To prepare or not to prepare:
A research into the individual crisis preparedness for a terrorist attack
Hilde Schueler S1686062
Master’s thesis Crisis and Security Management Leiden University
June 2020
Supervisor: dr. G.M. van Buuren Second reader: dr. G.G. de Valk
2
Abstract
We live in a society where one of the main concerns is handling risks such as terrorism. The responsibility for handling risks lays with the government as well as with the civilians. For the civilian, a part of this responsibility consists of preparing himself.
It is tried to get a better understanding as to why an individual does or does not prepare himself for terrorism. To do so, Paton’s social-cognitive model of natural disaster preparedness (Paton, 2003) is used as a theoretical framework and a questionnaire based on this model is held amongst people living in The Netherlands. The results show that overall, the individual crisis preparedness for terrorism is low. Moreover, Paton’s model only partly explains the process prior to the preparedness behaviour of an individual when it comes to preparing for terrorism. The last phases of the process, where self-efficacy leads to intention and intention gets converted into actual preparedness behaviour, can be understood using the model. However, no significant results were found for the relations between motivation, outcome expectancy and self-efficacy. Policy implications should therefore be targeted at the later phases of the model.
3
Table of Contents
1. Introduction___________________________________________________________p. 6 2. Theoretical Framework__________________________________________________p. 9 2.1 Terrorism______________________________________________________p. 9 2.2 Risk society and terrorism_________________________________________p. 10 2.3 Crisis preparedness______________________________________________ p. 11 2.4 Paton’s theory for preparedness behaviour____________________________p. 13 2.4.1 Motivation______________________________________________p. 13 2.4.2 Intention formation_______________________________________p. 14 2.4.3 Linking intention and preparation____________________________p. 15 2.4.4 Preparedness behaviour____________________________________p. 16 2.5 Hypotheses_____________________________________________________p. 17 3. Methodology__________________________________________________________p. 19 3.1 Research design_________________________________________________p. 19 3.1.1 Variables_______________________________________________p. 20 3.2 Data collection__________________________________________________p. 20 3.3 Participants_____________________________________________________p. 21 3.4 Operationalisation_______________________________________________p. 21 3.5 Data preparation_________________________________________________p. 26 3.6 Data analysis___________________________________________________p. 26
3.6.1 Method of analysis for hypothesis 1__________________________p. 26 3.6.2 Method of analysis for hypothesis 2 and 3_____________________p. 28 3.6.3 Method of analysis for hypothesis 4__________________________p. 28 3.7 Procedure______________________________________________________p. 30
4 3.8 Limitations_____________________________________________________p. 30 3.8.1 Reliability______________________________________________p. 31 3.8.2 Validity________________________________________________p. 31 4. Results_______________________________________________________________p. 33 4.1 Respondents____________________________________________________p. 33 4.2 Motivation_____________________________________________________p. 33 4.3 Intention formation_______________________________________________p. 34 4.4 Linking intention and preparedness__________________________________p. 35
4.4.1 Moderator variables_______________________________________p. 36 5. Analysis______________________________________________________________p. 37
5.1 Hypothesis 1____________________________________________________p. 37 5.1.1 Assumptions____________________________________________p. 37 5.1.2 Analysis and results_______________________________________p. 41
5.1.2.1 Best model fit____________________________________p. 41 5.1.2.2 Analysis for Model 3______________________________ p. 43 5.1.3 Conclusion hypothesis 1___________________________________p. 47 5.2 Hypothesis 2____________________________________________________p. 47
5.2.1 Assumptions____________________________________________p. 47 5.2.2 Analysis and results_______________________________________p. 47 5.2.3 Conclusion hypothesis 2___________________________________p. 49 5.1 Hypothesis 3____________________________________________________p. 49
5.3.1 Assumptions____________________________________________p. 49 5.3.2 Analysis and results_______________________________________p. 49 5.3.3 Conclusion hypothesis 3___________________________________p. 50
5 5.4 Hypothesis 4____________________________________________________p. 50 5.4.1 Assumptions____________________________________________p. 50 5.4.2 Analysis and results_______________________________________p. 53
5.4.2.1 Simple regression analysis__________________________p. 53 5.4.2.2 Regression analysis with moderator variables___________p. 55 5.4.3 Conclusion hypothesis 4___________________________________p. 60 6. Conclusion____________________________________________________________p. 61 7. Discussion____________________________________________________________p. 62 Reference list____________________________________________________________p. 66 Appendix 1: Questionnaire_________________________________________________p. 73
6
1. Introduction
On March 18th, 2019 The Netherlands is left shocked after random civilians are
attacked on a tram in the Dutch city Utrecht. A terrorist motive was considered and later confirmed by the government and temporarily the terrorist threat level was increased to the highest level (Gemeente Utrecht, 2019; Nationaal Coördinator Terrorismebestrijding en Veiligheid, 2019). Gökman, the perpetrator of the attack, got a lot of media attention on that day and in the year that followed. A year after the attack, he was convicted to life
imprisonment (De Rechtspraak, 2020).
The Dutch National Coordinator for Security and Counterterrorism (NCTV) is responsible for deciding on the threat level of the likeliness of a terrorist attack happening in The Netherlands. Since December 2019 the level has been scaled back from four to three, on a scale of five. Although it is now lower than it has been the past year, a threat level of three still means a terrorist attack is conceivable (Rijksoverheid, 2019). There is thus a
(conceivable) risk that a terrorist attack will happen in The Netherlands in the near future. Since we live in a risk society, the government is not the only actor who is responsible for dealing with the risks of contemporary society and the actual crises these risks may turn into, such as the risk of terrorism and an actual terrorist attack. The work of government emergency agencies is influenced by the acting and omitting of civilians (De Vries, 2005). Therefore, the government depends on the civilians to act up when it comes to preparing for and handling during a crisis situation. A civilian is in part responsible for preventing and reducing the effects of a crisis for himself and others around him.
It is not only in the best interest of the government that civilians prepare themselves for a crisis. Being prepared for a crisis, in specific a terrorist attack, has multiple advantages for the individuals themselves. First, individuals can help protect their physical safety and physical health by being prepared for an emergency situation that may be caused by a
(terrorist) attack. Second, preparedness may contribute to an individual’s resilience to trauma and overall mental health and lastly, preparedness levels of individuals might play a role in averting a terrorist attack. If a (potential) terrorist believes that the attack may be less
successful due to the high levels of preparedness in society, chances are the attack might not take place. However, it is important to note that these chances are small. The potential terrorist has to have knowledge on the preparedness levels and has to take this into account
7 when thinking about conducting the attack. Moreover, when he does so, it is also possible he will conduct the attack elsewhere (Eisenman et al, 2006).
Because of the reasons mentioned above, it is highly socially relevant to look further into the preparedness of people living in the Netherlands. First, the risk of a terrorist attack happening in The Netherlands is relatively high, Second, we as civilians are (partly)
responsible for dealing with this risk and the possible crisis situation that follows ourselves. Third, it is beneficial for the individual to be prepared and lastly it preparation might lower the chance of an attack happening.
There have been quite a few studies into the preparedness of a county, municipality, state, country or individual for a terrorist attack. Especially after the 9/11 attacks, there has been an increase in studies into the preparedness of the different actors in the United States, New York and specifically New York City (Boscarino, Figley, & Adam, 2003; Eisenman et al., 2006; Hurley-Hanson et al., 2008). However, there is a definite gap in knowledge when it comes to such research conducted in the Netherlands. There are only a few studies on the topic of crisis preparedness in general (Filippoupolitis, MacKinnon, & Bacon, 2015; t Hart, P, 1997). Furthermore, the only research that goes into the crisis preparedness of the Dutch civilians takes into account all disasters labelled as such by the ‘Leidraad maatramp’. (Laat, 2009; Ministerie van Binnenlandse Zaken en Koninkrijksrelaties, 2003).Not only is this guideline for disasters outdated, the existence of terrorism and the likeliness of an attack call for other levels of preparedness than the preparedness for a natural disaster or disaster in general. Terrorism preparedness is a highly specific component of general crisis preparedness. It calls for prudent levels of preparedness alongside the preparedness already recommended for other catastrophes (Eisenman et al, 2006). This paper will try and contribute to the existing body of knowledge and it will try to fill some of the gap in the knowledge by examining which factors and to what extent these factors play a role in predicting the level of preparedness of an individual for terrorism.
The theoretical model that will be used to examine the process prior to the
preparedness behaviour, is Paton’s social-cognitive model of natural disaster preparedness (Paton, 2003). Social‐cognitive models of an individual’s response to terrorism are of value to the development of risk management strategies. These models shine light onto psychological or behavioural issues that might emerge in face of a crisis. In this case, it might help the design of programs aimed at improving individual behaviour prior to and during terrorist
8 attacks. Panton’s model is designed for explaining people’s motivation, intentions and
behaviour when it comes to preparing for a (natural) disaster. The aim of the study is to test Panton’s model on preparedness in light of the preparedness for terrorism and see to what extent the (lack of) preparedness for a terrorist attack can be explained using the model. In essence: Is Panton’s model also applicable to explain the individual preparedness for terrorism? By studying this, more can be learned about the process prior to preparedness behaviour, specifically for a terrorist attack, and conclusions can be drawn on what to do the enhance this behaviour.
The research question that will be answered in this paper is: ‘To what extent can Paton’s social-cognitive model of disaster preparedness explain the level of preparedness for a terrorist attack of people living in the Netherlands?’. It is important to note that Paton’s
theoretical model has some severe shortcomings when it comes to the statistical testability. It is not possible to statistically test the model as a whole, and it is not possible to statistically test for the causation the model predicts. This is explained in more detail in the methodology chapter. It is therefore decided to divide the model into four different parts that on their own are testable. As a result, there is not one but four hypotheses underlying this study. In short, the first hypothesis states there is a positive relation between motivation and outcome expectancy. The second hypothesis states there is a positive relation between outcome expectancy and self-efficacy. The third hypothesis states there is a positive relation between self-efficacy and intentions. The last hypothesis states there is a positive relation between intentions and preparedness behaviour, however, this relation is moderated by several other variables. These hypotheses will be explained further at the end of the theoretical framework chapter.
First, the theoretical framework will be explained in detail. In this chapter, the working definition of terrorism is given and explained, some more information is given about the risk society to get a better understanding about the role terrorism plays in our society and the consequences of it for the civilians, the definition of crisis preparedness is discussed and Paton’s model is explained in detail. At the end of this chapter the hypothesis are explained. In the second chapter, the methodology, information is given about the design of the research, the data collection, the participants, the operationalisation of Paton’s model, the data
preparation and analysis, the procedure that is followed, and the limitations of the research method. In the fourth chapter the results of the questionnaire are presented, after which in the fifth chapter the outcomes of the different analysis are given and explained. The last chapter is
9 dedicated to the conclusions and discussion. Conclusion are drawn and the study, its results and limitations are discussed. Lastly, recommendations for further research are given.
2. Theoretical framework
In this chapter the theoretical framework that underlies this study will be further illustrated. To do so, first the different definitions of ‘terrorism’ are given. Second, the contemporary risk society will be further explained, since this gives some background information on the world we live in, and explains the role of terrorism and the role of the civilian in it. Third, it addresses the conceptual discussion and definitions of crisis
preparedness. Lastly, Paton’s social-cognitive model of natural hazard preparedness and the hypothesis that result from this model will be explained.
2.1 Terrorism
Terrorism is an ambiguous concept with a lot of different definitions. A questionnaire under leading academics on this topic revealed 109 different definitions of what constitutes ‘terrorism’. Taken as a whole, the various definitions of terrorism all include, to some extent, the following components: (a) the use of force or violence (b) by individuals or groups (c) that is directed toward civilian populations (d) and intended to instil fear (e) as a means of
coercing individuals or groups to change their political or social positions (Moghaddam & Marsella, 2004). However, the exact used definition differs per study, organization and government agency, depending on the focus of their work (Moghaddam & Marsella, 2004).
The Dutch National Coordinator for Security and Counterterrorism defines terrorism as ‘’the ideological threat of threatening, preparing or committing serious violence aimed at people, or acts aimed at causing socially disruptive damage to property, with the aim of causing social changes, bringing fear to the population, or influencing political decision-making’’ (Nationaal Coördinator Terrorismebestrijding en Veiligheid, 2008). This definition emphasises the ideological motive of the act, and takes into account not only violence directed at human life, but also at property.
Seen from a legal perspective, the Dutch Criminal Code describes in article 83a what is meant by ‘’terrorist intent’’: ‘’Terrorist intent shall be understood to mean the intention of causing fear in the population or a part of the population of a country, or unlawfully
10 acts or to tolerate certain acts, or of seriously disrupting or destroying the fundamental
political, constitutional, economic or social structures of a country or an international organisation.’’ (Article 83a, Dutch Criminal Code). Here the emphasise lays on the fear the terrorist behaviour causes. Terrorism is intended to create an extremely fearful state of mind (Kaplan, 1981). Moreover, this fearful state is not intended for just the victim(s) of the
terrorist attack, it is intended for an broader audience. This broader audience might not have a relationship with the victim(s) (Ruby, 2002).
The two definitions discussed above, show the meaning of terrorism from a moral or ideological perspective, and from a legal perspective. However, most important for this study is a third and final perspective, the behavioural perspective. The behavioural perspective defines terrorism purely by the behaviours or acts involved, regardless of the laws or morality of those who define it. As a result, with the behavioural perspective, the chances are high that different interpreters will draw the same conclusions as to whether or not a particular act counts as terrorism (Taylor as cited in Ruby, 2002) This makes the behavioural perspective the most reliable way of looking at terrorism, it is the only aspects of the three that entails a reliable operational definition of terrorism regardless of who measures it. The use of a legal or moral perspective may cause important variables that influence the development of terrorist groups and terrorist actions to become less distinguished. These two perspectives will also minimize the fact that different people are accountable to different laws and have different morals. Using these perspectives may increase the focus on the observer’s biases rather than on the act or behaviour themselves. With the behavioural perspective, acts of terrorism can be seen more clearly for what they are and what damage they cause (Ruby, 2002).
Not only is the behavioural perspective the most reliable way of defining terrorism, it is also the most suitable way of defining terrorism when it comes to this study. It is the terrorist behaviour and the effects of the behaviour for which individuals can prepare themselves. The moral and legal perspective do not really matter to a civilian, since these perspectives do not influence the act, the effects of the act for the individual or its
surroundings, or the ways in which an individual can be prepared for it. As the British Ambassador of the United Nations, Jeremy Greenstock, put it in 2001: ‘’Increasingly,
questions are being raised about the problem of the definition of a terrorist. Let us be wise and focused about this: terrorism is terrorism… What looks, smells and kills like terrorism, is terrorism.’’ (Schmid, 2011).
11
2.2 Risk society and terrorism
Risk used to be a concept that belonged to insurance companies and investors. Now, it is a common umbrella concept used for a variety of different social(-cultural) issues
(Boutellier, Drenth von Februar, Gude, Hajer, & De Vries, 2005). The first scholar to write about the risk society is Ulrich Beck. Beck’s risk society describes the contemporary society that, as a result of industrialisation and modernisation, is confronted with new threats and problems (Beck as cited in Toohey, 2008). Beck defines the risk society as: ‘’a systematic way of dealing with hazards and insecurities induced and introduced by modernisation itself’’ (Beck, Lash & Wynne, 1992. p. 21). With this modernisation, there has been a shift in society from modernity to ‘second modernity’. As a result, society is no longer concerned with the distribution of power, but with handling risks. Where, in the beginning of the twentieth century, the division between wealth and scarcity was central, the government is now for the most part occupied with the handling and division of risk (Boutellier et al., 2005).
Beck (1992) noted that there are three layers of danger that can be identified in the risk society. These three layers are: the ecological crises, the global economic crises and, most important for this study, terrorism. Terrorism eminently is a really powerful but also
unpredictable risk, and one that can have major consequences for society. Since we live in a society characterized by the desire to control chances and be secure, the terrorist power of uncertainty is enormous. It reminds us that there are limits to our risk assessment and management capabilities (Toohey, 2008).
These new, strong, and ongoing risks ask for a new way of governing. Where the responsibility to handle these kinds of risks used to lay fully in the hands of the government, this responsibility is now shifting more towards the members of the society (De Vries, 2005). The government is no longer capable of handling these risks on its own. Hence, individuals are (partly) responsible for their own safety and the government depends on the civilians to take action in preparing for a crisis as well as acting during a crisis. A part of this new found responsibility is thus to (try to) prepare yourself for potential crises, one of the crises with the most negative effects being a terrorist attack (De Vries, 2005).
2.3 Crisis preparedness
It seems to be a universal understanding in the literature of crisis management that there should not only be attention for what happens during and after a crisis, but also before a
12 crisis. Crisis preparedness is a big part of this and should be high on the crisis agenda (Boin, ‘t Hart, Stern, & Sundelius, 2005; McConnell & Drennan, 2006). Academic interest in crisis preparedness has increased since the end of the last century, which has resulted in both practical and academic depth and breadth of the practice of crisis preparedness. Although there is more and more research into the topic, as mentioned specifically after the 9/11 attacks, the definition of crisis preparedness remains ambiguous. Crisis preparedness is a sensitized concept, which means is does not lend itself to direct observation like physical phenomena do. Preparedness is embedded in the capacities and skills inherent to crisis management organizations as well as (the behaviour of) the general population (Staupe-Delgado & Kruke, 2017).
Empirical studies use indicators of preparedness, which function as a operationalised definition of preparedness instead of there being a clear definition beforehand. As a result, preparedness definitions vary enormously. To solve this, Staupe-Delgado and Kruke (2017) isolated four different types of crisis preparedness, based on the degree of control and the level of authority or governance. The four types that follow from this are: integrated preparedness (strong control and high level of governance), enacted preparedness (weak control and high level of governance), enforced preparedness (strong control and low level of governance) and personal preparedness (low control and low level of governance). The type that is relevant for this study is the personal preparedness. Personal preparedness includes preparedness measures undertaken voluntarily at an individual (or household, business, or community) level. The measures are taken without legislative requirements or coordination from a political level. Personal preparedness is therefore characterised by self-organization and individual agency (Staupe-Delgado & Kruke, 2017). At the individual level, crisis preparedness is a self-protective behaviour that is a response to potential losses to life and property (Mishra & Suar, 2012).
What can be defined more clearly is the concept ‘crisis’. There is a strong agreement that crises tend to be characterised by three sets of conditions. These are: severe and largely unexpected threats, high uncertainty, and the need for urgency in decision making (Rosenthal, Boin & Comfort, 2001; Boin et al. 2005). A terrorist attack falls within these conditions and can thus be seen as a crisis.
Lastly, a difference has to be made between time before a crisis and during a crisis. The time during a crisis is called the warm, repressive phase. During this phase, people seem
13 to able to handle the crisis situation quite well. Civilians are not waiting, but take action, bring themselves and others into safety, inform each other and do not panic (Sievers, 2009). The time before a crisis, when a crisis has not occurred yet, is called the cold, preventive phase. When civilians are preparing themselves for a crisis before it occurs and help prevent the crisis from occurring, the warm phase could be a lot less bad (Sievers, 2009). Crisis preparedness behaviour plays a role in the preventive phase.
2.4 Paton’s theory for preparedness behaviour
It is difficult to point out what factors cause individuals to engage in preparedness activities (Tierney, Lindell & Perry, 2013). If one thing is clear about the preparedness of an individual, it is that the substantial discontinuity between people’s risk beliefs and their level of preparedness suggests that there is a lot more to it than just the rational consideration of costs and benefits. Paton’s social-cognitive model of natural hazard preparedness includes a wide range of variables for predicting preparedness behaviour (Paton, 2003). The starting point of the model is that there is a relation between motivating factors and risk reduction behaviour. This relation is mediated by intentions. As a result, the process consists of three phase, all influenced by a specific set of variables. The first phase concerns factors that motivate individuals, the precursor variables. The second phase entails the variables that link this motivation with the formation of intentions. The last phase describes the relation between preparatory intentions and the actual preparation.
2.4.1 Motivation
In phase one, there are three precursor variables: risk perception, critical awareness and hazard anxiety. Risk perception is the first valid precursor. Risk perception entails an individual’s acceptation that a given hazard can pose a threat. However, even if an individual accepts the possible threat, the motivation to prepare may be tempered by other processes such as normalisation bias or unrealistic optimism. Normalisation bias describes how people conclude that, if they are able to cope with an objectively minor crisis, they are also capable to deal with any future occurrence (Paton, Smith, Daly, Johnston, 2008). Unrealistic optimism is the consistent belief of people that they are better prepared relative to other member of their community. This belief is characterised by people perceiving themselves as less vulnerable or more skillful than other around them (Paton, Smith, Daly, Johnston, 2008). To take these factors into account, two additional important precursor variables are added: critical awareness and hazard anxiety.
14 Critical awareness of a hazard is an important precursor in a response to adverse circumstances, like a crisis (Dalton & Wandersman, 2001). Critical awareness describes the extent to which individuals think and talk about a specific risk or hazard with their (personal) environment during normal, pre-disaster circumstances. It shows what individuals perceive as important. Critical awareness might be especially important since a terrorist attack is still somewhat of a rarity and individuals encounter adversity from various less rare sources, like crime, unemployment, and to lesser extent natural hazards (Paton, 2003). These more daily hazards get more media attention and are therefore more thought and talked about. A hazard like terrorism has to compete with that. Only when an individual perceives a terrorist attack as important or critical, he or she is likely to be motivated to take preparedness action.
Hazard anxiety means the anxiety or fear an individual has for a potential disaster. Disasters form a large source of anxiety, because of their unpredictable and uncontrollable nature and their potential of causing death and destruction. This anxiety might motivate individuals to prepare themselves, however it might have the opposite effect. It could also reduce the likeliness that individuals prepare for a disaster. For example, this anxiety could be reduced by ignoring information about, or actions related to the source of anxiety (Duval & Mulilis, 1999). This makes that including anxiety as a motiving or de-motivating factor is important (Paton, 2003).
The presence of risk perception, critical awareness of hazards and hazard anxiety motivates preparedness behaviour. Some level of presence of these factors is required for the process of preparing to begin. If the factors are presence at adequate levels, an individual will progress to the next phase, forming the intentions to prepare himself. However, this
progression, from motivation to intention formation, is influenced by another set of variables (Godin & Kok, 1996).
2.4.2 Intention formation
There are two predictors for the formation of intentions. Outcome expectancy and self-efficacy. Once people are motivated to think about hazards, people make a judgement whether their personal action will mitigate the effects of the hazard. The outcome of this judgement is called the ‘outcome expectancy’. If an individual has a positive outcome expectancy, whether or not they progress towards formation of the intention depends on their level of self-efficacy. efficacy is the extent to which a person thinks it is capable to realise these actions. Self-efficacy has proven to be a significant influence on behaviour when dealing with issues that
15 are perceived as less or not controllable. A terrorist attack can be seen as such an issue.
Besides these factors, one has to have the right resources, such as time, skills, money and social networks (Panton, 2003).
In the phase of intention formation, there are two other factors that play a role.
Problem-focused coping and response efficacy. Problem-focused coping is a predisposition to choose action directed at changing a certain situation. This predictor for disaster preparedness and resilience is mediated by the factor ‘response efficacy’. Response efficacy describes people’s perception of the availability of required resources and the perceived benefits of adoption. It is the perceived effectiveness of the recommended response to avoid the threat, in this case the perceived effectiveness of preparing yourself for a terrorist attack to avoid the (negative) consequences of an attack (Thrasher et al., 2016).
Problem-focused coping may predispose people to fix a problem, however these benefits may not be realised if people do not think they have the resources to do so (Paton, 2003).
2.4.3 Linking intention and preparation
Intention is not all. There are certain factors that affect whether intentions are converted into actual (preparedness) behaviour or not. The factors are: sense of community, perceived responsibility, timing of hazard activity, normative factors and, once again,
response efficacy. Sense of community stands for how attached an individual feels to certain people or places. People with strong feelings of attachment are more like to convert intentions into actual behaviour of preparedness. The extent to which someone feels responsible for their own safety can work in a similar way (Duval & Mulilis, 1999; Paton, 2003).
Disasters and crisis are often unpredictable and infrequent. This makes that the beliefs regarding the expected timing of the next damaging hazard event could moderate the relation between intention and behaviour. The longer no disasters take place, the less likely people perceive the urgency to act on their intentions.
Lastly, the intentions may be moderated by the normative factors or beliefs. These factors can reflect actual experience, but can also reflect beliefs and perceptions that are formed as a result of experience of others, for example trough interaction with others or the use (social) media. Two important aspects of the normative beliefs are trust in authorities or information sources and participation or empowerment in community issues (Paton, 2003).
16
2.4.4 Preparedness behaviour
A person can be prepared for a crisis in many different ways. Preparedness behaviour can be differentiated in regard to whether the behaviour calls or individual of collective action (Paton, 2003). In this study, only individual behaviour or action will be taken into account. These are actions one can do and does himself, without depending on others. When limiting the different kinds of behaviour to individual behaviour only, actions that require community participation or actions that require a political process can be disregarded (Paton, 2003). For example, having a WhatsApp neighbourhood watch requires (a part of) the neighbours to be involved, and it can there not be seen as individual behaviour.
It is difficult to say if taking preparedness actions means you are prepared for (the effects of) terrorism. How many actions should an individual take to be prepared? And how well does he have to perform these actions? Is it possible to be fully prepared for a terrorist attack? This study therefore only goes into the subjective preparedness level, as further described in the methodology chapter. It is assumed that someone who has taken more preparedness actions, is more prepared than someone who has performed less actions. The objective preparedness is left aside.
Figure 2.1
17
Figure 2.2. Theoretical framework for this study, adapted from Paton (2003).
2.5 Hypotheses
This study contains four hypotheses. The first hypothesis entails the relation between the various motivation variables and the outcome expectancy.
Hypothesis 1: There is a postive relation between motivation (consiting of critical awareness of hazards, risk perception and hazard anxiety) and outcome expectancy.
Hypothesis 1a: The relation between hazard anxiety and outcome expectancy is either postive or negative.
The hypothesis is visualized in Figure 2.3.
Hypothesis 2: There is a postive relation between outcome expectancy and self-efficacy.
The hypothesis is visualized in Figure 2.4.
Hypothesis 3: There is postive relation between self-efficacy and intentions. The hypothesis is visualized in Figure 2.5.
18
Figure 2.3 Visualization of Hypothesis 1 and Hypothesis 1a
Figure 2.4 Visualization of Hypothesis 2
Figure 2.5 Visualization of hypothesis 4
It is important to note that the hypotheses state relation, not causation. Although the theoretical model assumes there is a causal relation between the independent and dependent variables, it is not possible to statistically test for causality with this type of data.
Hypothesis 4: There is a positive relation between intentions and preparation. This relation is moderated by perceived responsibility, sense of community, timing of hazard acitivity, normative factors and response efficacy.
19
Figure 2.6. Visualization of Hypothesis 4
3. Methodology 3.1 Research design
To examine to which extent the factors of Panton’s model explain the individual crisis preparedness for a terrorist attack, a quantitative, cross-sectional, large-N study will be conducted using primary data.
Paton’s model is a theoretical model that is difficult to test statistically. First, the model has several different pathways, and a lot of different variables with various correlations and causations with each other. This makes it impossible to perform one statistical test to test or analyse the entire model. However, it is statistically not right to test one model or
hypothesis with more than one statistical test. It is therefore decided to divide the model into four parts, which will be tested and analysed separately from each other. This way,
conclusions can be drawn about the separate parts, but it is not possible to say something about the relation between the different parts of the model. Second, the model assumes causal relations between several of the variables. It is however not possible to statistically test for causality, specifically when the data is cross-sectional. Only in the case of time series, a series of data points indexed in time order, it is possible to say something about causality, although this method has its flaws as well. Moreover, data that come from questionnaires and other such methods, are observational and correlational, and cannot be used to determine causality.
The first part of the model that is tested, is the relation between the motivation variables, being critical of hazards, risk perception and hazard anxiety, and the outcome expectancy. The second part of the model is the relation between outcome expectancy and self-efficacy and the third part about the relation between self-efficacy and intention. Lastly,
20 in the fourth part, the relation between intention and preparedness behaviour is tested. The moderating variables mentioned in the model, being perceived responsibility, sense of
community, timing of hazard activity, normative factors and response efficacy, are taken into account as well.
Since the model is quite challenging to test statistically, it is decided to focus on the main path of the model, leading from the motivation variables, to outcome expectancy and self-efficacy, to intentions and finally to preparation. Problem-focused coping and response efficacy, which are part of the intention formation process, are not taken into account.
3.1.1 Variables
The overarching dependent variable in this study is the individual’s crisis preparedness for a terrorist attack. The independent variables are the variables mentioned in Panton’s model: critical awareness of hazards, risk perception, hazard anxiety, outcome expectancy, self-efficacy, intentions, perceived responsibility, sense of community, timing of hazard activity, timing of hazard, and two normative factors, being trust in authorities and participation or empowerment in community issues. However, as stated above, it is not possible to test the model in this way. The model is divided into different parts, and each part has its own dependent and independent variables, which are explained in more detail in the paragraphs on data analysis.
Two variables of Paton’s model are left out, being problem-focused coping and response efficacy, since the focus lays on the main path of the model.
3.2 Data collection
To collect the necessary data, an online questionnaire will be held. Online
questionnaires are a feasible and efficient way to reach a large amount of people in a short amount of time. For example, when using an online questionnaire, it is possible to reach different people in a short amount of time, despite possibly being separated by geographic distances. Moreover, it is cost efficient since it costs (almost) nothing to set up and spread an online questionnaire, especially in contrast to the traditional offline questionnaire (Wright, 2005). Using an online questionnaire is also good for the reliability of the study. Participants seem to trust and appreciate the anonymity that comes with addressing a computer screen rather than the researcher directly, and participants are more likely to fill in the survey truthfully (Rudestam & Newton, 2007).
21
3.3 Participants
The target population are people of the age of eighteen or older, living in the Netherlands. Since the questionnaire is in English, this will limit the study population to English speakers. It is important that the study population is a good representation of the target population, the large N will contribute to this. Ideally random sampling would be the best way to ensure that the study population represents the target population. Random sampling means that every member of the target population has an equal chance of being selected for the sample (Nolte, Shauver, & Chung, 2015). It is however not feasible to obtain a truly random sample. Participants will be reached via different social media platforms. Facebook will be the main platform to spread the survey. The survey will be posted on the personal page of the researcher, as well as in different Facebook groups to reach a wide variety of people, and not just the personal circle. To reach more people, the questionnaire can also be shared by others on their timeline. This way, a snowball effect will take place, in which the respondents will be used to find new respondents, to obtain a sufficient number of participants. The survey can also be send directly to people who do not have Facebook, but this will only be done to acquainted people. Moreover, the questionnaire will be shared in WhatsApp groups to reach even more people.
When calculating how many respondents is sufficient, the size of the target population, the confidence level and the margin of error are important, and can be used to calculate what number of respondents is sufficient. With a confidence level of 90%, a margin of error of 5% and a target population that is greater than 20000, 269 respondents would be sufficient (Qualtrics, 2019).
Apart from the questions about preparedness behaviour, motivation and intention formation, the respondents will also be asked about gender, age, ethnicity, highest level of education, annual income, form of employment and whether they live in a city or village, to get an idea of the study population.
3.4 Operationalisation
The concepts that need to be operationalised are: critical awareness of hazards, risk perception, hazard anxiety, outcome expectancy, self-efficacy, problem-focused coping, response efficacy, intentions, perceived responsibility, sense of community, timing of hazard activity, normative factors and the eventual crisis preparedness. Some of concepts will be
22 operationalised in more than one question, others will be operationalised in just one question. All questions will have the possibility to answer ‘do not know’. The full questionnaire is presented in the appendix.
Critical awareness. To measure the variable critical awareness, participants are asked
how often they think or talk about the possibility of a terrorist attack happening in the Netherlands. This is measured using a scale variable. The answers possibilities are: daily, frequently, sometimes, almost never and never.
Risk perception. Risk perception is measured using four different scale questions. The
first question is: ‘’How high do you perceive the risk of a terrorist attack happening in the Netherlands in the near future?’’. The answer possibilities range from ‘’really high risk’’ to ‘’no risk’’. The second question asks how the respondent feels about the possibility of a terrorist attack happening in The Netherlands. The answers possibilities are: ‘’really
worried’’, ‘’worried’’, ‘’somewhat worried’’ and ‘’not worried’’. The third question is: ‘’How likely do you think it is that the risk of a terrorist attack in the Netherlands will affect future generations (e.g. economically)?’’. The answer possibilities range from ‘’really likely’’ to ‘’not likely’’. The fourth and last question to operationalise risk perception asks whether the respondents thinks the risk of a terrorist attack happening in the Netherlands is decreasing, increasing or staying the same. The mean score is calculated and used in the analysis (Espina & Teng-Calleja, 2015).
Hazard anxiety. To measure hazard anxiety, respondents are asked how afraid they are
of a terrorist attack happening in the Netherlands (Laat, 2009). The answer possibilities are: ‘’really afraid’’, ‘’afraid’’, ‘’somewhat afraid’’ and ‘’not afraid’’.
Outcome expectancy. Outcome expectancy is measured by asking respondents how
they think preparing themselves will mitigate the effects of the terrorist attack. The possible answers range from ‘’really mitigates the effects’’ to ‘’does not mitigate the effects’’ (Paton, 2003).
Self-efficacy. Self-efficacy is measured using one question: ‘How capable do you think
you are in preparing yourself (sufficiently) for a terrorist attack?’’. The possible answers range from ‘’really capable’’ to ‘’not capable’’ (Paton, 2003).
23
Problem-focused coping.1 Problem-focused coping is a difficult concept to
comprehend in one question or a few questions. Most of the time, there is a whole
questionnaire dedicated to just this concept (Mead, 2020; Carver, 1997). The COPE inventory is a method that is often used to measure coping strategies (Carver, Sheier, & Weintraub 1989). However, this inventory is really long and is therefore not feasible for most research, including this one (Carver, 1997). Carver (1997) made a new method, based on the COPE inventory, to make it more feasible to measure coping mechanisms. The Brief COPE is intended to encourage a wider examination of coping in naturally occurring settings. The instrument does not need be used in an all-or-none fashion. Researchers who have a focused interest, or who have extreme time demands, can selectively use the scales that are of most interest to their study (Carver, 1997). Therefore, the problem-focused coping is be measured using only the active coping scale. This scale consists of two statements: ‘’When I have a problem, I concentrate my efforts on doing something about it’’ and ‘’In the past, I’ve taken action to try to make a bad situation better’’. The respondents are asked if the statement characterizes them. If this is the case, they should answer ‘’yes’’, if not, they should answer ‘’no’’.
Response efficacy. Response efficacy exists of different aspects and is therefore be
measured with two different questions, both with ‘’’yes’’ or ‘’no’’ answers. The questions are: ‘’Do you think you have the right resources (e.g. money, access to internet, storage space) available to you to prepare for a terrorist attack?’’ and ‘’Do you think it is useful to be
prepared for a terrorist attack?’’ (Paton, 2003; Trasher et al., 2016).
Intentions. An intention is something someone wants and plans to do. The question is
therefore twofold, ‘’Do you want to prepare yourself for (the effects of) a terrorist attack, in case one happens?’’ and ‘’Do you plan to prepare yourself (even more) for (the effects of) a terrorist attack, in case one happens?’’. Both are ‘’yes’’ or ‘’no’’ questions, however the last question can also be answered with: ‘’I have already prepared myself for a terrorist attack’’ (Paton, 2003).
Perceived responsibility. Perceived responsibility is measured by two questions:
‘’How responsible do you feel responsible for your own safety?’’ and ‘’How responsible do
1 The operationalisation of the concepts was done before it was decided to focus on the main line of the theoretical model.
Therefore, ‘problem-focused coping’ is still part of the operationalisation and the questionnaire. For the sake of completeness, the operationalisation is still included.
24 you feel for the safety of others (e.g. children, partner, parents, friends)? ‘’. The answers range from ‘’really responsible’’ to ‘’not responsible’’ (Paton, 2003).
Sense of community. Sense of community will be measures with the question: ‘’ How
attached do you feel to your community or neighbourhood?’’. Possible answers are: ‘’really attached’’, ‘’attached’’, ‘’somewhat attached’’ and ‘’not attached’’ (Paton, 2003).
Hazard timing. Hazard timing will be measures using one question: ‘’When do you
expect the next terrorist attack to happen in The Netherlands?’’. The answer possibilities are: ‘’this year’’, ‘’in the coming years, but not this year’’, ‘’this decade, but not in the coming years’’, ‘’after this decade’’, and ‘’never’’.
Normative factors. The most important normative factors are trust in authorities or
information sources and the participation or empowerment in community issues. Only those two factors will be taken into account. The questions that will be asked, are: ‘’How much do you trust the national authorities to handle a terrorist attack well?’’ and ‘’How actively do you participate in your community or neighbourhood when there are (community) issues?’’. For the first question, the possible answers range from ‘’really trust national authorities’’ to ‘’do not trust national authorities’’. For the second question, the answers range from ‘’really active’’ to ‘’not active’’ (Paton, 2003).
Crisis preparedness. There are many ways in which an individual can prepare for a
crisis, or in specific a terrorist attack. Therefore, crisis preparedness will be measured using one question with different answer possibility. Preparatory actions enable a person to make the appropriate responses in the event of an attack (Davis, LaTourette, Mosher, Davis, & Howell, 2003). A general way of preparing yourself is thinking or talking to someone about how you would act during a terrorist attack. Therefore, one of the answer possibilities is: ‘’Thought or talked about what I would do during a terrorist attack’’ (Paton, 2003). According to Davis et al. (2003) there are four steps of preparatory actions when it comes to preparing for a chemical, radiological, nuclear of biological terrorist attack. The first step is to gain understanding about the different types of terrorist attack. The second step is to facilitate response actions by making (communication) plans and gathering information in advance, for example information about potential shelter or building evacuation plans. The third step is to ensure an emergency kit. The last step is to enhance protection in more passive ways, for example by weatherizing your home or installing particulate filters (Davis et al., 2003). These steps are included in answer possibilities in the following way: ‘’learned to
25 recognize characteristics of different typed of terrorist attacks’’, ‘’made a (communication) emergency plan with family’’, ‘’made (communication) emergency plan with others’’, ‘’built or bought an emergency kit’’, ‘’enhanced home protection’’, ‘’discovered building evacuation plans’’, and ‘’discovered potential shelter in buildings’’.
The Dutch government has a website on how to be prepared for a crisis and how to act during a crisis, with special attention to terrorism (Nationaal Coördinator
Terrorismebestrijding en Veiligheid, 2020a; Nationaal Coördinator Terrorismebestrijding en Veiligheid, 2020b). Therefore, included in the answer possibilities are: ‘’Read the
Government's website about what to do during a terrorist attack’’ and ‘’ Read the
Government's website about what to do during a crisis’’. To make it more comprehensive, also the answer possibility ‘’Searched for (other) information about how to prepare for a terrorist attack’’ was added. On this website, there are mentioned multiple actions in specific, most of which are already covered. A preparatory action that is mentioned that is not yet covered is the installation of the NL-alert. NL-Alert is a cell broadcast alarm system that allows the authorities to quickly alert and inform citizens of certain dangerous of crisis situations. Authorities can send messages to all users or the users in specific areas by using specific cell towers to alert phones within their reach (Nationaal Coördinator
Terrorismebestrijding en Veiligheid, 2020c).
Another way of preparing yourself for a crisis that is mentioned in a lot of different studies is to buy extra (food) supplies (Eisenman et al., 2006). There are also answer possibilities that cover insurances with terrorism coverage: ‘’Insured consciously at an insurance company affiliated with the NHT (Nederlandse Herverzekeringmaatschappij voor Terrorismeschaden)’’ and ‘’Took a cancellation insurance with terrorism coverage’’.
Lastly, there is the possibility to answer ‘’other…’’, which obliges the respondent to fill in which other action(s) they have taken. There is also the option to answer that you have taken no action. The score will be generated by counting the number of ticked answers (Espina & Teng-Calleja, 2015). It is important to note there is are discrepancies between people’s expectation of their preparedness and their actual preparedness (Ballantyne, Paton, Johnston, Kozuch & Daly, 2000). Because of this, respondents will not be asked whether they are prepared for a terrorist attack, or how much they are prepared for a terrorist attack, but they will be asked to tick all the actions they have taken to prepare themselves. It is tried to make the list as extensive as possible.
26
3.5 Data preparation
Before the data can be used in the analysis, it was necessary to do some data preparation. First, ‘risk perception’, ‘response efficacy’, ‘intentions’, ‘problem-focused coping’, ‘perceived responsibility’, and ‘normative factors’, were operationalised using more than one question and were therefore converted into multiple variables. These variables were transformed into one variable. These new variables are used in the analysis.
Second, the results of the analyses are easier to interpret when a value of 0 stand for ‘no’ and a value of higher than 0 stands for something more than ‘no’. This was not the case in the original data, expect for ‘preparedness’. All other variables were recoded in a way that the values range from low to high.
Third, in the questionnaire the respondents were asked to tick off all measures they had taken to prepare themselves for a terrorist attack. In SPSS, all these option were single dichotomous variables with a ‘yes’ or ‘no’ answer. All these variables are transformed into one variable for ‘preparedness’. The more measures someone had ticked off, the higher the score for preparedness. The score for ‘preparedness’ was therefore a score between 0 and 15.
3.6 Data analysis
The collected data is statistically analysed using the program SPSS Statistics (version 26). First, an univariate analysis is performed to describe and characterize the sample on the basis of the demographic questions of the questionnaire. Subsequently, the different parts of the theoretical model are analysed separately. The method of analysis and the corresponding assumptions are discussed for each of the parts.
3.6.1. Method of analysis for hypothesis 1
The first part of the model, the relation between the several motivation variables and outcome expectancy, is analysed using a multinomial logistic regression. The multinomial logistic regression is used to predict an ordinal dependent variable given one or more independent variables (Leard Statistics, 2020a). In this case, the dependent variable is
‘outcome expectancy’, which has an ordinal level of measurement. The independent variables are ‘critical awareness’, ‘risk perception’ and ‘hazard anxiety’, which also have an ordinal level of measurement. It is preferred for the independent variables to have a ratio or interval level of measurement however the regression is also feasible when the independent variables have a ordinal level of measurement. It is important to note that having ordinal independent
27 variables instead of interval or scale variables, changes the way the outcome of the regression has to be interpreted. For ordinal independent variables, it is possible to interpret the odds that one group has a higher or lower value on the dependent variable, compared to the second (redundant) group (Leard Statistics, 2020a).
There are two other methods of analysis that are often used to analyse relations in ordinal data. The most obvious method would be to use an ordinal logistic regression. Initially, it was decided to use this method. One of the most important assumptions of the ordinal logistic regression is the assumption of proportional odds. Although this is one of the most important assumptions, this assumption is almost never met with real life data, the data of this study being no exception (Grace-Martin, 2020; UCLA, 2020). The test of the
proportional assumption has been described as “anti-conservative, that is it nearly always results in rejection of the proportional odds assumption, particularly when the number of explanatory variables is large, the sample size is large or there is a continuous explanatory variable in the model’’ (O’Connell, 2006, p29). Since the problem with the proportional odds assumption occurs so often, there are several solutions for this. However, these solutions make the analysis of the data even more complicated and are often specifically useful when the assumption was not met by a little. In this case, the assumption was far from being met. Therefore, it is decided not to try and solve the problem of proportional odds but rather use another method of analysis.
A second method that is commonly used for analysing ordinal data is transforming the ordinal variables into dummy variables and perform a (multiple) linear regression. Dummy variables are dichotomous, quantitative variables that represent the categorical data. The range of values of a dummy variable is small and regression results are easiest to interpret when dummy variables are limited to two specific values, 1 or 0. The possibility of computing dummy variables and performing a linear regression was considered, since this often makes it easier to analyse the outcomes. However, the number of dummy variables that is needed to represent a particular categorical variable depends on the number of values that the
categorical variable can assume. To represent a categorical variable that can assume k different values, there need to be k - 1 dummy variables (Stattrek, 2020). In this case, there would be four variables that needed to be transformed into dummy variables, with a total of twenty-six values. The large amount of dummy variables would make the analysis less feasible and less clear than the analysis of an ordinal logistic regression. It is therefore decided to perform the ordinal logistic regression instead of the multiple linear regression.
28 The multinomial logistic regression makes several assumptions. The first two
assumptions are about the level of measurement of the dependent and independent variables. The first assumption is that the dependent variable is measured at nominal level, or at ordinal level if the ordinal logistic regression is not a valid option. The second assumption is that the independent variables continuous, ordinal or nominal. It is important to note that ordinal independent variables must be treated as being either continuous or categorical. They cannot be treated as ordinal variables when running an ordinal logistic regression in SPSS. The third assumption entails that there has to be independence of observations and that the dependent variable should have mutually exclusive and exhaustive categories. The fourth assumption concerns the multicollinearity of the independent variables. It is assumed that the independent variables are not highly correlated with each other. The fifth assumption assumes a linear relationship between any continuous independent variables and the logit transformation of the dependent variable. Lastly, there should be no outliers, high leverage values or highly
influential points.
3.6.2 Method of analysis for hypotheses 2 and 3
The second part of model is about the relation between outcome expectancy and self-efficacy. In this case, the independent variable is outcome expectancy and the dependent variable is self-efficacy. Both these variables have an ordinal level of measurement. Spearman rank correlation is used to analyse the relation between the variables. Spearman rank
correlation is a non-parametric test that is used to measure the degree of association between two variables. The test holds no other assumptions aside from the level of measurement, which needs to be at least ordinal for both of the variables (StatisticsSolutions, 2020).
The third part of the model goes into the relation between ‘self-efficacy’ and
‘intentions’. The analysis for this part of the model is the same as the analysis for the second part of the model, meaning that the Spearman rank correlation test is used for the analysis. In this case, the independent variable is ‘self-efficacy’ and the dependent variable is ‘intentions’. Both variables have an ordinal level of measurement and thus are confirm the assumption.
3.6.3 Method of analysis for hypothesis 4
The last part of the model is the most interesting part, the relation between intentions and preparedness behaviour, moderated by perceived responsibility, sense of community, timing of hazard activity, normative factors and response efficacy. A linear regression is used
29 to analyse this relation. A linear regression allows a dependent variable, to be modelled as a linear function of one or more independent (predictor) variables (Calabrese, 2019). The dependent and the independent variable have to have an interval or ratio level of
measurement. Since de independent variable ‘intentions’ is an ordinal variable, it is necessary to create dichotomous dummy variables, which also can be used in a regression analysis. The variable ‘intentions’ can assume three different values. As stated before, to represent a
categorical variable that can assume k different values, there need to be k - 1 dummy variables (Stattrek, 2020). In this case, there is a need for two dummy variables. For the values ‘high intention’ and ‘medium intention’ a dummy variable has been created.
First, the regression is performed without the moderator variables. Afterwards, the moderator variables are taken into account. To do so, the extension PROCESS is added to SPSS. This extension allows the user to add up to two moderator variables into a linear regression, also called a moderator analysis. A moderator analysis is used to determine whether the relationship between an independent and a dependent variable depends on, is moderated by, the value of a third variable. It is not possible to test for more than two
moderator variables at once, since this would blur the effects a moderator variable has on the dependent variable. It would not be clear which moderator has which effect. Moreover, it should first be tested if the variable has a moderating effect at all. It is therefore decided to test the potential moderator variables one by one.
Another method that is commonly used to test for moderating variables is the addition of an (linear) interaction term in a multiple regression model. This method is however not suitable for this study, since it is necessary for creating an interaction term that the variables are continuous and not categorical (Leard Statistics, 2020b) This is not the case, and therefore it is decided to use the PROCESS extension, which does not hold such assumptions.
The linear regression analysis makes several assumptions. The first assumption is existence of the linear relationship. There must be a linear relation between the independent variable and dependent variable. Second is the assumption of multivariate normality. It is assumed that the residuals are normally distributed. Third, the multicollinearity of the independent variables. It is assumed that the independent variables are not highly correlated with each other (StatisticsSolutions, 2020). Fourth, homoscedasticity of the independent variables. This assumption states that the variance of error terms is similar across the
30 autocorrelation in the data. Autocorrelation occurs when the residuals are not independent from each other.
3.7 Procedure
The questionnaire was made using the online tool Qualtrics. This tool can be used for free when you are a university student. Qualtrics allows you to export the results to programs like SPSS. The questions in the questionnaire are obligatory. This prevents respondents from answering the survey partly and handing it in.
At the beginning of the questionnaire the potential respondents were informed about the research and their participation in this research. Most importantly, they were informed that participation in this research is voluntary, that their data will be handled confidentially, and that they have the right to stop the questionnaire at any moment without further consequences. Also, a short explanation was given to explain more about what being prepared means, since it often entails more or different actions than people think (Laat, 2009). Lastly, the contact details of the researcher were given. If the respondent continued, he or she agreed with the following statement: ‘’… you acknowledge that your participation in the study is voluntary, you are at least 18 years of age and that you are aware that you may choose to end your participation in the study at any time.’’.
The distribution of the questionnaire was done via Facebook and WhatsApp. The link to the questionnaire was posted on the personal Facebook page of the researcher and in several different Facebook groups. The network of the researcher was used to distribute the questionnaire even further, since different ‘Facebook friends’ decided to share the post. This way the audience not only grew, but also targeted a larger part of the target population.
Once enough respondents have filled out the questionnaire, the questionnaire was closed. The questionnaire was online from 26th of April, 3:45 PM until the 1st of May, 11:00 AM. As said before, the analysis of the results was done using the statistics program IBM SPSS Statistics 26.
3.8 Limitations
When assessing the limitations of this research, both reliability and validity need to be taken into account. The different limitations will therefore be discussed in light of these two concepts. First, the concept ‘reliability’ will be explained, after which the limitations related
31 to the reliability will be discussed. Subsequently, the same will be done with the concept ‘validity’.
3.8.1 Reliability
Reliability is concerned with the consistency of the results. Reliability entails that the results of the research should depend on chance as little as possible (Baarda & De Goede, 2001). The question that should be answered is: ‘’To what extent will the research get the same result if it is repeated at another date or place?’’ (Korzilius, 2000).
First, the reliability of the study is enlarged by the large number of respondents, 236 in total. The bigger the total number of respondents, the greater the chances are that the results do not depend on chance. Moreover, it is tried to reach a diverse audience that fully represents the target audience.
Second, most of the theoretical concepts that need to be operationalised are complex and complicated. This make the operationalisation a challenge and might reduce the reliability of the study. It is important that, if the research is replicated, the operationalisations stand the test of time. To do so, the operationalisations of the different variables is based on other scientific research.
Third, the limited guarantee of accuracy can also decrease the reliability (Korzilius, 2000). Respondents are able to complete the survey more than once, they can end the survey before all the questions are answered or they can lie while filling out the questionnaire. This is unavoidable. However, certain measures are taken to try and avoid this. First, it is checked if the questionnaire is not filled out more than one time from the same IP-address. Second, questionnaires that are only filled out partly are left out of the analysis. Third, all questions are made obligatory to answer, so it is not possible for respondents to skip a question, which would result in (unnecessary) missing data. Fourth, at the beginning of the questionnaire it is stated that it only takes a few minutes to complete the questionnaire and anonymity is
guaranteed. However, where most people are more likely to answer the questions truthfully in an online survey, others might also feel less pressure to tell the truth when the answers are online and anonymous.
3.8.2 Validity
Validity entails the extent to which the survey measures what it intends to measure (Adcock & Collier, 2001). It is the relation between the survey responses and the reality
32 (Nolte, Shauver, & Chung, 2015). Validity roughly exists of two kinds, external validity and internal validity. External validity describes the possibility to generalise the results to other people, place or times.
Important for high external validity is that the study population is representative for the target population. This is tried to accomplish by the somewhat random sampling. However, this will not be perfect because of the biases that come with online surveys. First, the selection bias. With an online survey, there will always be certain people more prone to complete the survey than others, for example elder people are less likely to be on the internet and thus fill in the survey. This creates a study population that is not representable for the entire target population. Second, the questionnaire is spread within the personal circle, once again this might create a bias towards younger, higher educated, female respondents. It is tried to avoid this by letting other share the questionnaire as well, to reach a wider and more
diverse audience.
Second, another limitation is the possibility of socially desirability bias. People might think that answering a question a certain way will put them in a favourable light. For example, saying they took certain measures to prepare themselves for a terrorist attack even if they did not. To prevent or at least minimize this bias, the anonymity and the confidential nature of the survey will be emphasized (Nolte, Shauver, & Chung, 2015).
Third, it is possible that the preparedness measures that respondents have taken, are taken coincidentally or for another purpose than preparing themselves for the effects of a terrorist attack. For example, buying extra (food) supplies is a way to prepare yourself for a terrorist attack, however it can also be done with other intentions or goals.
Lastly, single-item self-assessment of the different concepts might not fully capture the intended concepts (Eisenman et al., 2006). To counter this, several concepts are
operationalised in two or more questions. However, it is not desirable to make the
questionnaire too long, since this will increase the chance of a respondent stopping halfway through the questionnaire. Therefore, only concepts of which it is thought to be real
necessary, are operationalised using more than one question.
Internal validity describes the extent to which conclusions about a causal relationship between the variables can be made (Korzilius, 2000). Gay and Airasian (2000, p. 345) describe it as: ‘’the condition that observed differences on the dependent variable are a direct
