Framing cyber security: Do fear appeals and frames interact to improve online safety of elderly British people?
Kate E. Symons (11789166) The University of Amsterdam
Author Note
Supervisor: Dr Fam Te Poel MSc Communication Science: Persuasion
Words: 8,136
Abstract
This study examined the effects of fear appeals and message frames on the online safety intentions and behaviours of 260 British elderly people, age range in years, 64 to 81 (M = 68.64, SD = 3.52). A between-subjects experiment explored how fear appeals (neutral vs. fear appeal) and message framing (loss vs. gain) might interact to create a stronger persuasive effect. Results showed no effect of fear appeals or type of framing on intentions. However, loss frames and not gain frames were more effective for reducing risk of unsafe online behaviour. Fear appeal and framing did not interact. The study provides new empirical evidence for testing fear appeals and framing on an elderly target group in the context of helping to promote better online safety. If correctly designed, fear appeals are not less effective than neutral messages among the elderly. Furthermore, loss frames are recommended for helping to prevent less risky online behaviour among this target group.
Key words: Fear appeals, Framing theory, Elderly people, Cyber security, Socio-emotional selectivity theory
Framing cyber security: Do fear appeals and frames interact to improve online safety of elderly British people?
Internet-based technologies help us conduct almost every aspect of our daily lives. Internet company Cisco predicted that by 2020 up to 40 billion devices could be web-enabled (Friedman & Singer, 2014). While this brings people extraordinary benefits, there are also huge costs of human reliance on web-based applications.
Cybercrime is now one of the top five global risks in terms of likelihood (World Economic Forum, 2019). It can potentially affect all web-based devices. Cyberattacks reportedly cost the UK government up to £27 billion each year (UK Cabinet Office, 2011). Both organisations and
individuals face different threats including but not limited to: spyware, malicious software, DDoS attacks, and social engineering such as phishing and smishing. Friedman and Singer (2014) state, ‘a cyber problem only becomes a cybersecurity issue if an adversary seeks to gain something from the activity,’ (page 34). Confidentiality, integrity, accessibility and resilience of systems and devices can be compromised (Friedman & Singer, 2014).
Technology and online threats create new challenges for the British police, including resourcing. In an interview with the author of this current study, a United Kingdom (UK Police and Crime Commissioner said: “The amount of actual police intervention once a cybercrime has happened is limited. We need to prevent people being exploited.”
A more specific societal concern relates to the risks that older people face when using web-based technology (computers, phones and other internet devices). Grimes, Hough, Mazur and Signorella (2010) found that older people can be less knowledgeable about internet security than younger adults. Grimes et al., (2010) find differences in education explains why some senior citizens are more at risk because more educated people had more knowledge about internet threats (page 188). James, Boyle, and Bennett (2013) found that susceptibility and age are positively associated, and other factors such as decreased psychological well-being and lower literacy lead to greater risk of
financial scams and internet fraud. Australian law enforcement agencies furthermore believe that older people are appealing targets for potential identity fraudsters (Cross, 2017); perhaps because some pensioners have accumulated wealth and assets (FBI, 2018).
When it comes to technology, older people are a heterogeneous group with different abilities and preferences (Renaud & Biljon, 2010) but cognitive, affective, social and neural changes due to ageing can make older people more vulnerable (Spreng, Karlawish, & Marson, 2016). The UK also has an ageing population which provides another reason to investigate the effects of preventative communication in among older target group; more people over 60 years old live in the UK than those under 16 (General Register Office for Scotland 2002, quoted in Lee & Coughlin, 2015).
Cyber security academics have furthermore highlighted a pressing need for more research into how people can be better motivated to become cyber secure. Masthof, Collinson and Vargheese (2018) state in the introduction to the Symposium on Digital Behaviour Interventions for Cyber Security 2018, that more research needs to be properly rooted in behaviour change. This echoes Anderson & Agarwal (2010) who provide a chronological summary of some personal and organisational behavioural security literature. They assert: ‘No single stream of literature completely frames the home computer user security behaviour phenomenon’ (page 261).
Academic evidence moreover shows that older adults process persuasive messages differently to younger (Carstensen, Fung, & Charles, 2003; Chowdhury, Sharot, Wolfe, Düzel, & Dolan, 2014; Rösler et al., 2005; Van der Goot, Van Reijmersdal, & Kleemans, 2015). Furthermore, communicating the serious nature of threats and motivating people to adopt recommended safeguards can be difficult for policy makers (Shillair et al., 2015). This study attempts to address these issues by focussing on fear appeals and message framing; two examples of threat-based behaviour change communication.
Shen & Dillard (2014) offer the following definition of fear appeals: ‘Fear appeals are typically conceived as messages intended to frighten recipients into compliance,’ (page 96). The
threat component of a fear appeal highlights the severity of an issue (or behaviour) by describing ‘negative and undesirable’ consequences. Some academics argue that fear appeals are ineffective
(Peters, Ruiter & Kok, 2014) because ‘backfire effects’ can occur resulting in avoidant reactions (Witte, 1994). Although the use of ‘scare tactics’ divides academic opinion, plenty of evidence shows that when used correctly, fear appeals with strong self-efficacy statements, can
effectively change attitudes, intention and behaviour (Tannenbaum et al., 2015; Witte & Allen, 2000). Fear-based communication is usually presented as a loss frame because the focus is on negative outcomes (Jensen et al., 2018). This means that message content refers to bad things that will happen and good things that will not happen (Rothman, Bartels, Wlaschin, & Salovey, 2006). Conversely, gain frames set out the good things that will happen and the bad things that will not happen when the preventive behaviour is executed (Rothman et al., 2006).
The persuasive effects of loss and gain frames is known as ‘framing theory’ which, like fear appeals, is also a heavily researched and hotly debated area of communication science. Meta-analytic evidence from framing research, which uses self-report and brain scans, suggest that gain frames can be more successful than loss frames when persuading people to adopt health prevention behaviours (Covey, 2014; Falk & Scholz, 2018; Gallagher & Updegraff, 2012). With a considerable amount of research looking at how fear appeals and framing work separately in preventative communication, academics are now making the case for studying the combined effects (Jensen et al., 2018). By examining them together in the context of cyber security, it is possible to understand how these two persuasive mechanisms might work to reduce risk of online crime.
As Jensen et al. (2018) ask, could gain frames used within a fear appeal — instead of the usual loss-framed information — improve message effects by enhancing the self-efficacy element of a fear appeal (Jensen et al., 2018)? Few studies look at the interaction between fear and framing in the context of online safety. This study seeks to investigate these questions: RQ 1) To what extent do fear appeals and message framing affect preventative intentions and behaviour
to improve online safety among British people aged over 60? RQ2) Is the effect of fear appeals on intentions and behaviour affected by type of message framing?
Theoretical Framework
Fear Appeals
Over the decades of fear appeal research, numerous studies show evidence of a
‘boomerang effect’, an undesirable consequence of using scary communication. This claim is drawn from findings in health prevention literature backed up by a meta-analytic reviews (Kok, Peters, Kessells, Ten Hoor, & Ruiter, 2018; Kok, 2016; Ruiter, Kessels, Peters & Kok, 2014). Within health prevention and cyber security literature, two theoretical models are widely used to study how people respond to fear appeals, these are Protection Motivation Theory (Rogers, 1975) and the Extended Parallel Process Model (Witte, 1992). Both examine how threat, or fear-inducing
communications can urge people to avoid or take recommended action (Bartholomew et al., 2011). PMT focuses on two cognitive reactions, ‘danger control responses’, which are set off when a person sees a threatening or fear-inducing message. These are an evaluation of the threat (severity and susceptibility) and the ability to deal with that threat (self- and response-efficacy, a person’s perceived competence to perform an advised response and the apparent effectiveness of the suggested response). Together, these appraisals form a person’s ‘protection motivation’ that underpins intention to adopt advice. The model also incorporates potential rewards and reinforcement, as well as potential costs of punishment leading to dysfunctional reactions (Bartholomew et al., 2011; Rogers, 1975).
Within cyber security literature, of the 17 organisational studies reviewed by Anderson and Agarwal (2010), four use PMT as a theoretical framework and only one of these (Johnston & Warkentin, 2010) incorporates a deeper look at how fear appeals improve intentions and security behaviour. For the latter study, results indicate that fear appeals did influence the intentions of employees and students at a large university. Of the five personal computing
studies listed, only two are related to PMT and persuasion. One of these studies was based on a student sample and found that personal responsibility, self-efficacy and response- efficacy relate most strongly to intentions to engage in less risky online behaviour (LaRose, Rifon, & Enbody, 2008).
In a paper examining and explaining why cyber security campaigns fail to change
behaviour, Bada, Sasse, and Nurse, (2015) boldly state, ‘invoking fear in people is not an effective tactic, since it could scare people who can least afford to take risks,’ (page 9). This statement stems from Witte’s extensively tested EPPM model which better explains how avoidant reactions might occur than the PMT model (Witte, 1994).
Expanding on why fear appeals sometimes fail, Witte (1994) makes the point that prior fear appeal theories had ignored how people emotionally process or control fear when an emotional response might be stronger than a cognitive response (‘danger control’). If ‘danger control’ processes are weak, ‘fear control’ will take over where a person’s focus is on internal concern. When this happens, a reaction to the negative, fearful emotion will occur. Witte (1994) explains that conscious or unconscious fear control processes can take different forms
including, inhibiting thoughts of the danger (‘defensive avoidance’), and resisting the communication (‘message minimization’).
The Bada et al. (2015) statement previously referred to — that fear appeals are not
effective — contradicts extensive meta-analytic evidence. A key statement of Tannenbaum et al. (2015) is that there are not many situations in which fear appeals do not work. They go on in their paper to distinguish between two ‘efficacy statement hypothesis’ which are important for understanding the success of fear appeals (Tannenbaum et al., 2015). The ‘strong efficacy’ hypothesis outlines that without efficacy statements fear appeals will backfire. The ‘weak efficacy’ hypothesis states that without efficacy statements, fear appeals lead to weaker results compared with those that contain efficacy statements (Tannenbaum et al., 2015). A common meta-analytic finding is that fear appeals do effectively motivate behaviour change when they
contain strong self-efficacy statements (Tannenbaum et al., 2015; Witte & Allen, 2000). This is further supported by Kok (2016); in a meta-analysis only including experiments that met strict experimental criteria, participants changed their behaviour as recommended but this happened only when efficacy was high. If self-efficacy is weak in a strong fear appeal, behaviour change can be wrongly directed (Witte & Allen, 2000; Kok, 2016). Efficacy statements are of central importance to effective fear appeals because they provide recipients with reassurance
(Tannenbaum et al., 2015).
Within the cyber security field, a 2018 Dutch internet security study of 1,219 participants uses PMT in the context of fear appeal interventions to reduce the threat of Phishing attacks in a strong fear appeal, weak fear appeal and control condition. A key finding includes that self-efficacy and fear were the most influential factors that predicted protection motivation (Jansen & Van Schaik, 2018).
A person’s mental assessment of fearful emotions take place regularly (Witte, 1994). It is this cognitive assessment of the perceived threat and consequent emotion that causes
message acceptance or rejection. Although more fear can lead to better message acceptance, the EPPM shows that fear does not directly cause a message to be accepted or rejected, the nature of the mental appraisal does (Witte, 1994). Academic evidence shows that ageing processes can create an ‘optimism bias’, which could potentially reduce the likelihood of an older target group updating their beliefs with pessimistic emotional information. This could mean that during cognitive appraisal, when processing threat messages, an older target group might be less receptive to information when it is presented as a scary message than if it were more
neutral or positive. If so, should fear be a less frequently used emotion for communicating threats to older people? Findings from an established research stream within the ageing literature can help to answer this question.
Socioemotional selectivity theory. Sparks & Ledgerwood (2018) explain that Socioemotional Selectivity Theory (SST) ‘describes how motivational priorities change across the life span’ (page
3). The theory shows how tendencies and perceptions alter which can lead to older people preferring optimistic emotional over negatively-valenced messages. Using SST as a theoretical framework, self-report, neural studies and eye-tracking research, find support for the idea that older people process negatively-valenced emotions and information less favourably because of biases that develop as lifetime span decreases (Carstensen, Fung, & Charles, 2003; Chowdhury, Sharot, Wolfe, Düzel, & Dolan, 2014; Rösler et al., 2005; Van der Goot, Van Reijmersdal, & Kleemans, 2015). One study found that health messages that emphasized less pessimistic emotional goals were more
positively evaluated by older people (Zhang, Fung & Ching, 2009). Chowdhury et al. (2014) employed brain scans to test differences in how healthy older adults and younger people update beliefs about 45 emotionally-challenging events such as robbery. Results show that when older adults assess undesirable information, they are less likely to update their beliefs about the likelihood of that event happening. This creates an ‘age-related positivity bias’ which could impact how older people make decisions relating to personal issues such as health and financial security (Chowdhury et al., 2014).
A review of the ageing literature reveals a dearth of research looking directly at the effectiveness of fear appeals on older people. One study did not find different psychological responses to fear appeals among older and younger consumers (Benet, Pitts & LaTour, 1993). Another study found that, generally, elderly consumers are not responsive to fear appeals (Bailey, 1984). In their meta-analysis, Tannenbaum et al. (2015) examined 127 fear appeal studies with 248 statistically independent samples with a total of 27,372 participants. Participants ranged in age from eight to 87 years, the mean age was 22.77 years (SD = 9.24) which suggests that the effects of fear appeals on older people in the light of SST is far less researched.This study seeks to provide new information on this topic in the context of cyber security.Based on evidence of greater reluctance among an older audience to update beliefs with emotionally negative information (outlined by SST) this could create greater risk of fear control responses and we
H1 (main effect): Intentions to improve online safety and security among older people will be higher after exposure to a neutral message than strong fear appeal message.
Framing
When it comes to taking risks, losses loom larger than gains in the human mind (Tversky & Kahneman, 1981). This is a central premise of Tversky and Kahneman’s Prospect Theory (PT) which tells us that people make non-rational choices about risk (Tversky & Kahneman, 1981).Framing theory as set out by Rothman and Salovey (1997) argues that predictions about the effect of gain- and loss- frame messages on health behaviour could be based on the conceptual framework that
underpins PT. When applied to health promotion, this means that detection behaviours (e.g., breast examination, HIV test) involve the risk of an unpleasant outcome and loss-framed appeals should be more persuasive. Conversely, low-risk prevention behaviours are more persuasively promoted with gain frame information.
Meta-analytic evidence from studies that use self-reports and brain scans suggest that, generally, gain frames outperform loss frames when persuading people to adopt health prevention behaviours (Gallagher & Updegraff, 2012; Covey, 2014; Falk & Scholz, 2018). There is a lack of cyber-crime prevention studies looking at the effects of message framing in an older group of home computer users. Using a student population, Rosoff, Cui, and John (2013) found that gain frames increased respondents’ support for a safer cyber response option when dealing with a near miss scenario. This provides evidence that gain frame messages will work better in the field of cybercrime prevention.
In a review of neural evidence and traditional message research, Falk and Scholz (2018) state that people are less persuaded by facts and more by subjective value. If Carstensen’s previously mentioned socio-emotional selectivity theory applies, and older people value positive information more than negative, they are less likely to be persuaded by messages that are negatively framed (loss frames). Research conducted by Jayanti (2010) showed that ‘accentuating the positive’ is a
much more effective strategy when trying to engage older adults with health prevention
literature. Gain frames were more persuasive. Jayanti et al. (2010) also conclude that cognitive impairments, a preference for heuristic processing and a tendency to avoid negative emotions are possible factors contributing to their results. Based on the evidence reviewed, we arrive at the following hypothesis:
H2 (main effect): Intentions to improve online safety and security among older people will be higher after exposure to a gain frame message vs loss frame message.
Intentions: An Interaction Between Fear Appeals and Framing
Immense academic effort has gone into researching fear and framing separately. Due to a ‘large number of null findings’, more studies are needed to understand the effect of combining these techniques in what Jensen et al. (2018) call a new breed of ‘second generation loss/gain-framing research’ (page 246).
Traditionally within the field of health prevention research, academics have stated that a loss frame is different to a fear-arousing message as framing will not necessarily lead to an emotional reaction (Bartholomew et al., 2011). But Jensen et al. (2018) challenge this by asking: ‘Is it possible that loss-framed messages are a form of fear appeal as they often focus on susceptibility to
undesirable outcomes?’ (page 246). They cite evidence that supports this claim. More interestingly
for the purposes of this current study, they also ask: ‘If loss frames are threat appeals, then does it make sense to conceptualize gain frames as efficacy appeals?’ (page 256).
Because of academic thinking about how EPPM could possibly provide a framework for studying loss/gain-frames, instead of the traditional theoretical framework Prospect Theory (Jensen et al. 2018), this current study will look at loss/gain-framing as a moderator on the relationship between a fear appeal and the dependent variables, intentions and behaviour.
Two prevention studies have demonstrated an interaction effect between fear appeals and framing. Gerend, Maner, and Kazak, (2011) showed that student participants in a fear condition reported eating more servings of fruits and vegetables after exposure to a loss-framed message. Zimmerman, Cupp, Abadi, Donohew and Grossl (2014) studied the interaction between framing (loss vs. gain) and fear appeals (high vs. low threat) in a study about anti-marijuana public service
announcements. They found medium to large effect sizes and significant interaction effects for both threat and framing on message effectiveness and fear response. Although they fall under the
prevention umbrella, behaviours in these studies are very different to promoting online safety and new research needs to test interactions between fear and framing in the context of cyber security.
To recap, the key difference between loss/gain-framing and fear appeals is that the latter are intended to provoke an emotional response and it is the dose of fear which can (if the fear appeal is correctly designed) stimulate a cognitive danger control response which might lead to behaviour change (Witte, 1994). Academics continue to make a strong case against the use of fear appeals (Kok et al., 2018) but recent studies continue to find an optimistic role for fear in threat communication. Nabi, Gustafson, Risa & Jensen (2018) show that scary messages can stimulate climate change prevention intentions. This happened when a threat message was followed by information with gain-framed climate change efficacy statements which elicit more hope than equivalent loss-frame information.
Building on the previously mentioned importance of strengthening self-efficacy and response-efficacy in fear appeals, Jensen et al. (2018) explore some ideas about why gain frames might
enhance efficacy. Firstly, gain frames highlight the benefits of performing recommending actions, in so doing they promote the usefulness of an action (Jensen et al. 2018). Gallagher and Updegraff (2012) consider that gain framed messages transmit information that we know influences behaviour such as self-efficacy or positive emotions. Jensen et al. (2018) recognise that gain frames might be able to motivate audiences and also mention that optimistic messages can increase attention and motivation, particularly for older populations.
According to a cybersecurity study by Larose, Rifon, & Enbody (2008), ‘Efficacy has a direct
impact on safe behaviour, but also interacts with risk perceptions. Fear is most likely to work if the threat information is coupled with information about how to cope with them, since the coping information raises self-efficacy,’ (page 72). It follows that gain-framed prevention messages with strong self- and response-efficacy statements within a fear appeal will work to improve a person’s risk perception and ability to cope. Because gain frames present positive consequences of performing an action, when based within a fear appeal with self- and response-efficacy elements, this could increase the chance of triggering a cognitive response rather than message avoidance and fear control. Therefore, gain-framed information contained within a fear appeal with strong self- and response-efficacy elements could potentially make elderly message recipients more likely to accept rather than reject a prevention message promoting online safety.
We have established that an elderly target group are less likely to respond to loss frame messages because, in line with socioemotional selectivity theory, negativity bias weakens as age increases (Sparks & Ledgerwood, 2018). Given that fear can play a motivating role in behaviour change, SST poses interesting questions about how an older target group might respond when gain-framed information reinforcing benefits of preventative action is set within a fear appeal. With evidence that older people respond positively to gain frames (Jayanti et al. 2010), combined with research that highlights the effectiveness of fear appeals (Roberto, Mongeau & Liu, 2018; Borland, 2018; Witte & Allen, 2000), added to the fact that fear appeals need further testing among older people, we can hypothesise:
H3 (interaction effect): Fear level and framing type will interact such that exposure to a strong fear appeal has a positive effect on intention to improve online safety, but only when this fear appeal is gain framed.
Results will contribute evidence to the few studies that do combine fear and framing and prevention intentions (and behaviours) in older people.
Behaviour: Fear and Framing Interaction
Leading fear and framing researchers have made several strong criticisms that not enough studies test behaviour (Peters, Ruiter, & Kok, 2013; Updegraff & Rothman, 2013). This current study attempts to address this by presenting participants with a task related to cyber security risk which was adapted from the study by Rosoff, Cui, & John (2013). The task assessed how likely people are to choose a less risky or more risky option based on provided information. When the prevention information is framed differently, it is interesting to examine how the fear appeal vs neutral condition make participants assess the cyber risk. We arrive at the final
hypothesis:
H4: Fear level and framing type will interact such that exposure to a strong fear appeal has a positive effect on developing less risky online behaviour among older people, but only when this fear appeal is gain framed.
Methodology
Participants and Design
A sample of 268 participants completed the survey- embedded experiment between 15 to 22 May, 2019. Participants were drawn from British research panel, Prolific (Prolific;
https://prolific.ac/). An analysis of crowdsourced participant panels states that Prolific is an online participant panel for ‘sound scientific research’ (Palan & Schitter, 2018). A 2 x (fear vs neutral fear), 2 x frame (loss vs gain) between-subject design was used for testing the persuasive effect on
intentions and behaviours. Four conditions included, Fear loss, n = 63; Fear gain, n = 65; Neutral loss, n = 64; Neutral gain, n = 68. The approval statement was not completed for three responses and these were discarded. Of the 265 remaining participants, the average completion time was 457.41 seconds (M = 7 minutes, 62 seconds). The suggested completion time was 10 minutes; two participants took less than three minutes and were excluded and three participants took over 24 minutes, more than three times the average completion time. These participants may have left their
computers during completion and responses were removed. Full responses were received from 260 people ranging in age from 64 to 81 (M = 68.64, SD = 3.52). In total, 90 men (34.4%) completed the study. Participants were from all nine geographical regions of the United Kingdom, including
Scotland and Northern Ireland. In education, 22.1% (n = 58) had completed high school, 22.9% (n = 60) had completed further education, 34.0% (n = 89) had completed higher education, 13.4% (n = 35) had completed a Masters, 4.6% (n = 12) had completed a Doctorate, and 2.3% (n =6) marked the ‘Other’ category.
Procedure
Participants first read the ethics and approval statement with a cover note which says the study is about online activities. For randomisation purposes, demographic information was asked first. Participants then saw one of four randomised stimulus materials. To prevent skipping, they could only proceed to the next section after 20 seconds and then were presented with a set of Likert-style self-report items that help predict security intentions (Tsai et al., 2016). A question to help establish behavioural intentions in an online risk scenario was then asked.
Manipulation checks include a measure assessing Fear-related negativity and a question to check that participants recognised the gain vs loss framing. After this, four different Likert questions, each with three to seven items were included as control variables: Self efficacy, Response efficacy, Experience with safety hazards and Existing cyber security habits (Tsai et al., 2016). Finally, participants were debriefed.
Stimulus Material
Shen and Dillard (2014) say that fear appeals include a threat section (outlining severity and susceptibility) and action element (threat solutions, self-efficacy and response-efficacy elements); all these are incorporated in this study’s stimulus materials. Two cyber security professionals were consulted and suggested that password management and URL checking (for preventing Phishing attacks) were suitable to test with home computer users.
The fear appeal contains a real-life personal anecdote involving a female losing £10,000 in a Phishing attack. The money is stolen in Fear loss and nearly stolen in Fear gain. The fear conditions contain threat words such as ‘victim’ and ‘crime’ and softer words such as ‘people affected’ and ‘online incidents’ are used in the neutral stimulus. A quote outlines general information about how Phishing incidents can affect people in neutral, instead of an anecdote. Advice is based on recent guidance from the UK’s Stay Safe Online and National Cyber Security Centre. Text is either loss or gain-framed. See Appendix A for materials.
The Fear loss and Neutral loss conditions were pre-tested with 21 participants (age, M = 72, SD = 3.27. An independent samples t-test assessed Fear-related negativity (see Manipulation checks below), Fear loss (n = 11) and Neutral loss (n = 10). Levene’s test was significant, F = 5.75, p = .027, equal variances are not assumed. Fear Loss (MFearLoss = 4.82, SD = 0.70) is statistically significantly higher than Neutral Loss, (MNeutralLoss = 3.41, SD = 1.04), t (15.60) = -3.61, p = .002), 95% CI [-2.24, -0.58], d = 0.54. We can conclude that the stimulus materials are likely to induce the desired amount of Fear-related negativity.i
Manipulation Checks
Fear appeal. A scale called Fear-related negativity tests how scary the Fear appeal is and contains the following items: ‘scared’, ‘tense’, ‘annoyed’, ‘upset’, ‘uneasy’, ‘anxious’, ‘distressed’ and ‘irritated’ (adapted from Brown & West, 2015). Responses were made on 1–7 point, Likert scale anchored with ‘Strongly disagree’ and ‘Strongly agree’. An eight-item factor analysis was run with an oblique Direct Oblimin rotation. This indicated that the scale had two factors explaining 80.06% of the variance, Factor 1 = 64.54% of the variance; Factor 2 = 15.51%. The Principle Axis Factoring showed a Kaiser-Meyer-Olkin (KMO) measure of .86 and Bartlett’s test of sphericity was
statistically significant (p < .001). The two factors had an Eigenvalue larger than 1, (Factor 1, Eigenvalue = 5.16; Factor 2, Eigenvalue = 1.24) with an obvious point of inflection in the screeplot. Items loading on the second factor included ‘Irritated’ and ‘Annoyed’ are not directly related to being scared and the results analysis proceeds with six items with the following factor loadings: scared,
.866; tense, .883; upset, .676; uneasy, .809; anxious .984; distressed, .787. The Fear-induced negativity scale is internally consistent indicated by α = .94 (M = 3.50 , SD = 1.36).
Framing. To establish that participants saw the loss/gain condition, a manipulation check asks: ‘The information you read is focused on’, then the participant saw one of two options, ‘Money and data lost as a result of online crime’, or ‘Monday and data protected by trying to prevent online crime’.
Measures
Intention to improve online safety. Six items make up this scale (e.g. ‘I am highly likely to take security measures to protect myself on the internet’). Participants rated each item on a seven-point Likert scale, anchored with ‘Strongly disagree’ and ‘Strongly agree’. Higher ratings indicate a stronger intention to act safely. The security intentions scale was originally used in Tsai et. al, (2016). Items about URL checking were added to reflect stimulus material information. A Principle Axis Factoring showed a Kaiser-Meyer-Olkin (KMO) measure of .724 and Bartlett’s test of sphericity was statistically significant (p < .001). Two factors had an Eigenvalue 1; Factor 1, Eigenvalue = 2.87 and explained 47.90% of the variance, Factor 2, Eigenvalue = 1.37 and explained 22.85% of the variance with an obvious point of inflection in the screeplot. Because of this, a two-factor solution is applied to the results analysis. The first factor measures willingness to improve password behaviour; Intentions, passwords consists of four items with the following factor loadings: ‘I am highly likely to take security measures to protect myself on the internet in the future, .347; ‘I will change my
passwords more often’, .902; ‘I will use passwords that are harder to guess’, .741; ‘I will use
passwords that are made up of three random words’, .656. Intentions, password protection proves to be a reliable scale, α = .77 (M = 5.37, SD = 1.00).
The second factor measuring willingness to check URL’s consists of two items with the following factor loadings: ‘When I receive an electronic message (email, text message etc), I will know how to check the URL before responding,’ .724; ‘When I receive an email about a service,
even if it is from a recognised organisaiton, I will check the URL before clicking on links or entering informatino’, .860. The combined scale Intentions, URL is internally consistent, α = .74 (M = 5.98, SD = .88).
Online behaviour. A scenario about online behaviour used by Rosoff et al. (2013) is adapted
in this study to assess if `the stimulus material affects participants’ risk assessment. See Appendix B for full text. Participants then saw a security warning and were asked to ‘Indicate your level of agreement with the following statement: 1 = completely disagree; 7 = completely agree, “You will press ‘run’ and risk installing a plug-in that could be helpful but damage your computer.” Higher scores indicate riskier behaviour (M = 1.89, SD = 1.56).
Control Variables
Existing online habits. Five items on a 7-point scale assessed strength of existing cyber security habits (e.g. ‘Online security protection is something I do automatically’; Tsai et al. 2016). An extra item was added for this study, ‘Someone helps me with online security’ because a younger relative might help an older person perform computer security tasks. A Principal Axis Factor analysis with an oblique Direct Oblimin rotation was run on the six items and showed a Kaiser-Meyer-Olkin (KMO) measure of .87 and Bartlett’s test of sphericity was statistically
significant (p < .001). One factor was extracted which had an Eigenvalue 1(Eigenvalue = 4.11) and explained 68.53% of the variance, with an obvious point of inflection in the screeplot. A reliability analysis revealed that the six items combined to form an internally consistent, α = .76 which would be improved if the item ‘Someone help me with online security’ was dropped. Because of this, it is excluded from the study, leaving five items which combined to form an internally consistent scale of α = .94. A single scale was created (M = 5.76, SD = 1.01).
Self-efficacy. Six items make up the 7-point scale assessing how effectively and easily participants think they can deal with online security (e.g. ‘I feel comfortable taking measures to secure my internet devices’). A higher score indicates a higher level of self-efficacy. This scale was
originally used by Tsai et al. (2016). One of the items, ‘I feel nervous when thinking about online security issues’ was recoded. A Principal Axis Factor analysis with an oblique Direct Oblimin rotation was run on the six items and showed a Kaiser-Meyer-Olkin (KMO) measure of .849 and Bartlett’s test of sphericity was statistically significant (p < .001). One factor was extracted which had an Eigenvalue larger than 1 (Eigenvalue = 3.31) and explained 55.24% of the variance, with an obvious point of inflection in the screeplot. A reliability analysis revealed that the six items combined to form an internally consistent scale, α = .80. A single scale was created (M = 5.18, SD = 0.69).
Response-efficacy. Four newly developed items make up the 7-point scale assessing how effective participants think the recommended responses are for enhancing online security (e.g. ‘Carefully checking URLs will help protect me against Phishing attacks’). Two items were recoded and a higher score indicates a higher level of response efficacy. A Principal Axis Factor analysis with an oblique Direct Oblimin rotation was run on the four items and showed a Kaiser-Meyer-Olkin (KMO) measure of .590 and Bartlett’s test of sphericity was statistically significant (p < .001). Two factors were extracted which had an Eigenvalue larger than 1 and explained 68.27% of the variance; Factor 1, Eigenvalue = 1.71 and explained 42.92% of the variance; Factor 2 = 1.01 and explained 25.34% of the variance. The Pattern Matrix show three items in Factor 1, but one has a factor loading below 0.4, below an acceptable threshold and it was omitted from the study. Factor 2 contains one item and is also omitted for this reason. Response efficacy factor 1 therefore consists of the following items: ‘Carefully checking URLs will help protect me against Phishing attacks’, and ‘Recommended security measures to protect myself on the internet are effective for protecting against the threat of online crime’. A reliability analysis revealed that the two items combine to form a scale with a α = .64. This falls below an acceptable level of internal consistency of >.70 and results relating to this control variable should be interpreted cautiously. A single scale was created (M = 5.73, SD = 0.84).
Prior experience with safety hazards. This 7-item scale was originally used by Tsai et al.
study. However, the scale did not combine to form a reliable measure and was not included in the results analysis.
Results
Manipulation Check
An Independent samples t-test examined whether the Fear and Neutral stimulus material had a different effect on Fear-induced negativity. Levene’s F test for equality of variances was not
significant, F = 2.69, p = .102, and equal variances can be assumed. Results reveal a significant difference between the Fear (n = 127, M = 3.85, SD = 1.28) vs neutral condition (n = 133, M = 3.17, SD = 1.36), t(258) = -4.14, p < =.001, 95% CI [-1.00, -0.35], d = 0.51. We can conclude that the experimental stimulus material worked as intended; the higher mean and medium effect size show that the fear condition created a stronger negative effect. A Chi square test assessed whether participants recognised the Loss / Gain frame manipulation. Results show that the Conditions with Frame check as a dependent variable is statistically significant with a weak association: 2(3, N = 260) = 33.04, p = .000, Phi (1, 260) = .35. See table 1 for full results. We can conclude that the framing manipulation created sufficient difference between conditions.
Table 1
Manipulation check: condition*frame
Condition Fear gain Fear loss Neutral gain Neutral loss Total Frame check Loss Count 33 45 15 27 120 % 27.5% 37.5% 12.5% 22.5% 100% Gain Count 32 18 53 37 140 % 22.9% 12.9% 37.9% 26.4% 100% Total Count 65 63 68 64 260 % 25% 24.2% 26.2% 24.6% 100% Randomisation Checks
A two-factor ANOVA was carried out to assess that there was no difference between groups with regard to age. Results reveal a non-significant difference and very weak effect size between age and each of the in the four experimental conditions, F (3, 256) = 0.90, p < .000, η2 = 0.01. A chi-square analysis revealed that there is no difference between experimental groups with regard to gender, 2(3, N = 260) = 2.46, p = .481, Phi = .09. A further chi-square analysis revealed that there is no difference between experimental groups with regard to education, 2(15, N = 260) = 12.55, p = .637, Kendall’s Tau B = .00. See table 2.
Table 2
Randomisation check: Education*Condition
High school
Further education
Higher
education Masters Doctorate Other Total
Condition Fear gain Count 12 14 22 10 5 2 65
% within 18.5% 21.5% 33.8% 15.4% 7.7% 3.1% 100%
Fear loss Count 19 16 16 8 2 2 63
% within 30.2% 25.4% 25.4% 12.7% 3.2% 3.2% 100% Neutral Gain Count 18 14 23 8 3 2 68 % within 26.5% 20.6% 33.8% 11,8% 4.4% 2.9% 100% Neutral Loss Count 9 16 28 9 2 0 64 % within 14.1% 25.0% 43.8% 14.1% 3.1% 0.0% 100% Total Count 58 60 89 35 12 6 260 % within 22.3% 23.1% 34.2% 13.5% 4.6% 2.3% 100%
Online Safety Intentions
A two-way analysis of covariance was carried out to assess the influence of fear appeals vs neutral messages on the intentions of an older target group to improve online safety, in conjunction with the effect of gain or loss frame messages. Intentions to protect was measured with a two-factor dependent variable, 1) Intentions to protect passwords and 2) Intentions to check URLs. This analysis was conducted while controlling for self-efficacy (M = 5.18, SD = 0.70) and response-efficacy (M = 5.73, SD = 0.84) and existing online habits (M = 5.76, SD = 1.01).
Intentions to protect online safety. There is a non-significant effect among participants who were exposed to the fear vs neutral condition on Intentions to improve online password protection, F (1, 253) = 1.31, p = .250, η2 = 0.00, and Intentions to check URLs, F (1, 253) = .023, p = .879, η2 = 0.00. We reject H1, fear appeals did not affect Intentions to protect online safety. Furthermore, a
non-significant effect of framing was found among participants who were exposed to the loss vs gain frames on intentions to improve password protection, F (1, 253) = .538, p = .464, η2 = 0.00, and URL checking, F (1, 253) = 1.38, p = .240, η2 = 0.00. We reject H2 because gain frames did not affect Intentions to protect online safety. There was a non-significant interaction with no effect between message appeal and frames for checking passwords, F (1, 253) = .081, p = .783, η2 = 0.00, and URLs, F (1, 253) = 2.12, p = .146, η2 = 0.00. We also reject H3, fear level and framing type did not interact, the fear appeal does not have a stronger effect on Intentions when gain framed. The assumption of equal variances in the population has not been breached for password protection, Levene's F (3, 256) = 1.10, p = .347 or URL checking, Levene's F (3, 256) = .526, p = .665. Refer to Table 3, 4 5 and 6.
Table 3
Two-factor ANCOVA. Dependent variable: Intentions, password protection
SS df MS F p η²
Response-efficacy 4.82 1 4.82 4.53 .034 0.01
Self-efficacy 12.364 1 12.364 11.61 <.001 0.04 Existing online habits .301 1 .301 .283 .595 0.00 Fear / Neutral condition 1.41 1 1.41 1.331 .250 0.00 Loss / Gain condition .572 1 .572 .538 .464 0.00 Fear_neutral*Loss_gain .081 1 .081 .076 .783 0.00
Error 269.321 253 1.084
Total 7815.063 260
Note. N = 260 Table 4
Dependent variable: Intentions, password protection
M SD N Gain Fear 5.35 0.97 64 Neutral 5.33 1.01 67 Total 5.34 0.99 131 Loss Fear 5.50 1.26 63 Neutral 5.31 1.05 66 Total 5.40 1.16 129 Total Fear 5.43 1.12 127 Neutral 5.32 1.03 133 Total 5.37 1.07 260
Table 5
Two-factor ANCOVA. Dependent variable: Intentions, URL checking
SS df MS F p η²
Response-efficacy 5.701 1 5.701 10.419 <.001 0.02
Self-efficacy 2.778 1 2.778 5.077 .025 0.01
Existing online habits 14.102 1 14.102 25.775 <.001 0.06 Fear / Neutral condition .013 1 .013 .023 .879 0.00 Loss / Gain condition .760 1 .760 1.38 .240 0.00 Fear_neutral*Loss_gain 1.163 1 1.163 2.12 .146 0.00
Error 148.424 253 .569
Total 9530.222 260
Table 6
Dependent Variable: Intentions, URL checking
M SD N Gain Fear 5.92 0.83 64 Neutral 5.92 0.83 67 Total 5.92 0.83 131 Loss Fear 6.00 1.03 63 Neutral 6.11 0.82 66 Total 6.05 0.93 129 Total Fear 5.96 0.93 127 Neutral 6.01 0.83 133 Total 5.98 0.88 260
Risky Online Behaviour
A two-factor analysis of covariance was carried out to test the interaction between fear appeal and framing. We expected that older people who were exposed to the strong fear appeal with a gain frame displayed the least risky online behaviours. This analysis was conducted while controlling for Self-efficacy, Response-efficacy and Existing online habits.
We found a non-significant effect among participants who were exposed to the fear vs neutral condition on Likelihood of installing a risky computer plug-in, F (1, 253) = .256, p = .614, η2 = 0.00. The Likelihood of installing a risky computer plug-in was slightly higher among participants after exposure to the fear appeal (M = 1.98, SD = 1.64) vs the neutral message (M = 1.81, SD =1.51). A significant result and weak effect is found among participants who were exposed to the loss vs gain frames on Likelihood of installing a risky computer plug-in, F (1, 253) = 5.39, p = .021, η2 = 0.02. Likelihood of installing a risky online plug-in was higher among participants after exposure to the gain frame (M = 2.11, SD = 1.74) than the loss frame (M = 1.67 SD = 1.34). There was a non-significant interaction with no effect between message appeal and frames, F (1, 253) = 1.43, p = .232, η2 = 0.00, and so we reject H4, fear level and framing did not interact and exposure to a strong fear appeal that is gain framed did not have a stronger effect on developing less risky online behaviour. Results should be interpreted with caution because the assumption of equal variances in the population has been breached, Levene’s F (3, 256) = 3.58, p = .014. Refer to table 7.
Two-way ANCOVA results. Dependent variable: Likelihood of downloading risky plug-in
SS df MS F p η²
Response-efficacy 10.182 1 10.182 4.33 .038 0.01
Self-efficacy .116 1 .116 .049 .824 0.00
Existing online habits 4.438 1 4.438 1.89 .170 0.00 Fear_Neutral condition .600 1 .600 .256 .614 0.00 Loss_Gain condition 12.658 1 12.658 5.39 .021 0.02 Fear_neutral*Loss_gain 3.375 1 3.375 1.437 .232 0.00 Error 594.129 253 2.348 Total 1574.000 260 Note. N = 260 Discussion
This study set out to examine if fear appeals influence preventative intentions and behaviour to improve online safety among elderly British people, and whether this is moderated by type of message frame. An experiment was carried out to test the hypothesis that a fear appeal is only effective among an elderly target group when combined with gain frame information. An interaction effect was predicted showing a stronger persuasive effect because the gain frame information
presents positive consequence of performing an action. By reinforcing the efficacy elements of a fear appeal, which are critical elements of their success, gain-framed information, rather than loss-framed information which is seen in traditional fear appeals (Jensen et al. 2018), could potentially reduce a fear appeal’s backfire effect’s (Witte, 1994) by reinforcing the all-important efficacy statement.
Using gain frames to reinforce positive nature of messages is particularly important for an elderly target group considering that they are prone to ‘optimism bias’ as explained by SST.
When looking at main effects of appeal, it was predicted that a neutral message would be more effective than a fear appeal for encouraging stronger intentions and behaviour among an older target group. These hypothesis were not supported. It was furthermore predicted that gain frames would have a stronger positive effect on the intention to protect oneself online and actual protection behaviour. Results did not reveal such an effect for intention to protect. Results furthermore revealed no interaction effects between fear appeal and type of framing. However, results did show the loss frame and not the gain frame to be more persuasive for encouraging less risky behaviour, which is opposite to what was hypothesised. Participants who read the loss frame stimulus material were more likely to demonstrate safer online behaviour by reduced Likelihood of downloading a risky computer plug in.
The study possibly failed to find an interaction effect because the stimulus materials did not create a difference in intention to protect oneself online. This is despite the fear appeal manipulation check revealing a clear difference between the fear and neutral conditions with a medium effect size. On average, respondents had a relatively high intention, with all scores falling somewhere around five on a seven-point scale. Future research could compare participants to a ‘no information’ control group, and also make this a pretest-posttest design. This would make it easier to establish if the high level of intention is due to reading the stimulus materials, or not. Whether or not participants and high levels of digital literacy was not measured in this study but future research could measure digital literacy as a potential moderator of persuasive effects. It could be that participants in this study were already digitally very competent which could also explain the high intentions to protect themselves online.
Although speculative, not finding a difference in intention after exposure to a neutral versus fear-inducing message could be explained by the fact that password advice and guidance about URLs
has been commonly communicated. Evidence can be seen in comments to the pre-test of the stimulus material, one respondent said: ‘It (stimulus information) repeated concerns and advice that I have heard on various consumer programs.’ Future research could use cyber scenarios that are less familiar to participants because they might then be more naïve about how to respond which could lead to greater potential effect of reading the stimulus materials. Another explanation for not finding an interaction effect could be that the positive effect of the gain frame information cancels out feelings of anxiety or fright created by the Fear condition. Future research should include extensive testing of gain frames within fear appeals to provide further insights about how these two persuasive elements might interact.
Although no effects of framing on intention to protect oneself online were found, the loss frame proved to be more effective in affecting safer online behaviour. This result does not support SST, that older people are more likely to prefer positive information to negative. That loss frames were more persuasive could relate to prospect theory which states that when it comes to taking risks losses loom larger than gains (Tversky & Kahneman, 1981). The behavioural experimental scenario gave participants information about downloading a software plug-in that could save them money on food at local shops (see Appendix) but that could also damage their device. To refer directly to framing theory and Rothman et al. (2006), if participants are considering a behaviour which they understand involves risk of a bad outcome, as is the case with behavioural scenario in this study, loss-framed appeals should be more persuasive than gain frames. Loss frames might have been more persuasive in this study because they highlighted the negative nature of the risky scenario in participants’ minds, therefore persuading them that the risks outweigh the rewards.
Another possible explanation for loss frames being more persuasive for encouraging less risky behaviour in this elderly group is offered by Jensen et al. (2018); where self-efficacy is high, loss frames are more effective. Jensen et al. (2018) cite evidence from five studies supporting this idea. As a covariate, self-efficacy was not a statistically significant predictor of behaviour, only for intentions. Response efficacy is the only significant predictor of behaviour but this result should be
interpreted cautiously because of questionable internal consistency of this construct in the current study. Interestingly, exploratory analysis on the collected data did reveal a significant interaction effect between framing (loss vs. gain) and responses efficacy (low vs. high). Future research could further explore the relationship between framing and response efficacy.
Limitations.
This study could be improved by addressing the following limitations. Firstly, during pre-testing, the stimulus materials only focused on the Fear loss and Neutral loss conditions. Fear gain was tested separately with only four participants. It would be better to test all four experimental conditions to ensure that the gain-framed fear appeal induced sufficient anxiety and create a statistically significant difference between conditions, as the Fear loss condition did vs the Neutral loss message. Pre-testing all conditions could have strengthened the potential effect of the stimulus materials. Despite this limitation, the fear appeal manipulation did show a significant difference with medium effect size and results also revealed a significant difference in the behaviour measure.
Although improvements were made to the framing information at pre-testing phase, more thorough pre-testing of the gain and loss messages could have made it easier for participants to distinguish between these conditions. The manipulation check showed a significant difference between the gain and loss conditions but in the Fear gain condition, an equal number of participants recognised the frame as gain or loss, meaning that almost half of participants who were exposed to this condition incorrectly recognised text framing. In the Neutral loss condition, 26.4% of
participants marked the manipulation check as gain, only 22.5% correctly recognised that the text was loss framed. See table 1. As well as improving the wording in the stimulus material, the framing manipulation check would be better as an interval-level measurement, not nominal. Such a change would make the framing information easier for participants to interpret by using a 7-point scale (1 = completely disagree; 7 = completely agree) and asking if the message was more focussed on the positive or the negative consequences of preventing cybercrime.
Recommendations for practice
The health promotion literature clearly establishes that to reduce potential ‘boomerang effects’, all threat communication should include sufficiently strong self-efficacy and response-efficacy statements. Results from this study support the idea that fear appeals with a strong self-efficacy component are much less likely to create avoidant responses in participants. If this had been the case, intentions in the fear appeal condition would have been lower than the in the neutral
condition. Therefore, communication professionals who need to promote cyber security advice and guidance, should ensure that all materials contain clear and prominently featured self-efficacy and response-efficacy elements. When using neutral promotion material to encourage better online safety, results from this study show that loss frames are the most effective method for developing vigilant behaviour.
Conclusion
This study looked at the effect of fear appeals and message framing on an elderly target group in order to understand how a preventative message would influence online safety intentions and behaviour. Results revealed that fear appeals, when including information aimed at enhancing self-efficacy and response-self-efficacy, were not less effective than neutral messages among an elderly target group. Furthermore, loss frames are recommended for helping to develop less risky online behaviour among an elderly target group. There was no evidence of an interaction effect between fear and framing. The results from this study offer new empirical evidence that is deeply rooted in persuasion and behaviour change and therefore can contribute fresh insights to cyber security literature.
References
Anderson, C., & Agarwal, R. (2010). Practicing Safe Computing: A Multimethod Empirical Examination of Home Computer User Security Behavioral Intentions. MIS Quarterly, 34(3), 613–643. https://doi.org/10.2307/25750694
Bada, M., Sasse, A., & Nurse, J. (2015). Cyber Security Awareness Campaigns: Why do they fail to change behaviour? Retrieved from http://kar.kent.ac.uk/67511/1/csss2015_bada_et_al.pdf
Bailey, J. (1987). The Persuasibility of Elderly Consumers: A Study of Locus on Control and Responsiveness to Fear Appeals. Current Issues and Research in Advertising, 10(1-2), 213– 247. https://doi.org/10.1080/01633392.1987.10504920
Bartholomew, L., Parcel, G., Kok, G., Gottlieb, N., Fernandez, M., Gezondheidsvoorlichting, … RS: FPN WSP II. (2011). Planning health promotion program: an intervention mapping approach. Benet, S., Pitts, R., & LaTour, M. (1993). The appropriateness of fear appeal use for health care
marketing to the elderly: Is it OK to scare granny? Journal of Business Ethics, 12(1), 45–55. https://doi.org/10.1007/BF01845786
Borland, R. (2018). Misinterpreting theory and ignoring evidence: fear appeals can actually work: a comment on Kok et al. (2018). Health Psychology Review, 12(2), 126–128. https://doi.org/10.1080/17437199.2018.1445545
Brown, S., & West, C. (2015). Sequencing the threat and recommendation components of persuasive messages differentially improves the effectiveness of high‐ and low‐distressing imagery in an anti‐alcohol message in students. British Journal of Health Psychology, 20(2), 324–340. https://doi.org/10.1111/bjhp.12103
Carstensen, L., Fung, H., & Charles, S. (2003). Socioemotional Selectivity Theory and the
Regulation of Emotion in the Second Half of Life. Motivation and Emotion, 27(2), 103–123. https://doi.org/10.1023/A:1024569803230
Chowdhury, R., Sharot, T., Wolfe, T., Düzel, E., & Dolan, R. (2014). Optimistic update bias increases in older age, 44(9), 2003–2012. https://doi.org/10.1017/S0033291713002602 Covey, J. (2014). The Role of Dispositional Factors in Moderating Message Framing
Cross, C. (2017) ‘But I’ve never sent them any personal details apart from my driver’s
licence number...’: Exploring seniors’ attitudes towards identity crime. Security Journal, 30(1), pp. 74-88.
Falk, E., & Scholz, C. (2018). Persuasion, Influence, and Value: Perspectives from Communication and Social Neuroscience. Annual Review of Psychology, 69(1). https://doi.org/10.1146/annurev-psych-122216-011821
Fear appeals (2019). Research into the effectiveness of threatening communication. Retrieved from: https://fearappeals.com/
Federal Bureau of Investigation (2018) Scams and safety: Fraud against seniors. Retrieved from https://www.fbi.gov/scams-and-safety/common-fraud-schemes/seniors
Friedman, A., & Singer, P. W. (2014). Cybersecurity and Cyberwar: what everyone needs to know. New York, Estados Unidos: Oxford University.
Fung, H., & Carstensen, L. (2003). Sending Memorable Messages to the Old: Age Differences in Preferences and Memory for Advertisements. Journal of Personality and Social
Psychology, 85(1), 163–178. https://doi.org/10.1037/0022-3514.85.1.163
Gallagher, K., & Updegraff, J. (2012). Health Message Framing Effects on Attitudes, Intentions, and Behavior: A Meta-analytic Review. Annals of Behavioral Medicine, 43(1), 101–116. https://doi.org/10.1007/s12160-011-9308-7
General Register Office for Scotland. 2002. Scotland’s census 2001: 2001 population report Scotland. Available at: http://www.gro-scotland.gov
.uk/census/censushm/index.html quoted in Lee, C., & Coughlin, J. (2015). Perspective: Older Adults' Adoption of Technology: An Integrated Approach to Identifying Determinants and Barriers. Journal of Product Innovation Management, 32(5), 747-759.
Gerend, M., & Maner, J. (2011). Fear, Anger, Fruits, and Veggies: Interactive Effects of Emotion and Message Framing on Health Behavior. Health Psychology, 30(4), 420–423.
Grimes, G., Hough, M., Mazur, E., & Signorella, M. (2010). Older Adults’ Knowledge of Internet Hazards. Educational Gerontology, 36(3), 173–192.
https://doi.org/10.1080/03601270903183065
James, B., Boyle, P., Yu, L., & Bennett, D. (2013). Internet Use and Decision Making in Community-based Older Adults. Gerontologist, 53, 589–589.
Jansen, J., & van Schaik, P. (2018). Persuading end users to act cautiously online: a fear appeals study on phishing. Information and Computer Security, 26(3), 264–276. https://doi.org/10.1108/ICS-03-2018-0038
Jayanti, R. (2010). Accentuate the Positive: Elderly Responses to Health
Communications. Journal of Marketing Theory and Practice, 18(3), 263–274. https://doi.org/10.2753/MTP1069-6679180304
Jensen, J., Ratcliff, C., Yale, R., Krakow, M., Scherr, C., & Yeo, S. (2018). Persuasive impact of loss and gain frames on intentions to exercise: A test of six moderators. Communication Monographs, 85(2), 245–262. https://doi.org/10.1080/03637751.2017.1353699
Johnston, A., & Warkentin, M. (2010). Fear appeals and information security behaviors an empirical study. Management Information Systems : Mis Quarterly, 34(3), 549–566. https://doi.org/10.2307/25750691
Kok, G. J. (2016). Social psychology applied: Politics, theories, and the future. Valedictorian lecture, Maastricht: Maastricht University. Retrieved online:
http://interventionmapping.com/wpcontent/uploads/2016/04/Kok-2016-valedictory-lecture.pdf
Kok, G., Peters, G., Kessells, L., Ten Hoor, G., & Ruiter, R. (2018). Ignoring theory and
misinterpreting evidence: The false belief in fear appeals. Health Psychology Review, Epub, 1–15. Doi:10.1080/17437199.2017.1415767
Larose, R., Rifon, N., & Enbody, R. (2008). Promoting personal responsibility for internet
safety. Communications of the ACM, 51(3), 71–76. https://doi.org/10.1145/1325555.1325569 Liang, H., & Xue, Y. (2010). Understanding security behaviors in personal computer usage:
A threat avoidance perspective. Journal of the Association for Information Systems, 11(7), 394-413.
https://search-ebscohost- com.proxy.uba.uva.nl:2443/login.aspx?direct=true&db=buh&AN=56503617&site=ehost-live&scope=site
Lewis, I., Watson, B., & White, K. M. (2013). Extending the explanatory utility of the EPPM beyond fear-based persuasion. Health Communication, 28(1), 84-98. Masthof, J., Collinson, M., & Vargheese, J.P., (2018). Symposium on Digital Behaviour
Interventions for Cyber Security 2018. Retrieved from:
http://aisb2018.csc.liv.ac.uk/PROCEEDINGS%20AISB2018/Digital%20Behaviour%20Interve ntions%20for%20CyberSecurity%20-%20AISB2018.pdf
Nabi, R., Gustafson, A., & Jensen, R. (2018). Framing Climate Change: Exploring the Role of Emotion in Generating Advocacy Behavior. Science Communication, 40(4), 442–468. https://doi.org/10.1177/1075547018776019
Palan, S., & Schitter, C. (2018). Prolific.ac—A subject pool for online experiments. Journal of Behavioral and Experimental Finance, 17, 22–27. https://doi.org/10.1016/j.jbef.2017.12.004
Peters, G., Ruiter, R., & Kok, G. (2013). Threatening communication: a critical re-analysis and a revised meta-analytic test of fear appeal theory. Health Psychology Review, 7(sup1), S8–S31. https://doi.org/10.1080/17437199.2012.703527
Peters, G., Ruiter, R., & Kok, G. (2014). Threatening communication: A qualitative study of fear appeal effectiveness beliefs among intervention developers, policymakers, politicians,
scientists, and advertising professionals. International Journal of Psychology, 49(2), 71–79. https://doi.org/10.1002/ijop.12000
Removing Outliers From a Dataset. (2011, December 12). Retrieved from Youtube: https://www.youtube.com/watch?v=DDpym2j_ILY
Renaud, K., & Biljon, J. (2010). Worth-centred mobile phone design for older users. Universal Access in the Information Society, 9(4), 387–403. https://doi.org/10.1007/s10209-009-0177-9
Roberto, A., Mongeau, P., & Liu, Y. (2018). A (Re)defining moment for fear appeals: a comment on Kok et al. (2018). Health Psychology Review, 12(2), 144–146. https://doi.org/10.1080/17437199.2018.1445546
Rogers, R. (1975). A Protection Motivation Theory of Fear Appeals and Attitude Change1. The Journal of Psychology, 91(1), 93–114.
https://doi.org/10.1080/00223980.1975.9915803
Rösler, A., Ulrich, C., Billino, J., Sterzer, P., Weidauer, S., Bernhardt, T., … Kleinschmidt, A. (2005). Effects of arousing emotional scenes on the distribution of visuospatial attention: changes with aging and early subcortical vascular dementia. Journal of the Neurological Sciences, 229-230, 109–116. https://doi.org/10.1016/j.jns.2004.11.007
Rosoff, H., Cui, J., & John, R. (2013). Heuristics and biases in cyber security dilemmas. Environment Systems and Decisions, 33(4), 517–529. https://doi.org/10.1007/s10669-013-9473-2
Rothman, A., Bartels, R., Wlaschin, J., & Salovey, P. (2006). The Strategic Use of Gain‐ and Loss‐ Framed Messages to Promote Healthy Behavior: How Theory Can Inform Practice. Journal of Communication, 56((Supp.)), S202–S220.
https://doi.org/10.1111/j.1460-2466.2006.00290.x
Role of Message Framing. Psychological Bulletin, 121(1), 3–19. https://doi.org/10.1037/0033-2909.121.1.3
Ruiter, R., Kessels, L., Peters, G., & Kok, G. (2014). Sixty years of fear appeal research: Current state of the evidence. International Journal of Psychology, 49(2), 63–70. https://doi.org/10.1002/ijop.12042
Shen, L., & Dillard, J., P. (2014). Threat, Fear, and Persuasion: Review and Critique of Questions About Functional Form. Review of Communication Research, 2(1), 94–114.
https://doi.org/10.12840/issn.2255-4165.2014.02.01.004
Shillair, R., Cotten, S., Tsai, H., Alhabash, S., Larose, R., & Rifon, N. (2015). Online safety begins with you and me: Convincing Internet users to protect themselves. Computers in Human Behavior, 48, 199–207. https://doi.org/10.1016/j.chb.2015.01.046
Sparks, J., & Ledgerwood, A. (2019). Age Attenuates the Negativity Bias in Reframing Effects. Personality and Social Psychology Bulletin, 45(7), 1042–1056.
https://doi.org/10.1177/0146167218804526
Spreng, R., Karlawish, J., & Marson, D. (2016). Cognitive, social, and neural determinants of diminished decision-making and financial exploitation risk in aging and dementia: A review and new model. Journal of Elder Abuse & Neglect, 28(4-5), 320–344.
https://doi.org/10.1080/08946566.2016.1237918
Tannenbaum, M., Hepler, J., Zimmerman, R., Saul, L., Jacobs, S., Wilson, K., & Albarracín, D. (2015). Appealing to Fear: A Meta-Analysis of Fear Appeal Effectiveness and
Theories. Psychological Bulletin, 141(6), 1178–1204. https://doi.org/10.1037/a0039729 Tsai, H., Jiang, M., Alhabash, S., Larose, R., Rifon, N., & Cotten, S. (2016). Understanding
online safety behaviors: A protection motivation theory perspective. Computers & Security, 59, 138–150. https://doi.org/10.1016/j.cose.2016.02.009
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science (New York, N.Y.), 211(4481), 453–458.
https://doi.org/10.1126/science.7455683
UK Cabinet Office, D. (2011). The Cost of Cyber Crime: A Detica Report In Partnership With The Office Of Cyber Security and Information Assurance In the Cabinet Office. London: UK Cabinet Office, Detica
Updegraff, J., & Rothman, A. (2013). Health Message Framing: Moderators, Mediators, and Mysteries. Social and Personality Psychology Compass, 7(9), 668–679.
https://doi.org/10.1111/spc3.12056
Van ’t Riet, J., Cox, A. D., Cox, D., Zimet, G. D., de Bruijn, G.-J., van den Putte, B., de Vries, H., Werrij, M. Q., &Ruiter, R. A. C. (2016, advance online publication). Does
perceived risk influence the effects of message framing? Revisiting the link between prospect theory and message framing? Health Psychology Review, 1-13.
Van der Goot, M.J., Van Reijmersdal, E.A., & Kleemans, M. (2015). The effects of
emotional television advertising on older and younger adults. In P.W.J. Verlegh, H.A.M. Voorveld, & M. Eisend (Eds.), Advances in advertising research (Vol. VI): The digital, the classic, the subtle, and the alternative (pp. 115-124). Wiesbaden: Springer Gabler. Warkentin, M. C., Walden, E. W., Johnston, A., & Straub, D. (2016). Neural correlates of protection motivation for secure IT behaviors: An fMRI examination. Journal of the Association of Information Systems, 17(3), 194-215.
Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59(4), 329–349.
https://doi.org/10.1080/03637759209376276
Witte, K. (1994). Fear control and danger control: A test of the extended parallel process model (EPPM). Communication Monographs, 61(2), 113–134.
https://doi.org/10.1080/03637759409376328
Witte, K., & Allen, M. (2000). A Meta-Analysis of Fear Appeals: Implications for Effective Public Health Campaigns. Health Education & Behavior, 27(5), 591–615.
https://doi.org/10.1177/109019810002700506
World Economic Forum (2019). The biggest risks facing our world. Retrieved from
https://www.weforum.org/agenda/2019/01/these-are-the-biggest-risks-facing-our-world-in-2019/
Zhang, X., Fung, H., & Ching, B. (2009). Age differences in goals: Implications for health promotion. Aging & Mental Health, 13(3), 336–348.
https://doi.org/10.1080/13607860802459815
Zimmerman, S., R., Cupp, P., Abadi, M., Donohew, R., Gray, C., Gordon, L., & Grossl, A. (2014). The Effects of Framing and Fear on Ratings and Impact of Antimarijuana PSAs. Substance Use & Misuse, 49(7), 824–835.
