Shaked, comparing Israeli road accident records with attacks between 2001 and 2004, found fewer and less severe road accidents during periods of high fear in the first year of the intifada, before the relationship changed in later periods. It is a small study, but it cuts against any single rule.

How big the direct risk is

The point of the dread-risk research is not that terrorism is harmless. It is that the direct risk to an individual in most Western countries is small, and that the response can add harm of its own.

John Mueller and Mark Stewart have spent years putting the direct risk in the same units regulators use for every other hazard. Their conclusion for the United States, the United Kingdom, Canada and Australia is that the annual risk of death from terrorism is low enough to fall within the range regulators normally treat as acceptable. Their book on the costs of homeland security calculates that the increase in US domestic security spending since 2001 has exceeded one trillion dollars. To be cost-effective, even under assumptions tilted in its favour, it would have to deter or foil 1,667 otherwise successful attacks of the attempted Times Square type every year.

That is a contested political argument, and security specialists dispute parts of it. The opposing case is serious: Lisa Robinson and colleagues, working on how US agencies should value lives saved from terrorism, note that the standard value of a statistical life, about $6.5 million, was derived from workplace risks, and that the dread and ambiguity of terrorism may justify a higher value, possibly double. People are entitled to weigh some deaths more than others. What the research asks is that the weighing be done openly, with the indirect costs on the ledger.

The cost-effectiveness test

Mueller and Stewart’s calculation for US domestic homeland security spending since 2001.

$1 trillion+cumulative increase in US domestic security spending in the decade after 2001

1,667Times Square-type attacks a year it would need to prevent to be cost-effective

Mueller and Stewart, Terror, Security, and Money, Oxford University Press, 2011. Robinson and colleagues, Journal of Homeland Security and Emergency Management, 2010, argue for a higher value per life for terrorism risks.

The same pattern, in 2025

The mechanism is not confined to terrorism. After a run of aviation accidents in early 2025, including the mid-air collision over the Potomac near Washington in late January, an AP-NORC poll taken on 6 to 10 February found that the share of Americans who called air travel very or somewhat safe had fallen from 71% the previous year to 64%. The share calling it unsafe rose from 12% to about 20%.

No one has yet published an analysis of whether that shift moved any travel from air to road, and it would be premature to assume it did. The polls measure perception, not behaviour. But the shape is familiar: a few highly visible deaths change how safe something feels, while the far larger and steadier toll on the roads changes nothing about how driving feels.

How safe Americans said flying was

Share describing air travel as very or somewhat safe, AP-NORC polls.

202471%

February 202564%

AP-NORC Center for Public Affairs Research, poll of 6 to 10 February 2025, as reported by Fortune, 19 February 2025. The share calling it unsafe rose from 12% to about 20%.

What the evidence does not show

It does not show a precise number. Anyone who quotes 1,500 as a fact is quoting one estimate from a family of estimates that runs from roughly 350 to 2,300, built on different baselines and windows.

It does not show that fear always kills. Madrid is a clean counter-example, and the Israeli data move in different directions in different periods.

It does not show that avoidance is irrational. The surveys suggest people know the relative numbers and act on feeling anyway, and there are respectable arguments that weighting catastrophic events more heavily is not simply a bias.

What it does show, consistently, is that the harm from a dread event does not stop with the event. It travels through whatever alternative is at hand: the car in the United States, the bicycle in London. The most useful finding is Gaissmaier and Gigerenzer’s: the size of the second strike depended most on the opportunity to drive.

Questions people ask

Did 1,500 Americans really die because they were afraid to fly after 9/11?

That is Gerd Gigerenzer’s estimate for the twelve months after the attacks. Other published estimates for overlapping periods range from about 350 to 2,300. All find an increase in road deaths; they disagree on its size.

What is a dread risk?

A low-probability, high-consequence hazard that kills many people at once, such as a plane crash or a terrorist attack. People tend to fear these more than risks that kill as many or more people spread out over time, such as road crashes.

Is driving more dangerous than flying?

Per mile travelled in the United States, yes, by a wide margin. That is why substituting a long drive for a flight after the attacks was expected to raise deaths.

Did the same thing happen after other attacks?

Not in the same way. After Madrid in 2004 people avoided trains but road deaths did not rise. After London in 2005 cycling rose and cyclist casualties rose by 15.4% over six months, with no detectable rise in car crashes.

Is it irrational to fear terrorism more than car crashes?

Researchers disagree. Some treat it as a bias with measurable costs; others argue that reacting more strongly to catastrophic events can be sensible. Either way, the extra risk taken on to avoid it is real and can be counted.

The short version

• After 11 September 2001, fewer Americans flew and more drove. Published estimates of the extra road deaths range from about 350 to 2,300; the 1,500 figure is one of them.
• One reanalysis found most of the extra deaths on local roads, with pedestrians and cyclists over-represented, which does not fit a simple swap of flights for drives.
• The size of the effect varied by region and was best predicted by how easy it was to drive, not by closeness to New York.
• After Madrid in 2004, road deaths did not rise. After London in 2005, cyclist casualties rose by 214 over six months.
• In February 2025, after several air accidents, the share of Americans calling flying safe fell from 71% to 64%. No analysis yet shows a matching shift to the roads.

This article summarises published research on risk perception and travel behaviour. It is not security advice and it takes no position on any particular attack, government or policy beyond describing what the studies found and where they disagree. Official travel advice from your government and transport operators should take precedence over anything here.

Further reading. Gigerenzer, ‘Out of the frying pan into the fire: behavioral reactions to terrorist attacks’, Risk Analysis, 2006, is the twelve-month analysis behind the 1,500 figure. Sivak and Flannagan, ‘Consequences for road traffic fatalities of the reduction in flying following September 11, 2001’, Transportation Research Part F, 2004, is the reanalysis that found the deaths on local roads. Ayton, Murray and Hampton, ‘Terrorism, dread risk and bicycle accidents’, Judgment and Decision Making, 2019, covers London and is open access.

Sources

1. Gigerenzer G. Dread risk, September 11, and fatal traffic accidents. Psychological Science, 2004. doi:10.1111/j.0956-7976.2004.00668.x.
2. Gigerenzer G. Out of the frying pan into the fire: behavioral reactions to terrorist attacks. Risk Analysis, 2006. doi:10.1111/j.1539-6924.2006.00753.x.
3. Sivak M, Flannagan MJ. Consequences for road traffic fatalities of the reduction in flying following September 11, 2001. Transportation Research Part F, 2004. doi:10.1016/j.trf.2004.09.007.
4. Blalock G, Kadiyali V, Simon DH. Driving fatalities after 9/11: a hidden cost of terrorism. Applied Economics, 2009. doi:10.1080/00036840601069757.
5. Gaissmaier W, Gigerenzer G. 9/11, Act II: a fine-grained analysis of regional variations in traffic fatalities in the aftermath of the terrorist attacks. Psychological Science, 2012.
6. López-Rousseau A. Avoiding the death risk of avoiding a dread risk: the aftermath of March 11 in Spain. Psychological Science, 2005. doi:10.1111/j.0956-7976.2005.01551.x.
7. Ayton P, Murray S, Hampton JA. Terrorism, dread risk and bicycle accidents. Judgment and Decision Making, 2019.
8. Stecklov G, Goldstein JR. Terror attacks influence driving behavior in Israel. PNAS, 2004. doi:10.1073/pnas.0402483101.
9. von Winterfeldt D, Prager F. Estimating behavioral changes for transportation modes after terrorist attacks in London, Madrid, and Tokyo. 2010 (report, TRID record 935562).
10. Elias W, Albert G, Shiftan Y. Travel behavior in the face of surface transportation terror threats. Transport Policy, 2012 (online). doi:10.1016/j.tranpol.2012.08.005.
11. Cale MH, Shaked E. The effects of fear of terrorist attacks on driving behaviour and traffic safety. Conference paper, 2005.
12. Bodemer N, Ruggeri A, Galesic M. Why 100 once is worse than 10 times 10: dread risks versus “continuous” risks. Max Planck Society repository, 2013.
13. Gregory R, Mendelsohn R. Perceived risk, dread, and benefits. Risk Analysis, 1993. doi:10.1111/j.1539-6924.1993.tb01077.x.
14. Mueller J, Stewart MG. Terror, Security, and Money. Oxford University Press, 2011.
15. Robinson LA, Hammitt JK, Aldy JE, and colleagues. Valuing the risk of death from terrorist attacks. Journal of Homeland Security and Emergency Management, 2010. doi:10.2202/1547-7355.1626.
16. AP-NORC Center for Public Affairs Research poll, 6 to 10 February 2025, reported in Fortune, 19 February 2025.