The Impact of Negative Economic Prospect on
Municipal Counselors’ Policy Preferences and Risk
Perception
Evidence from Switzerland
Public Administration Master thesis by: Supervised by:
Marine MARTY Dr. Joris VAN DER VOET
Student Number: s2621096 Academic Year: 2019/2020
Table of content
1 Introduction ... 5
1.1 Research design and research framework ... 6
1.2 Data collection and method ... 7
1.3 Academic and societal relevance of the research ... 7
1.4 Content of the thesis ... 8
2 Theory ... 9
2.1 Independent variable ... 9
2.2 Dependent variable I: policy preference ... 10
2.2.1 The typologies of the State and administrative values ... 10
2.2.2 Hypothesis I ... 11
2.3 Dependent variable II: risk perception ... 11
2.3.1 The perception of risk theory and the prospect theory ... 12
2.3.2 Hypothesis II ... 12
2.4 Moderating variable: Risk perception ... 13
2.4.1 The performance feedback theory: ... 13
2.4.2 Hypothesis III ... 14
3 Data and Methods ... 15
3.1 Research framework ... 15
3.2 Research design ... 16
3.3 Unit of analysis and sample ... 16
3.4 Structure of the survey-experiment ... 17
3.4.1 Operationalization of the independent variable ... 18
3.4.2 Operationalization of the dependent variable I: policy preference ... 18
3.4.3 Operationalization of the dependent variable II: risk ... 19
3.5 Statistical tests ... 19
4 Analysis ... 22
4.1 Descriptive statistics of the data ... 22
4.2 Dependent variable 1 - Results ... 23
4.2.1 Preference for “directive state” ... 24
4.2.2 Preference for “hollow state” ... 24
4.2.3 Preference for “local-communitarian state” ... 24
4.2.4 Preference for “coping state” ... 24
4.2.5 Analysis of the results for the policy preferences ... 24
4.3 Dependent variable 2 - Results ... 27
4.3.1 Risk perception of the “directive state” ... 27
4.3.2 Risk perception of the “hollow state” ... 27
4.3.3 Risk perception of the “local-communitarian state” ... 27
4.3.5 Analysis of the results for risk perception ... 27
4.4 Moderator variable - Results ... 29
5 Conclusion ... 32
5.1 Limitations of the study ... 33
List of Figures & Tables
Figure 1: Research framework ... 7
Figure 2: Research framework with a moderator variable ... 15
Figure 3: Repartition of policy preferences between the experiment groups ... 25
Figure 4: Repartition of risk perception between the experiment groups ... 28
Figure 5: Repartition of policy preference and risk perception ... 30
Table 1: Demographic characteristics of the experiment groups ... 23
Table 2: Independent t-test results for policy preferences and risk perception ... 26
1 Introduction
“The virus, of course, is not comparable to world wars. But it has a strong impact on our lives. And it spreads great uncertainty. Perhaps it is precisely this uncertainty that is hard for us Swiss to bear. Because we are an island of stability in the midst of a historically tumultuous continent.”1
(Alain Berset, minister of interior of Switzerland, 2020)
For Lodge and Hood (2012), there are four ways in which a state can react to a financial crisis regarding the provision of public services. It can either expand (i.e. “directive state” option), outsource to the private sector (i.e. “hollow state” option), rely on the community and local organizations (i.e. “local-communitarian state” option) or reduce the provision of public services (i.e. “coping state” option). Uncertainty comes hand in hand with a financial crisis and decision-makers need to choose one of the options without being able to foresee the outcomes. The risk they associate with each of the policy, which can be defined as their perception of the probability that a policy would result in negative occurrence (Paek et al., 2017), must play an important role in the decision-making process. Because municipal counselors are at the front line in dealing with the effects of an economic crisis (Pauchard, 2020), we must take the opportunity to understand how they would tend to cope with this uncertainty, what their preferred policy option would be and how their risk perception and policy preference interacts with each other. Knowing whether our decision-makers anticipate a crisis and how they would tend to react and cope with the situation would advance the state-of-the-art of Public Administration and would enable societies to better anticipate any forthcoming crisis.
The main purpose of this thesis is to examine the interaction between municipal counselors’ risk perception, or which policy they deem risky, and their policy preferences. We expect that, in the prospect of an economic crisis, decision-makers would prefer the policy option which they associate a higher risk with. To investigate this claim, we first test which of the four options (“directive state”, “hollow state”, “local-communitarian state” and “coping state”) municipal counselors prefer in the context of a forecasted economic crisis. It must be highlighted here that there is no theoretical framework that enables us to predict which of the four options would be preferred by municipal counselors. In a second step, we test what level of risk they associate with those options. Based on existing theories, we expect municipal counselors to associate a lower risk with the “coping state option” in the context of a forecasted economic crisis than under normal circumstances. In overall, the thesis revolves around the two following questions: To what extent does the predicted economic crisis due to COVID-19 influence Swiss municipal counselors’ policy preferences and risk perception?
To what extent does Swiss municipal counselors’ perception of risk moderate the influence of the predicted economic crisis due to COVID-19 on their policy preferences?
1This sentence is retrieved from a speech that was given on the 1st of August 2020, on the occasion of the Swiss national Day, by the minister of interior of Switzerland in charge of handling the COVID-19 crisis. The original version is in Italian: “Il virus, ovviamente, non è paragonabile alle guerre mondiali. Però ha un forte impatto sulle nostre vite. E diffonde una grande incertezza. Forse, è proprio questa incertezza ad essere difficile da sopportare, per noi Svizzeri. Perché siamo un’isola di
In contrast with the performance feedback theory that analyzes the way decision-makers react to poor past performance, the focus is on the future, on predicted performance. The COVID-19 pandemic and the subsequent forecasted economic crisis offer a unique opportunity to answer these questions. Indeed, due to the outbreak of the virus, government were urged to implement drastic measures, from full to partial lockdowns, putting the world economy on hold. The employment rate, the public debt, and the GDP growth, among other indicators, are highly jeopardized by this crisis and the economic impacts are predicted to be long lasting (OECD, 2020a). We then use this special context to test how municipal counselors react to the negative economic prospect caused by the corona crisis and if their perception of risk is embedded in the projection.
1.1 Research design and research framework
To conduct the study, we choose to use a survey-experiment. We reached out to municipal counselors of five European countries: France, the Netherlands, the United Kingdom, Spain and Switzerland to invite them to participate in our study. This thesis focuses on the lattermost case.
A survey-experiment enables to randomly assign respondents to a treatment or control group. The treatment group receives an informative text about the predicted economic effects of the COVID-19 on the municipal level, whereas the control group receives an informative text about the currently known health effects of the virus. The two groups are then asked to state their policy preference, based on the terminology of the State proposed by Lodge & Hood (2012). Afterwards, the respondents are asked to scale the risk they associate with each of these policy options from 0 (risk null) to 10 (extremely risky).
Relying on a survey-experiment design offers many advantages. It enables to make use of the internal validity of experimental research and to combine it with the external validity of surveys (Schlueter & Schmidt, 2010). Moreover, through surveys, a considerable amount of data can be gathered, which offers a large sample. Additionally, by providing respondents with complete anonymity, we can be reasonably confident that, on average, the results reflect municipal counselors’ true opinion.
Figure 1 below illustrates our research framework. We utilize the COVID-19 economic crisis to give municipal counselors of the treatment group negative economic prospect (IV). The first goal is to measure the impact of receiving negative economic prospect (IV) on their policy preferences (DV1). As we ask respondents to distribute 100 points based on their preferences between four policy options (“directive state”, “hollow state”, “local-communitarian state” and “coping state”), our first dependent variable unfolds in four subcategories (arrow 1 of figure 1). In a second step, we evaluate whether the negative economic prospect (IV) affect the way respondents perceive risk (DV2). In other words, we test whether the control group and the treatment group associate the same level of risk with the four policy options (arrow 2 of figure 1). Finally, the goal of the third analysis is to determine whether the risk perception moderates the impact of receiving negative economic prospect on the policy preferences. This can only be done if the independent variable and the risk perception are not correlated. Thus, two outcomes need to be distinguished here to understand how the third and final analysis is conducted. The first one; if receiving negative economic prospect affects the risk perception, no further analysis can be conducted. The second one; if no correlation is observed, a moderation effect can be tested for (arrow 3 of figure 1).
1.2 Data collection and method
We rely on a survey-experiment to conduct this research. The link of the survey was sent to 1’434 Swiss municipalities’ general email address with a request to forward the link to the municipal counselors. On top of that, the link was sent to 495 municipal counselors’ private email address. After three weeks, 536 responses have been recorded. In total, our sample includes 264 valid answers. 142 respondents have been assigned to the control group and 122 to the treatment group.
We will measure the ATE (average treatment effect) by running independent t-tests for each of the policy option to check whether our independent variable impacts the policy preferences and the risk perception of the municipal counselors. Furthermore, to establish whether there is an interaction effect between the treatment and the risk perception of each policy, provided that the risk perception is exogeneous, a two-ways ANOVA test is run.
1.3 Academic and societal relevance of the research
On the academic level, this thesis follows up on Cepiku, Mussari & Giordano’s paper (2016) about how local governments manage austerity, i.e. which policy response is preferred by the governments and why. The conceptual framework the authors use in their research is quite similar to this thesis for many reasons. First, they also focus on municipalities. Second, they also use the four policies a state can adopt in response to austerity, as outlined by Lodge & Hood (2012), as a theoretical basis (i.e. they also use the “directive state”, “hollow state”, “local-communitarian state” and “coping state” terminology). Finally, they also utilize an economic crisis (the 2008 crisis) to operationalize their independent variable. For their research method, the authors decided to adopt a comparative case study and they collected data from 2007 to 2015 for six Italian municipalities. In overall, their research has unveiled the conditions that favor the implementation of a policy over another. For example, they established that a “greater levels of trust, good relationships with stakeholders and the community, and adequate political backing are good predictors of a strategic communitarian approach to austerity
outsource the provision of public services to the private sector as result of the crisis. Thus, this thesis widens the scope of Cepiku, Mussari & Giordan’s research by putting an emphasis on which of the four policies is preferred by council members in the prospect of a crisis and by adding a risk variable. Furthermore, by using a different research method and relying on survey-experiment results, we are able to reveal stronger effects. Additionally, this thesis, by analyzing how policymakers react to negative economic prospect, tests Shinkle’s theory according to which organizations are forward-looking and do not only consider past performance to make decisions (2012). Rather, this author states that organizations adapt their response to changes in the environment and take future prospect into account; the design of our experiment enables us to test this claim. Moreover, this thesis is a starting point for future research to test whether politicians actually implement the policy option they stated a preference for.
On the societal level, this thesis sheds light on municipal counselors’ policy preferences in times of crises. As a result of the COVID-19 pandemic, experts are confident that an economic crisis will occur, and this gives us a peculiar opportunity to explore how decision-makers anticipate and react to negative economic information. The use of a survey-experiment gives a direct insight on municipal counselors’ policy preferences and how those relate to the risk they perceive and associate with each of the policies. Furthermore, it reveals if their perception of risk is fixed or situational; do our elected officials perceive risk differently in times of crisis? Does it affect their policy preference? In overall, investigating these questions allows to better understand how policymakers would tend to handle a crisis and if their perception of risk is dependent upon the context. Thus, unveiling how our decision-makers would tend to behave enables societies to address future crisis better. Indeed, if politicians are aware of their own biases and tendencies, they are able to counteract and take better, more rational decisions. Moreover, knowing how politician would tend to handle a crisis enables voters to take better informed vote choices. In sum, there is a lot to gain in knowing how decision-makers tend to react to uncertainty.
1.4 Content of the thesis
The next section of this thesis presents all the relevant literature and theoretical framework that applies to our study and which our three hypotheses are based on. Our hypotheses are namely that receiving negative economic prospect impacts municipal counselors’ policy preference and that they would associate a lower risk with the “coping state” option. Moreover, we expect the risk perception to moderate the effect of receiving negative economic prospect on the policy preference in such a way, that municipal counselors would tend to prefer the policy which they associate a higher risk with. The third section explains the research design, the structure of the survey and the data collection. In the fourth section, the data are analyzed, and we observe that receiving negative economic prospect does not affect the policy preference of municipal counselors and we find no moderation relationship between the risk perception and the policy preference. Yet, we find that receiving negative economic prospect negatively affects municipal counselors’ risk perception of the “coping state”. As this thesis does not pretend to be fully unimpaired, the limitations of the study are presented in the conclusion section and suggestions for further research are laid out, namely, to investigate for a mediating effect between receiving negative economic prospect, the risk perception and the policy preferences.
2
Theory
In this part, we present the relevant literature surrounding the independent variable, the two dependent variables (policy preference and risk perception), and how we expect them to relate to each other. The hypotheses are then formulated.
2.1 Independent variable
By choosing “negative economic prospect” as an independent variable, we focus on the expected future performance of an organization and how it impacts decision-makers’ preferences.
Research in Public Administration has established that organizations set some desired performance level and aim at reaching these levels. Whenever they fail to meet the target, change is said to be enacted (see Greve, 1998; Greve, 2003; Cyert & March, 1963; Shinkle, 2012). This desired performance level, also referred to as “target”, “reference point”, or mostly “aspiration”, can be defined as “the smallest outcome that would be deemed satisfactory by the decision maker” (Schneider, 1992). The organization constructs its aspiration through two influences; a social influence and a historical influence. The social influence means that the organization will take peer’s performance into account and the historical influence suggests that the past performance of the organization will help determine the next desired target (Greve & Gaba, 2017).
According to Shinkle (2012), there are three frameworks through which organizational aspirations can be analyzed; the Behavioral Theory, Ansoff’s Strategic Management and
Strategic Reference Point Theory. In the Behavioral Theory, organizations are backward
looking, and they adjust their aspirations based on past experience, whereas in the Ansoff’s
Strategic Management and Strategic Reference Point Theory organizations are future-looking.
Thus, future performance expectations shape aspiration, but aspirations are also adapted in response to changes in the environment of the organization. In other words, according to this theory, negative economic prospect might influence decision-making to such extent that decision-makers are more likely to opt for an alternative course of action.
In sum, organizations set aspirations, which are desired level of performance. Municipalities set then performance targets which include economic and financial target. Understandably, having sound finances is one of the goals to reach and future economic condition would affect it directly. Municipal counselors, as agents within the organization, would also be dedicated to achieving this financial and economic balance. Hence, by facing decision-makers with negative economic prospect, we are implying that the aspirations will not be met, and we expect that decision-makers would want to enact change. We are then interested in the impact of negative prospect on municipal counselor’s policy preference and risk perception. The COVID-19 pandemic and the subsequent forecasted economic crisis offer a unique opportunity to operationalize negative prospect. As the OECD’s forecasts suggest, local government’s expenditures are highly likely to rise, while revenue would decrease. Furthermore, municipalities will have to face a surge in unemployment, municipal debt and business bankruptcies (OECD, 2020b). Thus, we analyze the impact of this information on municipal counselors’ policy preferences and risk perception.
2.2 Dependent variable I: policy preference
Our first dependent variable, the policy preference, is rooted in Lodge and Hood (2012) terminology of the state and Hood’s (1991) description of public management values. The two theories are presented and then the first hypothesis is formulated.
2.2.1 The typologies of the State and administrative values
Lodge and Hood (2012) outline four different responses a state can implement in the case of a financial crisis: the “directive state”, the “hollow state”, the “local-communitarian state” and the “coping state”.
In the “directive state”, the government provides most, or all, public services without any third party. The role of the state is then expanded and the intervention in the economy increased. In this setting, public servants are required to acquire fulfilled public managerial and organizational competences and they are rewarded for their long-term loyalty.
In the “hollow state”, the government hands over the provision of public services to the private sector. The intervention of the state in the economy is minimalized, as public services are turn into profit opportunities for private businesses. The concept of “revolving doors”, as outlined by Makkai & Braithwaite (1992), or the problems arising when civil servants leave the public sector to join the private sector, or vice versa, is no longer an issue at hand, since there is no conflict of interests anymore.
In the “local-communitarian state”, the government relies on local community and voluntary organizations to provide, or substitute, public services. Mediation and facilitation skills would be required from public servants, in order for them to ensure the existence of administratively competent organizations. Yet, the authors highlight that this method has its limitation, as it requires to have preexisting voluntary organizations that are financially stable. Plus, this organizational design might be unable to deal with large-scale and transboundary crises. Finally, the “coping state” reflects a government that shortens the provision of public services, without negotiating the take-over by any other business or organization. The number of public servants can be significantly scaled back.
We can relate these four concepts to Hood’s clusters of administrative values (1991). Hood suggests that different administrative values have different implications for fundamental aspects of administrative design. He differentiates between three sets of values in public management. The first one, sigma value, are axed toward ‘economy and parsimony’. The goals have to be met by matching the right resources at lowest cost and most efficient time. The emphasis is put on the output and success is measured in terms of money and time. The second one, theta-values, care about ‘honesty and fairness’. The process matters the most and goals need to be met by ensuring rectitude and uprightness. Finally, in lambda-values, the watchwords are ‘security and resilience’. The system needs to ensure redundancy and to be able to adapt in time of crisis.
The author points out that while those sets of values might overlap, mostly only two out of the three can be satisfied within an administrative design. We can thus draw a parallel with the four concepts of the state. For example, in the “directive state”, the ‘honesty & fairness’ and ‘security & resilience’ values can be guaranteed but at high cost whereas the “local-communitarian state” enables to save money, while impinging on the resilience and security of the system.
In sum, the first dependent variable is based on the concepts of the state, as outlined by Lodge and Hood (2012) and the study aims at evaluating whether the prospect of an economic crisis impacts the preferences of policymakers. If yes, the point is to find out which set of values, as outline by Hood (1991), they are willing to trade-off.
2.2.2 Hypothesis I
This enables us to formulate our first hypothesis. Lodge and Hood (2012) lay out that in response to a financial crisis, policymakers have four policy options (“directive state”, “hollow state”, “local-communitarian state” or “coping state”). We can then expect that policymakers who are faced with negative economic prospect will have a different preferred option than policymakers who are not given any economic information. As for which of the option will be preferred, we would be tempted to expect municipal counselors to prefer the option that displays the most efficiency, as a result of receiving negative economic prospect. But as Manzoor highlights, “efficiency in public administration is more than a technical relationship between resources and output; it has another dimension that incorporates outputs in relation to values and accountability as an inherent quality of democratic governance” (2014; p. 2) and the four options display different priorities on the kind of efficiency that needs to be achieved. There is then no theory that could enable us to predict which of the four options will be preferred by respondents who are faced with negative economic prospect. Moreover, we expect the preferences to be moderated by municipal counselors’ risk perception. Therefore, our hypothesis regarding which of the policy will be preferred will be elaborated with the risk as moderating variable. Hence, the first hypothesis stays open and goes as follow:
Hypothesis I: Receiving negative economic prospect impacts municipal counselors’ policy preference.
2.3 Dependent variable II: risk perception
By studying the impact of the economic prospect on the risk perception, we aim at evaluating whether the way respondents perceive risk, or which policy they deem as risky, is impacted by the contextual information we provide them with. Risk perception is defined as “people’s subjective judgments about the likelihood of negative occurrences, such as injury, illness, disease, and death” (Paek et al., 2017, p.1). In our case, we ask municipal counselors which of the four options of organization of the state they rate as having the highest likelihood of resulting in negative occurrence. Thus, this second dependent variable enables us to test whether policy makers’ risk perception is endogenous or exogeneous to the context.
Two theories elaborate on this question; the perception of risk theory by Paul Slovic (1987) and the prospect theory by Kahneman and Tversky (1979). It is important to highlight here that Slovic’s theory focuses on the risk perception whilst the prospect theory assess the conditions under which individuals take risks. Notwithstanding the fact that risk perception and risk taking are two different concepts and while acknowledging the distinction between them, we extrapolate the results of the prospect theory to estimate the relationship between negative economic prospect and the risk perception. We can indeed reasonably assume that the conditions that lead an individual to choose a risky option impact the way this individual perceives the risk in a similar manner.
The next section introduces those two theories. Then, based on the theories, the next hypothesis is formulated.
2.3.1 The perception of risk theory and the prospect theory
Based on Paul Slovic’s article on the perception of risk (1987), researchers have established a list of 14 factors that determine the way and extent to which individuals perceive risk (see for example Ropeik, 2002). The most relevant factors are uncertainty (the less we know about an event the more we fear it), familiarity (whether the respondent is familiar with the particular risk or if the risk is new), and awareness (whether the risk has been widely covered by the media). In other words, from this point of view, the way individuals perceive risk is determined by a list of fixed factors and is not conditional to the situation. In brief, risk perception is exogeneous.
The prospect theory, on the other hand, offers a contrasting view. This theory, outlined by Kahneman and Tversky (1979), models the way decisions are made under risk. In opposition to the expected utility theory, the theory assumes that people are risk-takers whenever they face losses and they are risk-averse whenever they face gains. In other words, one’s inclination to take risks is dependent upon the situation and the prospective outcome and individuals are more likely to take risks when the prospective outcome entails losses. Individuals in a situation of loss are more willing to take risks, which by extrapolation means they will tend to choose the option that they perceive as riskier. While this theory is originally used in Economics, it gained a prominent place in political science, as argued by Vis (2011) in her article “prospect theory and political decision making”. She claims that that this framework should be used in political science as it gives insight on the reasons why governments implement risky reforms. Moreover, in a study, Linde & Vis (2017) tested whether the prospect theory actually applies to politicians and their results indicate that decision-makers tend indeed to abide by the prospect theory’s assumptions.
2.3.2 Hypothesis II
In our case, we aim at evaluating whether receiving negative economic prospect impacts the way municipal counselors perceive risk. Hence, we ask municipal counselors to rate the risk they associate with each of the four options of the state (“directive state”, “hollow state”, “local-communitarian state” and “coping state”). Since we face council members with negative economic prospect, it is likely that they would consider potential personal losses attached to each of the policy option and they would be willing to take risks to avoid them. The “directive state” option suggests expanding the provision of public services, and by extension spending. While such policy might be risky during an economic crisis, it would not directly result in personal losses for politicians, such as job loss. At least in the short-term. The same goes for the “hollow state” and “local-communitarian state” options, which suggest decreasing the spending by delegating the provision of public services to external actors. Those options, although they promote cost-reduction, would not necessarily involve losses for council members, since they would still be needed to coordinate and overview the provision of public services by third actors. Again, these options do not involve personal losses, at least in the short-term. Conversely, the “coping state” policy option is the option that is the most likely to result in personal losses for municipal counselors, as it implies shortening the provision of public services and to scale back the number of civils servants. Since politicians are widely acknowledged to be short-term looking individuals (Pierson, 2000), we assume that they would endeavor to avoid short-term losses. This endeavor to avoid losses would push them to take risks, or to choose the options they deem risky, according to the prospect theory. By extension, it means that municipal counselors faced with negative economic prospects would tend to associate a high risk with the options that enable them to avoid losses, i.e. the “directive state”, “hollow state” or “local-communitarian state”, whereas they would associate a low risk with the “coping state” but would reject this option, as it entails losses. In other words, they are
willing to take risks, so as to avoid the “coping state” option. Consequently, on the basis of the prospect theory, our second hypothesis can be formulated as:
Hypothesis II: Municipal counselors would tend to associate a lower risk with the coping state option, as a result of receiving negative economic prospect.
Since the “directive state”, “hollow state” and “local-communitarian state” do not result in direct losses for municipal counselors whenever they are faced with negative economic prospect, we do not expect to find an impact of the independent variable on the risk perception associated with these three policy options. Hence, in those three cases, we can investigate if the risk perception plays a moderating role.
2.4 Moderating variable: Risk perception
After studying the impact of negative economic prospect on policy preferences and risk perception, we reach the core of the thesis, which aims at evaluating the extent to which the policy preferences interact with the risk perception. In other words, we are investigating for a moderation relationship. Many researchers have addressed the relationship between performance feedback and risk taking (see for example, Singh, 1986 ; Lant & Montgomery, 1987; March & Shapira, 1987 ; Bromiley, 1991; Miller & Chen, 2004), and we extrapolate their findings to theorize the reaction to negative economic prospect. We then base our third hypothesis on the performance feedback theory and on the prospect theory.
2.4.1 The performance feedback theory:
The performance feedback theory, which is rooted in the “Behavioral theory of the firm” (Cyert & March, 1963), models the way organizations enact change due to poor performance. Whenever the performance of the organization fails to reach the targeted aspiration, decision-makers will start searching for solutions and change. While the authors do not portray risk-taking as a specific outcome, implementing changes is said to be de facto riskier than maintaining the status quo. Greve (2003) also claims that the decision maker’s inclination to take risk is increased when the performance level is below the aspirations. This theory focuses on the effect of negative feedback on decision-makers and is backward looking. To analyze the effect of negative economic prospect on decision-makers, we extrapolate the results of the performance feedback theory and assume that decision-makers would also be more incline to take risks whenever they are faced with negative prospect.
As previously mentioned, aspirations are constructed through the influence of historical performance and peer’s performance. In the context of the economic crisis caused by corona, we can expect that municipalities’ performance will be lower than their past performances. Therefore, the aspirations will not be met, and we can expect decision makers to be more willing to take risks, i.e. to opt for policies that they formerly perceive as risky. As Greve (1998) highlights, organizational risk is highly sensitive to the performance level relative to the aspirations. In other words, by receiving negative economic prospect, municipal counselors can be confident that the future performance of the organization will not meet the aspirations. As a consequence, they would be more willing to take risks, which means they would tend to prefer the policy option they perceive as riskier.
This viewpoint is also supported by the aforementioned prospect theory. As mentioned, the prospect theory states that individuals are risk-takers whenever they face a prospective personal loss, whereas they are risk-averse whenever faced with a prospective gain. Thus, after receiving
negative economic prospect, decision-makers will choose to take risk and opt for the option that minimalizes their personal loss.
2.4.2 Hypothesis III
Following the feedback theory and the prospect theory, we can formulate this third hypothesis:
Hypothesis III: the risk perception moderates the effect of receiving negative economic prospect on the policy preference in such a way, that municipal counselors would tend to
3 Data and Methods
In this part, the way we conducted the research and collected the data, as well as the structure of the survey-experiment are detailed. We also elaborate on the statistical test used to answer our research questions.
3.1 Research framework
Our analysis is conducted in three steps. First, we investigate the relationship between receiving negative economic prospect (IV) and the policy preferences (DV1). In the second steps, we research the impact of receiving negative economic prospect on the risk perception (DV2). Finally, our third analysis depends on the results of the second step. Provided that we find no correlation between the negative economic prospect (IV) and the risk perception (DV2), we can analyze whether the risk perception variable moderates the impact of negative economic prospect on policy preference. For example, we analyze if the risk level associated with the “directive state” option interacts with the independent variable to explain the preference for the “directive sate” option. Yet, a moderator variable, which is defined as a variable that impacts on the strength or the direction of a relationship between an explanatory variable and an outcome variable (Glen, 2014), needs to be uncorrelated to the independent variable (Glen, 2020). Consequently, we need to distinguish two outcomes of the second analysis:
a) Dependent variable 2 is exogeneous: if the second dependent variable is exogeneous to the model, i.e. receiving negative economic prospect (IV) does not affect the way municipal counselors perceive risk (DV2), then we can analyze how the risk perception moderates the impact of negative economic prospect on policy preference. Thus, the risk perception becomes our moderator variable. Figure 2 below illustrates this case. For example, as receiving negative economic prospect (IV) does not affect municipal counselors’ perception of the “directive state” option (DV2), then we analyze if the risk perception of the “directive state” option interacts with the independent variable to explain the preference for this same policy option. This interaction effect can be investigated for the case of the “directive state”, “hollow state” and “local-communitarian state”, as the risk perception for these four policies are not impacted by the independent variable.
b) Dependent variable 2 is endogenous: if the second dependent variable is endogenous to the model, i.e. receiving negative economic prospect (IV) does affect the way municipal counselors perceive risk (DV2), then the independent variable and the moderator would be correlated, and the model would be biased. Since this correlation would jeopardize the validity of the results, we cannot inform on a potential interaction effect. This the case of the “coping state” option, as the risk perception associated with this policy is impacted by the independent variable.
3.2 Research design
We use a survey-experiment to conduct this study, which was sent to municipalities in Switzerland.
An experiment can be described as a form of research in which the researcher has some control over the environment and intervene in it so as to study a specific phenomenon. The goal of an experiment is to establish a causal link between the researcher’s intervention and the outcome. To do so, the researcher forms two groups of respondents, the treatment group and the control group and the intervention is made only in the treatment group, by administering a “treatment”. The outcomes of the two groups are then compared, and the researcher can establish a correlation if the outcome of the two group is on average significantly different. Yet, in order to be able to establish a causal link between the treatment and the outcome, the researcher must make sure that there is no other characteristics of the two groups that could affect the causal relationship. This is possible with randomization. Randomization ensures that on average, the two groups are exactly identical, and the only difference is the treatment (Toshkov, 2016). In Toshkov’s words, “a randomly assigned control group provides a counterfactual to the treatment group; a comparison case that is as similar as possible in all ways except for the experimental treatment” (2016, p. 172).
A survey experiment, then, is defined as a “deliberate manipulation of the form or placement of items in a survey instrument, for purposes of inferring how public opinion works in the real world” (Gaines et al., 2007, p. 4). In a survey experiment, the researcher does not have control over the environment in which the experiment takes place, but a treatment is administered to a random group of respondents. The results of the survey of the two group are then compared (Toshkov, 2016). The main advantage of a survey-experiment is to enable to make use of the internal validity of experimental research and to combine it with the external validity of surveys (Schlueter & Schmidt, 2010). Moreover, a large data sample can be gathered thanks to the facility of distribution. Additionally, it provides respondents with a complete anonymity, which allows us to reasonably assume that, on average, the results reflect municipal counselors’ true preferences. As Zwane et al (2011) outlined, respondents might alter their behavior if they participate in an experiment and might be incentivized to act in a socially desirable way if they know they are being surveyed. In our case, respondents do not even know they are being assigned to a control or treatment group and we cannot identify them through their answers. Thus, they are provided with full anonymity. However, survey experiment has disadvantages and one of them being the format of the questions. The way the questions are asked could influence the answer. For example, when asking respondents to scale the risk of a policy on a scale from 0 to 10, with 10 being the riskiest, each respondent would have his or her own interpretation of what 7 is. Moreover, when asking respondents which policy they prefer, different results would be yield whether we ask them to rank or to scale their preferences. In the experiment structure section, the choice of the format of our questions will be explained and justified.
In sum, we sent a survey link to municipalities in Switzerland and the respondents have been randomly assigned to the treatment group or the control group. The reasons why we chose municipalities as unit of analysis are explained in the next section.
3.3 Unit of analysis and sample
Focusing on municipalities so as to infer our causal relationship has three main advantages. First, municipalities are the smallest, but also the largest level of sub-national government, which enables us to have way more observations and a more consistent sample. Second,
municipalities are the most comparable unit of government between the five European countries that are part of the bigger study (France, Spain, the Netherlands, Switzerland and the United Kingdom). Indeed, their structure are the most similar to one-another, which is a prerequisite to send a survey to five different countries. Finally, according to the OECD, municipalities are “at the front line in addressing the COVID-19 pandemic, in terms of its containment, addressing the health challenges, and supporting populations in need”, which makes them a perfect subject to study the impact of a crisis on policy preference (OECD, 2020b).
This thesis focuses on the case of Switzerland. Swiss municipalities are well-suited to study the impact of economic prospect (IV) on policy preferences (DV1) and the risk associated with it (DV2), since those local governments are the most fiscally decentralized among OECD countries, as Thiessen (2005) highlights. The author finds out that Switzerland is the second most decentralized OECD country, both in terms of the share of local government’s expenditure over government’s expenditure and local government’s revenue over government’s revenue (Thiessen, 2005). Moreover, while the scope of the multipurpose municipal authority depends on the Canton, all local governments are required by the Swiss Constitution to adopt budgetary policies that take the national economic situation into account. Thus, municipal counselors’ policy preferences would be heavily affected by the negative economic forecasts of the COVID-19 pandemic. Finally, local politicians are acknowledged to be among the main actors dealing with the direct effects of the crisis (Pauchard, 2020) and they are also the decision-makers that are usually the most easily reachable.
Switzerland counts in 2020 2’202 municipalities. The number of council members in each municipality depends on the size and on the Canton of the municipality. The median number of inhabitants per municipality is 1’060. Therefore, and to ensure comparability between municipalities, our sample includes all local government with a population exceeding thousand inhabitants. This leaves us with 1’434 municipalities. On the 21st of June 2020, the survey was sent to the general email address of each municipality, with a request to forward the link to all council members. On top of that, the private email address of 495 municipal counselors was publicly available, so the survey could also be directly sent to them. After three weeks, 536 responses have been recorded. 285 of them were unfinished and our sample includes 264 valid answers. 142 respondents have been assigned to the control group and 122 to the treatment group.
3.4 Structure of the survey-experiment
In this section, the structure of the survey-experiment is detailed. Moreover, the way we operationalize the variables within the survey is explained.
To establish a causal effect of receiving negative economic prospect (IV) to the policy preferences (DV1) and risk perception (DV2), we conducted a survey-experiment that randomly affects respondents to the treatment or control group. The treatment group receives negative economic prospect, whereas the control group receives non-economic related information. Moreover, to ensure that the treatment group sees the pandemic as an economic crisis and the treatment group as a health crisis, we ask the respondents priming questions before giving them the treatment/control text. Priming is described as a psychological process in which a respondent, who is exposed to a concept, will use this already accessible concept to evaluate situations. Priming has then an impact on the person’s opinion by affecting the criteria she or he uses to evaluate the object or the situation (Lavrakas, 2012). Municipal counselors of the treatment group are then asked to define the extent to which the pandemic has affected the
control group are asked to define the impact of the pandemic on the wellbeing and health of the citizen of their municipalities.
3.4.1 Operationalization of the independent variable
The COVID-19 pandemic and the subsequent forecasted economic crisis offer a unique opportunity to operationalize the independent variable, i.e. “negative prospect”. The OECD forecasts suggest that local government’s expenditure will rise, while revenue will decrease (OECD, 2020b). Furthermore, municipalities will have to face a surge in unemployment, municipal debt and business bankruptcies. In Switzerland, the enforced and complete lockdown impacted the economic situation and experts forecast a long-lasting recession as a result of the quarantine.
The treatment group receives information about the possible economic consequences of the COVID-19 pandemic and we analyze the impact of this information on municipal counselors’ policy preference and risk perception. Thus, the treatment group received the following information, which, together with the priming question, is the treatment of the experiment:
The control group also receives COVID-19-related information, but this time, only the health aspects of the crisis are mentioned. The control text is the following:
3.4.2 Operationalization of the dependent variable I: policy preference
As mentioned, the dependent variable of the policy preference is based on Lodge & Hood (2012) terminology of the state. In the survey, the four options (“directive state”, “hollow state”, “local-communitarian state” and “coping state”) are presented, with their advantages and disadvantages, and the respondents are then asked to state their preferred option(s) by allocating 100 points. We choose the allocation of points over other methods (ranking or scaling) for the following reasons. First, we assume that some options will not even be considered by some participants, thus being allocated 0 point, and a ranking method does not enable to completely
“According to the Organization of Economic Cooperation and Development (OECD), local government organizations are at the frontline in bearing the financial and economic challenges of the COVID-19 pandemic. The OECD predicts that Local government organizations, as a consequence of the COVID-19 pandemic:
1. Are likely to experience a “scissors effect” of decreasing tax revenues and rising expenditures. 2. Must likely strongly increase expenditures regarding social protection and social welfare. (After the 2008 financial crisis, social protection spending by local governments in European countries increased by 6.4% in 2009 and by 3.5% in 2010).
3. Are likely to face rising unemployment, business bankruptcies and increased municipal debt.”
“According to the World Health Organization (WHO), the COVID-19 pandemic has had important consequences for the health and wellbeing of citizens. The WHO indicates that, as a consequence of the COVID-19 pandemic:
1. As of June 2nd, over 6 million people world-wide have contracted the virus, with over 373,000 confirmed deaths.
2. Particularly individuals of 60 years of age and higher are at risk, as well as individuals with health conditions such as lung or heart disease, diabetes, or conditions that affect their immune system.
3. In mental health terms, the main psychological impact to date is higher stress or anxiety, and levels of loneliness and depression are expected to rise.”
exclude one of the options. Second, the options entail policies that are at the antipodes. Preferring some options means eluding the others and a scaling system does not reflect the need for choice. Therefore, the municipal counselors are asked to the following:
3.4.3 Operationalization of the dependent variable II: risk
To be able to determine whether the treatment text affects the risk perception of the policies and the extent to which the risk perception moderates the policy preference, we ask respondents to assess the risk related to each of the four policy options presented by Lodge & Hood (2012). The four options are listed, and municipal counselors are asked to rate the risk of each policy on a scale from 0 to 10.
3.5 Statistical tests
To answer the first research question, which as a reminder is “to what extent does the predicted economic crisis due to COVID-19 influence Swiss municipal counselors’ policy preferences and risk perception?”, we evaluate from the survey the ATE (average treatment effect). We measure whether receiving negative economic prospect (IV) affects the policy preferences (DV1) and the risk perception (DV2) of municipal counselors. To do so, we first compare the number of points the control and the treatment group attributed to each of the policy options (i.e. “directive state”, “hollow state”, “local-communitarian state” and “coping state”), by running four independent t-tests for each option. An independent t-test is defined as a statistical test that enables to determine whether there is a significant difference between the means of two independent groups (Glen, 2015). Thus, we aim at testing the following null hypothesis:
Ho1: Mean (preference for the “Directive state” by the treatment group) –
Mean (preference for the “Directive state” by the control group) = 0 Ho2: Mean (preference for the “Hollow state” by the treatment group) –
Mean (preference for the “Hollow state” by the control group) = 0
Ho3: Mean (preference for the “Local-communitarian state” by the treatment group) –
Mean (preference for the “Local-communitarian state” by the control group) = 0
“After reading about the four positions the municipality can adopt, which one would you prefer in the current and future situation of the COVID-19 pandemics? Please distribute a total of 100 points based on your preference:
• An increased reliance on service provision by the municipality
• An increased reliance on service provision by private sector organizations
• An increased reliance on service provision by community and volunteer organizations
• A decrease in public service provision”
“After reading about the four positions a municipality can adopt, which one(s) do you deem risky? Please rate the options on a scale from 0 to 10, where 0 represents no risk and 10 is very high level of risk, based on your risk assessment.
• An increased reliance on service provision by the municipality
• An increased reliance on service provision by private sector organizations
• An increased reliance on service provision by community and volunteer organizations
Ho4: Mean (preference for the” Coping state” by the treatment group) –
Mean (preference for the “Coping state” by the control group) = 0
In a second step, we compare the level of risk respondents associate with each option, also by running four independent t-tests. Once again, we aim at testing the following null hypothesis:
Ho5: Mean (Risk perception of the “Directive state” by the treatment group) –
Mean (Risk perception of the “Directive state” by the control group) = 0 Ho6: Mean (Risk perception of the “Hollow state” by the treatment group) –
Mean (Risk perception of the “Hollow state” by the control group) = 0
Ho7: Mean (Risk perception of the “Local-communitarian state” by the treatment group) –
Mean (Risk perception of the “Local-communitarian state” by the control group) = 0 Ho8: Mean (Risk perception of the “Coping state” by the treatment group) –
Mean (Risk perception of the “Coping state” by the control group) = 0
To answer the second question (i.e. “To what extent does Swiss municipal counselors’ perception of risk moderate the influence of the predicted economic crisis due to COVID-19 on their policy preferences?”), provided that the risk perception is exogenous, we run a two-ways ANOVA test to evaluate the interaction effect of the treatment with risk perception variable. For example, we test if receiving negative economic prospect (IV) interacts with the risk perception associated with the “directive state” option (M) to explain municipal counselor’s level of preference for the “directive state” option (DV1). The statistical formula with an interaction term is the following:
! = # + %& + ɣ( + )&( + *
Where Y is the dependent variable, X is the independent variable, M is the moderator variable and XM is the interaction effect. β, ɣ and δ represent the coefficients of the relation between the variables X and M and the dependent variable. The two-ways ANOVA test makes a comparison of the means between the groups and tests if there is an interaction effect. In our equation, it tests if the ) is statistically significant (MacKinnon, 2011). The aim is to test the following four null hypotheses:
Ho9: There is no interaction effect between the treatment and the risk perception of the
“Directive state” policy option
Ho10: There is no interaction effect between the treatment and the risk perception of the
“Hollow state” policy option
Ho11: There is no interaction effect between the treatment and the risk perception of the
“Local-communitarian state” policy option
Ho12: There is no interaction effect between the treatment and the risk perception of the
To run the tests, this thesis uses the statistical software R and Rstudio [RStudio Team (2020). RStudio: Integrated Development Environment for R. RStudio, PBC, Boston, MA URL http://www.rstudio.com/] as well as Stata [StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC].
4 Analysis
In the first part of this section, the sample is presented, and the demographic characteristics of the control group and the treatment group are compared, so as to ensure that randomization worked and to ensure we can evaluate the effect of the treatment without any bias. Then, the results of the t-test and ANOVA tests, as aforementioned, are described. Finally, the results of the tests are analyzed with regards to the previously stated hypotheses.
4.1 Descriptive statistics of the data
The survey was sent the 21st of June 2020 to 1’434 municipalities and 495 municipal counselors’ private email address. After three weeks, we received 536 responses, roughly 50% of the them were incomplete. Thus, our sample includes 264 valid answers. It is unfortunately impossible to determine the response rate, as the exact number of municipal counselors who received the link to the survey is unknown. Indeed, we asked municipalities to forward the email to all council members, but we do not know how many of them actually did it.
Table 1 below summarizes the repartition of respondents between the control and the treatment group. Independent t-tests have been run to test if the groups are statistically different based on observable characteristics. We find no statistical difference between the sex, the age of the control group (M=52.84; SD=12.2914) and the treatment group (M=53.88; SD=10.64049); t(240)= -0.6997, p= 0.4848; p= 0.4848, and also no difference between the age of women and
men in the control and treatment group. However, there is a statistically significant difference between the groups when it comes to their political position, t(249)= -2.55, p=0.0013. The political position variable refers to municipal counselors’ political identification between “left”, “center-left”, “center”, “center-right” and “right”, each option being assigned the value 1, 2, 3, 4, 5 respectively. We can see that the control group is slightly more left-wing than the treatment group (M control=2.88 < M treatment =3.27). But the groups remain equal when it comes to their fiscal position, which refers to municipal counselors’ preference for an increase of the public debt on a scale from 1 to 10. Notwithstanding the fact that we observe a significant difference in municipal counselors’ political identification, which limits the validity of the experiment, we conclude that, in overall, random assignment of respondents to the control and treatment group has worked. We assume that our two groups are, on average, comparable in all aspects but the treatment administered.
TABLE 1. Demographic characteristics control/treatment group
p.value t.value Control Treatment
Obs mean min/ max
Obs mean min/ max N 142 122 Sexe .7752 .28583 Women 39 34 Men 90 85 Unidentified 13 3
TABLE 1. Demographic characteristics control/treatment group
p.value t.value Control Treatment
Obs mean min/ max
Obs mean min/ max Age .4848 -.6997 126 52.84 25/79 116 53.88 29/87 Age women .1958 -1.307 36 53.64 26/73 33 50.15 29/72 Age men .1662 1.3908 38 52.78 25/79 81 55.17 33/83 Political position .0013 -2.55 130 2.88 121 3.27 Left (1) 23 10 Centre-left (2) 31 22 Centre (3) 23 29 Centre-right (4) 44 44 Right (5) 9 16 Fiscal position .7119 .3968 State should augment public debt 62 52
State should not augment public debt
54 51
Ho: Mean (Control)- Mean (Treatment) =0
If p-value > 0.005, fail to reject Ho at 5% level
Table 1: Demographic characteristics of the experiment groups
4.2 Dependent variable 1 - Results
The first goal of the thesis is to assess the impact of receiving negative economic prospect (IV) on municipal counselors’ policy preferences (DV1). To do so, the number of points attributed to each of the policy option (“directive state”, “hollow state”, “local-communitarian state” and “coping state”) by respondents is compared between the treatment and the control group. With the independent t-test, we evaluate if the difference of preference between the groups is statistically significant. As a reminder, we assume that randomization worked, and that the control and the treatment groups are, on average, on all points equal, but the treatment received. The treatment is to receive negative economic prospect in the context of the COVID-19 crisis.
In the next sub-section, the results for each policy are presented and then discussed with regards to our theoretical framework. An overview of all the results of the tests is presented in Table 2 on page 26.
4.2.1 Preference for “directive state”
The first independent t-test was conducted to compare the number of points attributed to the “directive state” policy option in the control group and the treatment group. On average, the control group attributed 50.54 points to this option (SD=29.64) and the treatment 47.34 points (SD= 30.25). This is not a statistically significant difference; t(262)=0.8677, p=0.386. In other words, receiving negative economic prospect does not influence municipal counselors’ preference for the “directive state”.
4.2.2 Preference for “hollow state”
For the “hollow state” option, respondents of the control group, on average, gave to this option 16.25 points (SD=19.4), and the treatment group 18.79 (SD=18.48). Again, the difference is not significant, and the treatment did not have any effect; t(262) = - 1.0843, p=0.2792.
4.2.3 Preference for “local-communitarian state”
Municipal counselors of the control group attributed on average 23.95 points to the “local-communitarian” option (SD=19.95), for 23.50 points by the treatment group (SD=18.83). This tiny difference is not statistically significant; t(262) = 0.1873, p=0.8516.
4.2.4 Preference for “coping state”
Receiving negative prospect also did not affect the way respondents assess the “coping state” option, as the control group attributed 9.25 points, and the treatment group10.36, which is not a significant difference; t(262)= -0.5410, p=0.5889.
4.2.5 Analysis of the results for the policy preferences
We can see from the results that receiving negative economic prospect did not influence municipal counselors’ policy preference in none of the four cases. Figure 3 below illustrates the repartition of preferences for each of the policy option between the two groups. We observe that the repartition of preferences between the treatment and control groups is quite similar, thus, that receiving negative economic information does no impact council members’ favored option. In absolute terms, Swiss decision-makers prefer the “directive state” option, which implies expanding the scope of the provision of public services, in any context. The second favored option is to rely on the community and local organizations to provide public services. The first hypothesis, which, as a reminder, was the following, needs to be rejected. Indeed, receiving negative economic prospect did not impact municipal counselors’ policy preference.
Hypothesis 1 (rejected): Receiving negative economic prospect impacts municipal counselors’ policy preference
TABLE 2. Independent t-test results for policy preferences and risk perception
p.value t.value Control Treatment
Obs mean min/ max
SE 95% conf. interval
Obs mean min/ max SE 95% conf. interval Policy preference Directive state .386 .8677 142 50.54 0/100 2.48 45.62/55.46 122 47.34 0/100 2.73 41.91/52.75 Hollow state .2792 -1.0843 142 16.25 0/91 1.62 13.03/19.47 122 18.79 0/85 1.67 15.48/22.10 Local-communitarian state .8516 .1873 142 23.95 0/100 1.67 20.64/27.26 122 23.50 0/80 1.70 20.13/26.88 Coping state .5889 -.5410 142 9.25 0/100 1.42 6.42/12.06 122 10.36 0/100 1.47 7.43/13.28 Risk perception Directive state .5287 -.6310 115 3.85 0/10 0.28 3.28/4.41 109 4.10 0/10 0.27 3.56/4.64 Hollow state .3648 .9079 133 6.48 0/10 0.24 5.99/6.97 114 6.16 1/10 0.223 5.69/6.63 Local-communitarian state .8863 .1432 128 4.78 0/10 0.23 4.21/5.24 113 4.73 0/10 0.22 4.28/5.18 Coping state .0032 2.9774 127 7.03 0/10 0.27 6.48/7.57 106 5.79 0/10 .31 5.17/6.41
Ho: Mean (Control)- Mean (Treatment) =0
If p-value > 0.005, fail to reject Ho at 5% level
4.3 Dependent variable 2 - Results
The second goal of the thesis is to evaluate whether receiving negative economic prospect (IV) affects the way municipal counselors perceive risk (DV2). In other words, whether risk perception is fixed and exogeneous, or situational and endogenous. Independent t-tests have been conducted to compare the way respondent of the control group and the treatment group perceive risk.
In the next sub-section, the level of risk municipal counselors associate with each of the policy option (“directive state”, “hollow state”, “local-communitarian state” and “coping state”) is presented. An overview of the results of the four independent t-tests can be consulted in Table 2 above. The results are then discussed and compared to our hypotheses and theoretical framework.
4.3.1 Risk perception of the “directive state”
The independent t-test aimed at evaluating whether the risk level municipal counselors associate with the “directive state” options differs from the control to the treatment group. The average risk level, which is a scale from 0 to 10, hits 3.85 for the control group (SD=3.04) and 4.10 for the treatment group (SD=2.84). The difference is not statistically significant, thus, the way municipal counselors perceive the risk associated with the directive state is exogenous; t(222)= - 0.631, p=0.5287.
4.3.2 Risk perception of the “hollow state”
Respondents of the control group associated an average risk level of 6.48 with the “hollow state” option (SD=2.86), whereas it is 6.16 for the treatment group (SD=2.53). The difference is not statistically significant and the treatment did not affect the risk perception regarding this policy option; t(245)=0.9079, p=0.3648.
4.3.3 Risk perception of the “local-communitarian state”
There is no statistically significant difference in the way municipal counselors perceive the risk of the “local-communitarian state” between the control group (M=4.78; SD=2.64) and the treatment group (M=4.73; SD=2.39); t(239)=0.1432, p=0.8863. Thus, the risk perception of this policy option is exogenous and not affected by the treatment.
4.3.4 Risk perception of the “coping state”
The mean level of risk associated with the “coping state” policy option by the control group reaches 7.03 (SD=3.11), whereas it reaches 5.79 for the treatment group (SD=3.21). This is a statistically significant difference; t(231)=2.9774, p=0.0032. Municipal counselors associate a lower risk to the “coping state” option as a result of receiving negative economic prospect. We can then conclude that, in that case, the risk perception is endogenous and that receiving negative economic prospect negatively impacts the risk perception associated with the “coping state” option.
4.3.5 Analysis of the results for risk perception
The results suggest that receiving negative economic prospect does not affect the way municipal counselors perceive risk, except in the case of the “coping state” option. The difference between the way municipal counselors perceive the risk associated with the “coping state” in the treatment group and the control group is the only statistically significant difference we find in our model. Thus, when receiving negative economic prospect, municipal counselors associate a lower risk with the “coping state” option (Mean of the treatment group=5.79 < Mean of the
control group=7.03). In other words, receiving negative economic prospect negatively affects municipal counselors’ risk perception of the coping state a 5% level of confidence. Figure 4 below illustrates the difference of risk perception between the groups:
We can see that the repartition is more or less similar between the two groups for the risk associated with the first three policy options (“directive state”, “hollow state”, “local-communitarian state”) but when it comes to the last one, the “coping state”, we see that the risk level is more scattered for the treatment group, compared to the control group, which lowers the average risk associated with this policy options. Hence:
Hypothesis II (Approved): Municipal counselors would tend to associate a lower risk with the coping state option as a result of receiving negative economic prospect.
4.4 Moderator variable - Results
The final purpose of the thesis was to evaluate if the risk perception (M) interacts with the negative economic prospect (IV) to explain municipal counselors’ policy preferences (DV1). Two-ways ANOVA tests have been conducted and all the results are displayed in Table 3 on page 31.
As mentioned, the prerequisites for a moderator variable is to be uncorrelated to the independent variable (Glen, 2020). Therefore, we did not run a test to see if the risk associated with the “coping state” option interacts with the negative economic prospect, as we found that the two variables are correlated. Thus, the independent assumption of the two interaction terms is violated.
First, we tested if the risk perception associated with the “directive state” option interacts with the negative economic prospect to explain the preference for this policy option. The same test has been conducted to see if the negative economic prospect interacts with the “hollow state” option and the “local-communitarian” option to explain the corresponding level of preference. We find no statistically significant interaction effect in none of the three cases. The p-value of the interaction effect is always highly superior to .05, thus the Ho hypotheses of the interaction effect cannot be rejected (p=0.69; p=0.78; p=0.88 for the interaction effect of the negative economic prospect (IV) and the risk perception (M) of the “directive state”, “hollow state” and “local-communitarian state” respectively). In other words, the effect of receiving negative economic prospect on the policy preference is not moderated by the risk perception associated with it. Consequently, our last hypothesis needs to be rejected.
