Professional Documents
Culture Documents
Utilities Policy
journal homepage: www.elsevier.com/locate/jup
a r t i c l e i n f o a b s t r a c t
Article history: This research focuses on understanding the significance of information in affecting public opinion. Nu-
Received 5 January 2017 clear energy is among the most controversial of policy issues and the information-saturated mature civil
Received in revised form society of Hong Kong makes it an ideal location for this study. A novel field experiment is set up in which
8 April 2017
pro-nuclear and anti-nuclear information collected from local media reports is inserted into two iden-
Accepted 8 April 2017
Available online 14 April 2017
tical questionnaires, to which respondents are randomly assigned. The percentage of the public who
switched their opinion was estimated to be 5e37%, depending on the framing of the question.
© 2017 Elsevier Ltd. All rights reserved.
Keywords:
Information and public opinion
Nuclear energy
Field experiment
http://dx.doi.org/10.1016/j.jup.2017.04.001
0957-1787/© 2017 Elsevier Ltd. All rights reserved.
T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57 49
2006). It is sometimes suggested that public opposition is lower tendency to decline the facts about climate change, which
emotional and a result of ignorance, while the science is unques- means that the framing of solutions could influence public per-
tioned. On the one hand, some studies have found no significant ceptions of the problem (Pralle and Boscarino, 2011).
impacts of information to validate this assumption (Ramana, 2011). Pidgeon et al. (2008) also suggested that in the case of a major
The problem may also be on the “scientific” side when experts fail incident, public support would be withdrawn over concerns about
to understand the public's perceptions about risk. Taking the safety, another possible frame. As illustrated by Bird et al. (2014),
example of genetically modified maize in Mexico, Brunk (2006) Australian public opinion on nuclear power shifted more to the
argued that the conflict with the local farmers resulted from the negative side after the Fukushima incident. Mazur (1981) also
experts' failure to understand the social and cultural implications of suggested that on controversial issues like nuclear power, media
GM maize's “intrusion” into the local community. On the other coverage has a particular power in shaping public opinion, and the
hand, some studies have confirmed that information has discern- impact could be difficult to determine. Citing the example of the
ible impacts. For example, a study in South Korea used the Three Mile Island incident, Mazur argued that the media could not
contingent valuation method to estimate the social value of nuclear only report the event, but also generate and shape controversy,
energy, and better information was found to increase public sup- given that the percentage of public opposition towards nuclear
port (Jun et al., 2010). The study by Stoutenborough et al. (2013) power coincided with the amount of media coverage. He further
also suggested that a person who possesses more information argued that the public tends to take a more conservative stand
tends to support nuclear power. when faced with scientific controversies, citing that people who
From the perspective of journalism, there is no absolute “unbi- received both positive and negative information about fluoridation,
ased” or “undistorted” news since, according to Gans (1979), “the tended to oppose it compared to those who did not receive the
mere act of reproduction would constitute a distortion on that re- information. Even within scientific circles, support for nuclear po-
ality” and “non-distortion” of news can be judged only from a wer could be built by framing the issue in terms of “relieving the
“relational perspective”. In the real world, a single piece of infor- energy needs of the under-developed countries” and “utilising
mation can be more or less biased, one-sided, manipulated, or nuclear power for desalination”, as mentioned in Sovacool and
sometimes even wrong. The internal values of the journalist and Ramana (2015), showing that scientists are not “immune” to
the practices of news gathering would cause bias in the news these rhetorical devices.
(Entman, 1990). The media play an influential role in public opinion The issue of public risk perception further complicates the
formation on environmental issues through agenda setting, which problem. Renn (1998) summarised some of the major frameworks
is disclaimed by journalists, yet done (unintentionally) whenever of risk perception in a social scientific and technical analysis of risk.
they decide what to report and at what level of detail (McCombs, Likewise, public risk perception is affected by personal biases and
2004). Doyle (2011) examined how the media have contributed issue framing (Renn, 1998). Kahan et al. (2007) suggested that even
to the reframing of nuclear power as a low-carbon substitute to when the same piece of information is given, “identity-protective”
mitigate climate change in the UK and how this influences public cognition leads individuals to accept information when it is framed
understanding. Palfreman (2006) gave a more detailed account of in a way that affirms their personal values and commitments. In
how information can be distorted during its conveyance: the addition, scepticism about information varies depending on the
“factual distortions” due to the journalist's scientific illiteracy, the source of the information: the government; scientists; or the nu-
“narrative distortions” of reporting stories of readers' interest clear industry for instance (Greenhalgh and Azapagic, 2009). Renn
instead of the facts, and the “distortions of balance” under the (1998) proposed that there is a need for deliberative communica-
notion of “balanced coverage”. Instead of a one-way flow of infor- tions between experts, the general public, and industry regulators
mation from the media to the public, there exists a two-way flow for effective mutual learning and better understanding of social
because the media adjust their reporting frame based on the in- values.
terests of their readers or viewers. These literature provide examples of how framing influences
A study of the power of the media on public opinion towards public opinion formation. The focus was mainly on how framing
nuclear energy in the post-war USA up to the 1980s showed that can shift opinion, yet few studies have addressed how distorting
the media discourse is influential on the formation of public information about the same issue or frame might influence public
opinion on the issue (Gamson and Modigliani, 1989). In their work, opinion. The frames most commonly discussed were risk percep-
Gamson and Modigliani suggested that journalists play a double tion and environmental concerns, including climate change. Issues
role in opinion formation. How and what kinds of information are other than these two are rarely encountered in the available public
delivered would affect the audience. With commentary, journalists opinion polling analyses (e.g., Nealey et al., 1983; Rosa and Dunlap,
also react to the issue that has been partly framed, giving opinions 1994; Whitfield et al., 2008). Few studies have examined the rele-
that are shared by the audience. The framing of the issue is rec- vance of other significant frames for policy-makers and utility
ognised as an influential factor upon public opinion on the issue: regulators, including operating costs and reliability. Ansolabehere
whether nuclear power is regarded as a progressive means for and Konisky (2009) discussed the issues of cost, siting, and envi-
economic development or a devil's threat to safety would alter ronmental concern, but their work focused on the comparison of
public perceptions and support for the utilisation of nuclear power. different energy sources apart from nuclear power. Thus, research
A study by Bickerstaff et al. (2008) indicated that public attitudes relating to public opinion formation and nuclear power utilisation
towards nuclear power are not necessarily fixed, but could vary remains limited. Knowledge about public opinion in this area and
depending on the framing, although an increase in preference to- the influence of issue framing is valuable for both the government
wards nuclear power is probably a result of the reluctant accep- and civil society in order to facilitate collective decision-making
tance of the lesser evil compared to other problems, particularly about nuclear energy. An understanding of public opinion can
climate change. When put into the context of climate change also help utility regulators improve their practices, such as infor-
mitigation and energy security, an increase in support for nuclear mation disclosure. The relevance of a particular aspect of nuclear
power is possible, although it would be a conditional and reluctant power to the process of collective decision-making will be affected
acceptance of nuclear power (Bird et al., 2014; Corner et al., 2011; by public opinion, which is in turn shaped by information.
Pidgeon et al., 2008). Conversely, when nuclear power is pro- This paper aims to quantitatively measure “opinion switchers”, a
posed as a solution to climate change, the public would also have a group within the general public that could easily switch between
50 T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57
positive and negative opinions when receiving different informa- impacts of all the variables together, and this makes it difficult, if
tion. Whether public opinion is altered, the proportion of switched not impossible, to single out the impacts of one-sided information
opinion and the factors affecting the switch are examined. The alone. We designed a novel field experiment to effectively eliminate
mature media and numerous information channels, which allow the impacts of these control variables. Field experiment as a
easy information distribution and access, make Hong Kong a good research method has been applied more and more widely to
location for this study. In this context, the presentation of distorted conduct experiments on the natural environment (Harrison and
or one-sided information from one source should be unlikely to List, 2004). The said novel method is proposed to examine the
significantly influence public opinion formation due to scepticism impact of one-sided information on public opinion in a daily-life
and easy access to alternative information sources. This influence scenario, with the influence of respondent background and media
from one-sided information would be a bigger problem in other discourse about nuclear power isolated. The description of the
societies where access to multiple information sources is poorer. formulae that explains how this method works is shown in
Another reason for studying Hong Kong is the growing impor- Appendix 1.
tance of nuclear power in the fuel mix in light of the consequential Two questionnaires were designed, and every respondent was
exposure to the issue following the Fukushima incident. The issue randomly selected to answer one of them. These two question-
of nuclear power utilisation in Hong Kong has been largely un- naires have exactly the same set of questions. We inserted simple,
touched by the general public and the media after the commission easy-to-understand, yet one-sided information in front of opinion-
of the Daya Bay Nuclear Power Plant in neighbouring Guangdong soliciting questions. The information in each of the two question-
province in the early 1990s until the Fukushima incident. Public naires was consistently one-sided, either positive or negative for
opinion has also played an insignificant role in deciding the energy nuclear energy. The respondents would not be exposed to both
mix in Hong Kong. In September 2010, with the objective of sides of the story and only the impact of the information from a
replacing coal and reducing air pollutants and CO2 emissions, the particular side was recorded for each respondent. If the re-
Hong Kong Environment Bureau proposed to increase the share of spondents were genuinely randomly selected, the paired ques-
nuclear power in the fuel mix for the electricity supply from 23% to tionnaires will effectively control the impacts of independent
50% by 2020 by importing more nuclear electricity from the Daya variables other than the one-sided information.
Bay Nuclear Power Plant (Environment Bureau, 2010). However, Information in the two questionnaires was extracted from local
several months later, in March 2011, the Fukushima incident newspaper articles from the news search through the Wisenews
attracted broad media coverage, and public opinion towards nu- search engine. Articles from March 11, 2011 to February 4, 2013
clear power swung greatly to the unfavourable side. In order to about “nuclear power” were collected. The starting date marked the
facilitate collective decision-making, there was a major focus to Fukushima incident, which prompted media attention on many
provide more information to the general public and to enhance nuclear-related issues, including the Daya Bay Nuclear Power Plant
“nuclear literacy” on the basic features of nuclear power operation, and the role of nuclear power in Hong Kong's fuel mix. The search
safety systems, and evacuation procedures (Loh, 2012). In results were further screened in three steps by using content
September 2012, the Hong Kong Nuclear Society was established analysis, focusing on information about five nuclear-power-related
with this purpose clearly in mind (Hong Kong Nuclear Society, issues: stability; cost; safety; pollution; and impact on climate
2014). Given that Hong Kong remains at the infancy stage of pub- change. Firstly, 588 passages were filtered out in accordance with
lic discussion about the utilisation of nuclear energy, there is thus their relevance to the issues based on the topics. Secondly, key
much interest in how information provision might affect public points were extracted and categorised. Thirdly, anti-nuclear infor-
opinion formation in this area. mation and pro-nuclear information were paired in a way that the
Public opinion is commonly solicited as a single indicator in two sides form a “balanced report” on the issue. For instance, the
opinion surveys, with various aspects of issues mixed together. In information “nuclear fuel costs less than fossil fuel” is given in the
this study, we argue that the public opinion could vary significantly pro-nuclear survey while the information “nuclear fuel cost would
depending on issue framing and that it is crucial to examine the increase with the rising demand” is given in the anti-nuclear sur-
underlying factors contributing to the opinion. We also argue that vey. Put together, the two form a balanced report on fuel cost:
public opinion could be shaped, though this is not likely for values “nuclear fuel currently costs less than fossil fuel, but its price would
that have been deeply entrenched, and the proportion of opinion increase in the future with rising demand” (see Table 1 for details).
switchers would vary for different aspects and values. These issues The information was provided without references to eliminate the
should be handled with caution by utility regulators because public influence of the information sources.
pressure could ultimately affect decisions about utility operations, The questions in the questionnaires were structured into three
especially in democratic societies, and regulators are also in a po- categories (Table 2). In the first category, the respondents were
sition to provide the information that influences opinion. Thus they asked to express their general impression and concerns about nu-
deserve a closer investigation. The organisation of this paper is as clear power by using an 11-point scale: from 0 to 10, in which
follows. Section 2 introduces data and methodology. Section 3 0 denotes a positive impression or not concerned at all whereas 10
analyses the data to measure the proportion of opinion switchers. denotes a negative impression or highly concerned. Six questions in
Section 4 explains the results and identifies key influential factors. total ask about general attitude and concerns about reliability, cost,
Conclusion and discussion are provided in Section 5. greenhouse gas emission, safety, and pollution from nuclear energy
(see Category 1 of Table 2). The purpose was to lead respondents
2. Data and methodology into the topic, and, more importantly, to test the distribution
randomness of the two questionnaires. If the randomness was well
2.1. The design of the questionnaires achieved, the opinions of the two groups of respondents should
resemble each other statistically. In the second category, one-sided
In an attempt to clearly distinguish the relationship between information was given to the respondents before the questions on
public attitude, as the dependent variable, and one-sided infor- the five aspects of nuclear power were asked, by again using an 11-
mation, as the independent variable, one major obstacle is to point scale (see the first five items in Category 2 of Table 2). Re-
control many other potentially influential independent variables. spondents were revealed to the information that showed only part
The directly solicited data of public opinion generally mix the of the story in the particular aspects, either supporting the anti-
T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57 51
Table 1
Paired, one-sided information in the two questionnaires.
Reliability Hong Kong has been importing electricity from the Daya Bay Nuclear Nuclear power plants may shut down ahead in case of an oncoming
(POWER_CUT) Power Plant for 20 years. No major blackout has occurred. typhoon, to avoid damage to the facilities.
Electricity tariff For generation cost, using nuclear fuel is cheaper than using fossil fuel. The demand for nuclear fuel is expected to increase by 4% annually,
(TARIFF) which could lead to substantial increase in price.
Safety risk of nuclear
The Daya Bay Nuclear Power Plant is situated at a geologically stable Facing geologic cycles and climate change, effective designs for safety of
power (SAFETY) location, with no record of magnitude 6 or above earthquakes. A huge nuclear power plants in the past may be no longer applicable today.
barrier is built to protect the plant from tsunamis.
Environmental No fossil fuel is burnt during nuclear power generation. Therefore, it does Nuclear waste is highly radioactive, which is harmful to human health.
impact not cause air pollution. The extraction and processing of nuclear fuel also generate significant
(POLLUTION) amounts of chemical waste.
Impact to climate Compared to a coal-fired power plant of the same scale, the Daya Bay Consider the whole process of construction, fuel processing and power
change (CLIMATE) Nuclear Power Plant can reduce CO2 emission equivalent to planting generation, the greenhouse gas emission of nuclear power is three times
100,000 ha of forest. that of other renewable energies.
Development of Whether Hong Kong is purchasing nuclear power from the mainland Hong Kong's purchase of nuclear power from Guangdong Province would
nuclear power in would not affect the development strategy of nuclear power in China. encourage the development of nuclear power in the province. A high
mainland China Hong Kong should buy more nuclear power to increase the participation concentration of nuclear reactors would largely increase the chance of a
in and influence on nearby nuclear power plants. nuclear accident.
Table 2
Questionnaire design and results.
Question Variables Explanation of variables One-sided information given before Mann- Asymp. Sig. (2-
category question Whitney tailed)
Ua
Category 2 POWER_CUT The risk of power cut in Hong Kong if using more nuclear Yes 26919.5 0.000
power
TARIFF The impact on electricity tariff of more nuclear power in the Yes 24079.5 0.000
fuel mix
SAFETY The safety risk of the Daya Bay Nuclear Power Plant to Hong Yes 27256.5 0.000
Kong
POLLUTION The pollution of nuclear power to the world Yes 18512.5 0.000
CLIMATE The significance of nuclear power in mitigating greenhouse Yes 18994.0 0.000
gases
JUDGE The overall judgment on the pros and cons of nuclear power No 25144.5 0.000
MIX_NUCLEAR Nuclear power's share in the fuel mix Yes 24673.5 0.000
nuclear or the pro-nuclear stance. The information includes the without receiving one-sided information, was set in the current
significance of nuclear power on greenhouse gas mitigation, the study. These two settings restrict the result analyses in several
electricity tariff, the risks of Fukushima-type accidents, and the ways, and this will be discussed in the respective sections below.
nuclear development plan in Mainland China (as shown in Table 1).
Besides answering specific opinion questions, the respondents 2.2. Questionnaire validation, data collection, and processing
were asked for their overall judgment on the relative strength of
nuclear power's pros and cons as well as their ideal fuel mix (see To validate the questionnaires and to check their comprehen-
the last two items in Category 2 of Table 2). To ensure the re- sibility before the main survey was conducted, a pilot survey was
spondents' close attention, the high values in the scale corre- conducted in the form of both a street survey in metro stations and
sponded to a pro-nuclear stance in some questions but to an anti- an online survey through email and social media. A total of 144
nuclear stance in others. In the third category, demographic data questionnaires were collected, namely, 80 from the street survey
were solicited (see Category 3 of Table 2). If the two questionnaires and 64 from online. In the pilot survey it is shown that public
were genuinely distributed randomly, the two groups of re- opinion could be intervened by one-sided information. Based on
spondents would show similar demographic characteristics. the opinion of the respondents, the phrasing of the information
No question asked for the respondent's impressions about the provided in the questionnaires was fine-tuned to reduce the
five specific issues before the reception of information to avoid the complexity for easier comprehension.
impression of suggestive interrogation which would distort the The main survey was conducted in the form of a street survey in
result. Owing to time and resource limitations, no control group, the autumn of 2015. Public parks and metro stations were selected
52 T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57
for the survey and they covered all major regions of Hong Kong, INCOME in Table 2, p > 0.4 for all) of the respondents showed
including the New Territories, Kowloon, and Hong Kong Island. No statistically significant differences between the pro-nuclear and the
particular population group was targeted in this study and the anti-nuclear questionnaire respondents. This indicates that the two
design of the collection method was mainly for randomness. The groups of respondents responding to the two questionnaires were
two types of questionnaire were distributed alternately to ensure a statistically non-differentiable.
random and even selection of population. A total of 534 responses One-sided, paired information was given before the Category 2
were collected in the main survey, but not all responses could be questions. It was found that the one-sided information was effec-
used. tive in tilting public opinion, with a systematic, although not sub-
The validity of the current study relies on the sampling process stantial, change. For the two groups of respondents on average,
that an unbiased random sampling should be conducted, to isolate there was a highly statistically significant difference between the
the impact of one-sided information on the deviation in judgment scores given in the two questionnaires for all Category 2 questions
on nuclear power, if any. As intentionally designed into the ques- (p < 0.001) (see Table 2). On whether nuclear power could enhance
tionnaires, the randomness could be analysed with data from the electricity stability, lower the electricity tariff, operate safely,
Category 1 and Category 3 questions (as shown in Table 2). The mitigate pollution, and ease climate change (POWER_CUT, TARIFF,
Category 1 questions capture each respondent's impression SAFETY, POLLUTION, and CLIMATE in Table 2), the mean scores of
regarding nuclear power before receiving the one-sided informa- the pro-nuclear questionnaire were 0.87e1.99 lower: tilted to-
tion, and any difference between the respondents is independent wards the side of positive impact, than those of the anti-nuclear
from the current study. The results from these questions should not questionnaire. The general attitude towards nuclear power was
be statistically significantly different between the two groups of not tilted to either the pro- or anti-nuclear side before receiving the
respondents who answered the different questionnaires, given the one-sided information (IMPRESSION), with an average score of 5.34
same population for sampling and the sampling method. Any re- under a scale of 0e10, in which 0 ¼ positive, 5 ¼ neutral, and
sults that yield a statistically significant difference in this part have 10 ¼ negative. After receiving the one-sided information, the av-
to be discarded because this intergroup difference would add noise erages (JUDGE) became 4.57 and 5.53 for the pro-nuclear and anti-
to the final analysis. nuclear questionnaires, respectively: tilted towards the two sides,
A Mann-Whitney U test was conducted to evaluate the differ- respectively. The information also generated statistically signifi-
ences between the two datasets. This test often outperforms the t- cantly different opinions on revising the share of nuclear power in
test in analysing ordinal-level scale data (Nanna and Sawilowsky, Hong Kong's fuel mix (MIX_NUCLEAR in Table 2). On average, a
1998). In addition, t-tests and ANOVA cannot be used because slight increase in the share of nuclear power in the energy mix from
normal distribution was not observed in the data (Norman, 2010). the current 23%e24.2% was given in the pro-nuclear questionnaire,
While the Kolmogorov-Smirnov test (KeS test) tests the distribu- whereas a decrease to 17.5% was given in the anti-nuclear
tional difference between the groups (Dytham, 2011), it is not questionnaire.
recommended for samples that consist of a large number of ties or Digging deeper into the individual responses, three types of
more categorical data, which was the case for this study, and this respondents can be perceived: (i) firm opinion holders who
was not utilised. The results showed that there was no significant showed a more pro-nuclear stance despite receiving anti-nuclear
difference concerning the demographic factors, but a significant information, and vice versa; (ii) the inert who showed no change
difference in some of the Category 1 questions. in preference after receiving the information; and (iii) the opinion
Based on the results of the Mann-Whitney U test, the data in 75 switchers. The opinion switcher is the focus of the current study
questionnaires collected during the first day of the survey were showing how much public opinion one-sided information can in-
excluded due to the intergroup difference in concern about nuclear fluence, whereas the inert represent the inability of the one-sided
power between the respondents answering the two question- information to influence public opinion. The firm opinion holders
naires. On that day, incidentally, the respondents who answered represent an interesting phenomenon that one-sided information
the pro-nuclear questionnaire gave a significantly lower impor- would lead to deeper entrenchment of their existing perceptions of
tance to the issues of electricity tariff and safety in considering the the issue. There were some restrictions in analysing them in detail
usage in nuclear power compared to those who answered the anti- in the current study due to the limitation from the questionnaire
nuclear questionnaire. The exact cause of this before-information- design. The change in opinion can be measured for the general
reception intergroup difference remains unknown. This was attitude only, not the five specific issues and preferred energy mix;
possibly a sampling bias due to the limited number of samples this is because no base-line opinion was asked for. Furthermore, the
collected within a day, although such a difference was not found in proportion of opinion switchers cannot be explicitly calculated due
the remaining samples, both in total and by date. Nevertheless, this to the lack of a control group. An alternative approach was utilised
biased before-test impression had to be eliminated from the in the analysis. The percentage difference between the proportion
dataset to ensure that any difference in public opinion after of respondents having a pro-nuclear stance/preferring an increase
receiving the one-side information was caused by the information in nuclear share in the pro-nuclear questionnaire and that having a
itself. The final number of valid responses was 459, in which 236 pro-nuclear stance in the anti-nuclear questionnaire was consid-
received pro-nuclear information and 223 received anti-nuclear ered as the proportion of opinion switchers which the pro-nuclear
information. information can induce. As summarised in Fig. 1, the proportion
varied for different questions.
3. Result on the proportion of opinion switchers
3.2. The opinion switcher
3.1. Influence on general public opinion
For the change in general attitude, the attitude of the re-
After the randomness check, neither the subjective opinions spondents before and after receiving the one-sided information
(IMPRESSION (p > 0.05), CON_STABILITY (p > 0.5), CON_TARIFF was examined, looking at the percentage difference between the
(p > 0.05), CON_GHG (p > 0.5), CON_SAFETY (p > 0.05), and two questionnaires of the respondents who changed their
CON_POLLUTION (p > 0.5) in Table 2) nor the objective de- impression towards nuclear power at the end of the survey. The
mographic variables (GENDER, AGE, EDUCATION, DISTRICT, and one-sided information was considered to be inducing the change as
T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57 53
long as the respondent's impression changed at the end of the denoted neutrality; and 7e10 denoted a negative opinion towards
survey, regardless of whether the original stand was pro-nuclear or that certain aspect of nuclear power. Opinions of the respondents
anti-nuclear. on the five aspects of nuclear power were sought with one-sided
There was a statistically significant difference in the general information given: risk of power cut; electricity tariff; safety;
attitude change between the two groups after receiving the one- pollution; and impact on climate change. The percentage of opinion
sided information, whereas the degree of impact varied. switchers varied significantly between different aspects from 5% to
Regarding a ± 1 point from the original attitude as no change, to 37% (as shown in Fig. 1). The pro-nuclear information induced
reduce the rate of error, the percentages of respondents having an 5e23% more respondents to give a positive opinion, whereas the
opinion echo with the one-sided information were 32% and 22% in anti-nuclear information induced 12e37% more respondents to
the pro-nuclear and anti-nuclear questionnaire groups, respec- give a negative opinion. In all aspects except the risk of power cut,
tively, whereas the opposite groups: respondents who favoured the proportion of respondents induced by the anti-nuclear infor-
nuclear more after receiving anti-nuclear information and vice mation to change to the negative side was greater than that
versa, were 20% and 14%, respectively (as shown in Fig. 1). This induced by the pro-nuclear information to change to the positive
suggests that the impact of the pro-nuclear questionnaire was more side. The percentage of opinion switchers was the smallest for the
significant, possibly due to the more negative impression towards safety issue: 5% and 17% were influenced by pro-nuclear and anti-
nuclear power in general in the first place. Upon calculation, 12% of nuclear information, respectively. The percentages for the risk of
respondents were moved by pro-nuclear information to support power cut were 18% and 12%, respectively, whereas those for
nuclear power, whereas 8% of respondents were moved by anti- electricity tariff were 13% and 20%, respectively. The percentages for
nuclear information to oppose nuclear power. Therefore, the the pollution problem were 22% and 34%, respectively, whereas
range of the opinion switcher percentage was 8e12%. In compari- those for the issue of climate change were 23% and 37%, respec-
son, the percentages of respondents having no change in attitude tively. This suggests that negative information would be more
were 53% and 57% in the pro- and anti-nuclear questionnaires, influential upon the impression of people on nuclear power.
respectively.
For the issue of energy fix preference, the cutting point was at 3.3. The firm opinion holder
the current percentage of nuclear power in the energy mix of Hong
Kong, 23% (Environmental Bureau, 2014), which was provided to all A significant proportion of respondents maintained a stance
the respondents. Regarding a 2% difference in preference from the that was opposite to what the one-sided information implied. Some
current 23% share as no change, the percentage of respondents respondents became more pro-nuclear, in terms of attitude or en-
preferring more nuclear, no change in nuclear, and less nuclear ergy mix preference, despite receiving anti-nuclear information,
were 39.0%, 18.2%, and 42.8%, respectively, in the pro-nuclear and vice versa. For simplicity, the two groups are regarded as
questionnaire, and 23.8%, 9.9%, and 66.4%, respectively, in the “firmly pro-nuclear” and “firmly anti-nuclear” hereafter. For gen-
anti-nuclear questionnaire (as shown in Fig. 1). 15% of respondents eral attitude change, 20% of the respondents were firmly pro-
were moved by pro-nuclear information to support nuclear power, nuclear and 14% were firmly anti-nuclear. For energy mix prefer-
whereas 24% of respondents were moved by anti-nuclear infor- ence, 24% and 43% were firmly pro-nuclear and anti-nuclear,
mation to oppose nuclear power. Therefore, the range of opinion respectively. For the five specific issues, the percentages of re-
switcher percentage was 15e24%, whereas the range of opinion spondents who were firmly anti-nuclear were more than those
inertia was 10e18%. who were firmly pro-nuclear in perception towards the safety,
For different issues of nuclear power, respondent opinion was pollution, and climate change issues of nuclear power, using the
given in a scale of 0e10, and positive, negative, and neutral readings methods stated in the opinion switcher section. For the safety,
were defined as follows: 0e3 denoted a positive opinion; 4e6 pollution, and climate impact issues, the firmly anti-nuclear
54 T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57
percentages were 37%, 28%, and 17%, respectively, whereas the level of concern of the five issues, namely, reliability, electricity
firmly pro-nuclear percentages were 18%, 13%, and 16%, respec- tariff, safety, pollution, and climate issues were 7.76, 7.25, 9.08, 8.48,
tively. The opposite result was revealed for reliability and electricity and 7.71, respectively. While the pairing for the safety issue
tariff. The firmly pro-nuclear percentages were 35% and 21%, matched the prediction: highest concern and highest rigidity in
respectively, whereas the firmly anti-nuclear percentages were 9% opinion formation, the correlation was generally not as great as
and 17%, respectively (as shown in Fig. 1). expected. The correlation between the level of concern and the pro-
These results have some interesting implications. Theoretically, nuclear and anti-nuclear opinion switchers were 0.32 and 0.42,
the one-sided information should induce the respondents to give respectively. The data indicate that public opinion on an issue of
an opinion that is inclined to what the information provokes: the very high concern would be relatively rigid and would not be
respondent should give a more pro-nuclear stance after receiving influenced by a brief twist of the information provided, but the
pro-nuclear information. In reality, however, providing only one general trend is less obvious.
side of the story has led to exactly the opposite result, namely, the The impact of one-sided information on impression changing is
strengthening of some of the respondents’ existing perceptions of mixed when the issue is analysed in different ways. For the energy
the issue. This is illustrated with the result of general attitude mix preference, 43% preferred less nuclear power in the pro-
change, namely, that a clear shift in attitude to the opposite side nuclear questionnaire: the firmly anti-nuclear proportion,
from what the one-sided information should provoke for some whereas only 24% preferred more nuclear in the anti-nuclear
respondents was observed. questionnaire: the firmly pro-nuclear proportion. In contrast, the
The mechanism behind this could possibly be explained by the percentage of respondents who were firmly pro-nuclear was larger
personal perception of the information, with reference to previous than that of those who were firmly anti-nuclear: 20% and 14%,
knowledge, identity affiliation, and even scepticism about the respectively, when inspecting the change in general attitude after
intention of the information provision. It is possible that the in- receiving the one-sided information. We found no correlation be-
formation arouses alertness of previous knowledge that is in tween the two factors: a change in attitude did not guarantee a
contrast to the given information, and that the new information change in preference of the actual usage of the energy, and vice
might also conflict with the respondent's own value judgment. In versa. The results from firm supporters, opinion switchers and
these cases, the respondent would decide to discard the new in- general impression change altogether suggest that the impact of
formation. It is also possible that the provision of consistently lop- the one-sided information on the general impression would be
sided information would lead to scepticism from the respondent on mixed.
the purpose of the study, and they would intentionally respond in To conclude, the provision of one-sided information appears to
the opposite way. This has been reported during the data collection systematically influence public opinion on the issue. Depending on
process, that the respondent explicitly questioned the purpose of the degree of the public's concern about the issue, the proportion of
the study, and thus the record had to be discarded. Further studies opinion switching also varies. The relationship between the degree
should be conducted to explore these issues. of concern and the ability to influence public opinion through
The results for the five specific issues may not have the same manipulated information is generally negative. Public opinion can
implications since they were not derived from a beforeeafter vary greatly on issues of little concern yet information has only a
comparison. It is possible that the one-sided information provided marginal impact for issues of high concern. In other words, the
was simply insufficient to provoke the respondent to give a significance of one-sided information in influencing public opinion
different opinion from their original perception. Instead, the results is limited. It is even possible for the provision of one-sided infor-
here suggest the significance of each of the issues on influencing mation to induce deeper entrenchment of the public in terms of
general public attitude towards nuclear power, that safety, pollu- their own perception, thus widening the split between the two
tion, and climate issues are major concerns which prevent the views. Nevertheless, information could still play a critical role when
public from supporting nuclear power, whereas the strengths of the two views are evenly matched in strength and when the fate of
nuclear power in terms of reliability and electricity tariffs are much the decision rests with the opinion switchers.
appreciated by the public.
4. Explaining the proportion of opinion switchers
3.4. Discussion
Public opinion is typically given and captured in opinion polling
Taking both the firm opinion holders and the opinion switchers as a single output. What lies between the information input and the
into account, the negative image of nuclear power over the safety opinion output is a black box in which the individual assigns weight
issue is somewhat rigid and it is unlikely to change with a brief to the importance of information about different aspects of the
twist in the information provided. In contrast, the positive image of issue. Discovering the critical factors that would influence the
nuclear power in providing a reliable electricity supply is also personal impressions about nuclear power would therefore be
strong, but is less unlikely to change should the other side of the important to policy-makers and interested parties who want to
story be told compared to the safety issue. While a significant understand and possibly shape public opinion. Multiple regressions
proportion of people held a negative view towards the impact of were conducted to examine the degree of impact from different
nuclear power on climate change and pollution, it is more likely to factors and aspects of nuclear power on opinion formation and
see a substantial proportion of people change their side should impressions of nuclear power. Three factors were set as the
different information be given, and this could twist the proportions dependent variables indicating the general impression towards
of supporter and opponent. The opinion on electricity tariff would nuclear power: the personal impression after receiving the one-
be unpredictable, since nearly half of the respondents held a sided information (JUDGE); the degree of change in general atti-
neutral perspective and neither side would dominate even when tude caused by the one-sided information (CHANGE_ATTITUDE);
one-sided information is presented; therefore, policy justification and the proposed share of nuclear power in the energy mix
would have to rely on influencing the neutral group. (MIX_NUCLEAR). The first two items directly measure the correla-
The concern about the individual aspect could be an effective tion between each aspect of nuclear power and the general
predictor of the rigidity of respondents’ opinion on the particular impression, in terms of absolute degree and the change along the
aspect. From the results of the Category 1 questions, the average impression spectrum compared to the original stand, respectively.
T.F. Kwok et al. / Utilities Policy 46 (2017) 48e57 55
The last one measures the inclination of the respondent to reduce, more significant in influencing the preferred share of nuclear po-
increase, or make no change in nuclear power usage, with reference wer in the energy mix; for every increased point in concern, the
to the current energy mix of nuclear power in the local electricity share of nuclear power is increased by 0.825%, with p < 0.01, which
supply. Demographic factors were also added into the regression is larger than that of perception of impact (p < 0.05). This suggests
models to examine their importance. Concerns about the different that a low electricity tariff with the use of nuclear power is
aspects of nuclear power were also added to isolate the interper- perceived, and thus those who prefer a lower electricity tariff prefer
sonal difference of perception of the issues. The results are shown more nuclear power.
in Table 3, in which the original model showing all factors (Model 1) The safety issue is revealed to be the most influential factor
and the final model after removing the insignificant factors in affecting the general impression of nuclear power (p < 0.01), but
backward operation (Model 2) are presented. not for the impression change in which the original impression
In all three measurements, perception of the risk of power cut (IMPRESSION) is a determining factor, as expected. For every point
with the utilisation of nuclear power is revealed to have no sig- increase in considering nuclear power as dangerous, personal
nificant influence on the general impression of nuclear power, judgment is 0.271 points more negative, impression change is
regardless of the type of one-sided information received. In 0.230 points more towards the negative side, and the preferred
contrast, concern about the stability of the electricity supply ap- share of nuclear power in the energy mix is 1.564% lower. For the
pears to significantly influence the impression of nuclear power in pro-nuclear questionnaire only, an increase in the concern about
the measurements of judgment and attitude change. For every the safety issue was associated with more nuclear power in the
point increased in the concern about stability, the judgment on preferred energy mix; people who are more concerned about the
nuclear power is 0.097 points more positive, with a 0.104 points safety issue actually prefer more nuclear power, by 0.784% for every
change in general attitude towards the positive side (p < 0.01). This point increase in the concern about safety (p < 0.1). Because people
supports the claim made in Section 3.3 that the positive impression receiving pro-nuclear information preferred more nuclear power at
of the stability of nuclear power is strong. This impression is the end of the survey, the impact was projected onto the personal
conveyed into the overall judgment about nuclear power and it is valuation in the regression model.
unlikely to be influenced by one-sided information. However, this Among the five aspects of nuclear power, the problem of
concern for stability does not significantly influence the preferred pollution is statistically significantly, it having a moderate impact
energy mix. on the general impression of nuclear power. For every point in-
The impact of the electricity tariff is revealed to have a statis- crease in perceiving nuclear power as polluting, personal judgment
tically significant influence on the general impression of nuclear is 0.115 points more negative (p < 0.01), impression change is 0.145
power, but with a lower significance level and a smaller degree of points more negative (p < 0.01), and the preferred share of nuclear
impact. For every point increase in perceiving an increase in elec- power in the energy mix is 0.472% lower, at a lower significant level
tricity tariff with the use of nuclear power, personal judgment is (p < 0.1). The concern about the pollution issue is more significant
0.065 points more negative, impression change is 0.077 points in influencing the preferred energy mix. For every point increase in
more towards the negative side, and the preferred share of nuclear concern, the preferred share of nuclear power in the energy mix is
power in the energy mix is 0.58% lower. Compared to the percep- 0.892% lower (p < 0.01), thereby suggesting that nuclear power is
tion of the impact, the initial concern about the electricity tariff is perceived to be polluting, before receiving the one-sided
Table 3
Influence of individual factors on public opinion on nuclear power.