Confidence in a Crisis: The Effect of Framing Feedback on Councilor Preferences

Confidence in a Crisis

Timo van der Kraan

Acknowledgements

First and foremost, I would like to thank my supervisor, Dr. van der Voet, for his patience, his guidance, and the contributions he made to the topic, structure and research of this

thesis.

Second, I would like to thank my dear friend Victoria Isabella Cornelia Smit, without whose unwavering support and unrelenting supervision, this thesis would not have come to

fruition.

Finally, I would like to thank my mother, Karen Poot and my beloved Skye Tjong-Ayong – Visser for their help in proofreading and finalizing this thesis.

Abstract

Some of the most influential politics happen in response to a crisis. As a result, it is important to study how a crisis impacts a politician’s thoughts, preferences and frame of mind. Understanding whether negative information during a crisis causes a shift in preferences between blame avoidance and credit claiming can help us understand how crises impact politicians’ willingness and ability to act. By using a survey experiment conducted amongst municipal councilors across 5 European countries; France, the Netherlands, Spain, Switzerland, and the United Kingdom, the goal of this study is to investigate whether negative financial prospects have an effect on politicians’ preferences for blame avoidant or credit claiming policies. Level of abstraction (abstract or concrete) was used as a signal for either blame avoidant or credit claiming policies allowing this research to add a quantitative dimension and scope to the literature on credit claiming and blame avoidance, performance feedback, and crisis accountability. Using six linear regression models, the experiment showed that in general, councilors are more likely to support than to not support policy goals. There was a small but significant treatment effect as councilors in the treatment group rated both blame-avoidant and credit claiming policies lower than their control group peers. However, treatment did not create the expected shift in preference from credit claiming to blame avoidance or vice versa. Rather, it was found that both groups rated blame-avoidant policies substantially higher than credit claiming policies. Given these preferences, strategic use of blame avoidance can help politicians to gather support for difficult but necessary policies.

List of Tables and Figures

Tables

Table 1 Survey Responses 12 Table 2 Gender Demographics of Survey Respondents 12 Table 3 Distribution of Control/Treatment Group 12 Table 4 Summary of Responses 13 Table 5 Regression Table 14

Figures

Table of Contents

Acknowledgements i

Abstract ii

List of Tables and Figures iii

Introduction 1 Literature Review 3 Theoretical Framework 7 Methodology 9 Results 12 Analysis 14

Discussion and Limitations 18

Conclusion 22

Bibliography 24

Introduction

Politicians are strange creatures. One day they will claim credit for a program while strongly denying any affiliation with it the next. Under normal conditions, that is just a part of politics as you know it. Do you claim credit for expenditure at the risk of affiliating yourself with an unpopular policy or do you avoid affiliation, at the cost of credit you could have claimed for passing a popular policy? Political credit claiming and blame avoidance are based on interpretations of voter sentiment. With blame avoidance being a proven strategy to pass difficult cuts and credit claiming helping to shed light on the value generated by politicians (Wenzelburger, 2014; Weaver, 1986). However, during crises, what voters favor becomes subject to rapid change. Can you claim credit for the immense sums allocated towards first relief without associating yourself with the unavoidable economic blow? As a party, how do you navigate the difficult choices that must be made during crisis times while maintaining (at least a semblance of) your political support? This question is perhaps too broad to answer as a whole, however, one part of the puzzle worth looking at is the relationship between the negative financial prospects faced by councilors during a crisis and a potential shift in preferences between credit claiming and blame avoidance. Understanding whether a crisis might cause a shift in preferences can help us understand how blame avoidance and credit claiming apply during crises. To help answer this question, this research, through a survey experiment of municipal councilors during the COVID-19 pandemic, aims to find whether negative financial prospects have an effect on politicians’ preferences for blame avoidant or credit claiming policies.

Why are blame avoidance and credit claiming in crisis conditions relevant for political support? One example of this problem, and the societal relevance of this research, can be found in the Dutch elections following the political response to the 2008 financial crisis. The labor party (PVDA) in the Dutch coalition government suffered unprecedented losses that they are still recovering from (van Houten, 2017). Associating themselves with the unpopular austerity policies that got the Netherlands through the economic collapse was not rewarded by voters. During the same time, the liberal party (VVD) was involved with the same policies but was not punished by voters, in fact, they delivered the prime minister for the next 3 cabinet formations (NOS Nieuws, 2020). This is an example of how strategic credit claiming and blame avoidance can make or break a party for years to come.

As for academic relevance, much has been written about how blame avoidance affects constituent attitudes. However, only few studies focus on the circumstances that influence a politician’s choice between credit claiming and blame avoidance (Wenzelburger, 2014; Grimmer, Messing, & Westwood, 2012) The experiment by Baekgaard and Serritzlew (2015) comes closest but only studies the impact of performance information on councilors’ spending attitudes. This study adds comparative quantitative data through an experiment on the impact of performance information on councilors’ preferences between blame avoidance and credit claiming. This will add another dimension to both the theories of performance feedback, and the theory of credit claiming and blame avoidance.

In sum, this research uses a survey experiment conducted amongst municipal councilors across 5 European countries; France, the Netherlands, Spain, Switzerland, and the United Kingdom, wherein they were asked to rank various policy goals. The policy goals are distributed among 2 axes, economic or social, and credit claiming or blame avoidant. This research investigates the relationship between negative financial prospects and a preference for credit claiming and blame avoidance, contributing to the literature on crisis accountability, credit claiming, and blame avoidance. Through a set of linear regression models, this research will add a quantitative dimension and scope to discussions about credit claiming and blame avoidance theory. This thesis continues as follows, first, a general overview is provided of the literature and the relevant theoretical framework. Second, the methodology, the experimental set-up, data collection, and useable models will be established. Third, an analysis of the results of this research. Fourth, the discussion and limitations, and finally, the conclusion.

Literature Review

Two relevant theories together form the framework within which this research will take place. The theory of performance feedback by Jordan and Audia and the theory of blame avoidance and credit claiming by Paul Pierson (Jordan & Audia, 2012; Pierson, 1996). This research uses these two theories because they complement each other, as will become clear in the subsequent literature review. The literature review will consist of an introduction to the theory of blame avoidance and credit claiming, an introduction to the theory of performance feedback, and supporting theories like particularistic spending, crisis accountability, and framing. Subsequently, the assumptions for this theoretical framework will be explained further in the next section.

Credit Claiming and Blame Avoidance

The theory of credit claiming and blame avoidance was first coined in Weaver’s (1986) essay “The politics of blame avoidance.” It has since been adopted, revisited, and tested by the large community of public administrations researchers in particular in the research on welfare state retrenchment (Pierson, 1996; Giger & Nelson, 2010; Bonoli & Natali, 2012). The theory describes two strategies for politicians to position themselves in relation to a policy or policy proposal. They can either claim credit or avoid blame. The theory, in essence, poses that politicians are generally motivated to minimize blame, for fear of losing re-election, even more so than they are motivated to seek credit for popular policies (Weaver, 1986). However, unclear benefits and clear risk mean that legislators’ tendency to claim credit varies depending on many factors beyond the scope of Weaver’s research. Weaver attributes this general tendency to the economic theory of negativity bias, wherein people generally respond more strongly to potential loss than to potential gain (ibid.).

Indeed, many have observed that politicians, national and local, use strategies to minimize blame (Houlberg, Leth Olsen, & Holm Pedersen, 2016; Nielsen & Baekgaard, 2015). Whether they do this out of fear of electoral punishment is unclear. Wenzelburger (2014) found no evidence to support the practice of electoral punishment in response to retrenchment, though his study was unable to account for the successful implementation of blame avoidance strategies. Wenzelburger suggests that, because of the difficulty in operationalizing ‘successful blame avoidance’, an experimental setting might help clarify this connection (Wenzelburger, 2014). For now, blame avoidance is believed to be tied to perceptions of risk, rather than empirical evidence of electoral punishment (Wenzelburger, 2014; Giger & Nelson, 2010). Claiming credit for particularistic spending can help build support from constituents (Grimmer, Messing, & Westwood, 2012). Particularistic spending is the discrepancy that occurs between expansion and retrenchment policies, which will be discussed further in this review. Legislators have been found to use a credit claiming message to generate a belief that they were responsible for allocations of funds towards constituents’ needs (ibid.) (Houlberg, Leth Olsen, & Holm Pedersen, 2016). Conversely, when politicians fear a negative public response like when passing difficult cuts, they can instead attempt to avoid blame by using several strategies like using layers of abstraction to avoid blame. The section on framing will

In recent research, much of the emphasis remains on the use of blame avoidance in retrenchment, linking it to everything from decentralization to Brexit (Mortensen, 2011; Hansson, 2019). The theory of credit claiming by comparison is largely left to supranational challenges (Madama, 2018). This is surprising, as new opportunities for credit claiming, such as a substantial budgetary surplus in the Netherlands, and the New Growth Deals in the UK went largely unresearched (UK Government, 2019; Rijksoverheid, 2020).

Performance Feedback

The theory of performance feedback evolved from Cyert and March’s (1963) ‘Behavioral Theory of the Firm.’ Cyert and March responded to the then-common theory of rational choice, which supposed that firms operated to maximize profit under conditions of perfect knowledge. Cyert and March rejected these assumptions and argued that decision-makers worked to ‘satisfice’; to achieve good enough results as defined by a compromise between relevant managerial and worker layers. From this, they argued that decision-makers set aspirational levels of performance and worked to improve towards these standards. Based on whether or not they achieved that standard, they argued that, given negative feedback, leaders are more likely to search for alternatives, implement changes and choose strategies with higher associated risks (Jordan & Audia, 2012). Given positive feedback, they will be less inclined to search, implement, and choose. This theory is based on an expectation of bounded rationality, an organizational theory that accounts exceptions of rationally expected behavior to cognitive limitations on the part of the decision-maker. In short, according to this classical theory, goals are set prospectively and remain fixed until a changing environment or strategy demands that they are changed.

Jordan and Audia (2012) however, argue that the theory should be expanded to account for further behavioral phenomena observed in practice. What they noticed was that decision-makers would sometimes retrospectively change their assessment of the situation to provide a more favorable view of what would otherwise constitute as low performance. This is what they call the mode of ‘self-preservation’, with the behavior expected under classical theory retroactively being named ‘problem-solving.’ The mode of problem-solving is consistent with previously mentioned literature by Cyert and March, where leaders search, implement and choose based on the level of feedback. Problem-solving requires clear performance goals and moderately ambitious aspiration levels. The mode of self-preservation, by comparison, is characterized by responding to negative performance feedback by revising the standards of evaluation, such as revising the priorities of stated goals, increasing the level of abstraction of the goals, or comparing not to prior or comparable performance but a speculative alternate scenario. The authors suggest that exceptions to rationally expected behavior are not attributed to limited information or capacity, but to the need to protect a sense of self for the responsible leader. It can be incited in various ways, though Jordan and Audia suggest that further research should be done into the situations that magnify self-enhancement. Depending on factors such as level of narcissism, a belief of ability, accountability to an audience, and how outcome-oriented their audience is, they are more or less likely to respond to negative performance feedback by seeking self-preservation rather than problem-solving. The implications for the theory of performance feedback are that leaders may not respond to

negative performance feedback with the level of urgency and care that would be expected in classical theory. Self-enhancement may lead to behavioral rigidity, reducing search, change, and risk-taking actions, thus ultimately inhibiting necessary performance improvements. Employing this new theory, it may be favorable to shape performance feedback strategically. Framing carefully so as to not incite the feeling of an attack on a sense of self, or reducing opportunity for reframing policy goals would improve the consistency of the effect performance feedback has on decision-making.

Particularistic spending

Under normal circumstances we assume that politicians avoid blame when cutting budgets and claim credit when they are spending (Pierson, 1996). This assumption was the basis for Pierson’s (1996) ‘The New Politics of the Welfare State.’ Contrary to Weaver’s assumption that politicians are just naturally more inclined towards blame avoidance as a result of negativity bias, Pierson attributes this to ‘the crisis of the welfare state.’ A concept that concludes that in the last 20 years (at the time of his writing), the limits of the welfare state in the face of demographic changes had required politicians to scale back the welfare state’s generosity and coverage to stay fiscally sustainable. This called for welfare state retrenchment. Before Pierson’s seminal work on the welfare state, little was known about the patterns of particularistic spending and cutting. Economic growth and decline do not occur in the same context at the same time, making theories on retrenchment and expansion difficult to test against one another (Pierson, 1996). Under previous theories, retrenchment was considered politically impossible as a result of organized interest groups and negativity bias. Because postwar expansion served a large, clearly defined audience of beneficiaries, the diffuse concerns about taxes by opponents of new policy were relatively easy to manage (Pierson, 1996). This large, clearly defined audience of beneficiaries then works against retrenchment by organizing effectively against the proposed policy, while the beneficiaries, taxpayers, are far more diverse and difficult to mobilize. It was Pierson who concluded that it was possible because the programs that politicians choose to cut in times of scarcity are not the same they choose to invest in once the economic climate improved (Pierson, 1996). During retrenchment, through careful selection of the least popular, least served policies, combined with effective blame avoidance and framing techniques, retrenchment was made possible, even popular, without the parties involved suffering great electoral losses. From this, Pierson theorized that spending and cutting follow distinct patterns that are different from each other and proposed a framework for how the enactment of unpopular retrenchment policies works in contrast to previous theories adapted from the postwar enactment of popular welfare state expansion. Retrenchment is more incremental, preferably done less visibly, and dependent on successful blame avoidance strategies. Pierson wrote his theory during what he called a time of ‘permanent austerity.’ A period that stands in contrast to the ‘thirty glorious years’ of economic growth and welfare state expansion in the postwar era. A definition that might well shift as a result of unprecedentedly cheap credit and investment by central banks across the world (The Economist, 2020).

Crisis accountability

In times of crisis, the theory of blame avoidance operates differently from normal, there is a shift from accountability to crisis accountability. A sudden, attention-grabbing event that serves as a trigger for policy change, commonly referred to in the literature as a focusing event, marks this shift (Birkland, 1998). An event of this nature, like an economic collapse, natural disaster, or pandemic, has a great impact on the policy agenda that follows and may influence the results normally expected from previous experiments on the theories used. Time is the deciding factor between management under regular versus crisis conditions. In an event of a major crisis, the cost of inaction outweighs the cost of unpopular action (Bonoli & Natali, 2012). This creates conditions under which blame avoidance has different purposes than under regular conditions. Crisis accountability literature suggests that in this situation, blame avoidance serves two purposes; to channel public emotions and to provide a new sociopolitical equilibrium after a crisis (Brändström & Kuipers, 2020). As for performance feedback, in a way, a crisis is on its own a form of negative performance feedback that can lead to a range of behaviors from problem-solving to self-enhancing. (Antonacopoulou & Sheaffer, 2014) As the acute response leaves little time for investigation and lengthy policy debate, the actions taken by a minister are limited to three major themes: rhetoric, substantive action, and symbolic action. Rhetorical strategies, such as effective framing and crisis communications have been found to be most influential to the future careers of involved ministers (Brändström & Swinkels, 2015) Substantive action, such as major reforms, usually require more time to be worked out. This leaves symbolic action. This includes, for example, the firing of subordinates and the passing of largely symbolic reforms. In this thesis, we will focus mostly on rhetorical strategies.

Framing

Framing involves using specific language and linguistic structures to evoke the desired emotions and images in an attempt to shape the public perception and the debate surrounding issues. Tversky and Kahneman (1981) first found this in a survey experiment on college students to test the validity of rational choice theory. They found predictable shifts in preference when the same problem was framed differently, noting that negative framing usually results in more risk-taking and positive framing in reduced risk-taking. This connects to the theory of performance feedback, which finds a similar effect (Jordan & Audia, 2012). As explained in the section on performance feedback theory in this literature review. The literature on framing allows us to create intentionally biased frames, through which we can test our hypotheses. Linguistic discourse analysis provides a framework that makes this possible in a policy sense using only discursive differentiation. The linguistic concept called level of abstraction is useful. Level of abstraction describes how close or distant a description is to things, events, and properties that we can perceive directly (Turney, Neuman, Assaf, & Cohen, 2011). In the linguistic setting, the use of concrete or abstract phrasing determines the perception of interpersonal distance (Reitsma-van Rooijen, Semin, & van Leeuwen, 2007). Negative abstract messages created the highest level of interpersonal distance, compared to concrete negative, abstract positive, and concrete positive messages. As Hansson mentions, this interpersonal distance is used by politicians to avoid blame for policies (Hansson, 2015).

Theoretical Framework

Within this thesis, the following theoretical framework is used. First of all, this research assumes that the period in which the experiment is carried out matches the description of a crisis as described in the literature on crisis accountability (Bonoli & Natali, 2012). At the time of writing, the Covid-19 pandemic in Europe has had a substantially disruptive effect on the everyday lives of citizens, businesses, and government. This circumstance is a central part of this thesis, as it allows us to study the political response to a policy problem that is more or less relevant to all respondents, across the political spectrum, in all surveyed nations. Second, it is assumed that blame avoidance is connected to the mode of self-enhancement as described by Jordan and Audia (2012). People are more likely to process positive information about themselves more fluently and attribute failures to other factors. When confronted with a setback, such as negative economic feedback, politicians are more likely to seek self-enhancement as a form of self-preservation. Self-self-enhancement is also more likely when audiences hold coercive power, like in the case of an electorate. This is relevant for this thesis because the population of interest are municipal councils, which are bodies that have such an electorate. Under these conditions, given negative feedback, this research proposes that politicians will be inclined to take the mode of self-enhancement and seek to avoid blame. Third, this thesis applies the theory of abstract and concrete policy preferences to the concepts of credit claiming and blame avoidance. In an attempt to keep the policy goals as similar as possible to each other, the experiment uses framing to differentiate between blame avoidant and credit claiming behavior. Jordan and Audia (2012) find self-enhancement leads to the employment of three primary strategies, of which preferring an increase in the level of abstraction in policy goals is one. This is supported by the theory on framing and level of abstraction, which gives us the linguistic framework to link our relevant variables to abstractness and concreteness (Hansson, 2015). Therefore, a preference for abstractly phrased policy goals may be read as a signal to indicate a preference for blame avoidant behavior. Similarly, this research proposes that a preference for concretely phrased policy goals may be read as a signal to indicate a preference for credit claiming behavior. While in the literature, credit claiming has not been connected to discursive analysis in the same way blame avoidance has. Intuitively it would make sense that if abstract policy creates distance, concrete policy creates proximity. This reasoning forms the basis for the use of concrete policy goals as a signal for credit-claiming behavior.

Finally, acceptance of public debt and risk-acceptance are included as control variables. Given that central bank policies have created affordable credit, this research proposes that increased acceptance of public debt is positively correlated with concrete policies and, therefore, credit claiming (The Economist, 2020). Low acceptance of public debt could explain a lower overall score for all forms of policy. Additionally, given the propensity for leaders to accept more risk when they are problem-solving and less risk when they are self-enhancing, this research proposes that a higher level of risk explains a lower level of blame avoidance and a higher level of credit claiming.

Operationalization

Six concepts need to be defined to make them measurable in our experiment: abstract and concrete policy goals, blame avoidant behavior, acceptance of public debt, risk acceptance, and negative financial prospects.

Concrete policy is defined as a policy goal that meets all S.M.A.R.T. guidelines, with S.M.A.R.T. standing for Specific (simple, sensible, significant), Measurable (meaningful, motivating), Achievable (agreed, attainable), Relevant (reasonable, realistic and resourced, results-based), and Time-bound (based, limited, time/cost limited, timely, time-sensitive) (Government Finance Officers Association, 2020). S.M.A.R.T. guidelines were designed in the private sector to improve the quality of goal setting. Since then, some governmental agencies like the US Government Finance Officers Association have also adopted the framework as a test of policy. As such, because it offers a clear framework for designing concrete policies, it is suitable for helping define what a concrete policy goal is. Conversely, abstract policy goals are defined as a policy goal that is imprecise and unaccountable according to S.M.A.R.T. guidelines (ibid.). Failing these criteria of the S.M.A.R.T. framework makes a policy abstract.

Blame avoidant behavior is defined as an expressed preference for abstract policies that are designed to avoid blame (Hood C. , 2011). This will be measured as the average rating of all abstract policies on a scale between -2 and 2, compared to the average rating of all concrete policies on the same scale. Preference is expressed as a higher average for the abstract, compared to the concrete policies.

Acceptance of public debt is defined as a belief that the state should increase its public debt. This will be measured using a self-reported score from 0 to 10 to the question “To what extent must the state increase its public debt?” Similarly, risk acceptance is defined as a self-reported level of risk-taking in daily life. This is measured using a self-reported grade from 0 to 10 of the question “To what extent do you take risks in your daily life?”

Negative prospects are defined as a statement containing short factual information from the Organisation for Economic Co-operation and Development (OECD) about negative economic prospects (OECD, 2020). This primes respondents to think about the future negatively, thus prompting a more defensive stance towards policy developments. The control group receives a message with widely distributed information about the coronavirus from the World Health Organization (WHO) that they are likely familiar with (World Health Organization, 2020). This should result in a neutral stance. Both messages are available for review in the appendix.

Hypotheses

From this, we can draw two competing hypotheses and a null hypothesis: the main hypothesis is that information on negative financial prospects leads to more blame avoidant behavior. The alternative hypothesis is that information on negative financial prospects leads to more credit claiming behavior. For the null hypothesis, information on negative financial prospects does not affect preferences for either blame avoidant or credit claiming policies.

Methodology

This research sought to answer the question of whether negative financial prospects have an effect on a preference for blame avoidant or credit claiming policies. The hypotheses to be tested were that either, information on negative financial prospects leads to more blame avoidant behavior, or whether information on negative financial prospects leads to more credit claiming behavior. To do so, an online survey experiment was performed, based on the experiment by Baekgaard and Serritzlow (2015). Our experiment was performed on a sample of municipal councilors from France, the Netherlands, Spain, Switzerland, and the United Kingdom. Our randomly assigned treatment group was presented with a biased frame of negative economic prospects, and a control group with neutral information, they were then asked to rate potential policies that were either designed to be blame avoidant or credit claiming. After which, the data was compiled, and multiple Ordinary Least Squares regression models were created using the responses.

The survey experiment was chosen for several reasons. First, to ensure a controlled environment that would let us reduce factors that would otherwise confound the results. Second, to be able to approach a large number of municipal councilors so that we would be able to make effective use of statistical methods. Third, because it is difficult to test a theory ground in intentional explanations, an experiment was envisioned to examine the theory teleologically. By limiting the experimental context, there are clear pathways through which we can establish a change in preference, all other variables being equal (Toshkov, 2016). Finally, the survey experimental approach had previously been successfully applied by Baekgaard and Serritzlew (2015) to answer a similar question on the effect of performance feedback. This provided a peer-reviewed framework from which to work on.

Data collection

The survey was modeled after an experiment by Baekgaard and Serritzlew (2015), which similarly used manipulated performance feedback to measure its effect on the spending attitudes of municipal councilors. Our team of researchers designed new questions, using this same basic experimental setup, to create a dataset that could be used to answer the four research questions for the respective theses being written.

Municipalities were chosen arbitrarily based on the languages spoken by the research team. This led us to approach municipal councilors in France, the Netherlands, Spain, Switzerland, and the United Kingdom. Distributing translations of the survey in all official languages, such as French, Dutch, Spanish, Swiss-French, Swiss-Italian, and Swiss-German. These countries have diverse administrative interpretations of what defines a municipality; however, the selected municipal systems did all have a concept of municipal councilor and were, therefore, suitable for this research (Higonnet & Drinkwater, 2020; VNG International, 2007; Direccion General de Cooperacion Local, 2012; Debela & Meissner, 2020; Politics.co.uk, 2020). As part of our research team, we had the ambition and resources to perform our study on a larger sample of municipal councilors from across Europe. This resulted in a diverse selection of respondents for this research, making sure blame avoidance or credit claiming was not just a

Where individual addresses were not available, we approached the general municipal e-mail address with a request to forward the survey experiment link to their municipal councilors. For Dutch municipalities, a database of e-mail addresses was made available from earlier research.

The treatment group was assigned randomly by the Qualtrics XM surveying software. Those in the treatment group were biased using two methods: one short response priming question and one short text containing negative economic information. The priming question stimulates a concept in memory that remains active for the duration of subsequent questions (Lavrakas, 2008). Similarly, the biased frame is based on Tversky and Kahneman’s (1981) suggestion that the same problem framed differently results in a predictable change in preference, with negative frames resulting in more risk-taking. The control group meanwhile was exposed to a more neutral priming question and short text. Their question focused on societal welfare rather than economics and the short message containing neutral factual information on the coronavirus pandemic from trustworthy sources that the surveyed audience was likely already familiar with. Both texts were first independently written by the involved researchers and then reduced and finetuned through an iterative approach until the final version was reached, which can be read in the appendix to this paper.

The subsequent questions were also written using an iterative approach and were finetuned until all researchers agreed on their content. For this thesis, the relevant question asked respondents to rate twelve potential policy goals, designed to vary along two dimensions: economic or social, and concrete or abstract. There were a total of three policies of each kind, providing enough variation to filter out individual differences in preferences across each of the two dimensions without creating an overly burdensome survey experience. Because the ability to distinguish between credit claiming and blame avoidance depended on the framing of the policies to be ranked in our policy experiment, we used only linguistic differentiation of the same general policy goal between the concrete and abstract versions of a policy goal. That means that a concrete and abstract goal had the same general target and goal, like reducing bankruptcy amongst small businesses, but differed in their level of abstraction. This was to reduce substantial changes to individual policies that might have created external factors that were not related to either blame avoidance or credit claiming but to personal and political preference. This research analyses only the variation in abstract or concrete policies to answer the research question.

Processing

The survey was sent starting on the 14th _{of June 2020 with submissions closed on the 10}th _of July. Data was then downloaded on the 10th _{of July 2020 and processed for further analysis.} After this, the survey responses from each language were exported from Qualtrics XM. As we had created a survey per nation and language, the data needed to be concatenated into a single file. After filtering out the questions from the other members of the research team, all variables were translated to English and recoded into a symmetrical equidistant scale from -2 to +-2 where --2 represents the answer “totally disagree” and +-2 represents “totally agree” to the question of whether they think a policy goal should be implemented. Incomplete responses

to the relevant questions were discarded during the initial analysis. Selecting complete cases from the original 2186 observations left us with 1269 observations to perform our analysis on. We chose to do a complete case analysis because there was an ample number of observations to permit us to do so. An alternative to discarding incomplete cases would have been to fill in the missing data using imputation. This was not done as it would have added complication and theoretical inference that would have diminished the clarity of our data. By only using complete cases, we avoided this path and were able to generate a simple, commonly accepted statistical analysis.

To equalize the effects that the phrasing of individual questions may have had on the preference expressed by individual councilors, the answers to the questions designed to be abstract and the answers to the questions designed to be concrete were aggregated. This resulted in a new scale from -12 to 12, where -12 represents a total rejection of all policies, and 12 represents a total agreement with all policies.

Analysis

We chose an Ordinary Least Squares regression for two reasons. First, from the legion of statistical tests, this one is simple, commonly accepted, and is suited to the hypotheses we were trying to test. Second, we wanted the freedom to add more control variables and to ensure the robustness of our model, which ruled out the use of an ANOVA. These additional control variables resulted in six models. Three for each summed set of abstract and concrete policies.

Model 1 was the OLS regression, with each subsequent model adding a variable to test the explanatory power and robustness of the model. Creating this set of models allowed us to examine the individual contributions made by adding specific variables and analyzing how they contribute to the overall R-squared and P-values.

Model 2 added acceptance of public debt as a variable. We expected politicians more accepting of public debt to score all policies higher on average than those that did not accept. We wanted to test whether more acceptance of public debt would explain some of the results we would have otherwise attributed to credit claiming (The Economist, 2020).

Model 3 added risk acceptance as a variable. Given the influence risk-taking played in both performance feedback theory and the theory of framing, we verified that our results could not be explained by an increase in risk-taking alone (Jordan & Audia, 2012; Tversky & Kahneman, 1981).

Software

This thesis used Qualtrics XM for data gathering and statistical software R and R-Studio, with libraries car, readr, and ggplot for the subsequent analysis (Qualtrics, (c) 2020; R Core Team, 2020; Fox & Weisberg, 2019; Wickham, Hester, & Francois, 2018; Wickham H. , 2016). Additionally, e-mails were collected using Email Extractor (Unified Address Book, 2020).

Results

Summary statistics

Our survey was sent to a list of publicly available e-mail addresses of municipalities and municipal councilors. For France and Switzerland, the general e-mail addresses of the municipality in question were approached with the request to distribute the survey among their municipal councilors. For the Netherlands and the UK, e-mail addresses of individual councilors were available, so councilors were approached directly. For Spain, a combination of general and direct e-mail addresses was approached. Because of this mixed approach, we can only estimate the total response rate.

Table 1 shows that the mixed response rates reflect a diverse range of willingness to participate in our unsolicited survey. Two countries stand out in particular: the Netherlands and Switzerland. The high number of Dutch responses may reflect familiarity with Leiden University, the institute at which our research was conducted. Additionally, the e-mail addresses used to approach Dutch councilors had previously been approached for a different survey, which may increase their perception of the legitimacy of this survey. The high Swiss response rate may reflect a high compliance rate with our request to forward the survey to their municipal councilors.

Table 1 Survey Responses

France Netherlands Spain Switzerland UK Total

E-mails sent 967 8216 5200 1434 12.993 28810

Responses Received 87 2.231 563 564 368 3813

Response rate 9% 27% 11% 39% 3% 13%

Table 2 Gender Demographics of Survey Respondents

Female Male Undisclosed Other

25,6% 63,9% 10,3% 0,2%

The distribution of male to female respondents shows an imbalance between female and male respondents. Males are strongly overrepresented, in part because men are overrepresented in politics in all surveyed nations as well (Uberoi, Watson, & Kirk-Wade, 2020; Bouwmans, 2018; SwissInfo.ch, 2018). Finally, some councilors chose not to report their gender and some third gender councilors participated in this study.

There is a slight imbalance between our treatment and control responses, most likely attributed to the software used, in conjunction with the 431 incomplete surveys.

Control Treatment

47% 53%

Table 3 Distribution of Control/Treatment Group

Table 4 Summary of Responses to Survey Question 31 and 33 on Whether Policy Proposals Should Be Implemented in the Respondent’s Municipality from (-2) Totally Disagree to (2) Totally Agree

Min Q1 Median Mean Q3 Max

Econ_1 Con -2 0 1 0,7746 2 2 Econ_2 Con -2 -1 -1 -0,4334 0 2 Econ_3 Con -2 1 2 1,227 2 2 Econ_4 Abs -2 0 1 0,5556 2 2 Econ_5 Abs -2 1 2 1,289 2 2 Econ_6 Abs -2 0 1 1,029 2 2 Soci_1 Con -2 1 1 1,152 2 2 Soci_2 Con -2 0 1 1,08 2 2 Soci_3 Con -2 0 1 0,9976 2 2 Soci_4 Abs -2 0 1 1,047 2 2 Soci_5 Abs -2 1 2 1,441 2 2 Soci_6 Abs -2 1 2 1,459 2 2

The above questions reflect one of four possible policy positions: Economic Concrete, Economic Abstract, Social Concrete, and Social Abstract. Within these policy positions, three policy goals were presented, which can be seen in the included appendix under questions 31 and 33. The range of all the above questions is from -2 to 2, as expected. Except for question Econ_2 Con, the interquartile range varied between 1 and 2. Similarly, excluding question Econ_2 Con, all questions had a slightly higher mean indicating respondents on average were more inclined to agree than to disagree with the provided statements, as will become clear in table 4, the regression table.

The distributions were slightly skewed with questions Econ_2 Con and Soci_1, 2, and 4 all slightly right-skewed and all others slightly left-skewed as the mean and median were not the same.

From the graphical summary we can see that overall, most questions follow a “stair” pattern, with the most frequent response being “totally agree”. Some exceptions to that are question Soci_1 and Soci_2 Con, and Soci_4 Abs, showing a flatter profile indicative of less enthusiastic support for these policies.

Figure 1 Graphical Summary of Responses on Survey Questions 31 and 33, whether Policy Proposals Should Be Implemented in the Respondent’s Municipality from (-2) Totally Disagree to (2) Totally Agree

Analysis

Table 5 Regression Table

Abstract Model 1 Model 2 Model 3 Concrete Model 4 Model 5 Model 6

Intercept 7,0538 *** (0,1425) 7,17122 *** (0,29231) 6,71875 *** (0,36848) 5,1126 *** (0,1427) 3,58875 *** (0,28791) 3,53911 *** (0,36445) Treatment -0,4395 * (0,1955) -0,43883 * (0,19560) -0,44425 * (0,19538) -0,5933 ** (0,1959) -0,61048 ** (0,19317) -0,61075 ** (0,19325) Public Debt -0,02116 (0,04599) -0,03244 (0,04627) 0,26130 *** (0,04306) 0,01017 (0,04576) Risk 0,08832 * (0,04388) 0,26013 *** (0,04340) N 1269 1269 1269 1269 1269 1269 R2 adj. 0,003186 0,005364 0,004964 0,006407 0,03372 0,033 P val. F test 0,02475 0,01224 0,02561 0,002501 < 0,001 <0,001

Note. Ordinary Least Squares regression on data gathered from a survey experiment on municipal councilors in France, the

Netherlands, Spain, Switzerland and the United Kingdom. The policies rated were summed together into a scale from -12 to 12, indicating totally disagree to totally agree with all policies. The treatment was exposure to negative economic prospects.

Model 1 and 4 – OLS Regression

Model 1 is a basic regression of treatment versus control group on councilors’ preference for abstract, blame avoidant policies. The intercept was significantly different from 0 at the 0,05 level. On average, on a scale of -12 to 12, the control group scored the abstract policies a 7,05. The treatment group by comparison scored statistically significantly lower, with a mean difference of 0,44. That represented on average about half a point lower for the treatment group; a small but statistically significant difference. This gives sufficient statistical evidence to state that there is a difference between the treatment and control group. It appeared that information on negative economic prospects did have an effect on preferences for blame avoidant policies. However, hypothesis one (negative economic prospects lead to more credit claiming) was rejected as well. While from the theory we would have expected negative economic prospects to lead to a higher preference for blame avoidant policies, we saw from the results that the treatment group scored those policies lower than the control group. On the side of concrete, credit claiming policies, Model 4 shows that its intercept was also significantly different from 0 at the 0,05 level. This meant that the control group scored the concrete policies a 5,11 on average. The treatment group by comparison scored these policies significantly lower, with a mean difference of 0,59. Hypothesis 2, that negative economic prospects lead to more credit claiming behavior, was rejected as the treatment group ranked these policies lower than the control group as well. After reviewing both models, there are two noteworthy results:

The interesting difference between Model 1 and Model 4 is that there was a difference between the intercepts for both models, on average, concrete policies were scored about two points lower than their abstract counterparts. While we did not see the distinction between the treatment and control group that we had expected, the groups did show a strong preference for blame avoidant policies overall.

Additionally, Models 1 and 4 showed that the treatment group consistently ranked policies lower than their control group counterparts. This showed that while the experimental setup did not have the expected result of dividing the control and treatment group opinion between blame avoidant and credit claiming behavior, it did produce a statistically significant effect on the willingness of councilors to pursue policy initiatives.

With regards to the fit of the model, looking at the adjusted R-squared, which is given to take into account extra parameters in the R-squared calculation, model 1 explains 0,3% of the variance. This is rather small; however, the P-value of the F-test of the model shows that Model 1 is significantly better than an intercept only model. Model 4 explains 0,6% of the variance. This is still rather small; however, the P-value of the F-test of the model still shows that model 4 is also significantly better than an intercept only model. Given the difficulties of excluding all exogenous variables in a social science experiment, a low R-squared is not a cause for concern (Rose & McGuire, 2019).

Model 2 and 5 – Public Debt Acceptance

Model 2 added acceptance of public debt as a variable. Similar to Models 1 and 4, it showed that the intercept was significantly different from 0 at the 0,05 level. The control group in this model ranked the abstract policies on average with a 7,17. The treatment group by comparison ranked these policies significantly lower by 0,44, which meant the model remained robust. The addition of the variable on public debt was not statistically significantly different from 0 for abstract policies.

Model 5 similarly added acceptance of public debt as a control variable but did show a statistically significant effect. A higher acceptance of public debt was statistically significantly correlated with a higher rating for concrete policies with a mean difference of 0,26. That meant that while for abstract policies, acceptance of public debt did not make a statistically significant difference to a councilors rating of a policy, for concrete policies it did. For model 5, the intercept remained statistically significantly different from 0 and dropped to a control group average rating of 3,6. By comparison, the treatment group scored these policies significantly lower with a mean difference of 0,61. We had expected that acceptance of public debt would have a positive effect on the acceptance of both policy types, however, it appeared that only when rating abstract policies did councilors show this behavior. This could potentially be attributed to the possibility that concrete framing may have invoked a clearer sense of the cost involved than it did for abstract policies.

In terms of model fit for Model 2, the adjusted R-squared improved to 0,5% of the variance. Model 5 increased this substantially to explain 3% of the variance. The P-value of the F-test of the model still showed that both model 2 and model 5 were significantly better than an intercept only model.

Model 3 and 6 – Risk acceptance

Model 3 added risk-acceptance as a control variable. Again, the intercept was significantly different from 0 at the 0,05 level with the control group ranking abstract policies an average of 6,72. The treatment group still scored these policies statistically significantly 0,44 lower, which indicates that the model remains robust. A higher acceptance of public debt remained statistically insignificant (-0,03), but acceptance of risk did show a slight positive effect of 0,08 towards the rating of blame avoidant policies.

For Model 6, the intercept remained statistically significantly different from 0 to a control group average rating of 3,54. By comparison, the treatment group scored these policies significantly lower with a mean difference of 0,61. While public debt acceptance was no longer statistically significantly correlated with a higher rating for concrete policies, a higher risk-acceptance did lead to a statistically significantly lower rating for these policies with a mean average of 0,26.

From the theory, we would expect that risk acceptance would increase preference for concrete over abstract policies. While we see from the changing effect size that overall, higher risk acceptance has a stronger effect on concrete (0,26) than on blame avoidant (0,08) policies, both policies are scored higher when risk acceptance is higher. This would suggest that while risk acceptance is more effective in predicting support for concrete policies, risk accepting individuals also have a slightly higher chance to rate other policies higher as well.

As for the model fit, for Model 3, the adjusted R-squared dropped slightly to 0,49% of the variance, but the P-value of the F-test improved to 0,03. At the same time, Model 6 still explained 3% of the variance. The P-value of the F-test of the model showed that both model 3 and model 6 were significantly better than an intercept only model. Conclusions

All of the above models showed a small but significant treatment effect. Policies were structurally rated lower by the treatment group than by the control group, with additional variables nuancing the effect size only marginally. As for the difference between abstract and concrete policies, it appeared that the impact of negative economic information primed our councilors to be more pessimistic in general, rating both abstract and concrete policies lower than their control group peers. For both groups, however, there was a noticeably higher rating for abstract policies rather than concrete policies on average. This showed that a preference for blame avoidance may have reflected more structural than situational preference than we expected for our councilors.

At the same time, we did see smaller differences between the control group and the treatment group in the abstract policies than in the concrete policies. The effect size of 0,44 on a twelve-point scale changed to 0,61 when analyzing concrete policies, with the treatment group rating these policies even lower than they did for the abstract policies compared to the control group average. This shows that there is a slight difference in how the treatment affected preferences for credit claiming policies as compared to blame avoidant policies. While that on its own was not enough to prove our first hypothesis, it provided the groundwork upon which further experimental studies could build.

For all of the above models, the adjusted R-squared remains low. This indicates that external factors that have not been included thus far could have played a role. Nevertheless, it was already expected that the treatment variable would make only a modest impact on this study as the theory it is based on remains difficult to test empirically (Rose & McGuire, 2019). However, the effect was still noticeable, and the analysis produced other valuable insights.

Discussion and Limitations

The research problem was to determine whether there is empirical evidence that the use of framing could influence a councilor’s support towards a certain type of policy. In this case, whether negative financial prospects would have an effect on a preference for blame avoidant or credit claiming policies. The answer to the research question is as follows: the results indicate that negative financial prospects do not appear to generate a shift in preference from credit claiming to blame avoidant policies. Neither do negative financial prospects generate a shift in preference from blame avoidant to credit claiming policies. Nevertheless, the analysis did result in four additional conclusions that offer valuable insight into the interaction between framing and credit claiming and blame avoidance. First, a negative economic frame makes councilors less likely to prefer policies as a whole. Second, councilors are more likely to prefer than not to prefer policy goals. Third, councilors as a whole prefer abstract, rather than concrete policy goals. Finally, risk acceptance and public debt have a smaller effect on preferences for abstract than for concrete policies. While the research may not have proven our hypotheses, overall, the research still showed other results that may be useful for future study.

First, the hypotheses laid out in the theoretical framework predicted that, information on negative financial prospects leads to a preference for more abstract policy goals, negative financial prospects leads to a preference for more credit claiming policies, or negative financial prospects do not have an effect on the preference for either abstract or concrete policy goals. None of the hypotheses are completely true. Overall, the treatment did show an effect, just not the one that was expected. Given negative financial prospects, councilors consistently scored proposed policies more negatively than when given neutral prospects. When confronted with negative financial information, councilors were on average half a point more pessimistic about the provided policy goals. The research is consistent with the literature on framing and showed that it does impact a councilors overall perception of the desirability of proposed policy goals. However, it did not create a marked shift in preference from credit claiming to blame avoidance or vice versa. When councilors were exposed to treatment, they ranked policies lower on average. It just happened across the board, instead of delineated by abstract or concrete.

Second, the average rating for all models was higher than zero. This shows that councilors are in general supportive of the presented policy goals. From the literature, we can infer that the topic of a crisis makes councilors more willing to try various policies, as the focusing event, rising corona infections, brings with it a sense of urgency (Birkland, 1998). This information supports what Jordan and Audia (2012) found about the impact of negative performance metrics on legislator’s propensity to search for, implement, and choose options.

Third, councilors do show a clear preference for abstract, rather than concrete goals across the board, with an intercept that is on average 2 to 2,5 points higher. This is consistent with Weaver’s (1986) prediction that councilors, in general, tend towards blame avoidance more than towards credit claiming. This research adds to the literature by providing the estimated effect sizes of these influences. From the results, we can see that negative prospects have a modest but present effect on councilors’ preference ranking. In line with the literature by Jordan and Audia (2012) on negative performance feedback, given the framing of this study, councilors are likely to respond favorably to abstract policies during times of crisis.

Finally, acceptance of public debt was added as a control variable to potentially explain higher policy ratings that would have otherwise been attributed to our treatment effect. While for abstract policies we did not find a statistically significant effect, there was a small effect for concrete policies. This suggests that concrete policies are more budget-sensitive than abstract policies. Intuitively, abstract policies allow for more interpretation and can be interpreted to suit any budgetary preference, whereas concrete policies have a more definite range of associated prices. Risk acceptance was added as a final control variable to check whether our treatment effect may be correlated to the amount of risk an individual takes in their daily life. We found a small effect for both abstract and concrete policies, suggesting that taking risks does lead to a higher likelihood of supporting concrete policies. This makes sense as concrete policies allow for more credit claiming but are also subject to more risk of blame assignment. Both effect sizes were small, though consistently smaller for abstract policies than for concrete ones. This adds to the literature on framing through the level of abstraction by showing that level of abstraction can be used not just by politicians to gain support from constituents, but also to gain support from politicians for certain causes by strategically using layers of abstraction.

Limitations

Any research faces limitations, including this study. Five areas could be improved to strengthen the generalizability of the results: the explanatory power of the model, the surveyed demographic, the survey settings, the operationalization of blame avoidance and credit claiming, and the focusing event.

First, the explanatory power of the model has room for improvement. The adjusted R-squared remained low throughout our models, explaining only up to 3% of the variance by our variables. This is, statistically speaking, quite low. However, low explanatory power plays a role in many social sciences models, giving rise to discussions about the weight to be given to the R-squared (Rose & McGuire, 2019). Given the results, the largest factor may be the suitability of the treatment variable. Because it was a shared survey, certain choices had to be made to universalize the treatment variable. This may mean that the treatment may not have been optimally suited for proving the exact hypothesis that this research posed. The negative financial prospects were expected to trigger the mode of self-preservation. While it has successfully generated a difference in results between our treatment and control group, it was not demarcated across our concrete or abstract policy goals as hoped. That means that councilors faced with negative financial prospects may not have been sufficiently primed to distinguish between these blame avoidant and credit claiming policies. A future study may want to use additional variables to incite the mode of self-enhancement to see if the results replicate the ones portrayed in this study.

Second, our surveyed demographic may have reduced the generalizability of the study. While our country selection was arbitrary, it was not random. The country selection could have contributed to a skewed sample. Different councilors might have been more or less subject to re-election pressures at the time of the survey. On top of that, our non-response rate, combined with the varying amounts of survey invitations may have also caused some response bias. From the perspective of constituents, much has been written about the demographics of those who are less and more likely to participate in surveys. However, little is known about the specific demographics and characteristics of non-respondents in surveys among politicians. Overall, research does suggest that non-respondents differ from respondents in terms of demographics and personality (Porter & Whitcomb, 2005). As a result, these differences may have contributed to a selective sample of respondents, which may lead to selective results as well. This is an important note for a potential follow-up survey. Instead of a convenience sample, future researchers may want to acquire an administrative probability sample by asking the government to collaborate on the survey.

Third, because respondents were free not to respond to every question, this led to several incomplete responses. The way we designed our regression eliminated these incomplete survey responses to provide the clearest view possible. While there are some methods to correct for non-response, such as imputation methods or multiple imputation methods, these come at the cost of added complexity (Rubin, 2004). In a subsequent or follow-up survey, future researchers could improve the completion rate without such methods by requiring respondents to fill out all questions on the survey. Fourth, using linguistics to determine what constitutes credit claiming and blame avoidance is a novel approach that could be strengthened by comparing it to other definitions of credit claiming and blame avoidance. In this study, the choice was made to use linguistic differentiation to keep the intention behind a policy goal similar enough that councilors would only distinguish between the concrete and abstractness of a policy goal, not based on other merits. In subsequent studies, future researchers may want to explicitly test the relationship between layers of abstraction and credit claiming and blame avoidance to support the theoretical arguments for the relationship with empirical data.

Finally, both a strength and a limitation of this study was its use of the coronavirus epidemic as a focusing event. On one hand, this synchronized and universalized the pressures faced by municipal councilors, on the other hand, a result of this design is that the outcome and results reflect these trying times and may at times differ from what would be theoretically expected under more conventional circumstances. A future study may need to be performed under different circumstances to understand how much of the effect can be attributed to crisis conditions.

Nevertheless, the results offer a first look into the combination of performance feedback theory with blame avoidance and credit claiming theory. To improve the generalizability of this study, a follow-up experiment could be performed under regular conditions, using a finetuned treatment variable, presented to an administrative sample of councilors with mandatory question completion. This would allow us to compare whether the effects shown in this research remain stable when these factors are changed.

Conclusion

To reiterate, this thesis set out to determine whether negative financial prospects would have an effect on a preference for blame avoidant or credit claiming policies. It combined the theories of performance feedback, credit claiming and blame avoidance, and crisis accountability to construct a survey experiment capable of testing the hypotheses that negative performance feedback either increases a preference for blame avoidant or credit claiming policies.

In the research, it became clear that preferences do not appear to change substantially between blame avoidance and credit claiming when exposed to negative economic prospects. So, the short answer is no, preferences do not appear to shift substantially towards blame avoidance or credit claiming when exposed to negative economic prospects. However, the long answer is that negative economic prospects do influence a councilors perception of proposed policy goals. Overall, politicians who were confronted with negative financial news were slightly more pessimistic about enacting policies of any kind. Also, for both treatment and control groups, preference for abstract, blame avoidant policies was substantially higher than for concrete, blame avoidant ones. At the same time, councilors’ attitudes, on the whole, were not neutral but generally favorable towards the presented policy goals.

Performance feedback theory offers insights into framing performance feedback constructively so that politicians take the necessary actions to improve their performance. This research adds empirical results to support the conclusion that under uncertain focusing event circumstances and faced with negative performance feedback, councilors may become more hesitant to take action. Compared to councilors faced with neutral information, who are more supportive of policies in general. We propose that blame avoidance can help alleviate this deadlock and could help enact unpopular but required policies.

To come back to our introduction, how does this research help the PvdA, or for that matter any political party? This study has provided another look into crisis accountability, performance feedback, and the theory of blame avoidance and credit claiming. It integrates the idea of blame avoidance as a part of the policy feedback cycle, which helps us understand both in a more nuanced way. A way that allows us to deploy it strategically, to manage public emotions, and to create opportunities for politicians to enact the policies that are necessary to overcome the crisis at hand. As we know from the literature, blame avoidance is valuable for politicians, but this study argues that it is also useful for the public. It can play a big role in productively managing public emotion in a way that allows politicians to confidently enact the policies necessary to get out of a crisis. Understanding this value to the public in a crisis accountability scenario may well help destigmatize its use.

At the same time, while blame avoidance is a powerful tool for passing unpopular spending cuts, credit claiming is a powerful tool for building support from constituents. Credit claiming might well be encouraged by the improved borrowing conditions that allow for substantial lending at little cost. This could well bring an end to the era of permanent austerity (Pierson, 1996; The Economist, 2020). An interesting experiment could be done with positive economic prospects to test this hypothesis further. Based on these conclusions, performance feedback providing organizations like the relevant bureau of statistics or the applied policy research institutes should consider framing performance feedback carefully taking into account the risk of triggering the mode of self-preservation. This may help politicians make decisions more easily when they are most needed.

While future studies may wish to use a different variable to coax councilors into the mode of self-enhancement or use different variables to define blame avoidance and credit claiming, this research provides a promising basis for further exploration into connecting public administration theories like blame avoidance and performance feedback theory into a comprehensive understanding of political reality.

Bibliography

Antonacopoulou, E., & Sheaffer, Z. (2014). Learning in Crisis: Rethinking the Relationship Between Organizational Learning and Crisis Management. Journal

of Management Inquiry, 5-21.

Baekgaard, M., & Serritzlew, S. (2015). Interpreting Performance Information: Motivated Reasoning or Unbiased Comprehension. Public Administration

Review, 73-82.

Birkland, T. A. (1998). Focusing Events, Mobilization, and Agenda Setting. Journal

of Public Policy, 53-74.

Bonoli, G., & Natali, D. (2012). The Politics of the New Welfare State. Oxford, UK: Oxford University Press.

Bouwmans, H. (2018). Vrouwen in het lokaal bestuur. Den Haag: Nederlandse Vereniging voor Raadsleden.

Brändström, A., & Kuipers, S. (2020). Accountability and Blame Avoidance After Crises. Oxford Research Encyclopedia of Politics, 1-24.

Brändström, A., & Swinkels, M. (2015). Crisis accountability and career management in the Netherlands. In N. Schiffino, L. Taskin, C. Donis, & J. Raone, Organizing

After Crisis: The Challenge of Learning (pp. 131-158). Berlin: Peter Lang

Publishing Group.

Cyert, R., & March, J. (1963). A Behavioral Theory of the Firm. Mansfield Centre, CT: Martino Publishing.

Debela, K., & Meissner, R. (2020). Local governance in Switzerland: Adequate municipal autonomy cumintergovernmental cooperation? Cogent Social

Sciences.

Direccion General de Cooperacion Local. (2012). Local Government in Spain. Madrid: Ministerio de Administraciones Publicas.

Fox, J., & Weisberg, S. (2019). An {R} Companion to Applied Regression, Third

Edition. Thousand Oaks, California, United States of America: Sage.

Giger, N., & Nelson, M. (2010). The electoral consequences of welfare state retrenchment: Blame avoidance or credit claiming in the era of permanent austerity? European Consortium for Political Research.

Government Finance Officers Association. (2020, June 23). S.M.A.R.T. Goals. Retrieved from Government Finance Officers Association: gfoa.org/smart-goals Grimmer, J., Messing, S., & Westwood, S. J. (2012). How Words and Money Cultivate a Personal Vote: The Effect of Legislator Credit Claiming on Constituent Credit Allocation. The American Political Science Review, 703-719.

Hansson, S. (2015). Discursive strategies of blame avoidance in government: A framework for analysis. Discourse & Society, 297-322.

Hansson, S. (2019). Brexit and Blame Avoidance: Officeholders' Discursive Strategies of Self-Preservation. In V. Koller, S. Kopf, & M. Miglbauer, Discourses of Brexit (pp. 12-24). Routledge.

Hendrickx, F. (2020, Juni 25). Het draait weer om geld in de Kamer: verdienen ‘onze zorghelden’ loonsverhoging? de Volkskrant.