Upload
vuonglien
View
217
Download
3
Embed Size (px)
Citation preview
Title: The American Public’s Preference for Preparation for the Possible Effects of Global Warming: Impact of Communication Strategies
Journal Name: Climatic Change
Authors:Bo MacInnisJon A. KrosnickAdina AbelesMeg CaldwellErin PrahlerDebbie Drake Dunne
Corresponding Author:Jon A. Krosnick, Stanford UniversityE-mail: [email protected]
Online Resource
1 Trends in Use of Terms Describing the Purpose of Preparation in Newspaper Stories
Separately for each year between 2000 and 2013, searches were conducted in April 2014 using
Lexis Nexis to gauge the numbers of newspaper articles published containing the phrase “global
warming” or “climate change” and containing each of the six target phrases. In Lexis Nexis, the
“source” was set to “newspapers”, and full text search terms were the following:
(“climate change” or “global warming”) and “reduce risk”
(“climate change” or “global warming”) and “reduce vulnerability”
(“climate change” or “global warming”) and “increase resilience”
(“climate change” or “global warming”) and “increase preparedness”
(“climate change” or “global warming”) and “increase readiness”
(“climate change” or “global warming”) and “prevent maladaptation”
An additional search was conducted for each year to count the total number of newspaper articles
containing phrase “global warming” or “climate change” and containing any of the six target
phrases. The source was again set to “newspapers”, and full text search terms were the
following:
(“climate change” or “global warming”) and (“reduce risk” or “reduce vulnerability” or
“increase resilience” or “increase preparedness” or “increase readiness” or “prevent
maladaptation”)
For each year, Figure S1 shows the number of newspaper articles published containing the
phrase “global warming” or “climate change” and containing each of the six target phrases
divided by the total number of newspaper articles published containing the phrase “global
warming” or “climate change” and any of the six target phrases.
2 General Trust in Endorsers
General trust in endorsers of preparation can be gauged by national surveys of Americans
conducted in the recent past. For example, according to the Allstate/National Journal Heartland
Monitor Poll of the American public conducted between May 19 and 23, 2012, 37% of
Americans trusted “scientists and academics” “a great deal”, in contrast to 22% for “clergy and
other religious leaders”, 4% for “elected officials”, and 3% for “leaders of major corporations”
(Allstate/National Journal 2012). Similarly, in the 2012 General Social Survey, 40% of
Americans expressed “a great deal” of confidence in the “scientific community”, 20% said so
about “organized religion”, 17% said so about “major companies”, 14% said so about the
“Executive Branch of Federal Government”, and 7% said so about “Congress” (Smith and Son
2013). And in the Northeastern University Innovation in Higher Education Survey, conducted
between October 13 and 18, 2012, 29% of respondents expressed “a great deal” of trust in
information from “college and university professors”, as compared to 6% in information from
“leaders of major corporations” and 3% for “elected public officials” (Northeastern University
2012).
3 Familiarity and Simplicity of Terms
According to Google searches conducted on September 13, 2013, “reduce risk” appeared on
about 4 million web pages, compared to about 287,000 for “reduce vulnerability,” 164,000 for
“increase resilience,” 25,000 for “increase readiness,” 19,000 for “increase preparedness,” and
1,000 for “prevent maladaptation.”
Among the phrases describing the purpose of preparation that we examined, all involved
two words, but they varied in the numbers of letter and syllables. “Reduce risk” contains the
fewest of each (3 syllables and 10 letters), compared to “reduce vulnerability” (8 syllables and 19
letters), “increase resilience” (5 syllables and 18 letters), “increase readiness” (5 syllables and 17
letters), “increase preparedness” (5 syllables and 20 letters), and “prevent maladaptation” (7
syllables and 20 letters). This, too, might lead us to expect “reduce risk” to have an advantage in
inducing people to endorse preparation.
The same sort of logic can be applied to the other language choices we examined. For
example, we explored whether people reacted differently to “prepare for” vs. “plan for” to
describe preparation. According to Google searches done on August 19, 2013, “plan for” and
“prepare for” appeared on about the same number of webpages (see Table S1), and they have the
same numbers of words (2) and about the same number of syllables (2 vs. 3) and letters (7 vs.
10), so these phrases might be equally likely to induce positive evaluations.
We also applied this analytic approach to phrases describing not preparing: “adapt to,”
“adjust to,” or “respond to.” They are identical in the number of words (2) and the number of
syllables (3), and very similar in terms of the numbers of letters (7 vs. 8 vs. 9). However, as
shown in Table S1, Google searches indicated that “respond to” appeared on considerably more
webpages than did “adapt to” or “adjust to”. Therefore, “respond to” might be the most familiar
of these phrases and might therefore be the most likely to dissuade people from endorsing
preparation.
4 Merriam-Webster (Merriam-Webster 2014) Definitions of Words
Risk: 1. the possibility that something bad or unpleasant (such as an injury or a loss) will happen;
2. someone or something that may cause something bad or unpleasant to happen.
Vulnerable: 1. easily hurt or harmed physically, mentally, or emotionally; 2. open to attack,
harm, or damage.
Resilience: 1. the ability to become strong, healthy, or successful again after something bad
happens; 2. the ability of something to return to its original shape after it has been pulled,
stretched, pressed, bent, etc.
Preparedness: 1. the fact of being ready for something; 2. the state of being prepared.
Ready: 1. prepared to do something; 2. properly prepared or finished and available for use; 3.
almost about to do something.
Maladaptation: poor or inadequate adaptation
Respond: 1. to say or write something as an answer to a question or request; 2. to do something
as a reaction to something that has happened or been done; 3. to have a particular reaction
to something; 4. to have a good or desired reaction to something
Adapt: 1. to change your behavior so that it is easier to live in a particular place or situation; 2. to
change (something) so that it functions better or is better suited for a purpose; 3. to
change (a movie, book, play, etc.) so that it can be presented in another form
Adjust: 1. to change (something) in a minor way so that it works better; 2. to change the position
of (something); 3. to change in order to work or do better in a new situation
Prepare: 1. to make (someone or something) ready for some activity, purpose, use, etc.; 2. to
make yourself ready for something that you will be doing, something that you expect to
happen, etc.; 3. to make or create (something) so that it is ready for use.
Plan: 1. to think about and arrange the parts or details of (something) before it happens or is
made; 2. to intend or expect to do (something); 3. to expect something to happen.
5 Question Wordings and Codings of Measures
Sex: Respondents’ sex (male or female) was recorded by interviewers. Dummy variable Female
was set to 1 if the respondent was female and 0 otherwise.
Age. Respondents were asked, “In what year were you born” and “Have you already had a
birthday this year?” People who refused to report their year of birth were asked to select an age
range from among the following categories: under 25, 25-34, 35-44, 45-54, 55-64, 65-74, and 75
or older. Based on these questions, we constructed the following dummy variables. Dummy
variable Age 18 to 24 was set to 1 if the respondent was between age 18 and 24 and 0 otherwise;
dummy variable Age 25 to 34 which was set to 1 if the respondent was between age 25 and 34
and 0 otherwise; dummy variable Age 35 to 44 which was set to 1 if the respondent was between
age 35 and 44 and 0 otherwise; dummy variable Age 45 to 54 which was set to 1 if the
respondent was between age 45 and 54 and 0 otherwise; dummy variable Age 55 to 64 which
was set to 1 if the respondent was between age 55 and 64 and 0 otherwise; dummy variable Age
65 or older was set to 1 if the respondent was age 65 or older and 0 otherwise; dummy variable
Age missing was set to 1 if the respondent did not answer and 0 otherwise.
Race and ethnicity. Respondents were asked this ethnicity question, “Are you of Hispanic
ethnicity?” Among respondents who answered “Yes” were asked the race question: “Are you
White, Black, Asian, American Indian, or other?” Dummy variable Hispanic was set to 1 if the
respondent answered “Yes” to the ethnicity question and 0 otherwise; dummy variable Not-
Hispanic white was set to 1 if the respondent answered with “No” to the ethnicity question and
“White” to the race question and 0 otherwise; dummy variable Not-Hispanic black was set to 1 if
the respondent answered with “No” to the ethnicity question and answered with “Black” to the
race question and 0 otherwise; dummy variable Not-Hispanic other race(s) was set to 1 if the
respondent answered with “No” to the ethnicity question and answered with “Asian”, or
“American Indian” or “Other” to the race question and 0 otherwise; dummy variable Race and
ethnicity missing was set to 1 if the respondent did not answer the ethnicity question and/or did
not answer the race question and 0 otherwise.
Education. Respondents were asked, “What is the last year of school you completed – Grade
school or some high school, Completed high school, Some college but did not finish, Completed
a two year college degree, Completed a four year college degree, or Completed a post-graduate
degree such as Master’s or PhD?” Dummy variable Less than high school graduate was set to 1
if the respondent answered with “Completed Grade school or some high school” and 0
otherwise; dummy variable High school graduate was set to 1 if the respondent answered with
“Completed high school” and 0 otherwise; dummy variable Some college was set to 1 if the
respondent answered with “Some college but did not finish” or “Completed a two year college
degree” and 0 otherwise; dummy variable College graduate was set to 1 if the respondent
answered with “Completed a four year college degree” or “Completed a post-graduate degree
such as Master’s or PhD” and 0 otherwise; dummy variable Education missing was set to 1 if the
respondent did not answer and 0 otherwise.
Income. Respondents were asked, “Now, I am going to read a list of income ranges. When I get
to the income range that best describes your household income from all sources in 2012, please
stop me. Was your household income for 2012 under $25,000, $25,000 to less than $55,000,
$55,000 to less than $75,000, $75,000 to less than $100,000, or $100,000 or more?” Dummy
variables Under $25,000 was set to 1 if the respondent answered with “under $25,000” and 0
otherwise; dummy variable $25,000 to less than $55,000 was set to 1 if the respondent answered
with “$25,000 to less than $55,000” and 0 otherwise; dummy variable $55,000 to less than
$75,000 was set to 1 if the respondent answered with “$55,000 to less than $75,000” and 0
otherwise; dummy variable $75,000 to less than $100,000 was set to 1 if the respondent
answered with “$75,000 to less than $100,000” and 0 otherwise; dummy variable “$100,000 or
more” was set to 1 if the respondent answered with “$100,000 or more” and 0 otherwise.
Dummy variable Income missing was set to 1 if the respondent did not answer and 0 otherwise.
Region. The firm that generated the sample of telephone numbers to dial (Survey Sampling
International) provided the region in which each respondent’s residence was located based on the
area code. We created dummy variables identifying these four regions: Northeast, Midwest,
South, and West.
Political Party Identification. Respondents were asked, “Do you consider yourself a Democrat, a
Republican, an Independent, or none of these?” Dummy variable Democrat was set to 1 if the
respondent answered with “Democrat” and 0 otherwise; dummy variable Republican was set to 1
if the respondent answered with “Republican” and 0 otherwise; dummy variable Independent set
to 1 if the respondent answered with “Independent” or “none of these” or didn’t answer and 0
otherwise.
6 Survey Methodology
Studies 1, 2, 3 and 4, were random digit dial (RDD) telephone surveys of representative
samples of American adults. Phone numbers to be called were drawn from landline and cellular
telephone number frames by Survey Sampling International. Landline telephone numbers were
drawn with equal probabilities from active blocks (area code + exchange + two-digit block
number) that contained one or more residential directory listings. The cell phone sample was
generated through random sampling from 1000-blocks dedicated to cellular service according to
the Telcordia database. The questionnaire was written in English and then translated into Spanish
by one translator to permit administration in both languages. Interviews were administered in
English and Spanish.
Study 1 was a survey of 1,004 U.S. adults conducted between January 3 and 6, 2013 (804
people via landline telephones, and 200 via cell phones) by Ipsos. The AAPOR Response Rate 3
was 4.2%. Study 2 was a survey of 1,000 U.S. adults interviewed between February 7 and 11,
2013 (800 respondents via landline telephones, and 200 via cell phones) by Ipsos, with an
AAPOR Response Rate 3 of 1.7%. Study 3 was a survey of 1,003 U.S. adults conducted between
March 7 and 11, 2013 (801 via landline telephones, and 202 via cell phones) by Ipsos. The
AAPOR Response Rate 3 was 3.3%. Study 4 was a survey of 1,003 U.S. adults conducted
between April 4 and 8, 2013 (800 via landline telephones, and 201 via cell phones) by Ipsos,
with an AAPOR Response Rate 3 of 2.9%.
Weights were computed to account for unequal probabilities of selection and to post-
stratify to population proportions in terms of age and sex cross-tabulated, education, ethnicity
and race cross-tabulated, Census region, and metro density.
The distributions of unweighted and weighted demographics of the respondents and of
the nation are displayed in Table S2 for Study 1, Table S3 for Study 2 and 3, and Table S4 for
Study 3. We used the Feb 2013 Current Population Survey as the benchmark for all our studies,
and since demographic change extremely little from month to month, we use just one month’s
CPS figures throughout these tables. These distributions show that the survey samples were
similar to the American population before the weights were applied and were more similar after
the data were weighted. The results reported in the paper were generated using weighted data.
7 Spanish Translations
Study 1:
“Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Estos científicos creen que el calentamiento en el pasado y futuro tendrá efectos sobre las personas y el medio ambiente, y que cambiará la forma en que las personas vivirán sus vidas. Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse a ellos.
Sin embargo, para reducir los riesgos a personas y propiedades, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios. Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y organizaciones?
Esperar a que los cambios sucedan y después adaptarse a ellos” o “Tratar de prepararse ante estos cambios para reducir los riesgos a personas y propiedades”
In the second version of the question, “las personas y organizaciones” was replaced by “muchos
líderes empresariales piensan que las personas y organizaciones”. In the third, fourth, and fifth
versions of the question, “líderes empresariales” in the second version was replaced with “líderes
religiosos”, “muchos funcionarios gubernamentales”, and “muchos investigadores
universitarios”, respectively.
Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Estos científicos creen que el calentamiento en el pasado y futuro tendrá efectos sobre las personas y el medio ambiente, y que cambiará la forma en que las personas vivirán sus vidas.
Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse a ellos.
Sin embargo, para reducir riesgos a personas y propiedades, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios.
Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y organizaciones? Esperar a que los cambios sucedan y después adaptarse a ellos, o tratar de prepararse ante estos cambios para reducir riesgos a personas y propiedades.
The second, third, fourth, fifth and sixth versions of the questions were constructed by replacing
“reducir riesgos ” in the first version with “reducir la vulnerabilidad ”, “aumentar la resistencia ”,
“aumentar la preparación ”, “aumentar la preparación ”, and “prevenir la mal adaptación ”,
respectively.
Study 2:
The first version was:
Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Estos científicos creen que el calentamiento en el pasado y futuro tendrá efectos sobre las personas y el medio ambiente, y que cambiará la forma en que las personas vivirán sus vidas.
Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse a ellos. Sin embargo, para reducir riesgos a personas y propiedades, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios.
Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y organizaciones? Esperar a que los cambios sucedan y después adaptarse a ellos, o tratar de prepararse ante estos cambios para reducir riesgos a personas y propiedades.
The second version of the question was:
Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Y los científicos creen que el calentamiento pasado y futuro causará más lesiones, enfermedades y muertes.
Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse a ellos. Sin embargo, para reducir riesgos a la salud de personas, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios.
Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y
organizaciones? Esperar a que los cambios sucedan y después adaptarse a ellos, o tratar de prepararse ante estos cambios para reducir riesgos a la salud de personas
The third version of the questions was:
Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Y los científicos creen que el calentamiento pasado y futuro destruirá playas y propriedades a lo largo de la costa, reducirá el suministro de productos del mar, y desbaratará la vida silvestre.
Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse a ellos. Sin embargo, para reducir riesgos a las playas, propriedades a lo largo de la costa, el suministro de productos del mar y la vida silvestre, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios.
Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y organizaciones? Esperar a que los cambios sucedan y después adaptarse a ellos, o tratar de prepararse ante estos cambios para reducir riesgos a las playas, propriedades a lo largo de la costa, suministro de productos del mar, y vida silvestre.
Studies 3 and 4
Muchos científicos que estudian el clima mundial creen que la Tierra se ha estado calentando durante los últimos 100 años y que continuará haciéndolo en el futuro. Estos científicos creen que el calentamiento en el pasado y futuro tendrá efectos sobre las personas y el medio ambiente, y que cambiará la forma en que las personas vivirán sus vidas.
Una opción que tienen las personas y organizaciones es esperar a que estos cambios sucedan y después adaptarse. Sin embargo, para aumentar la preparación de las personas y propiedades, las personas y organizaciones podrían tomar medidas pronto para prepararse ante estos cambios.
Si tuvieras que escoger, ¿qué preferirías que hicieran las personas y organizaciones? Esperar a que los cambios sucedan y después adaptarse a ellos, o tratar de prepararse ante estos cambios antes de que sucedan.
In the second and third versions of the question, the word “adaptarse” in the phrase “esperar a
que estos cambios sucedan y después adaptarse” in the first version was replaced with the word
“ajustarse” or “responder”, respectively. The fourth, fifth, and sixth versions of the question
were formulated by replacing the word “prepararse ante” in the first, second, and third version
with the word “planificar”, respectively.
8 Analytical Methods
To permit parsimonious analysis of the data, we stacked the data from the multiple studies into a
single dataset. Specifically, the original datasets for Studies 1, 2, 3, and 4 contained 1004, 1000,
1003, and 1001 rows, respectively, one for each of the 1004, 1000, 1003, and 1001 respondents
in each study, respectively. The final dataset contained 1004+1000+1003+1001 = 4008 rows.
The original dataset included variables (one per column) representing the respondent's answers to
the survey questions, as well as the sample weight for the respondent and a series of 20 dummy
variables indicating the experimental conditions in which the respondent fell (e.g., which frame,
which endorser, which of each set of language). After the data from all respondents were
stacked into a single dataset, 4 dummy variables were created to identify in which study the
respondent participated.
9 Predictors of Support for Preparation
This investigation explored the predictors of attitudes toward preparation. Past research on
environmental attitudes suggests that demographic groups differ in their support for efforts to
protect the natural world (for a review, see, e.g., Daniels et al. 2013). Often, women have been
found to be more pro-environment than men, perhaps due to women’s socialization attaching
greater value to nurturing or altruism (Daniels et al. 2013). Some past work suggests that racial
minorities might be more supportive of environmental protection, perhaps due to increased
perceptions of vulnerability to environmental problems, but this pattern has not been consistently
observed (Daniels et al. 2013). Some past studies suggest that younger adults are more pro-
environment, perhaps because they expect to live longer with the environment than will older
individuals (Daniels et al. 2013). Formal education has been observed as a predictor of
environmental attitudes in only a minority of past studies that tested for it (Daniels et al. 2013).
Similarly, income has rarely been found to be a predictor. More reliably, Democrats have been
more pro-environment than have Republicans (Daniels et al. 2013). We used the survey data to
gauge the associations of these variables with attitudes toward preparation.
A logistic regression identified various predictors of preference for preparation. First,
consistent with past research, preference for preparation was the greatest among Democrats and
lowest among Republicans, with Independents in between. Democrats’ preference for
preparation was 16 percentage points greater than that of Independents’ (p < .01), and
Republicans’ preference was 11 percentage points lower than that of Independents’ (p < .01).
Also in line with some past research, preference for preparation was greater among
females than among males, a significant difference of 8 percentage points (p < .01). Preference
for preparation was generally higher among younger adults—compared to adults age 65 or older,
preference for preparation was 8 percentage point higher among adults age 18-24 (p < .01), 9
percentage point higher among adults age 25-34 (p < .01), 7 percentage point higher among
adults age 35-44 (p < .01), and 6 percentage point higher among adults age 54-64 (p < .01).
Hispanics were 11 percentage points more likely to prefer preparation than non-Hispanic whites
(p < .01). Preference for preparation was generally higher among higher income respondents—
compared to respondents with less than $25,000 income, preference was 5 percentage points
higher among respondents with income $25,000 to less than $55,000 (p = .10), 6 percentage
points higher among respondents with income $55,000 to less than $75,000 (p = .03), and 5
percentage points higher among respondents with income $100,000 or more (p = .05).
Preferences did not differ significantly by education or by region of residence.
10 Discussion of Findings
Language choices Perhaps the term preparedness is appealing because it (1) embraces
the notion that undesirable events are inevitable with regard to global warming and (2)
emphasizes the notion of taking actions in advance to minimize damage, rather than simply
responding effectively or avoiding responding ineffectively after the fact. Yet this logic raises a
puzzle: if “increase preparedness” is appealing to people, why is “increase readiness” any less
appealing, since the latter’s meaning is extremely similar if not identical to the former’s?
Interestingly, “preparedness” has been a term of choice more often than “readiness” in
various specific fields of risk management. For example, a Google search on April 3, 2014, for
preparedness and public health yielded 11 million webpages, whereas a search for readiness and
public health yielded 6 million webpages. Likewise, a search for preparedness and homeland
security yielded 4 million webpages, whereas a search for readiness and homeland security
yielded 1 million webpages. And a search for preparedness and disaster yielded 39 million
webpages, whereas a search for readiness and disaster yielded 6 million webpages. Thus,
between these two terms, various risk-related fields seem to have a preference for preparedness,
and our evidence justifies that preference if the goal of word choice is to generate support for the
activity.
Yet the reason or reasons explaining why “preparedness” is more appealing or motivating
than “readiness” is not illuminated by the current research. One interesting possibility is that the
American public is more used to hearing the term “preparedness” with regard to threats
generally, so the explanation lies in familiarity of the specific relevant uses of the term rather
than the public’s exposure to the term generally in all language. Clearly, future research is
needed to understand this phenomenon, and the findings reported here provide justification for
this research.
Endorsers It is interesting to consider one possible explanation for why endorsement of
preparation efforts by government officials and university scientists did not alter public
preference for preparation. Although we have not conducted a comprehensive content analysis
of media coverage and public statements, it seems that public endorsements of the existence and
threat of global warming and a need to prepare for it have come to the public more often from
government officials and university researchers than from religious leaders or business leaders.
So perhaps when our survey respondents heard that preparation was endorsed by government
officials or university scientists, the respondents learned nothing now, so this information had no
impact on their attitudes. That is, respondents may already have known that government
officials and university researchers thought preparation was wise. Future research might explore
this possibility as well.1
Emphasis frame Our finding of equal impact of various emphasis frames might seem at
first to contradict evidence reported by Myers et al. (2012), who examined three “emphasis
frames” regarding global warming: (1) public health, (2) national security, and (3) the
environment. These investigators concluded that a public health frame aroused more hopeful
emotions about climate change than did the other two frames. Since we did not examine the
same two alternate frames, our results should not be viewed as directly contradicting theirs.
Furthermore, their measures of emotional reactions are not comparable to our measures of 1 In retrospect, it would have been valuable to measure respondents’ trust in the various endorsers in general and specifically with regard to statements addressing global warming. Unfortunately, our resources did not permit making such measurements.
attitudes toward preparation, so this is another reason not to necessarily expect their results to
match ours.2
Implications for survey methodology It is valuable to view the present findings as
relevant to survey research methodology, illuminating how different question wordings can
produce different results in surveys. A huge literature has explored such phenomena (e.g.,
Schuman and Presser 1981). Although our experiments showed some instances in which a
change in wording altered the distribution of respondents’ answers, we also observed many
instances in which wording changes did not alter the distribution of answers. This suggests that
survey results are not necessarily so easily manipulated by wording changes or as fragile as some
critics have suggested.
The evidence here that question wording alterations affected answers is in line with a
variety of other studies showing that people sometimes react differently to different words
expressing the same idea. For example, Dawes (2000) asked respondents about the likelihood
that they would be “changing from” (or, alternatively “renewing with”) their present insurance
provider. We might expect that probability judgments would be similar for the two question
wordings, but this was not the case. More people said there was no chance they would defect on
car insurance (51%) than said they were certain they would renew (41%), a significant
difference.
2 Although Myers et al.’s (2012) publication did not report any tests of statistical significance, their online supplementary materials showed that the public health frame yielded significantly more hope than the environment frame in just two of six subgroups of the population that were examined, no difference in three subgroups, and the reverse difference in the sixth subgroup. The public health frame also yielded significantly more hope than the national security frame in only three subgroups. And the public health frame yielded significant less anger than did the environment frame or the national security frame in three of six subgroups. Furthermore, the frames did not have consistent effects on both dependent variables in the same subgroups. Thus, those data do not support the general claim that the public health frame was consistently most impactful.
Similar results were reported in a study of evaluations of a person who can be referred to
as the committee “chair” or the committee “chairperson.” Although these might seem to be
synonyms, McConnell and Fazio (1996) found that a person labeled “chair” was perceived as
more masculine than the same person labeled “chairperson,” but only if the person was said to
have chosen the label for himself/herself, regardless of gender. But when respondents knew that
the label was chosen by someone other than the person to whom it was attached, the label had no
impact on inferences of his/her personality. Our findings contribute to the literature on such
effects by documenting some wording changes that were consequential and some that were not.
Our findings suggest humility when guessing whether a change in question wording will alter
responses, because some changes that seemed likely to alter reactions did not.
Generalization We should be careful about generalizing the present results to non-survey
settings. In this study, questions were asked by survey interviewers who respondents are likely
to have perceived as unbiased with regard to the issue of global warming. When a message is
delivered by representatives of other organizations, different responses might be observed,
though we see no strong reason why this should occur. Furthermore, the results reported here
may apply most readily to situations in which people hear a message aurally and form opinions
during a paced conversation, as happens during survey interviews. If respondents read a
message and/or determine how long to think about a message before rendering a judgment and
perhaps talk to other people, different results might be observed.
References
Allstate/National Journal (2012) Allstate/National Journal heartland monitor poll XIII. http://syndication.nationaljournal.com/communications/Allstate%20National%20Journal%20Heartland%20Monitor%20XIII%20TOPLINE.pdf. Accessed 8 October 2013.
Daniels DP, Krosnick JA, Tichy MP, Tompson T (2013) Public opinion on environmental policy in the United States. In: Kraft M, Kamieniecki S (eds) Handbook of U.S. environmental policy. Oxford University Press, New York.
Dawes J (2000) The impact of question wording reversal on probabilistic estimates of defection/loyalty for a subscription product. Mark. Bull.11(Research Note 1):1-9.
McConnell AR, Fazio RH (1996) Women as men and people: Effects of gender-marked language. Personality and Soc. Psychol. Bull. 22:1004-1013.
Merriam-Wesbter (2014) http://www.merriam-webster.com/. Accessed in April 2014.
Northeastern University (2012) Northeastern University – Innovation in higher education survey toplines. http://www.northeastern.edu/innovationsurvey/pdfs/survey-results.pdf. Accessed 8 October 2013.
Schuman H, Presser S (1981) Questions and answers in attitude surveys: Experiments on question form, wording, and context. Academic Press, San Diego.
Smith TW, Son J (2013) Trends in public attitudes about confidence in institutions. NORC at the University of Chicago, Chicago, Illinois.
Table S1: Frequency of Various Phrases Appearing on Webpages
Phrase Frequency of Occurrences
Frequency of Occurrences(along with “global warming” or
“climate change”)
Plan for 79 million 12 million
Prepare for 61 million 8 million
Respond to 161 million 20 million
Adapt to 21 million 7 million
Adjust to 12 million 1 million
Notes: Presented are the numbers of web pages, rounded to the nearest million, containing the phrases, obtained using Google on August 19, 2013.
Table S2. Demographics of the Sample in Study 1 and the Current Population Survey (CPS)
Demographic GroupJanuary 2013 Survey
(unweighted)January 2013 Survey
(weighted) February 2013 CPSJanuary 2013 Survey
(weighted) – CPS
Gender
Male 47.4% 48.6% 48.1% .5%
Female 52.6 51.5 51.9 -.5
Total 100.0% 100.0% 100.0%
(N = 1004) (N = 1003) (N = 100,497)
Age
18-24 7.1% 11.6% 12.6% -1.0%
25-34 11.6 19.2 17.5 1.7
35-44 12.8 13.2 16.8 -3.6
45-54 20.8 23.0 18.4 4.6
55-64 23.6 15.3 16.5 -1.2
65+ 24.1 17.6 18.2 -.6
Total 100.0% 100.0% 100.0%
(N = 998) (N = 998) (N = 100,497)
Race and Ethnicity
Non-Hispanic White 68.7% 68.0% 66.0% 2.0%
Non-Hispanic Black 6.2 11.2 11.6 -.4
Hispanic 20.6 13.9 15.0 -1.1
Other 4.5 6.9 7.3 -.4
Total 100.0% 100.0% 100.0%
(N = 1004) (N = 1004) (N = 100,497)
Education
Less than HS 11.2% 13.9% 12.2% 1.7%
HS graduates 22.6 33.9 29.9 4.0
Some college 29.1 24.5 29.0 -4.5
College or higher 37.1 27.8 29.0 -1.2
Total 100.0% 100.0% 100.0%
(N = 994) (N = 994) (N = 100,497)
Region
Northeast 18.9% 18.5% 18.2% .3%
Midwest 22.3 21.9 21.4 .5
South 36.4 37.0 37.1 -.1
West 22.4 22.7 23.3 -.6
Total 100.0% 100.0% 100.0%(N = 1004) (N = 1004) (N = 100,497)
Table S3. Demographics of the Samples in Study 2 and 3 and the Current Population Survey
(CPS)
Demographic Group
February 2013 Survey
(unweighted)
February 2013 Survey (weighted)
March 2013 Survey
(unweighted)
March 2013 Survey
(weighted)February 2013 CPS
Difference of February
2013 Survey
from CPS (weighted)
Difference of March
2013 Survey
from CPS (weighted)
Gender
Male 53.2% 48.1% 48.2% 48.6% 48.1% .0% .5%
Female 46.8 51.9 51.8 51.5 51.9 .0 -.4
Total 100.0% 100.0% 100.0% 100.0% 100.0%
(N = 1000) (N = 1000) (N = 1003) (N = 1003) (N = 100,497)
Age
18-24 7.6% 15.9% 8.1% 13.9% 12.6% 3.3% 1.3%
25-34 9.3 15.0 10.6 16.9 17.5 -2.5 -.6
35-44 10.9 12.2 10.4 13.5 16.8 -4.6 -3.3
45-54 18.3 24.1 17.3 22.7 18.4 5.7 4.3
55-64 25.3 15.3 24.3 13.7 16.5 -1.2 -2.8
65+ 28.5 17.6 29.5 19.2 18.2 -.6 1.0
Total 100.0% 100.0% 100.0% 100.0% 100.0%
(N = 987) (N = 987) (N = 994) (N = 994) (N = 100,497)
Race and EthnicityNon-Hispanic White 75.7% 67.9% 69.2% 67.1% 66.0% 1.9% 1.1%
Non-Hispanic Black 10.6 10.8 10.5 11.5 11.6 -.8 -.1
Hispanic 8.3 13.8 13.9 13.8 15.0 -1.2 -1.2
Other 5.4 7.5 6.5 7.6 7.3 .2 .3
Total 100.0% 100.0% 100.0% 100.0% 100.0%
(N = 1000) (N = 1000) (N = 1003) (N = 1003) (N = 100,497)
Education
Less than HS 6.7% 10.3% 7.4% 11.7% 12.2% -1.9% -.5%
HS graduates 21.5 37.4 22.3 36.1 29.9 7.5 6.2
Some college 31.1 24.5 33.5 24.5 29.0 -4.5 -4.5
College or higher 41.4 27.7 36.8 27.7 29.0 -1.3 -1.3
Total 100.0% 100.0% 100.0% 100.0% 100.0%
(N = 988) (N = 988) (N = 990) (N = 990) (N = 100,497)
Region
Northeast 19.0% 18.5% 18.5% 18.5% 18.2% .3% .3%
Midwest 22.5 21.9 21.9 21.9 21.4 .5 .5
South 36.2 37.0 37.0 37.0 37.1 -.1 -.1
West 22.3 22.7 22.7 22.7 23.3 -.6 -.6
Total 100.0% 100.0% 100.0% 100.0% 100.0%(N = 1000) (N = 1000) (N = 1003) (N = 1003) (N = 100,497)
Table S4. Demographics of the Sample in Study 4 and the Current Population Survey (CPS)
Demographic GroupApril 2013 Survey
(unweighted)April 2013 Survey
(weighted) February 2013 CPS
Difference of April 2013
Survey from CPS (weighted)
Gender
Male 47.4% 48.6% 48.1% .5%
Female 52.7 51.4 51.9 -.5
Total 100.0% 100.0% 100.0%
(N = 1001) (N = 1001) (N = 100,497)
Age
18-24 6.5% 14.2% 12.6% 1.6%
25-34 10.6 16.7 17.5 -.8
35-44 10.3 13.2 16.8 -3.6
45-54 18.1 23.1 18.4 4.7
55-64 24.1 14.0 16.5 -2.5
65+ 30.4 18.9 18.2 .7
Total 100.0% 100.0% 100.0%
(N = 977) (N = 977) (N = 100,497)
Race and Ethnicity
Non-Hispanic White 71.0% 66.5% 66.0% .5%
Non-Hispanic Black 8.2 11.5 11.6 -.1
Hispanic 13.9 13.7 15.0 -1.3
Other 6.9 8.3 7.3 1.0
Total 100.0% 100.0% 100.0%
(N = 1001) (N = 1001) (N = 100,497)
Education
Less than HS 6.4% 11.8% 12.2% -.4%
HS graduates 21.3 36.0 29.9 6.1
Some college 34.2 24.5 29.0 -4.5
College or higher 38.1 27.8 29.0 -1.1
Total 100.0% 100.0% 100.0%
(N = 973) (N = 973) (N = 100,497)
Region
Northeast 19.0% 18.5% 18.2% .3%
Midwest 22.4 21.9 21.4 .5
South 36.3 37.0 37.1 -.1
West 22.4 22.7 23.3 -.6
Total 100.0% 100.0% 100.0%(N = 1001) (N = 1001) (N = 100,497)
Table S5. A Summary of Manipulations in Studies 1-4
Manipulation Study 1 Study 2 Study 3 & 4
Endorsers Manipulated No endorser No endorser
Policy Objective Manipulated Manipulated “Increase readiness”
Consequence
“People, the environment, and property”
Manipulated Manipulated
Verb Describing Preparing “Prepare” “Prepare” Manipulated
Verb Describing Waiting “Adapt” “Adapt” Manipulated
Figure S1: Of All Newspaper Articles Containing the Phrases “Global Warming” or “Climate Change” and Any of the Six Target
Phrases, the Proportion That Contained Each of the Six Target Phrases
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 20130%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
"reduce risk" "reduce vulnerability" "increase resilience""increase preparedness" "increase readiness" "prevent maladaptation"