Int J Biometeorol (2007) 52:43–55
DOI 10.1007/s00484-006-0073-4

CLIMATE AND PERCEPTION

The social impacts of the heat–health watch/warning system
in Phoenix, Arizona: assessing the perceived risk
and response of the public
Adam J. Kalkstein & Scott C. Sheridan

Received: 26 May 2006 / Revised: 20 October 2006 / Accepted: 9 November 2006 / Published online: 30 January 2007
# ISB 2007

Abstract Heat is the leading weather-related killer in the
United States. Although previous research suggests that
social influences affect human responses to natural disaster
warnings, no studies have examined the social impacts of
heat or heat warnings on a population. Here, 201 surveys
were distributed in Metropolitan Phoenix to determine the
social impacts of the heat warning system, or more
specifically, to gauge risk perception and warning response.
Consistent with previous research, increased risk perception
of heat results in increased response to a warning. Different
social factors such as sex, race, age, and income all play an
important role in determining whether or not people will
respond to a warning. In particular, there is a strong sense
of perceived risk to the heat among Hispanics which
translates to increased response when heat warnings are
issued. Based on these findings, suggestions are presented
to help improve the Phoenix Heat Warning System.
Keywords Heat . Weather warnings . Weather . Climate .
Perception

Introduction
The Centers for Disease Control estimate that from 1979
through 1999, 8,015 deaths were directly caused by the heat

(Centers for Disease Control 2002). However, the actual
total is much higher since extreme heat results in increased
mortality for many other causes of death, such as
cardiovascular and respiratory disease (Shen et al. 1998;
Kilbourne 1997). The 1995 heat wave in Chicago was
responsible for over 800 deaths alone, and estimates for the
2003 heat wave across Europe range from 22,000 to
upwards of 50,000 deaths (Whitman et al. 1997; WHO
2004; Brucker 2005).
As a result of these recent disasters, a large amount of
research has begun to focus on heat and its impacts on
humans. Furthermore, a new heat warning system has been
developed based upon actual human health responses to the
heat. However, despite the increase in heat–health research,
there have been very few studies examining the social
response to heat and heat warnings. For example, how does
the public perceive heat and are they aware that heat can be
an extremely dangerous weather-related phenomenon?
What is the public response to extreme heat conditions or
a National Weather Service (NWS) heat warning? Do
people alter their daily habits as a result? How concerned is
the public about heat; do they feel threatened or worried?
These are all vital questions, since previous research has
shown that perceived risk from a threat often leads to action
to help mitigate potential consequences.
Perceived risk

A. J. Kalkstein (*)
Department of Geography, Arizona State University,
Tempe, AZ, USA
e-mail: Adam.Kalkstein@asu.edu
S. C. Sheridan
Department of Geography, Kent State University,
Kent, OH, USA
e-mail: ssherid1@kent.edu

Research has suggested that perception of risk and
mitigating action are closely related, and Janis (1962)
provides an interesting counter-example. An event is
examined in which people were killed by carbon monoxide,
an odorless, colorless, but extremely dangerous gas.
Although some warning signs existed such as nausea and
dizziness, people did not respond immediately since the

44

threat was “highly ambiguous” and because the danger
signs appeared when people were not expecting danger.
Janis (1962) concludes that this unusual combination of
circumstances is what led to the disaster. Not surprisingly,
heat can act in a very similar fashion, especially in Phoenix,
where the population can be lulled into a false sense of
security by afternoon temperatures that average over 103°F
(39.4°C) throughout much of the summer. The threat of
heat seems ambiguous, and many people perceive little or
no threat from it since, as many Phoenix residents report,
“it always seems hot in the summer.”
Perry and Lindell (1997) note that a large amount of
research has been conducted on human responses during
and following disasters while very little research exists
examining responses to disaster warnings. However, of the
research that has been conducted on warnings, most
scientists agree that the leading factor that dictates behavior
is a perceived risk (Sims and Baumann 1983; Janis 1962;
Perry et al. 1982).
Perceived risk to a disaster can best be obtained in two
ways: (1) education, and (2) personal experience with the
disaster (Vitek and Berta 1982). Often, due to a lack of
education, people are simply unaware that they are at risk
for a natural disaster and do not realize they are in danger
(O’Riordan 1986). As a result, when disaster does strike,
most people do not take action and are surprised or caught
off guard. Even worse, those who are uneducated about the
disaster often conclude that it could not happen again, thus
putting themselves at unnecessary risk for a future event
(Vitek and Berta 1982).
Janis (1962) notes that people are more likely to respond
to a warning if someone they know has been previously
affected by the disaster. This personal experience with the
disaster tends to reduce apathy, indifference, wishful
thinking, and denial, all of which can lead to an underreaction to a warning. Furthermore, a personal experience
with the disaster allows a warning to be reinforced socially,
through friends, neighbors, and community networks (Perry
and Lindell 1997). Sims and Baumann (1983) note that
“while people get information from the media, that
information is given meaning and importance through
being processed by friends and neighbors,” thus leading to
increased risk perception and an increased likelihood of
responding to a warning.
The media also play an important role in the public
perception of disasters, and Hiroi et al. (1985) suggest that
the media tend to exaggerate damage in disasters. As a
result, the public often believes that more people die from
natural disasters than is actually true (Whyte 1986).
However, this pattern seems to be reversed when it comes
to heat; both the public and the scientific community have a
consistent pattern of under representing the actual impacts
of heat. For example, in a recent article specifically about

Int J Biometeorol (2007) 52:43–55

the progress of natural disaster warning systems across the
United States, Sorensen (2000) examines 12 disaster
warning systems. Surprisingly, heat is not among them
despite the fact that it has been shown to be the leading
weather-related killer in the United States (Sheridan and
Kalkstein 2004). Perhaps this general disregard for heat in
both academia and the public is partially due to lack of
media reporting. Heat has been called the “silent killer” in
that it lacks the “awe” factor associated with other natural
disasters. As Sandman (1994) notes, “journalists are in the
news business, not the education business or the health
protection business”, and this could be a possible cause of
the apparent perception that heat is not among the most
severe or dangerous natural disasters.
Risk response
The public response to a natural disaster warning is
surprisingly complex and, often, the human response to a
warning is not as expected. It has been shown that even if
the public is informed of a risk and knows what to do, it
will often not do it. However, “sometimes, under highly
specified conditions, and if properly executed, with certain
target publics, information may lead to awareness and
awareness may lead to behavior” (Sims and Baumann
1983). Thus, the question is: what exactly are the specified
conditions, properly executed, with certain target audiences
that may lead to a public response to a warning?
First, Mogil (1980) notes that the terminology of the
warning is extremely important. The ideal warning will
describe the hazard, its probable time of impact, and will
then suggest appropriate adaptive behaviors. Furthermore, more specific messages tend to produce an
increased level of warning belief and, thus, warning
response (Perry et al. 1982). Warnings must also be clear
and perceived as emanating from a credible source (Sims
and Baumann 1983).
Next, there needs to be a clear connection between the
threat and what someone is supposed to do about it. Withey
(1976) notes that a person is more likely to respond to a
warning if the message clearly states what adaptive
behavior is required. However, if the warning provides no
way to substantially reduce or eliminate the threat, the
population is much less likely to act (Withey 1962).
Research has also shown that the public is often
skeptical of warnings, and only a small percentage of the
population takes action when the NWS issues a warning
(Carter 1980). This could be partially due to the “cry wolf
effect”, in that many people think the NWS over-warns for
weather-related disasters (Mogil 1980). Furthermore,
Atwood and Major (1998) suggest that if a warning expires
without a disaster, both protective behavior and the
perceived importance of the threat decrease. Initial warn-

Int J Biometeorol (2007) 52:43–55

ings have been associated with increased anxiety and a
higher likelihood of a response, while repeated warnings
result in less perceived risk (Hansson et al. 1982). Even
more confusing is the fact that threats by officials about the
disaster are often ineffective, but, ironically, explicit statements by officials minimizing the risk such as “don’t
worry” were often found to be more alarming (Sandman
1994). Finally, Sims and Baumann (1983) conclude that to
elicit the greatest warning response from the public,
officials must “convince the skeptics” that there is a real
and credible threat. Unfortunately, this can be exceedingly
difficult because, according to Sims and Baumann (1983),
“people want intellectual closure; thus, once opinions are
formed (regardless of the representativeness of the evidence
on which they are based), they tend to become fixed and
new evidence is made to conform even if it requires
considerable distortion.”
Recent technologies have allowed scientists to better
predict natural disasters and, thus, have led to more
effective warnings systems. Sorensen (2000) documents
these improvements for numerous warning systems (heat is
not included) over a 20-year span beginning in the late
1970s. He suggests that there have been “major” improvements in hurricane forecasts and the integration of these
forecasts into warnings. Despite these improvements, recent
disasters such as Hurricane Katrina illustrate that, despite
better technology, many people still cannot or do not heed
the warning. Perhaps with the devastation observed with
Hurricane Katrina, people will be more likely to take action
in the future.
It is important to note that various portions of the
population are more likely to heed warnings than others,
and age, gender, and ethnicity are all important factors in
warning response (Perry and Lindell 1997). Perry et al.
(1982) suggest that death rates for minorities are higher
than those for Whites during natural disasters. To support
their findings, they note that, in one case, MexicanAmericans were more skeptical about believing warning
messages. Furthermore, Mexican-Americans interpreted the
same warning messages as indicating lower levels of
personal danger and, thus, they were less likely to take
protective action. Also, minorities tend to be more fearful
of technological disasters such as a nuclear radiation leak,
while Whites are generally more concerned about natural
disasters (Vaughan and Nordenstam 1991). It is likely the
racial differences in risk perception and response are
partially due to varying life experiences, such as increased
exposure to poverty or crime, which cause many to become
more fearful of a man-made disaster. Finally, despite the
fact that many believe the elderly are less likely to respond
to a warning, Perry and Lindell (1997) find otherwise and
note that “age alone is not a useful predictor of warning
compliance.” In fact, they find that in certain cases, the

45

elderly sometimes respond in higher percentages than other
age groups.
Heat warning system
To help protect citizens from the negative heath impacts
associated with extreme heat, many NWS offices across the
United States have implemented a heat warning system.
Most of these systems issue excessive heat warnings based
upon the following criteria: “if the heat index is expected to
exceed 105–110°F (40.5–43°C) (depending on local
climate) for at least two consecutive days” (National
Weather Service 2006). However, there are several problems with these guidelines. First, the thresholds for calling
warnings are arbitrary; there have been no studies suggesting that human health begins to deteriorate above a 105°F
heat index. Next, the same thresholds are applied for
numerous cities across the United States despite the fact
that human responses to a 105°F heat index would be very
different in places such as Chicago and Atlanta (Kalkstein
et al. 1996b).
Beginning in 1995, the NWS began implementing a new
heat warning system, which was adopted by the Phoenix
NWS Office in the summer of 2002 (Sheridan and
Kalkstein 2004). This warning system is based upon the
identification of “oppressive” air masses, which have been
associated with increases in human mortality. A Spatial
Synoptic Classification is used to place each day at a
specific location into one of eight different air mass types,
based upon the location’s temperature, dew point, cloud
cover, wind direction and wind speed (Kalkstein et al.
1996a; Sheridan 2002). In Phoenix, two particularly hot
subsets of the dry tropical air mass were found to be
correlated with increases in human mortality, and the
Phoenix heat warning system is based upon the occurrence
of these two oppressive air masses. Thus, this new heat
warning system is based upon actual human-health
responses as opposed to arbitrary heat index thresholds.
Depending on the specific weather conditions present,
the NWS Office in Phoenix has specific instructions for
their forecasters on whether or not a warning should be
issued, along with whether the warning should be an
“excessive heat warning” (dangerous conditions are
expected within 24 h), an “excessive heat watch” (usually
issued up to 48 h before an oppressive air mass is forecast),
or a “heat advisory” (less severe than an excessive heat
warning). These different categories are used to inform city
health officials of the severity of the heat wave and the
potential human-health response.
The last important component of the heat warning
system is intervention by the city and health department
to help prevent any deterioration in human health. The city
of Philadelphia has a particularly elaborate set of interven-

46

tion plans, and recent research has suggested that many
lives have already been saved as a result (Ebi et al. 2004;
Kalkstein et al. 1996b). For example, when an excessive
heat warning is issued in Philadelphia, the following steps
are taken: media announcements of the warning, a “buddy
system” goes into effect where volunteers check on the
elderly, a telephone hotline called the “Heatline” is
activated, the Department of Public Health makes home
visits to susceptible portions of the population, utility
suspension is halted, medical service staff are increased,
outreach for the homeless increases, and designated airconditioned facilities are opened for the public (Kalkstein et
al. 1996b). Unfortunately, Phoenix does not have an
organized intervention plan with regard to excessive heat
warnings, and only after many consecutive days of
excessive heat warnings and heat advisories in the summer
of 2005 did Phoenix officials begin to distribute water and
open air-conditioned shelter areas. However, this only
occurred after numerous deaths had been reported. In all,
the summer of 2005 was particularly bad for Phoenix, and
at least 18 people were killed as a direct result of a severe
heat wave from 12 to 20 July (Villa 2005).
By examining the social response to the heat warning
system in Phoenix, a wide range of information can be
obtained which will serve to help improve heat warning
systems not only in Phoenix, but in other places across the
country. Thus, the goal of this research is to evaluate the
social impacts of the heat warning system in the Phoenix
Metropolitan Area. This study will focus primarily on the
public’s perceived risk and response to heat and heat
warnings, since perceived risk is often associated with
increased action.

Materials and methods
The data were obtained by distributing questionnaires at
various locations across the Phoenix Metropolitan Area
(Appendix A). Phoenix, the capital of Arizona, is located
in the Desert Southwest and provides an interesting
location to conduct this study considering it is the hottest
major city in the United States. Phoenix and its surrounding cities including Mesa, Scottsdale, Tempe, Gilbert,
Chandler, and Peoria, are situated in the Salt River Valley,
where summertime temperatures frequently exceed 105°F
(40.5°C). The Phoenix Metropolitan Area is composed of
only two counties, Maricopa and Pinal, although the vast
majority of the population resides in Maricopa County,
where all surveys were distributed.
The text of the surveys focused heavily on behavioral
changes as a result of heat warnings. A total of 201 surveys
were distributed at eight different locations across the
research area from 21 through 24 December 2005. This

Int J Biometeorol (2007) 52:43–55

study was conducted as part of a graduate research and field
examination for Arizona State University, thus making a
summertime distribution of surveys impossible. Only one
person handed out surveys, and each survey was filled out
individually by the participant. As with all self-survey
studies, this methodology has several limitations. For
example, there is likely a small social desirability bias, in
which respondents tend to write what they perceive to be
the socially accepted answers, not necessarily presenting an
accurate picture of the individual’s thoughts. Also, the
survey does not poll a completely random sample; those
who choose not to complete a survey are excluded. Finally,
to save time, some respondents might check similar
answers for each question rather than spending the
necessary time to think about each one. Still, despite these
drawbacks, survey distribution continues to be a common
method used to gather information about a population, and
a recent study by Sheridan (2006) conducted a phone
survey across four cities to gain insight into the perception
of heat among the elderly.
The specific survey locations were chosen to obtain data
that best represent the diverse nature of the Phoenix
Metropolitan Area and were often located in front of
shopping centers where numerous people were present. Of
the eight survey locations, two were located in the East
Valley, two were located South of Phoenix, two in the
North Valley, one in a northwest suburb (predominantly
elderly) and two in downtown locations (predominately
Hispanic). Totals of 30 surveys were distributed in each
location with two exceptions; 12 surveys were distributed
in one downtown location and 9 were distributed in one of
the northern locations. It is important to note that the survey
was only conducted on English-speaking participants, and
as a result, the Spanish-speaking portion of the population
is likely underrepresented in this study. However, the
percentage of Hispanic respondents is representative of
the local population. Surveys were only distributed to
people who live in the Phoenix Metropolitan Area
throughout the summer months, thus excluding those who
spend only the winters in the city. Only one survey was
distributed per household.
The participants in this study consisted of 73 males,
122 females, and 6 unidentified; the higher number of
females is attributed to the fact that they were more
likely to fill out surveys compared to males. In addition,
many surveys were distributed during the early afternoon,
when the majority of people running errands or shopping
were female. Of the 201 surveys, 120 listed their racial
or ethnic background as White, 42 Hispanic, 14 Black,
8 Native American, 4 Asian, 8 other, and 5 unidentified.
The age of the respondents was slightly skewed towards
the younger age groups, which was likely caused by
several factors. First, many of the elderly only live in the

Int J Biometeorol (2007) 52:43–55

47

city during the winter months, thus excluding them from
the study. Next, younger people were more likely to stop
and fill out a survey compared to the elderly. Finally,
according to the 2000 census, less than 11.9% of the
population in the Phoenix Metropolitan Area is over 65.
The breakdown among age groups is as follows: 55 of
the respondents were 18–29, 43 were 30–41, 48
respondents were 42–53, 36 were 54–65, and 14
participants were over 65 with 5 unidentified. Finally,
the annual income for the entire household, as reported
by the respondents were: 31 participants made under
$20,000, 44 made from $20,000 to $40,000, 38 from
$40,000 to $60,000, 28 from $60,000 to $80,000, 15
from $80,000 to $100,000, and 29 reporting over
$100,000 per year with 16 unidentified. Most of the
unidentified were located in Scottsdale, a wealthy suburb
of Phoenix, and it is possible those who are wealthy are
less likely to disclose their income. These demographic
data varied tremendously from one location to the next,
but, as a whole, they represent a very accurate sample of
the Phoenix Metropolitan Area, according to the 2000
Census, with the notable exception being the male to
female ratio.

Results
Warning awareness
Of the 201-person sample, the vast majority of respondents
reported that they were aware of days when the NWS in
Phoenix issued either an excessive heat warning or a heat
advisory. In fact, only 28 participants (13.9%) were
unaware that warnings had been issued. However, several
participants were steadfastly convinced that Phoenix did not
have a heat warning system, with one respondent even
stating, “Now why would Phoenix need any weather
warning systems?”
Surprisingly, the level of heat warning system awareness
varied significantly across different demographic categories
(Table 1). For example, 90.2% of females were aware of the
heat warning system while only 75.3% of males reported
being aware of the system. Furthermore, age and income

seemed to play large roles as well. While only 67.3% of
respondents between the ages of 18 and 29 reported being
aware of the system, 92.9% of those over 65 had heard of
it. Similarly, only one respondent making over $100,000
had not heard of the system (3.4%), while 35.5% of those
making under $20,000 were not aware that warnings had
been issued (Fig. 1). Finally, the differences among races
were smaller, but Whites (90.5%) were slightly more likely
to have heard of the heat warning system compared to
Hispanics (81.0%).
Perceived risk
The perceived risk of heat among the participants was
examined with the following question: “How dangerous
do you think the heat is for you?” Over 25% of the
respondents reported that the heat was “very dangerous”
for them and 44.7% reported that the heat was
“somewhat dangerous” for them. Only 7.6% felt that
the heat was “not at all dangerous”. Not surprisingly, the
responses varied somewhat among different demographic
segments of the population. First, women were slightly
more likely than men to report that the heat was very
dangerous for them. Furthermore, while only 17.9% of
White respondents felt that heat was very dangerous for
them, 46.3% of Hispanics felt similarly, clearly indicating an elevated sense of perceived risk among Hispanics.
Finally, 32.1% of youngest age group felt that heat was
very dangerous for them, more than any other age
category.
Another potential indicator of perceived risk was
measured in the question: “Which natural disaster do you
think is the biggest threat to you?” 50.6% of the participants
believed that heat was the biggest threat to them, although
this value is likely biased since many of the preceding
questions pertained to the heat. Of the remaining disasters,
19.4% felt that thunderstorms and lightning were the
biggest threat, while hurricanes, tornadoes, earthquakes,
floods, and dust storms each had less than 10%. Once
again, race seemed to play a major role. For example,
27.0% of Whites reported that thunderstorms/lightning
were the biggest threat to them, while none of the 31
Hispanics reporting felt similarly. In addition, 64.5% of

Table 1 Percent respondents who were aware of the heat warning system
Sex

Percent aware
n

Race

Age (years)

Male

Female

White

Hispanic

Black

Other

18–29

30–41

42–53

54–65

>65

75%
73

90%
122

88%
120

81%
42

79%
14

80%
20

67%
55

91%
43

96%
48

86%
36

93%
14

48

100
90
Percentage Aware of Warning System

Fig. 1 Average annual income
versus the percentage of
respondents who were aware of
the Phoenix heat warning
system

Int J Biometeorol (2007) 52:43–55

80
70
60
50
40
30
20
10
0
< $20

$20 - $40

$40 - $60

$60 - $80 $80 - $100

> $100

Annual Household Income (in thousands)
Hispanics believed that heat was their biggest threat
compared to only 49.0% of Whites.
To examine the “cry wolf” effect, an important factor in
perceived risk, participants were asked whether they
thought the NWS issued the proper number of warnings
for all weather-related phenomenon. The majority (64.7%)
believed that the NWS issued the proper number of
warnings, 21.6% said the NWS issued too few warnings,
and only 13.7% felt there were too many warnings.
However, there were several interesting differences among
the respondents. For example, Whites were more likely
than Hispanics to believe that the NWS issued the proper
number of warnings. Age also played a role: while less than
50% of those between 18 and 29 felt that the proper number
of warnings was issued, 85.7% of those over 65 felt
similarly. Furthermore, only 41.0% of those making under
$20,000 per year thought the NWS issued the correct
number of warnings compared to 89.0% for those making
over $100,000. Surprisingly, this question seemed to have
little impact on how seriously individuals took the warning.
For example, only 6.7% of those who reported not taking
the warning at all seriously felt that the NWS issued too
many warnings. Most interesting, however, is that among
those who were unaware heat warnings had ever been
issued in Phoenix, half felt that the NWS issued too few
warnings. Clearly, there are individuals within the metro-

politan area who are not getting any weather warning
messages.
Warning response and social impacts
Much of the questionnaire was dedicated to the social
impacts and warning response of the participants. Several
examples are: “Did you do anything different on these days
as a result of the warning or advisory?”, “Were you worried
when you heard about the excessive heat warning or heat
advisory?”, and “Overall, how much did your daily
activities change as a result of the excessive heat warning
or heat advisory?” Not surprisingly, answers to these
questions varied considerably across the different demographic categories.
Of the 169 participants who reported that they were
aware of the heat warnings and advisories, 84 (49.7%) said
they did something different on those days while 80
(47.3%) respondents did not. Females were slightly more
likely than males to change their behavior as a result of the
warning; 53.7% of females reported doing things differently
on days with warnings while 45.5% of males acted
similarly. There were also differences reported between
Hispanics and Whites: 61.8% of Hispanics reported
changing their behavior on warning days compared to only
44.7% of Whites. Most surprising were the differences

Int J Biometeorol (2007) 52:43–55

49

100
90

Percentage of Respondents

80
70
60
50
40
30
20
10
0
18-29

30-41

42-53

54-65

>65

Age

Fig. 2 Age versus the percentage of respondents who reported
changing their behavior during days with heat warnings or advisories
Fig. 3 “How dangerous is the
heat?” versus the percentage of
respondents who changed their
behavior on days with a heat
warning or advisory

between age groups. For example, only 32.4% of those
between 18 and 29 reported doing things differently on
warning days compared to 66.7% between 42 and 53
(Fig. 2).
One question in the survey directly asked each participant how much their activities changed as a result of the
warning or advisory. Of the participants, 29.4% reported
not changing their behavior at all, while the remainder
reported changing their activities a lot, some, or a little,
although only 8.9% reported changing their activities a lot.
Surprisingly, there were no notable differences for this
question across demographic categories.
The strongest indicator of whether or not an
individual altered their behavior on warning days was
perceived risk. For example, Fig. 3 illustrates that the
likelihood of whether or not a person changed behavior on
warning days is highly dependent on how dangerous that
individual perceived heat. While 83.3% of those who
felt heat was very dangerous changed their activities on
days with warnings or advisories, only 22.2% of those
who felt heat was not at all dangerous changed. The results
were similar for those who were worried about heat
warnings. For example, 87.5% of those who were very

100

Percentage Who Changed Behavior

90
80
70
60
50
40
30
20
10
0
very

somewhat

slightly

How Dangerous is the Heat?

not at all

50

Int J Biometeorol (2007) 52:43–55

Fig. 4 Actions taken by respondents who reported they
altered their behavior after
hearing there was a heat
warning or advisory issued

100

Percent Respondents

80

60

40

20

There were 84 respondents who were aware that
warnings had been issued and altered their behavior as
a result. Of these 84 participants, the most common
behavioral changes reported were drinking more fluids
(83.3%), avoiding the outdoors/sun (66.7%), and staying
indoors or seeking an air-conditioned location (64.3%)
(Fig. 4). Although the responses of avoiding the outdoors/
sun and staying indoors/seeking an air-conditioned location are similar, the latter suggests more initiative by the
respondent in that they were actively seeking an airconditioned location.
The 80 respondents that were aware of the warnings yet
chose not to do anything different on those days provided
reasons for their inaction in their survey, and not surprisingly, the most common response was: “It always seems hot
in the summer.” Thirty-four people (42.5%) marked this as
a reason why they chose not to act during an excessive heat
warning or heat advisory. The next most common response
was that they alter their behavior anyway during the

em

ot

he

r

r
em

de
el
do

n'

tr

k
ec

be

rly

ly
te
ria
ch

op
pr
ap

dr

es

s

ov
d
oi

av
Behavioral changes

n
ta
er

ex
er

ay
st

worried when a heat warning was issued changed their
activities, and only 30.8% of those who were not at all
worried changed.

tio

s
or
do
in

e
or
m

av

oi

d

ou

td

flu

oo

id

rs

s

0

summer, independent of any warnings. Others noted that
their work required them to be either indoors or outdoors,
and a warning or advisory would not affect this. Finally, 14
participants (17.5%) listed that, “I wasn’t worried; there
always seem to be weather warnings issued”, revealing the
possible impacts of the cry wolf effect.
Overall, 117 respondents (60%) reported changing their
behavior on excessively hot days, independent of any
excessive heat warnings or heat advisories. Of these, 77.8%
reported drinking more fluids, 72.6% avoided the outdoors/
sun, and 71.8% went to an air conditioned location or
stayed indoors. Of the remaining 78 participants who did
not change their behavior on excessively hot days, 55.1%
noted that it always seems hot in the summer, by far the
most common response (Fig. 5).

Discussion
Awareness, perceived risk, and action
This study confirms that the majority of people living within
the Phoenix Metropolitan Area are aware when excessive heat
warnings or heat advisories are issued. Respondents over
65 years old reported the highest level of awareness with

Int J Biometeorol (2007) 52:43–55

51

Fig. 5 Reasons given as to why
respondents didn’t take action
on excessively hot days

100

Percent Respondents

80

60

40

20

th

It'

s

al

w

O

s
ay

er

t
ho

m
ct
ffe
ta
n'

es
do
H

ea
t

ti
ea
H

nearly 93% stating that they had heard of heat warnings in the
past. This is especially important since that segment of the
population is most at risk for heat-related illness. Thus, those
who are most vulnerable to the heat are receiving the warning
message. However, that does not necessarily mean that those
over 65 are most likely to act during a warning. In fact, less
than 50% of that age segment reported changing their
activities as a result of the warning. This is worrisome and
implies that although the message is going out, it is not always
creating the desired response.
As expected, this study illustrates that perceived risk is
highly correlated with mitigating action. This is in
agreement with previous research suggesting that those
who feel most at risk from a disaster are more likely to act
when a warning is issued. Among those who choose to act
when a warning is issued, there are notable social
repercussions including an elevated sense of worry and
noteworthy changes in daily activities.
One of the more surprising results and contrary to
previous research, this study suggests that minorities have a

e

e
nc
en
co

In

't
sn

nv

ng
da

tw
ri
ea
th
n'
id
D

ie

er

as

ou

ho

s

t

0

greater sense of perceived risk; Hispanics were far more
likely than Whites to believe that heat is very dangerous to
them. Furthermore, more Hispanics believed that heat is the
biggest threat to them compared to Whites. Finally,
Hispanics were far more likely than Whites to take an
excessive heat warning “very seriously”. A possible
explanation for these results is the prevalence of outdoor
workers in Phoenix who are predominantly Hispanic. Even
if an individual does not work outdoors himself, it is likely
that the threat of heat is reinforced socially, through friends
and family, leading to a community with increased heat
awareness. The increased perceived risk among Hispanics
translated into increased action, as 61.8% of Hispanics
changed their behavior on days with warnings compared to
only 44.7% of Whites. Thus, it is highly likely that social
influences are responsible for increased perceived risk, and
thus action, within the Hispanic community.
Another interesting finding is that age and annual
income were good indicators of perceived trust of the
NWS. The older age groups, along with those making

52

higher salaries, were more likely to feel that the NWS
issues the proper number of warnings. This is important
since previous research has illustrated that the cry wolf
effect, the idea that too many warnings are issued,
decreases perceived risk and, thus, mitigating action. Also,
this study further confirms the cry wolf effect; only 36.4%
of those who felt the NWS issued too many warnings
changed their behavior on warning days. The fact that those
with lower incomes were prone to the cry wolf effect could
be a potential problem since that segment of the population
might have less access to air conditioning. Furthermore,
those with lower annual incomes are more likely to use
evaporative coolers, which have been shown to become less
effective during the North American monsoon season in the
desert Southwest. The monsoon takes place in the late
summer resulting in elevated dew points which reduce the
efficiency of evaporative coolers, creating a potentially
dangerous situation.
Social impacts
While it is evident that the majority of participants in this
study alter their behavior throughout the summer, it is less
clear what role the Phoenix heat warning system plays in
these behavioral changes. Were the warnings the only
thing responsible for participants drinking more fluids or
avoiding the outdoors? For example, of the 84 respondents who were aware of warnings and reported altering
their behavior as a result, how many would have changed
their behavior anyway based solely on the weather
forecast? How would their behavior differ if no warnings
were issued? While the warnings surely had some
influence, it is impossible to gauge the precise impact
that the warning system had on behavior, and it is
unlikely that the warnings were the only thing responsible
for behavioral changes.
Similarly, another important aspect of this study is
determining what constitutes a behavioral change as
reported in the questionnaires. If someone changes their
daily routine during June, July, and August, how would this
be reported, and would the respondents consider this a
behavioral change? There were 80 participants who
reported hearing about the heat warnings, but not changing
their behavior as a result. Despite their responses, there is
no doubt that many of these 80 people exhibited some sort
of behavioral change throughout the summer, including
warning days. For example, 42.5% of those respondents
reported that “it always seems hot in the summer” as the
primary reason for not altering their behavior. It is possible
that many of these responses could indicate that if it always
seems hot in the summer, their behavior gets altered on a
daily basis anyway. Other respondents reported that they
change their behavior anyway on hot days, independent of

Int J Biometeorol (2007) 52:43–55

any advisories or warnings. These responses suggest that
behavioral changes exist on a daily basis and that the vast
majority of Phoenix residents do, in fact, change their
behavior on particularly hot days, even if it is not a direct
result of a heat warning or advisory.
It should be noted that while heat is an extremely deadly
natural phenomenon, the mitigating action required to
prevent heat-related mortality is often less memorable than
that of other natural disasters. Thus, compared to other
studies examining the social impacts of natural disaster
warnings, this one likely underestimates the actual social
response to the heat and, more specifically, to the Phoenix
heat warning system. For example, a study on hurricane
evacuations would yield more consistent results since
evacuations are unusual, traumatic, and highly memorable.
Here, it is possible that at least some of the respondents did
not remember exactly what they were doing during a heat
warning or advisory. Perhaps people took mitigating action
as part of a daily routine or without realizing it and simply
failed to note any change in behavior for this study. Heat is
part of the daily summer life in Phoenix, and it is likely that
some of the mitigating actions have become so routine, that
many of the participants failed to mention them in the
questionnaire.
Suggestions
This study illustrates that although the vast majority of
people are receiving the message about heat warnings, only
around half of the population report changing their actions
on those days. Considering heat can be an extremely deadly
weather-related phenomenon, the warning is not as effective as it needs to be. Many people who did not act during a
heat warning responded that “it always seems hot in the
summer.” Thus, the language in the warning issued by the
NWS, along with the media outlets, need to make it
exceptionally clear that days with excessive heat warnings
are not “typical” summer days in Phoenix and can be
extremely dangerous. Furthermore, since research has
shown that people are more likely to act if they have clear
instructions, residents should be urged in straightforward
language to drink more fluids, seek air-conditioned locations, and participate in other mitigating actions during the
warning days.
Equally important are the responses by the city of
Phoenix and the Health Department during a warning or
advisory. Phoenix should use Philadelphia as an example,
setting up telephone hotlines, checking on the homeless and
elderly, opening air-conditioned shelters, and distributing
water. The summer of 2005 demonstrated what happens
when relatively few steps are taken during a warning; only
after many consecutive days of excessive heat warnings did
officials begin to act, distributing water, and helping the

Int J Biometeorol (2007) 52:43–55

homeless. Unfortunately, these actions came too late, and
18 people lost their lives as a result.
The homeless are particularly prone to heat-related illness,
and 14 of the 18 heat-related deaths reported during the July
2005 heat wave were homeless. This portion of the population
has little access to television, radio, and the internet, which are
the main outlets for the National Weather Service to broadcast
heat warnings and advisories. Clearly, more needs to be done
to alert the homeless when the weather is forecast to be
dangerous and potentially life threatening. When warnings
are issued, the city of Phoenix needs to act immediately to
open air conditioned shelters that welcome the homeless.
While these findings are particularly relevant for NWS and
health officials in Phoenix, this information can be used by
other NWS offices across the United States to better gauge the
effectiveness of their weather warning systems. In Phoenix,
the majority of respondents were aware of the heat warning
system, although those making under $20,000 along with the
younger age groups were less likely to have heard of the
warnings. With these exceptions, the NWS and media outlets
are doing a good job transmitting important information to
most of the public. However, too few people are altering their
behavior as a result of the warnings. This is a particularly
dangerous situation, especially if individuals are also not
responding to other weather warnings. Recent disasters such
as Hurricane Katrina confirm these fears, as many people
could not, or chose not to evacuate, despite the fact that
warnings issued by the NWS stressed the potential danger of
the situation. Considering many of the residents who
remained in New Orleans during the hurricane were likely
making under $20,000, this study supports the fact that poorer
segments of the population are simply not getting the
message, possibly explaining why so many people did not
evacuate. For those who do receive the proper warning
messages, but do not alter their behavior as a result, NWS
offices across the country need to emphasize increased
education programs so that individuals will act when
warnings are issued.

Appendix A
Survey Text
Adam Kalkstein, a PhD student from the Geography
Department at Arizona State University, is conducting a
project to help improve weather warning systems in
Phoenix. All information you share is strictly anonymous;
there will be no association between you and the information you give. Furthermore, I will never ask for your name,
address, phone number, etc. This study has been approved
by the ASU Institutional Review Board and shouldn’t take
more than five minutes to complete. Thanks for your
participation!

53

54

Int J Biometeorol (2007) 52:43–55

Int J Biometeorol (2007) 52:43–55

References
Atwood LE, Major AM (1998) Exploring the “cry wolf” hypothesis.
Int J Mass Emerg Disasters 16:279–302
Brucker G (2005) Vulnerable populations: lessons learnt from the
summer 2003 heat waves in Europe. Eurosurveillance 10:147
Carter MT (1980) Community warning systems: the relationship
among the broadcast media, emergency service agencies and the
National Weather Service. In: Disaster and the Mass Media.
National Academy Press, Washington, D.C., 214–228
Centers for Disease Control (2002) Heat-related deaths in four states,
July–August 2001, and United States, 1979–1999. Morb Mort
Wkly Rep 51:567–570
Ebi KL, Teisberg TJ, Kalkstein LS, Robinson L, Weiher RF (2004)
Heat watch/warning systems save lives: estimated costs and
benefits for Philadelphia 1995—1998. Bull Am Meteorol Soc
85:1067–1074
Hansson RO, Noulles D, Bellovich SJ (1982) Knowledge, warning,
and stress: a study of comparative roles in an urban floodplain.
Environ Behav 14:171–185
Hiroi O, Mikami S, Miyata K (1985). A study of mass media
reporting in emergencies. Int J Mass Emerg Disasters 3:21–50
Janis IL (1962) Psychological effects of warnings. In: Man and society
in disaster. Basis Books, New York, pp 84–86
Kalkstein LS, Barthel CD, Greene JS, Nichols MC (1996a) A new
spatial synoptic classification: application to air mass analysis.
Int J Climatol 16:983–1004
Kalkstein LS, Jamason PF, Greene JS, Libby J, Robinson L (1996b)
The Philadelphia hot weather health watch/warning system:
development and application, summer 1995. Bull Am Meteorol
Soc 77:519–1528
Kilbourne EM (1997) Heat waves and hot environments. In: Noji EK
(ed) The public health consequences of disasters. Oxford
University Press, Oxford, pp 245–269
Mogil HM (1980) The weather warning and preparedness programs:
the national weather service and the mass media. In: Disaster and
the mass media. National Academy Press, Washington, D.C., pp
205–213
National Weather Service (2006) Heat wave: a major summer killer
http://www.nws.noaa.gov/om/brochures/heat_wave.shtml
O’Riordan T (1986) Coping with environmental hazards. In: Kates
RW, Burton I (eds) Themes from the work of Gilbert F. White:
geography, resources and environment, vol 2. University of
Chicago Press, Chicago, pp 272–309

55
Perry R, Lindell M, Greene M (1982) Crisis communications: ethnic
differences in interpreting and acting on disaster warnings. Soc
Behav Pers 10:97–104
Perry RW, Lindell MK (1997) Aged citizens in the warning phase of
disasters: re-examining the evidence. Int J Aging Hum Dev
44:257–267
Sandman PM (1994) Mass media and environmental risk: seven
principles. Risk Health Safety Environ 5:251
Shen TF, Howe HL, Alo C, Moolenaar RL (1998) Toward a broader
definition of heat-related death: comparison of mortality estimates from medical examiners’ classification with those from
total death differentials during the July 1995 heat wave in
Chicago, Illinois. Am J Forensic Med Pathol 19:113–118
Sheridan SC (2002) The redevelopment of a weather-type classification scheme for North America. Int J Climatol 22:51–68
Sheridan SC (2006) A survey of public perception and response to
heat warnings across four North American cities: an evaluation of
municipal effectiveness. Int J Biometeorol (n press)
Sheridan SC, Kalkstein LS (2004) Progress in heat watch-warning
system technology. Bull Am Meteorol Soc 85:1931–1941
Sims JH, Baumann DD (1983) Educational programmes and human
response to natural hazards. Environ Behav 15:165–189
Sorensen JH (2000) Hazard warning systems: review of 20 years of
progress. Nat Hazards Rev 1:119–125
Vaughan E, Nordenstam B (1991) The perception of environmental
risks among ethnically diverse groups. J Cross-Cult Psychol
22:29–60
Villa J (2005) Heat wave claims 18 lives in 5 days. The Arizona
Republic July 21
Vitek JD, Berta SM (1982) Improving perception of and response to
natural hazards: the need for local education. Journal of
Geography: 225–228
WHO (2004) Health and global environmental change, Series No.2.
124
Whitman S, Good G, Donoghue ER, Benbow N, Shou WY, Mou SX
(1997). Mortality in Chicago attributed to the July 1995 heat
wave. Am J Public Health 87:1515–1518
Whyte AVT (1986) From hazard perception to human ecology. In:
Kates RW, Burton I. (eds) Themes from the work of Gilbert F.
White: geography, resources and environment Vol. 2. University
of Chicago Press, Chicago, pp 240–271
Withey SB (1962) Reaction to uncertain threat. In: Man and society in
disaster. Basis Books, New York, pp 93–123
Withey SB (1976) Accomodation to threat. Mass Emerg 1:125–
130