Out of the Frying Pan into the Fire : Behavioral Reactions to Terrorist Attacks

Risk Analysis, Vol. 26, No. 2, 2006
DOI: 10.1111/j.1539-6924.2006.00753.x
Out of the Frying Pan into the Fire: Behavioral Reactions
to Terrorist Attacks
Gerd Gigerenzer?
A low-probability, high-damage event in which many people are killed at one point of time
is called a dread risk. Dread risks can cause direct damage and, in addition, indirect damage
mediated though the minds of citizens. I analyze the behavioral reactions of Americans to
the terrorist attacks on September 11, 2001, and provide evidence for the dread hypothesis:
(i) Americans reduced their air travel after the attack; (ii) for a period of one year following
the attacks, interstate highway travel increased, suggesting that a proportion of those who did
not ?y instead drove to their destination; and (iii) for the same period, in each month the
number of fatal highway crashes exceeded the base line of the previous years. An estimated
1,500 Americans died on the road in the attempt to avoid the fate of the passengers who were
killed in the four fatal ?ights.
KEY WORDS: Dread risk; fear of ?ying; public reactions to terrorist attacks; September 11; traf?c
accidents
1. INTRODUCTION
ple’s anxiety about ?ying, the job loss in the tourism
industry, as well as peculiar consequences such as the
The 9/11 Commission’s report unfolded the
increase in criminal suspects being involuntarily ex-
chronology of the terrorist attacks on September 11,
amined for psychiatric hospitalization.(2,3) Note that
2001, which cost the lives of some 3,000 people and
these misfortunes are not a necessary consequence of
billions of dollars in property damage.(1) It focused
terrorist action; they are of psychological origin, and
on how al-Qaeda terrorism evolved, the possible fail-
could in principle be prevented, once individuals and
ures of intelligence agencies to detect and avoid the
institutions realize that terrorists target minds as well
attack, and potential diplomatic, legal, and techno-
as bodies.
logical measures to prevent future attacks. The report
concerned the origins and prevention of what I re-
fer to as direct damage, that is, the immediate conse-
2. DREAD RISKS
quences of terrorist action. In this article, I deal with
Low-probability, high-damage events in which
a second source of harm caused by terrorist action,
many people are killed at one point in time are called
which I refer to as indirect damage. Indirect damage
dread risks. As opposed to situations in which a simi-
is not under the control of terrorists; it is mediated
lar number of people or more are killed over a longer
through the minds of citizens. In the case of Septem-
period of time, people tend to react to dread risks with
ber 11, known indirect damages include the ?nancial
avoidance behavior.(4) The crash of the four planes in
damages in the aviation industry fueled by many peo-
the terrorist attack on September 11 exempli?es such
a catastrophic event. In contrast, the estimated 44,000
?
to 98,000 people who die every year in U.S. hospi-
Max Planck Institute for Human Development, Center for Adap-
tive Behavior & Cognition, Lentzeallee 94, Berlin, Germany
tals because of documented and preventable medical
14195; tel: +49-30-824 06 461; gigerenzer@mpib-berlin.mpg.de.
errors do not constitute a dread risk.(5) Even after
347
0272-4332/06/0100-0347$22.00/1 C 2006 Society for Risk Analysis

---

348
Gigerenzer
learning about the dangers, few people would avoid
travel, which left airlines and travel agencies ?ying
hospitals. One potential evolutionary account of this
into the red. For instance, the national revenue pas-
specialized avoidance behavior is in terms of pre-
senger miles decreased by 20%, 17%, and 12%, in Oc-
paredness, that is, human minds are prepared to learn
tober, November, and December 2001, respectively,
the association between dread risk and avoidance
compared with the same months in 2000.(10) Data for
behavior in one trial. The suggested reason is that
the second part of the dread hypothesis, in contrast,
for our evolutionary ancestors, living in small bands
are dif?cult to obtain because there is no record of
of hunter-gatherers, the loss of many members at one
how many people decided not to ?y and took their car
point of time could have brought the group beyond
instead. Indirect evidence can be obtained from the
a critical threshold that threatened their survival.(6)
Of?ce of Highway Policy Information, which reports
A further account is the lack of proper information
the number of vehicle miles driven before and after
about risks among the general public of many West-
the attack. To establish whether there was an increase
ern societies. For instance, few people are aware that
in driving, three conditions must be met. First, there
the probability of losing one’s life is about the same
must be a sudden increase in the individual monthly
for driving 12 miles by car as for a nonstop ?ight, say,
miles traveled in the months following the attack com-
from Boston to Los Angeles.(7) That is, if one arrives
pared to the monthly miles of the previous year. Sec-
safely by car at the airport, the most dangerous part
ond, this increase must not be observed in the months
of the trip may be over. A third account is in terms
before the attack, and ?nally, the increase must fade
of control: people fear terrorist attacks because they
away at some point, when the pictures of the attack
have no control, whereas people believe to be in con-
fade out of people’s minds.
trol while driving. While there is some truth in each
In the eight months before the attack (January
of these accounts, none by itself seems to be suf?cient
to August 2001), the individual monthly vehicle miles
(for instance, although the driver has some control,
traveled in 2001 (all systems) were on average 0.9%
the person sitting next to the driver has little control,
higher than in 2000—which is normal given that
yet he or she typically also feels little fear). My point
miles traveled increase from year to year in the
here is not to provide an explanation for the tendency
United States. Immediately after the attack and in the
to avoid dread risks, but rather to draw attention to
12 months following, the miles traveled increased sub-
avoidance behavior as a potential cause for the in-
stantially. In the three months after the attack (Octo-
direct damages of terrorism, mediated through our
ber to December 2001), the increase tripled to 2.8%.
minds.
In the ?rst three months of 2002, the increase was
3.1%, and then 2.9% in the subsequent six months
(April to September 2002), compared to the previous
3. AVOIDANCE BEHAVIOR AFTER
year. Thereafter, in the next six months (October 2002
SEPTEMBER 11
to March 2003), this ?gure declined to 0.5%; that is,
In this article, I investigate a possible mediated
the increase in road traf?c after September 11 dimin-
death toll of the attack on September 11. This pos-
ished after one year.
sibility has gone virtually unnoticed, although it was
The hypothesis that more people chose to drive
hypothesized shortly after the attack.(8) In earlier re-
rather than ?y after the attack has another testable
search, I collected preliminary data limited to three
implication. The increase in miles driven should be
months after the attack;(9) here, I provide a compre-
most pronounced on the rural interstate highways,
hensive analysis of the 18 months after the attack.
where much of long-distance driving occurs, rather
My hypothesis is as follows: if (i) Americans reduced
than in urban areas. Speci?cally, before the attack,
their air travel after the attack, and (ii) a proportion
the increase on the rural interstate highways should
of those who did not ?y instead drove to their desti-
be similar to the 0.9% increase on all road systems,
nation, then (iii) a number of Americans died on the
but thereafter rise above it. Consistent with this hy-
road in the attempt to avoid the fate of the passengers
pothesis, the increase in the eight months before the
who were killed in the four fatal ?ights. I call this the
attack was similar to that for all traf?c systems, 1%.
dread hypothesis for short. Is there evidence for such
In the three months following the attack, the vehicle
a mediated toll of lives?
miles increased by 5.2%. In the ?rst three months of
The ?rst part of the dread hypothesis—the re-
2002, the increase (compared to the previous year)
duction in air travel following the attacks—is well
was 3.7%, and in the following six months, 2.2%. One
documented. Millions of Americans reduced their air
year after the attack, the increase of miles driven on

---

Behavioral Reactions to Terrorist Attacks
349
rural interstate highways stopped and even reversed
through August 2001, shows that in the months before
to a slight decrease of an average of –0.2% in the
the attack, the number of fatal accidents consistently
six months following (October 2002 to March 2003).
followed the pattern of the preceding ?ve years. On
Thus, vehicle miles increased after September 11,
average, there were only nine (!) additional fatalities
most strongly on rural interstate highways, for a pe-
per month (out of some 2,500 to 3,500 each month),
riod of about 12 months.
and the number of fatal accidents always remained
Did this change in travel behavior go hand in hand
within the minimum and maximum values of the ?ve
with a surplus in fatal road accidents? To test the third
previous years.
part of the dread hypothesis, I compare the fatal road
This regularity broke down in the months follow-
accidents after September 11 with two baselines: ?rst,
ing September 2001 (Fig. 1). For a period of 12 months,
with the average number of fatal road accidents in the
October 2001 to September 2002, the number of fa-
?ve years preceding the attack (1996–2000, the zero
tal accidents exceeded the ?ve-year baseline every
line in Fig. 1), and second, with the number of fatal
month, as well as the baseline adjusted by the aver-
crashes in 2001 before the attack. The ?rst baseline is
age increase of nine fatal crashes in pre-September
meaningful because the number of fatal traf?c acci-
2001. In the majority of months, the surplus exceeds
dents had been very stable over those ?ve years. The
the maximum value of the preceding ?ve years, as
total monthly number of fatal traf?c accidents varied
shown by the bars in Fig. 1. This is exactly the same
between about 2,500 in February and 3,500 in Au-
period in which the passenger miles showed a marked
gust, while the maximum deviation from these ?gures
increase. The surplus death toll was highest in January
during the ?ve years was, averaged across all months,
and March of 2002. After one year, fatal crashes re-
only about 115 accidents, which amounts to 3–4% of
turned to the baseline before the attack, at the same
the monthly average. The second baseline, January
point in time when the road traf?c returned to normal
Fig. 1. The number of fatal traf?c accidents in the United States increased after the terrorist attacks on September 11, 2001, for a
period of 12 months. Numbers are expressed as deviations from the ?ve-year base line 1996–2000 (the zero line). The error bars
(shown for the 12 months following the terrorist attacks) specify the maximum and the minimum numbers for each month of the
base line. Before September 11, the average of the monthly numbers of fatal traf?c accidents for 2001 was close to the zero line,
and the monthly values were always within the maximum and minimum of the previous ?ve years. Yet in the 12 months follow-
ing the terrorist attacks (October 2001 to September 2002), the number of fatal traf?c accidents every month was higher than the
zero line, and in most cases exceeded the maximum of the previous years. Data are taken from the U.S. Department of Transporta-
tion, Federal Highway Administration: http://www-fars.nhtsa.dot.gov/FinalReport.cfm?stateid=0&title=crashes&title2=time&year=2002;
http://www-fars.nhtsa.dot.gov/?nalReport.cfm?stateid=0&year=2003&title=Crashes&title2=Time.

---

350
Gigerenzer
(see above). This consistent pattern after the attacks
5. COUNTERTERRORISM SHOULD
provides support to the hypothesis that the terrorist
ADDRESS REDUCTION OF
attacks caused a mediated secondary death toll.
MEDIATED DAMAGES
How many fatalities resulted from people’s de-
How to react to the emerging global terrorism?
crease in ?ying and increase in driving? To estimate
As a response, the 9/11 Report demands “the use of all
this number, I will use the ?ve-year baseline as the
elements of national power: diplomacy, intelligence,
comparison standard, corrected by the average in-
covert action, law enforcement, economic policy, for-
crease of nine fatalities per month. For the 12 months
eign aid, public diplomacy, and homeland defense.”
following the attack, one obtains a surplus of 317 fa-
(1, pp. ci–cii). A national counterterrorism center
tal crashes for October through December 2001, and
should coordinate these means and strive for defeat-
of an additional 1,188 for January through Septem-
ing terrorism anywhere in the world. The present anal-
ber 2002, which totals 1,505 fatal crashes. Given that
ysis indicates that there is a second goal, to defeat the
the ratio between fatalities and fatal traf?c acci-
effects of terrorism acting through our minds. Ter-
dents in 2001 and 2002 was consistently 1.06, the to-
rorist attacks are hard to prevent, even with costly
tal estimated number of Americans who lost their
diplomatic and military strategies and controversial
lives on the road by trying to avoid the risk of ?y-
surveillance systems. It would be comparatively easier
ing is 1,595. I want to emphasize that this num-
and less expensive to invest at least part of the efforts
ber is an estimate, since a nonexperimental study
in reducing the mediated death toll. The ?rst measure
cannot control for all alternative explanations. This
to achieve this goal is to make the issue an issue. The
estimate is six times higher than the total num-
psychological aspect has not yet entered public policy
ber of passengers (256) who died in the four fatal
awareness to the same degree that the technological
?ights.
side of ?ghting terrorism has. Yet there are a number
of measures that can be taken and tested, from mak-
4. THE MADRID ATTACKS
ing people aware of the fact that terrorists can strike a
second time through their minds, to disseminating rel-
Does the dread hypothesis generalize to other
evant information to the public, such as that a dozen
cultures? On March 11, 2004, exactly two and a half
miles of driving result in the same risk of dying as one
years after 9/11, the bombings of four commuter trains
nonstop ?ight. Factual information will not change ev-
during the Madrid rush hour killed about 200 peo-
eryone’s behavior, but by knowing the facts, people
ple and wounded 1,460. The evolutionary prepared-
can understand their immediate emotional reactions
ness hypothesis would predict that Spaniards would
and better control them. Such an extended countert-
then avoid riding trains. And indeed, Spaniards re-
errorism policy can save lives. Otherwise, history may
duced their train travel after the bombing, although
repeat itself after the next attack, if another should
the effect was smaller and shorter (only two months)
happen.
than the American reaction after September 11.(11)
But there are two other parts to the dread hypoth-
esis. The second, an increase in highway traf?c, did
REFERENCES
not happen in Spain; in fact, there was a decrease.
1. The National Commission on Terrorist Attacks Upon the
As a consequence, the fatal highway accidents also
United States. (2004). The 9/11 Report. New York: St. Martin’s
decreased rather than increased, and no secondary
Press.
loss of lives mediated through Spaniards’ minds was
2. Fainstein, S. S. (2002). One year on. Re?ections on Septem-
ber 11th and the “War On Terrorism”: Regulating New York
observed. Like Americans, the Spaniards avoided the
City’s visitors in the aftermath of September 11th. Interna-
dread risk, but unlike Americans, they did not take the
tional Journal of Urban and Regional Research, 26, 591–595.
next step and jump into their cars. Why is that? One
3. Catalano, R. A., Kessell, E. R., McConnell, W., & Pirkle, E.
(2004). Psychiatric emergencies after the terrorist attacks of
possible factor is that there is less of a car culture in
September 11, 2001. Psychiatric Services, 55, 163–166.
Spain than in the United States. A second is the avail-
4. Slovic, P. (1987). Perception of risk. Science, 236, 280–285.
ability of a better developed public transport system
5. Kohn, L. T., Corrigan, J. M., & Donaldson, M. S. (Eds.). (2000).
To Err is Human: Building a Safer Health System. Washington,
that allowed Spaniards to resume train travel quickly.
DC: National Academy Press.
A ?nal possible reason is that Spain has been exposed
6. Cosmides, L., & Tooby, J. (1992). Cognitive adaptions for so-
to decades of terror attacks, so that one terror attack
cial exchange. In J. H. Barkow, L. Cosmides, & J. Tooby (Eds.),
The Adapted Mind: Evolutionary Psychology and the Gener-
provides less of a dread risk and more of a calculated
ation of Culture (pp. 163–228). New York: Oxford University
risk instead.
Press.

---

Behavioral Reactions to Terrorist Attacks
351
7. Sivak, M., & Flannagan, M. J. (2003). Flying and driving after
(http://www.fhwa.dot.gov/ohim/tvtw/nov tvt/table1.htm);
the September 11 attacks. American Scientist, 91, 6–8.
2001 data: Table 1—Estimated Individual Monthly Mo-
8. Myers, D. G. (2001). Do we fear the right things? American
tor Vehicle Travel in the United States [2002-2003 data].
Psychological Society Observer, 14, 3.
(http://www.fhwa.dot.gov/ohim/tvtw/02dectvt/table1.htm);
9. Gigerenzer, G. (2004). Dread risk, September 11, and fatal
2002 data: Table 1—Estimated Individual Monthly Mo-
traf?c accidents. Psychological Science, 15, 286–287.
tor Vehicle Travel in the United States [2002–2003 data].
10. U.S. Department of Transportation, Federal Highway Admi-
(http://www.fhwa.dot.gov/ohim/tvtw/03dectvt/table1.htm);
nistration. (1996–1999). Table 1—Estimated Individual
2003 data: Table 1—Estimated Individual Monthly Mo-
Monthly Motor Vehicle Travel in the United States [2000–2001
tor Vehicle Travel in the United States [2002–2003 data].
data].
(http://www.fhwa.dot.gov/ohim/tvtw/archive/index.
(http://www.fhwa.dot.gov/ohim/tvtw/04aprtvt/page3.htm).
htm); 2000 data: Table 1—Estimated Individual Monthly
11. L ´opez-Rousseau, A. (2005). Avoiding the death risk of avoid-
Motor Vehicle Travel in the United States [2000–2001 data].
ing a dread risk. Psychological Science, 16, 426–428.

---