E. Mark Warr

Faculty Research AssociatePh.D., University of Arizona

Professor of Sociology
E. Mark Warr



Crime, Delinquency, Deviance.


Mark Warr received his Ph.D. in sociology in from the University of Arizona and was a postdoctoral fellow at the University of Washington and Washington State University before joining the faculty at Pennsylvania State University and, later, the University of Texas.  Professor Warr has served as an advisor, consultant, or author to the National Science Foundation, the National Academy of Sciences, the National Institute of Justice, the Bureau of Justice Statistics, the American Statistical Association, the British Home Office, the Canadian Ministry of Justice, and a variety of other foreign and domestic agencies.  He served on the NSF panel to establish the National Consortium for Violence Research and the NIJ committee to revise the measurement of rape in the National Crime Victimization Survey.  He has served on the editorial boards of the American Journal of Sociology, the American Sociological Review, and Criminology, and has been appointed to numerous committees of the ASC and ASA.  Professor Warr is a recipient of the President's Associates Teaching Excellence Award, the Eyes of Texas Award, and other teaching awards.  His book Companions in Crime: The Social Aspects of Criminal Conduct (Cambridge University Press, 2002) won the 2005 Michael J. Hindelang Award from the American Society of Criminology.  He was named a Fellow of the American Society of Criminology in 2012 and was recorded for their Oral History Project in 2015.



Fear of Crime in the
United States: Avenues
for Research and Policy
by Mark Warr
Fear of crime affects far more people in the United States than crime
itself, and there are sound reasons for treating crime and fear of crime
as distinct social problems. After assessing the state of knowledge on
fear, this chapter considers whether public fear of crime can and
ought to be controlled, and the moral and practical implications of
doing so. The discussion draws on the literatures of risk perception
and risk communication, as well as research on the etiology of fear
and public beliefs about crime. A final objective of the chapter is to
identify the most pressing unanswered questions about fear confronting
investigators today.
Mark Warr is a Professor in the Department of Sociology at the University of Texas
at Austin.
Criminal events capture the attention of the general public in a way that
few other events can (cf. Skogan and Maxfield 1981). One reason is that
crimes receive extraordinary emphasis in the mass media, from news coverage
to feature films to television dramas to crime fiction (Graber 1980; Skogan and
Maxfield 1981; Warr 1994). But even without this sort of amplification, crimes
are intrinsically interesting events. As condensed and emblematic accounts of
human conflict, they raise profound questions about the nature and sources of
human motivation, the misfortune of fellow humans, the ability of the state to
maintain social order, and, ultimately, the presence or absence of justice in
human affairs.
There is another, perhaps more crucial, reason that crimes generate such acute
public interest. Criminal events, at their most elemental level, are frightening
events. They are reminders to all that the world is not a safe place, that danger can
strike at any time or location, and that life, in the end, is tenuous and precious.
Judging from the attention it elicited from criminologists, public fear of crime
was regarded as a trivial consequence of crime through most of the history of
criminology. None of the great names of 19th-century criminology gave the matter
much attention, and the situation changed relatively little during the first half
of the 20th century. Many investigators, it seems, adopted the commonsensical
but questionable notion that fear is directly proportional to objective risk, and
assumed that strategies to control crime are ipso facto strategies to control fear.
Although the serious personal consequences of criminal victimization were
apparent to criminologists, no one allowed that fear alone could be debilitating.
Some three decades ago, however, the President’s Commission on Law
Enforcement and Administration of Justice (1967, 3) offered this brief but
trenchant observation: “The most damaging of the effects of violent crime is
fear, and that fear must not be belittled.” That statement prefigured a fundamental
shift in the way that criminologists think about the consequences of
crime, one that was to heavily influence the course of criminological research
in years to come. To fully understand the social consequences of crime, criminologists
came to realize, investigators cannot focus merely on those who
become direct victims of crime. Important as these individuals surely are,
researchers must also concentrate on those who suffer forms of indirect victimization
(Conklin 1971, 1975), the most egregious of which is fear of crime.
The wisdom and awful implications of this insight were quickly borne out by
survey research demonstrating that fear of crime in the United States is far
more prevalent than actual victimization (often by orders of magnitude), and
that Americans react to this fear through a variety of precautionary behaviors
so pervasive and normative that they form a significant and defining element of
American culture (Warr 1994).
Since the days of the President’s Commission, hundreds of studies of fear of
crime have been conducted, and the topic regularly appears in the journals of
the field. For reasons that remain elusive, however, the study of fear seems to
have stalled at a rudimentary phase of development, a situation that is in danger
of turning into outright stagnation. Investigators continue to revisit the same
well-worn issues, and, even after three decades, the meaning of the term “fear”
remains a matter of controversy.
This chapter has three principal purposes. One is to identify the most pressing
unanswered questions about fear of crime, giving proper recognition to existing
lines of research and traditions in the field. Another goal is to consider the merits
and prospects for controlling public fear of crime, recognizing that fear has
beneficial as well as deleterious consequences, that individuals can be too
unafraid as well as too afraid, and that fear depends in part on subjective factors
for which there is no objective standard or valuation. The logistic and ethical
complexities of controlling fear have thus far deterred researchers from any
protracted discussion of the matter, but it is far too important an issue to ignore
or defer. A final purpose of the chapter is to offer a brief recounting of the history
of research on fear of crime to those unfamiliar with the field.
The Nature of Fear
Despite decades of research and debate, investigators have yet to settle on a
definition of fear of crime. Over the years, the phrase has been equated with a
variety of emotional states, attitudes, or perceptions (including mistrust of others,
anxiety, perceived risk, fear of strangers, or concern about deteriorating
neighborhoods or declining national morality). Even those whose work is
otherwise laudable seem to have trouble defining fear of crime. Ferraro and
LaGrange, for example, initially defined fear as “negative emotional reactions
generated by crime or symbols associated with crime” (1987, 73). Under that
definition, however, it would be difficult to distinguish fear from sadness,
anger, despair, or resignation.
Much of the confusion over the meaning of fear seems to arise from a failure to
recognize elementary distinctions between perception, cognition, and emotion.
Notwithstanding the claims of some, fear is not a perception of the environment
(an awareness or experience of sensory stimuli), but a reaction to the perceived
environment. Although fear may result from the cognitive processing or evaluation
of perceptual information (e.g., a judgment that an approaching male is
armed, or that a sound signals danger), fear is not itself a belief, attitude, or
evaluation. On the contrary, fear is an emotion, a feeling of alarm or dread
caused by an awareness or expectation of danger (see Sluckin 1979). This
affective state is ordinarily (though not invariably) associated with certain
physiological changes, including increased heart rate, rapid breathing, sweating,
decreased salivation, and increased galvanic skin response (Thomson 1979;
Mayes 1979).
Were it not such a serious matter, the disarray among criminologists over the
meaning of fear might be amusing. Whatever confusion criminologists may
suffer, however, the concept of fear is routinely and profitably used in psychology
and the life sciences, with considerably less dispute as to its meaning. In
everyday life, the emotion of fear is a common experience for most human
beings, for whom it is no more mysterious than anger, joy, or despair. For their
part, criminologists continue to exhibit a tendency to isolate or compartmentalize
“fear of crime,” to assume that it differs in some fundamental way from
other ordinary fears, such as fear of traffic accidents, fear of falling, or fear of
disease. But there is no evidence that fear of crime is qualitatively different
from other forms of fear. What differentiates one from another is merely the
object or stimulus of fear.
One common source of confusion when it comes to defining fear of crime
occurs when investigators equate fear of crime with the perceived risk of victimization
(i.e., the subjective probability of victimization). However, there
are compelling reasons (among them predictive accuracy, convergent validity,
and logical necessity) to believe that perceived risk is a proximate cause of
fear—not fear itself (see Warr and Stafford 1983; Warr 1984, 1985, 1991, 1994;
Ferraro 1995). And there is corroborating evidence that measures of fear and
measures of perceived risk do not measure the same phenomenon and do not
behave similarly with respect to other variables (Rountree and Land 1996; Ferraro
1995). In short, fear is not perceived risk; by all indications, it is its consequence.
Fear of crime may be aroused by an immediate danger, as when an individual
is confronted by an armed attacker or is verbally threatened with harm. This
type of intense, immediate experience appears to be what some have in mind
when they speak of fear of crime. As sentient and symbolic beings, however,
humans have the ability to anticipate or contemplate events that lie in the future
or are not immediately apparent. Hence people may experience fear merely in
anticipation of possible threats or in reaction to environmental cues (e.g., darkness,
graffiti) that imply danger. Psychologists commonly use the terms fear
and anxiety to differentiate reactions to immediate threats (fear) from reactions
to future or past events (anxiety). This terminological clarity has not been
adopted in research on fear of crime, but it appears that most measures of fear
are designed to capture anxiety rather than fear of victimization. This practice
evidently rests on the assumption that anxiety about future victimization is much
more common among the general public than fear associated with actual
encounters with crime, a reasonable assumption (see Warr 1994). Hereafter,
I will draw the distinction between fear and anxiety where it is heuristically
useful or otherwise appropriate.
By its very nature, the term fear seems to imply a deleterious emotional or psychological
condition. Unlike love, pleasure, or happiness, fear is not a state that
people (thrill-seekers aside) ordinarily pursue. To assume that fear is therefore
dysfunctional for an organism, however, is to commit a serious error. On the
contrary, the presence of fear in virtually all animals is no accident. Without
fear, prey animals would walk amid predators, and humans would stroll across
busy freeways, knowingly ingest toxic substances, or leave their infants unprotected.
From an evolutionary point of view, animals that lacked fear would be
unlikely to live long enough to reproduce, suggesting that fear is a potent natural
selection factor (Russell 1979; Mayes 1979).
Fear, then, is not intrinsically bad. It is when fear is out of proportion to
objective risk that it becomes dysfunctional. We will return to that issue later
as we consider the control of fear.
Fear of crime can be characterized according to a
number of properties, including intensity (the English
language recognizes many degrees of fear: terror,
worry, alarm, apprehension, dread), prevalence (the
proportion of a population that experiences fear during
some reference period), and duration, both among
individuals and within social units (e.g., communities,
cities, nations). Because actual criminal events or
exposure to immediate signs of danger are usually
brief, episodes of fear (strictly defined) are usually
brief as well. Anxiety, on the other hand, can become
a chronic or obsessive condition (Sluckin 1979).
When individuals face an ostensibly dangerous environment, they may naturally
experience fear for their own personal safety. At the same time, they may
also fear for other individuals (e.g., children, spouses, friends) whose safety
they value. It is essential, therefore, to distinguish personal fear (fear for oneself)
from altruistic fear (fear for others). The prevalence and power of altruistic
fear are illustrated by the enormous public reaction that often attends crimes
committed against children (e.g., Polly Klaas, Columbine High School). Such
reactions surely reflect not only distress for the victim, but also parents’ profound
concern for the safety of their own children.
The prevalence and
power of altruistic
fear are illustrated
by the enormous
public reaction that
often attends crimes
committed against
One of the strongest indictments that can be leveled against research on fear
of crime is the continuing failure of investigators to collect systematic data
on altruistic fear, or even to recognize its existence. It is entirely possible that
altruistic fear is as prevalent as personal fear (perhaps more so) and has consequences
that are distinct from or amplify those arising from personal fear.
Research on altruistic fear could also provide insights into the sociometry of fear
in social organizations. For example, in family households, do wives fear for
their husbands as much as husbands do for wives? Do they share equal fear for
their children? How does the age or sex of children affect their parents’ fear?
The Measurement of Fear
Fear of crime can be measured by soliciting self-reports from individuals or by
monitoring physiological processes associated with fear. The emotion of fear is
ordinarily accompanied by certain involuntary physiological changes, and these
can be used as indicia to measure the presence or intensity of fear (Thomson
1979; Mayes 1979). One of the potential advantages of physiological measures
of fear is that it enables the measurement or monitoring of fear as it is occurring,
that is, in real time in natural settings. Because fear is often a fleeting emotion
and may occur at inopportune times or places (e.g., late at night in a downtown
parking lot), this is no minor advantage. Another related benefit of physiological
measures of fear is that they eliminate many of the problems associated
with self-reports, including errors in recall, demand effects, or reluctance to
disclose emotions.
Physiological measures of fear have certain limitations, however. They cannot
directly reveal the source of fear, i.e., the persons, things, or events to which
the subject is reacting. Furthermore, they cannot distinguish fear of crime from
other forms of fear (e.g., fear of accidents or threatening weather). These limitations
may present few problems in controlled laboratory experiments (when,
for example, subjects are presented with dangerous or innocuous scenes)
because the cues or stimuli of interest can be isolated and confounding cues
eliminated or controlled. However, the number and variety of cues that appear
in natural settings suggest that physiological measures of fear may be of limited
value in nonexperimental research. Moreover, the physiological changes commonly
associated with fear can accompany other emotional states as well
(Thomson 1979; Mayes 1979). Thus, for example, there appears to be little
physiological basis for distinguishing between persons who react to a violent
threat with anger and those who react with fear. Still another problem is that
feelings of fear and physiological reactions appear to be more strongly coupled
under some circumstances (when, for example, fear is intense) than others
(Mayes 1979).
Despite these possible pitfalls, there is a pressing need to explore the uses of
physiological measures of fear, because the payoff in knowledge is potentially
great. Consider some of the questions that might be answered using a continuous,
unobtrusive measure of fear:
n Does fear follow a reliable daily or weekly periodicity?
n What microenvironments—blocks, businesses (e.g., bars), neighborhoods—
are most conducive to fear?
n How is fear affected by the presence or absence of companions or
n Do certain kinds of persons (minority members, the homeless) evoke fear
among some individuals?
n Is being alone in public places more frightening than being with strangers?
n In which types of routine activities—school, work, shopping, or home—is
fear most pronounced?
n Does carrying a weapon outside the home reduce or actually exacerbate fear?
n What are the less obvious precautionary behaviors that people undertake
in response to fear (e.g., scheduling habits, monitoring the whereabouts
of others)?
Survey measures of fear
Survey research on fear of crime is extensive, but a truly bewildering variety
of questions have been used by investigators over the years to measure fear of
crime (see Ferraro 1995; Ferraro and LaGrange 1987; DuBow, McCabe, and
Kaplan 1979). Much of this diversity stems from variation in the context stipulated
in survey questions. Some questions ask about fear during the day; others,
about fear at night. Some pertain to fear at home, whereas others question
respondents about their fear in their own neighborhood or in their city. Still
others ask respondents about their fear when alone or with others. Such sensitivity
to context among researchers is commendable, but it is of little value
unless such contextual variables are systematically varied and their effects
evaluated. Unfortunately, that is rarely the case.
One item, however, has become the de facto standard for measuring fear of
crime: “Is there anywhere near where you live—that is, within a mile—where
you would be afraid to walk alone at night?” The item has become conventional
not because it was chosen by social scientists, but because it has been routinely
used by the Gallup Organization and the National Opinion Research Center
(NORC) to measure fear since the 1960s. During the past three decades, approximately
40 to 50 percent of Americans surveyed each year have responded
affirmatively to this question (for a review, see Warr 1995a).
The Gallup/NORC item has been criticized (e.g., Ferraro 1995) on many grounds:
it is hypothetical (how afraid would you be), is limited to nighttime, does not
mention crime, and only crudely measures intensity. In fairness, the measured
prevalence of fear obtained with this item is not radically different from that
measured in other national surveys (see Warr 1995a), and the routine use of the
item permits longitudinal comparisons of fear, if only in relative terms.
Much deeper issues are raised by questions of this kind, however. Almost two
decades ago, Warr and Stafford (1983) asked residents of Seattle to report their
everyday fear, not of “crime” in general, but of a variety of specific offenses
ranging from violent crimes like homicide, rape, and robbery to various property
and public order offenses. Even today, the rank order of offenses that emerged
from their analysis remains startling to many. Murder, for example, was low on
the list of fears, while residential burglary outranked all other offenses on fear.
Warr and Stafford demonstrated that these findings were not anomalous or even
counterintuitive. Contrary to common assumption, they showed, fear is not
determined simply by the perceived seriousness of an offense. Instead, the
degree of fear attached to particular crimes is a multiplicative function of the
perceived seriousness and perceived risk of the offenses. To generate strong
fear, an offense must be perceived as both serious and likely to occur. Residential
burglary is the most feared crime in the United States because it is viewed as
both relatively serious and rather likely. Murder, on the other hand, is perceived
to be very serious but unlikely to occur.
Since the publication of Warr and Stafford’s findings, only scattered offensespecific
data on fear have been gathered (see Warr 1995a; Ferraro 1995;
Haghighi and Sorensen 1996). These data generally corroborate the hierarchy
of fears observed by Warr and Stafford (insofar as they use comparable offenses),
but fear continues to be monitored primarily through generalized, omnibus
measures of the sort used by Gallup and NORC. As a consequence, important
questions about fear remain unanswered today. For example, when respondents
report fear of “crime” in social surveys, what specific offenses do they have in
mind? Are those offenses similar across individuals? The answer is that they
almost surely are not; fear of rape, for example, is very pronounced among
women but presumably not among men (see Warr 1985). On a different question,
are particular precautionary behaviors, such as spatial avoidance and timeshifting
(carrying out the same activity at an ostensibly safer time), linked to
fears of particular crimes?
Another limitation of current survey data is that there are no time-series data
on fear of individual offenses. Is rape feared today more than in the 1980s? Do
offense-specific fears follow offense-specific trends in the incidence of crime,
or do they respond to some “master” offense? Geographic variation in fear
should also be considered. Do residents of large cities fear the same offenses
as suburbanites or small-town residents? How far does the fear inspired by a
particular incident spread, and how does it vary with the nature of the offense?
Without systematic, offense-specific data on fear of crime, questions of this
sort cannot be answered.
None of this is necessarily to discount the value of omnibus measures of fear. It
is not unreasonable to assume that individuals can report an overall assessment
of their fear about “crime” as a category of risk, any more than it is possible to
list and measure all conceivable offenses that individuals might fear. Omnibus
measures, in short, are useful as a complement to, but not a substitute for,
offense-specific measures of fear.
Behavioral indicators of fear
Nearly all those who have investigated the emotion of fear agree that fear reveals
itself through behavior, from the myriad responses of nonhuman species (distress
cries, freezing, defecation, tonic immobility, or feigning death) to the complex and
sometimes subtle avoidance behaviors of humans (Sluckin 1979). A major problem
with behavioral indicators of fear in humans, however, is the difficulty of
ascertaining exactly what people are not doing (or are doing) out of fear, and convincingly
linking it back to fear. Is it obvious that a person with fear of heights is
intentionally avoiding tall buildings, bridges, or amusement rides? Is it evident that
a person who fears drowning showers rather than bathes for that reason?
However difficult it may be to establish, the link between fear and behavior
underscores one of the great ironies of fear: Those most profoundly affected
by fear—fear of flying, fear of automobile accidents—may rarely experience
it because they have taken extraordinary measures to avoid the source of their
fear (Kenny 1963). In the end, then, behavior may be the best indicator of fear,
but the behaviors through which fear makes itself known are not always easily
identifiable or detectable.
Transient public episodes of fear
When it comes to measuring fear of crime, what are the appropriate units of
analysis (individuals, neighborhoods, cities, nations)? And what is the appropriate
time interval for measurement (hourly, daily, monthly, yearly)? The answers
depend, of course, on the question to be answered, but the conventional method
for measuring fear—annual surveys employing national samples—is apt to
overlook crucial aspects of fear.
Consider an example. No close observer of American society will fail to notice
that certain horrific criminal events seize the attention of the general public
and become matters of almost universal discussion, speculation, and concern.
Although some gain national attention, most remain matters of more local concern,
affecting a particular city or portion of a city. In my own city of Austin,
Texas, the gruesome murder of four teenage girls in a yogurt store created
something approaching mass hysteria in the city and remained the lead story
on local television news for many weeks thereafter. Landlords reported that it
was difficult to lease apartments in the part of town where the crime occurred,
and nearby business owners reported significant declines in revenues.
The annual national surveys ordinarily used to measure fear of crime are too
coarse, both spatially and temporally, to capture these sorts of events. Put
another way, the scale of such surveys simply does not match the scale of certain
events that ought to be measured. Because of this, little is known about the
natural history of localized urban “panics,” even though they are perhaps the
most common social outbreaks of fear. How long do events of this kind ordinarily
last? Does fear decay gradually or subside suddenly? Once initial media
attention wanes, does fear decline as well? Does the arrest of a suspect affect
the course of fear? Does fear decay at different rates among various segments
of the population (young and old, male and female)? Even after fear fades, do
events of this kind become part of the collective memory and lore attached to
districts of a city? (“The south side is dangerous—remember that girl who was
killed?”) Put differently, is there a permanent “residue” of fear that remains
behind after such events?
Filling such gaps in existing knowledge will require surveys of social units
much smaller than the nation as a whole. And because the events that incite
such incidents are unpredictable and require pre-event (i.e., baseline) measures
and repeated postevent measures, the only feasible research strategy is to
implement routine surveys in selected jurisdictions. The best design would be
a series of small, periodic (e.g., monthly) sample surveys in perhaps a dozen
cities over several years. Capturing one transient episode is not enough to adequately
describe such events because the duration and intensity of events is
likely to depend on characteristics of the crime. (What age and sex were the
victim and offender? Was the event provoked?)
Why is it important to measure smaller scale events of this sort? Apart from
determining the frequency and geographic dispersion of such events, initial
public reactions to criminal events often seem to be based on local media
reports that are sketchy, hasty, and so short on critical details (Did the victim
know the suspect? Is the suspect still at large? Has this happened before?)
that an intelligent evaluation of the event is all but impossible. In such situations,
some individuals will assume the worst and act accordingly. Understanding
the features of criminal events that determine the intensity and duration of
public panics might lead to more judicious reporting on crime and less
unnecessary fear.
Regulating Public Fear of Crime
Social scientists are inclined to approach fear of crime by asking the sorts of
questions they raise about other human phenomena: What are its causes? What
are its consequences? What are its contemporary and historical parameters
(incidence, prevalence, social distribution)? Policymakers may be interested
in these questions as well, but ultimately they must confront other immediate
issues, some empirical and some normative. Can this phenomenon be controlled,
and, if so, at what expense? Assuming it can be controlled, should it be
controlled? Are there harms as well as benefits associated with intervention?
These questions will frame our discussion of this complicated issue.
Should fear be controlled?
Imagine for a moment that we possessed a magic dial through which we could
control or regulate fear of crime in the United States. Turn the dial to the left
and fear immediately decreases proportionately; turn it to the right, and fear
rises proportionately.
With the public interest in mind, no doubt our first inclination would be to reduce
fear substantially by rotating the dial far to the left. Suppose, however, that the
risks of crime are in fact real and substantial. Were we to greatly reduce fear of
crime with our device, we would concomitantly increase the chances that individuals
would fail to take necessary precautions to protect their own safety (or the
safety of others), and thereby increase the risk of victimization.
Reducing fear, in other words, is not necessarily an unqualified good or costfree;
by relieving fear, we stand the chance of increasing public injury. On the
other hand, were we to turn the dial too far to the right, people would engage
in needless precautions and unnecessarily constrain their own lives. At the
extreme is a “fortress society” in which citizens withdraw from public life
altogether and everyday social intercourse is sharply curtailed.
Which way should the dial be turned, then, or should it in fact be touched at
all? In the real world, of course, there is no magic dial or any direct way to
manipulate an emotion like fear (unless one were to propose dispensing sedatives
or other pharmacological agents to the public). Instead, we must attempt
to control fear by controlling its causes. As noted earlier, research by Warr and
others consistently suggests that the proximate cause of fear for any one crime
(that is, ignoring differences in the seriousness of crimes) is the perceived risk
of that crime. Altering fear, then, requires altering perceptions of risk.
Reformulating the question somewhat, then, which way should the “perceived
risk” dial be turned? The answer ultimately depends on the relation between
perceived risk and objective risk. Plot a in exhibit 1 illustrates a situation in
which perceived risk precisely matches objective risk; any increase or decrease
in the latter (over time or place or across crimes) is always matched by a
change in the former. In plot b, however, perceived risk always exceeds objective
risk by a fixed amount; the public consistently overestimates the risk of
victimization. In plot c, exactly the opposite is true; objective risk is greater
than the public realizes. In the final plot, d, the relation between objective and
perceived risk is more complex. When objective risk is low, the public overestimates
risk; when objective risk is high, the public underestimates risk.
What are the policy options implicit in these examples? If the world truly operated
as in plot a, the policy choice would be abundantly clear: Do not touch the
perceived risk dial; public perceptions are accurate and existing levels of fear
are justified. As noted earlier, many criminologists tacitly assume a tight connection
between perception and reality when it comes to crime, and thus feel
free (indeed, obliged) to set aside perceptions in favor of concentrating on
crime reduction itself. The same seems to be true of politicians, who often
advocate crime reduction without questioning whether public beliefs about
crime are accurate or in need of alteration. Such approaches rest on an implicit
but untested assumption, to wit, that any reduction in objective risk will be
noticed and appreciated by the public.
If the world were as in plot b—the public overestimates risk and is needlessly
afraid—then one would want to turn the dial down and reduce what is clearly
unnecessary fear. This is surely the most desirable and morally unequivocal
situation because it results in a gain in the public good without any increased
exposure to danger. This situation would be difficult to ignore once it is recognized,
and it is every social engineer’s dream. If, conversely, plot c were true,
we would be obligated to raise perceived risk so that the public would be suitably
afraid. This is a morally defensible but politically difficult task, one in
which the objective is literally to frighten the public. Though it might seem
extreme, it is not without precedent; public campaigns against smoking,
teenage pregnancy, AIDS, lead paint, dietary salt, and prescription drug
risks are but a few examples (e.g., Slovic, Fischhoff, and Lichtenstein 1982;
Objective risk
Plot a
Plot b
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Objective risk
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Perceived risk
Perceived risk
Exhibit 1. Perceived risk versus objective risk
0.1 0.2 0.3 0.4 0.5 0.6 0.7 1.0
Perceived risk
0.0 0.8 0.9
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Objective risk
Plot c
Plot d
Objective risk
Perceived risk
Exhibit 1 (continued)
Fischhoff, Bostrom, and Quadrel 1997). Plot d presents
a more complicated case, but the policy implication
is no different from the preceding instances:
take steps to bring objective and perceived risk into
One might perhaps want to entertain other policy
options, of course. If fear reduction were the only
consideration, one might argue that no change is
required for plot c; in other words, ignorance is bliss.
As we have seen, however, ignorance in this case is
not the absence of harm, and there is no morally
defensible theory that would justify that strategy.
Similarly, the situation depicted in plot b would in all
likelihood produce the lowest possible rate of criminal victimization, but at
a personal and social cost that would be difficult to justify.
In the end, then, the question comes down to this: Which of these plots describes
the real world? The weight of contemporary evidence suggests that the general
public probably exaggerates the risk of serious criminal victimization in a way
that resembles plot d.
What is this evidence? A small but persuasive body of studies in cognitive psychology
(see Lichtenstein et al. 1978; Slovic, Fischhoff, and Lichtenstein 1979,
1980, 1982, 1987) indicates that individuals tend to significantly exaggerate the
risk of rare lethal events (that is, causes of death like tornadoes, homicide, floods,
fire, accidents, or botulism), while underestimating the risk of common lethal
events (e.g., deaths due to heart disease, diabetes, or cancer). Slovic, Fischhoff,
and Lichtenstein (1980, 1982) attribute this tendency to a common error of judgment
arising from the availability heuristic (Tversky and Kahneman 1982) or the
tendency to judge the frequency of events by the ease with which they can be
recalled or imagined.
Why would individuals readily imagine or remember what are actually rare
causes of death? Slovic, Fischhoff, and Lichtenstein (1980, 1982) cite evidence
from Combs and Slovic (1979) showing that public perceptions concerning the
frequency of causes of death closely match the frequency with which those
causes are reported in newspapers. Newspaper accounts, in turn, are glaringly
at odds with reality:
[M]any of the statistically frequent causes of death (e.g., diabetes, emphysema,
various forms of cancer) were rarely reported by either paper during
the period under study. In addition, violent, often catastrophic, events such
Findings suggest
that the public is
likely to exaggerate
the frequency of
rare, serious crimes
and underestimate
the frequency of
more common,
less serious ones.
as tornadoes, fires, drownings, homicides, motor vehicle accidents, and all
accidents were reported much more frequently than less dramatic causes
of death having similar (or even greater) statistical frequencies. For example,
diseases take about 16 times as many lives as accidents, but there
were more than 3 times as many articles about accidents, noting almost 7
times as many deaths. Among the more frequent events, homicides were
the most heavily reported category in proportion to actual frequency.
Although diseases claim almost 100 times as many lives as do homicides,
there were about 3 times as many articles about homicides as about disease
deaths. Furthermore, homicide articles tended to be more than twice
as long as articles reporting disease and accident deaths. (Slovic,
Fischhoff, and Lichtenstein 1982, 468)
These investigators did not insist on a causal connection between media reports
and public perceptions, but they suggested that the pattern of errors in the two
is much too similar to be coincidental.
How do perceptions about one class of hazards (causes of death) relate to
another (the risk of criminal victimization)? Because the most serious crimes
(homicide, rape, robbery) are also the rarest forms of crime, the preceding findings
suggest that the public is likely to exaggerate the frequency of rare, serious
crimes and underestimate the frequency of more common, less serious ones.
In the early 1980s, Warr (1980; see also Bordley 1982) presented direct evidence
of this phenomenon, showing that the objective and perceived incidence
of offenses in four cities were related by a power function (y=aXb). That is,
people tended to systematically overestimate the frequency of rare offenses
while underestimating the frequency of common ones. Public perceptions
were remarkably accurate as to the relative frequency
of different crimes (for example, people recognize
that homicide is less common than burglary), but
considerably less accurate as to absolute frequencies.
Aside from these findings, there is an another reason
to suspect that the true relation between perceived
and objective risk probably resembles plot d (or perhaps
b). When the general public is asked where they
obtain most of their information about crime, the
resounding answer is the mass media, especially
news coverage of crime. Graber (1980), for example,
reported that 95 percent of respondents in her survey
identified the media as their primary source of information
on crime, although 38 percent cited other
sources as well (conversations or, more rarely,
The mass media
are thus a powerful
amplifying mechanism
when it comes
to crime; information
known only to a few
can within hours or
days become known
to thousands
or millions.
personal experience). Skogan and Maxfield (1981)
found that more than three-quarters of respondents in
the three cities they surveyed reported watching or
reading a crime story on the previous day (44 percent
had read a newspaper crime story, 45 percent had
watched a crime story on television, and 24 percent
had done both). The mass media are thus a powerful
amplifying mechanism when it comes to crime; information
known only to a few can within hours or days
become known to thousands or millions.
What is the image of crime presented in the mass media? A number of forms
of distortion in news coverage of crime have been identified and documented,
distortions that tend to exaggerate the frequency and the seriousness of crime.
In the real world, for example, crimes occur in inverse proportion to their seriousness;
the more serious the crime, the more rarely it occurs (e.g., Erickson
and Gibbs 1979). Thus, in the United States, burglaries occur by the millions,
robberies by the hundreds of thousands, and homicides by the thousands. In
news coverage of crime, however, the emphasis is on “newsworthiness,” and
a key element of newsworthiness is seriousness; the more serious a crime, the
more likely it is to be reported. By using seriousness as a criterion, however,
the media are most likely to report precisely those crimes that are least likely
to occur (Skogan and Maxfield 1981; Sherizen 1978; Sheley and Ashkins 1981;
Roshier 1973), or exactly the same pattern outlined previously for lethal events.
Among other things, this “mirror image” depiction of crime means that the
media place extraordinary emphasis on violent crime. Skogan and Maxfield
(1980) reported that homicides and attempted homicides constituted one-half
of all newspaper crime stories in the cities they examined, even though homicides
are but a minute fraction of all offenses. Furthermore, the number of
homicide stories reported in city newspapers, they found, did not closely match
the actual homicide rates of the cities examined, suggesting that the amount
of space devoted to crime has more to do with the “newshole” allocated to crime
by editors than with the true crime rate.
News coverage of crime has been criticized on other grounds as well, including
the practice of using crime news as “filler” when other news is slow, the use of
crime news (“If it bleeds, it leads”) to attract larger audiences, and the tendency
to report trends in crime using numbers rather than rates, thereby ignoring
changes in population (see Graber 1980; Warr 1980, 1994, 1995b).
The fact that the media present a distorted image of crime, of course, is no guarantee
that the public believes or heeds what it sees, hears, and reads. And public
The media are most
likely to report precisely
those crimes
that are least likely
to occur.
perceptions (or exaggerations) of the incidence of crimes do not necessarily
translate into estimates of personal risk (Slovic, Fischhoff, and Lichtenstein
1982). Still, the evidence on public perceptions of crime and media distortion of
crime news is strikingly corroborative, and it is difficult to believe that the media
have little or no effect on perceptions, especially when the public cites the media
as their primary source of information on crime and spends so much time watching,
reading, and listening to the media (Skogan and Maxfield 1981).
Given the gravity of fear as a social problem and the presumptive role of the
mass media, it is truly astonishing to find that there is virtually no systematic
research that assesses the impact of the media on public perceptions of crime
or fear of crime. To be sure, there is known to be a positive correlation between
fear of crime and the number of hours spent watching television (Skogan and
Maxfield 1981), but the causal direction is unclear and the correlation may well
be spurious with respect to age and other viewer characteristics.
In the end, the causal influence of media crime coverage cannot be established
without simultaneous measurements of (1) media content, (2) public exposure to
that content, and (3) the postexposure effects of media communications. Such
research is difficult to conduct in natural settings because of the enormous
quantity and variety of media and interpersonal messages on crime to which
the public is exposed (e.g., Graber 1980). Remarkably, a great deal of Federal
money is spent today to document the extent of crime and violence on television,
but rarely is there any accompanying research on the effects of such televised
violence on those who are exposed to it. It is, in essence, a research
design with no dependent variable.
One study that approaches an ideal design and points the way for future research
was conducted by Heath (1984). She questioned samples of newspaper readers
in 36 cities and examined their fear of victimization in light of the characteristics
of the newspapers they read. Heath found that fear was higher among readers
of newspapers that emphasized local crimes and crimes that were sensational
(bizarre or violent) or random (apparently unprovoked). However, reports of
sensational or random crimes evidently reduced fear if those crimes were not
local. Apparently, readers were reassured by learning that such crimes were
occurring to other people in other places.
Can fear be regulated?
The discussion thus far leads to a tentative conclusion that the public exaggerates
the risk of serious criminal victimization. It is worthwhile to reiterate that
this situation, if true, is the most desirable problem to remediate; fear can legitimately
be reduced without any attendant increase in the risk of victimization.
Let us turn now to the second question posed earlier. Assuming that fear ought
to be regulated, can it be regulated, and, if so, how?
There are two general approaches to this problem. One might be called selfcorrective,
meaning that it focuses on means to alter the way that crime is
presently depicted in the mass media. The other approach could be described
as counteractive, meaning that it attempts to discount or replace messages
promulgated through the mass media.
Let us begin with the former approach, bearing in mind that media news coverage
is the public’s primary source of information about crime. Assuming that
such coverage substantially affects public perceptions of crime, how might
news coverage of crime be changed for the better?
Consider the characteristics of everyday news coverage. Crime is ordinarily
reported in the form of isolated, discrete incidents (“three young adults were
injured today in a standoff with police”) or occasionally as counts (“thus far
this year, 13 robberies have been reported to the police”). (See, for example,
Graber 1980.) These reported events are not a complete or exhaustive list of
all crimes. Instead, they are selected from a much larger pool of crime events
available for reporting. What is more, the selection process is not random or
representative, but quite the opposite. Not only are violent crimes disproportionately
emphasized (particularly homicides), but crimes may be selected
merely because they are odd or unusual, involve prominent persons or public
figures, or fit a preestablished journalistic theme like “crimes against the elderly”
or “careless tourists” (Skogan and Maxfield 1981; Gordon and Heath
1981; Sherizen 1978; Ericson, Baranek, and Chan 1987; Fishman 1978, 1981).
To imagine the consequences of such reporting practices, consider this question:
Could an individual reliably estimate the magnitude and causes of population
growth in a city through isolated, incomplete, and nonrepresentative
interviews with those who have left or
have moved in?
From a public information perspective, what often
seems to be missing in news coverage of crime is not
raw information on criminal events (on the contrary) but
an informed perspective about crime risks. Only occasionally
are criminal events presented as populationbased
rates, from which one could estimate personal
risk or risk to loved ones. And rarely are such rates
placed in any sort of seasonal, historical, demographic,
or geographic context. From the point of view
Trying to detect
patterns or achieve
valid inferences
about crime from
isolated, sporadic
news accounts
is an exercise
in futility.
of readers or viewers, trying to detect patterns or achieve valid inferences about
crime from isolated, sporadic news accounts is an exercise in futility.
What is seriously lacking in news reporting, and might be of greatest benefit
to the public, is information about the risk of criminal victimization relative to
other aversive or benchmark life events. To illustrate, what is the probability
that I will be robbed this year compared with the chances that I will be
involved in a serious automobile collision, eat contaminated food in a restaurant,
contract an infectious disease at work or school, or suffer a heart attack?
How does my age, sex, racial/ethnic identity, or location affect my chances?
Many Americans, including journalists, might be surprised to learn that they
are more likely to be a victim of suicide than homicide, that automobiles kill
more individuals than all violent crime, or that, as a group, children face
greater danger from their parents than from strangers.
The didactic value of what I have called informed perspective can be illustrated
by comparing two possible television news accounts of the same hypothetical
event. The first reads:
A homicide occurred late last evening at 223 East Lansing. The victim,
a 23-year-old male, was stabbed twice and died shortly thereafter at Our
Lady General Hospital. According to the police, no arrest has been made.
Now add these words:
Fewer than 1 in 10,000 persons living in our city are victims of homicide
each year. Most, as in this case, are young males who die in alcohol-related
arguments with persons they know. The number of homicides thus far
this year—27—is no higher than average for the previous 5 years, and
two-thirds of those homicides occurred within the same three census tracts
of the city. For more information, contact the Metro Police Department
at 366–8942.
Journalists may object that the latter version is too long and dull, but a crime
worth reporting is surely worth properly contextualizing. In defense of journalists,
police crime reports often are sketchy or incomplete, and there is the
pressure of deadlines. On the other hand, it is not the police who are eager to
release incomplete reports, and reporters’ professions about public safety are
sometimes little more than transparent ruses to be the first to release the story.
In any event, audiences are likely to fill in sketchy or missing information by
assuming the worst, which is all the more reason to place reports of crime
within a larger factual context.
Counteractive measures
Aside from (or in addition to) changing media crime-reporting practices, information
on crime can be promulgated through alternative channels. Messages
about crime can be disseminated through a variety of means, including pamphlets;
billboards, transit ads, and other signage; magazine and newspaper ads; Web
sites; and oral presentations at public gatherings. Such a strategy might seem
inconsequential compared with the awesome power of television and newspapers,
but messages of this type have figured heavily in public campaigns about
smoking, heart disease, and other health risks (see generally, National Research
Council 1989).
When it comes to crime, at least two public agencies are a natural choice for
conveying such messages. One is the municipal police department. Most modern
police departments have a public information office to dispense information
on crime to persons (often journalists) on request. There is no major logical or
logistical jump in moving from a reactive function of this type to a more proactive
version of public information. The logical connection between crime and
the police makes the police a perfect agent for crime communications, and,
notwithstanding occasional scandals, the police enjoy enormous public support
in the United States (e.g., Warr 1995a). I witnessed this function of the police
when they embarked on a door-to-door campaign to inform residents of one
Austin neighborhood that several rapes had recently occurred there. Where the
costs of distribution pose a problem to police departments, there is no lack of
civic organizations and volunteers willing to place pamphlets or booklets in
mailboxes or on front porches.
Slovic, Fischhoff, and Lichtenstein (1982, 484) have also argued that a proper
setting for communicating risks is the school:
Informing people, whether by warning labels, package inserts, or extensive
media programs, is but part of the larger problem of helping people cope
with the risks and uncertainties of modern life. We believe that some of the
responsibility lies with our schools. Public school curricula should include
material designed to teach people that the world in which they live is probabilistic,
not deterministic, and to help them learn judgment and decision
strategies for dealing with that world. [T]hese strategies are as necessary for
navigating in a world of uncertain information as geometry and trigonometry
are to navigating among physical objects.
Although modern schools are swamped with demands for their time and suggestions
about their curriculums, a case can be made that the risks of crime
are of sufficient size and gravity that at least some time ought to be devoted
to them, if only to alleviate unnecessary fear throughout life. In occasional
lectures to schools and churches, I have been struck by the almost desperate
hunger of many people for objective information on crime and its risks.
No matter who the messenger is, what is the content that messages about crime
should convey? In recent years, an entirely new field of research in science has
emerged known as risk communication (cf. National Research Council 1989).
Concerned with the methods, problems, and efficacy of communicating risk to
the general public, the field has concentrated largely on new technological risks
(nuclear power, pesticides, toxic waste disposal), medical/health risks (recombinant
DNA, smoking, seat belt use, high cholesterol, alcohol abuse, cancer) and
both natural and manmade disasters (hurricanes, floods, aircraft crashes, lightening,
tornadoes, earthquakes). No one, to my knowledge, has grappled with
risk communication about crime, but the lessons of this field are nonetheless
useful and enlightening.
One lesson brings to mind the primary obligation of the physician: First do
no harm. The effect of a communication can only be ascertained empirically.
Untested messages can have unintended consequences, and therein lies the
Poor risk communications may cause more damage than the risks they
are intended to control. They can lead to wrong decisions by omitting key
information or failing to contradict misconceptions. They can create confusion
by prompting inappropriate assumptions or emphasizing irrelevant
information and produce conflict by eroding the audience’s faith in the
communicator. They can cause recipients to be unduly alarmed or complacent
or to undertake ineffective actions. Because communicators’ intuitions
about recipients’ perceptions cannot be trusted, there is no substitute for
empirical validation. (Bostrom et al. 1994, 796)
Fischhoff, Bostrom, and Quadrel (1997, 993) similarly observed:
Effective risk communication requires careful empirical research. Poor
risk communication can often cause more public health (and economic)
damage than the risk it attempts to describe. One should no more release
an untested communication than an untested medical device.
In short, risk communications must be pretested before they are disseminated.
Fischhoff (1989) argues that an essential prerequisite for designing risk communication
is the need to know what the public does not know, a matter that
ordinarily requires empirical research and cannot simply be presumed. In that
connection, Fischhoff, Bostrom, and Quadrel (1997) offer an ingenious idea
for altering public perceptions of risk. In keeping with evidence that people are
often highly overconfident about the information or beliefs they hold, these
investigators suggest that one function of communications is to give people
“the appropriate degree of confidence in their beliefs,” especially in cases
“where people confidently hold incorrect beliefs that could lead to inappropriate
actions” (1997, 997). As an example of misplaced confidence, they cite an
investigation showing that a majority of teenagers are not aware that a single
beer affects driving ability as much as a shot of vodka and that mistaken
teenagers are usually very confident about their incorrect information. Studies
of this kind are particularly useful for identifying misconceptions that need to
be targeted in messages.
How should communications about risk be constructed? What elements should
they contain? Information about risk is often highly technical, but technical
terms and examples should be avoided (Covello, von Winterfeldt, and Slovic
1987; Fischhoff 1989). Risks can be presented in everyday terms using alternative
examples. For example, the proportion of Americans who are murdered
each year in the United States (fewer than 1 in 10,000 per year) is roughly the
same as 1 day in 27 years, or 1 inch in 833 feet, or 1 gallon in a home swimming
pool. By contrast, the crude probability that a household will be burgled
in the United States is about 1 in 10–20 per year, or roughly the chance of
drawing two consecutive cards of the same suit from a fresh poker deck. Crosshazard
comparisons (where the risk of, say, murder is compared with the risk
of an auto accident, disease, or lightning strike) can be useful for illustrating
risk as well, but they can be difficult to interpret or understand if not properly
constructed (Slovic, Fischhoff, and Lichtenstein 1982).
Information in communications must also be relevant to the audience:
Poorly chosen information can . . . be seen as wasting their time (indicating
insensitivity to their situation), . . . can take up the place (in the media
or school) that could be filled with pertinent information (imposing an
opportunity cost), and . . . can lead them to misunderstand the extent
of their knowledge. (Fischhoff, Bostrom, and Quadrel 1997, 997)
Communicators must also ensure that events or risks discussed in the communication
have the same meaning to recipients as to themselves (e.g., Do respondents
know what a burglary is? Do they confuse it with robbery?), and they
need to be honest and straightforward about the limitations of their own information
(Fischhoff 1994; Fischhoff, Bostrom, and Quadrel 1997). The public
is often skeptical of experts and government officials, and honesty about the
accuracy of their estimates can help to offset mistrust.
It is impossible to fully survey the literature on risk communication here, but
it is useful to offer two summaries of effective communications by prominent
investigators in the field. These are not merely seat-of-the pants recommendations
but careful statements based on extensive research and experience. First,
in an appendix to a National Academy of Sciences conference on risk assessment,
Covello, von Winterfeldt, and Slovic (1987, 117–118) offer the following
advice on risk communication:
n Use simple, graphic, and concrete material, avoiding technical or
specialized language wherever possible.
n Compare risks within a carefully defined context that is relevant to the
target audience.
n Avoid comparisons of risk that may appear to the audience to be noncomparable
because of different qualitative characteristics—for example,
the risk of smoking compared to that of living near a nuclear power
n Understand and recognize qualitative concerns, such as concerns
about catastrophic potential, dread, equity, and controllability.
n Identify and explain strengths and limitations of different risk measures,
and present (whenever possible) alternative indexes of risk—for example,
measured or expected fatalities or incidences of diseases for the
entire population and for the most- and least-exposed individuals.
n Identify, acknowledge, and explain uncertainties in risk estimates.
n Provide opportunities for people to learn how to interpret risk information.
n Relate on a personal level—that is, when people ask personal questions
such as “Can I drink the water?” respond in a personal way without
minimizing risks and uncertainties.
n Recognize the power of subtle changes in the way that information is
presented and use such knowledge responsibly.
n Understand and recognize that health and environmental debates often
involve much broader considerations, including political values and
In addition, Fischhoff, Bostrom, and Quadrel (1997, 998) offer this advice:
Once information has been selected, it must be presented in a comprehensible
way. That means taking into account the

Curriculum Vitae

Profile Pages