Aiming the Blame: How Negative Performance Feedback Influences Municipal Councillors’ Responsibility Attribution

(1)

Aiming the Blame:

How Negative Performance Feedback Influences Municipal

Councillors’ Responsibility Attribution

Inge van de Scheur, S1526367 Master thesis

Public Administration: Public Management Leiden University

Supervisor: Joris van der Voet 5 July 2020

(2)

Chapter 1: Introduction

1.1 Responsibility attribution

Responsibility attribution is a topic important for democratic accountability to function, especially as bureaucrats are increasingly confronted with performance information about their functioning (Nielsen & Moynihan, 2017). As Hamilton discusses, “responsibility refers to a decision about liability for sanctions based on a rule” (Hamilton, 1978, p. 316).

Previously, most research that has been done has focused on the attribution of responsibility and blame by voters, such as Tilley and Hobolt (2011) and Shabad and Slomczynski (2011). Other research, like ‘the Blame Game’ by Christopher Hood, has focused on blame avoidance. This relates to responsibility attribution, because part of

attributing responsibility is seeing whether people view themselves as responsible or whether they ‘avoid blame’ – and whether this is contingent on their or their organisation’s level of performance. In Nielsen and Moynihan’s research (2017), politicians were more likely to attribute responsibility to bureaucrats when performance was low. This points to a negativity bias also acknowledged by Hood (2011). The above named research that has focused on responsibility attribution by decision makers, assumes that these decision makers are rational actors who act as problem solvers, an idea that stems from organisational theory (March & Simon 1958). Jordan and Audia (2012) dispute this assumption and propose the possibility of decision makers being self-enhancers – people who put their own interests and image first – rather than being problem-solvers solely. In this thesis, responsibility attribution by decision makers, and in particular municipal councillors, will be researched, and the possibility that decision makers are not always problem-solvers – but can also be self-enhancers – will be explored.

Furthermore, responsibility attribution consists of two parts: the people who attribute responsibility judge not only what the judged actor did (causal responsibility), but also what

(5)

that actor was supposed to do (role/functional responsibility) (Hamilton, 1978). In this thesis there will also be a differentiation in what ‘direction’ municipal councillors attribute

responsibility. The introduction of upward and downward responsibility attribution is new in (behavioural) public administration and as such, it is a theoretical innovation this study attempts to explore and prove. The question, with the addition of direction, is if the

respondents do not see themselves as responsible for the results presented in the performance feedback, will they attribute responsibility ‘down’ to the bureaucrats or organisations who are executing political decisions, or will they rather attribute responsibility ‘up’ to the national government, which influences the political decisions that are being made? When such questions are asked, upward and downward responsibility attribution can become embedded in existing literature and the direction in which these relationships occur can be researched.

1.2 Empirical context: budget shortages in youth care

The domain of youth care was chosen as the empirical context for this research. From 2015 onwards, youth care was decentralised, shifting the primary responsibility for youth care from the national government to individual municipalities. This decentralisation was realised because it was thought that municipalities would be able to help youngsters better and quicker than the national government. This would prevent problems for the youth and having early interventions for existing problems would (in turn) both prevent the worsening of problems as well as preventing long waiting times (Jeugdwet, art. 2.1). Since the idea was that larger problems would be prevented, the budgets for youth care were cut each year from 2015 to 2018. Thus, simultaneous to the decentralisation shifting the primary responsibilities for youth care to the municipalities, the budgets available for youth care were cut. In total, fifteen percent of the youth care budget that was available before the decentralisation was cut (Van der Kaaden, 2019).

(6)

However, as municipalities devoted more time to prevention, they found more youngsters who needed help (Van Zanen and Bruls, 2019). This meant that although the budget was cut, the demand for youth care increased. Hence, the new budget was insufficient for the majority of the municipalities. Therefore, the association of Dutch municipalities (VNG) sent a letter to Parliament in May 2019 declaring the urgency of increased funding for youth care so that municipalities could (continue to) carry out their responsibilities adequately (Van Zanen & Bruls, 2019). In October 2018, the Minister of Health, Welfare and Sports, Hugo de Jonge, had already ordered research into the budgetary struggles of municipalities concerning youth care. At the end of April, he sent this research to Parliament with an

accompanying letter (De Jonge, 2019). Research was conducted on three topics; an analysis of the municipalities who requested help from an emergency fund for municipalities, an analysis of the development of the volume of youth care, and a benchmark analysis of youth care spending in 26 municipalities. The overarching goal was to find causes for budget shortages concerning youth care in municipalities (Batterink et al., 2019). Taking this research into account, the VNG increasingly worried that the municipal budgets would not be sufficient to incorporate the increasing volume and costs of youth care, and they worried that they would have to return their obligations to the state – which would nullify the efforts to provide a better quality of youth care (Van Zanen & Bruls, 2019). As Minister de Jonge realised from the research, municipalities were in need of increased funding and hence, the budget for youth care was increased. As a consequence, Parliament promised to supplement the municipalities’ youth care budgets by 420 million euros in 2019 and 300 million euros in both 2020 and 2021 (De Jonge and Blokhuis, 2019).

That the majority of the municipalities has overspent their budget makes the domain of youth care suitable to carry out a research project that is interested in the consequences of negative performance feedback. In this study, negative performance feedback consists of

(7)

actual information on the (over)spending of municipalities on their youth care budget. Two types of negative performance feedback will be distinguished in this study: Historical and social performance feedback. These types of performance feedback will be further elaborated on in chapter 2. Another reason that youth care is a policy domain suited to research

responsibility attribution and the impact of (negative) performance feedback is that youth care encompasses various levels or audiences to which responsibility can be attributed. The

municipalities namely decide on the budget for youth care, but they are dependent both on the national government for the allocation of subsidies and on bureaucrats and youth care

providers who execute youth care policy. Hence, all three actors could potentially be seen as responsible, as they all have influence on youth care, albeit in different ways. The

involvement of these different audiences in the policy making and execution of youth care thus allows for the introduction of direction in the allocation of responsibility.

1.3 Scope and research question

This thesis researches if negative performance feedback affects responsibility attribution by municipal councillors, and if so, to which actor do they attribute responsibility? The study is directed towards municipalities, as they are by law responsible for many aspects of youth care (Jeugdwet, art. 2.6), and because the shortages exist at the municipal level. Youth care is thus a case that is used to look at the more general topic of responsibility attribution. The survey is not directed at all employees or otherwise involved people within a municipality, but is especially targeting municipal councillors, as one of the tasks of municipal councillors is to approve the municipal budget (VNG). Nielsen and Moynihan discovered a negativity bias in their research of responsibility attribution by municipal politicians, which will also be considered in this thesis, albeit in a different way. This leads to the following research question:

(8)

How does (social and historical) negative performance feedback affect responsibility attribution by decision makers and how does the decision makers’ membership identity influence this relationship?

1.4 Relevance and importance

1.4.1 Scientific relevance

The first contribution of this thesis is that it provides more evidence on responsibility attribution by decision makers. As mentioned previously, many studies have focused on responsibility attribution by voters: In these studies, voters were asked to judge which politicians they deemed responsible for certain issues. An example is the study by Tilley and Hobolt, in which people were given information about the government and then had to state how positive or negative this information was (2011). Nielsen and Moynihan finally did examine how politicians attributed responsibility (2017). They surveyed municipal politicians and found that these politicians mainly attributed responsibility to bureaucrats executing policy when said policy failed. This means that they found a negativity bias (2017). Nielsen and Moynihan thus found that municipal politicians attribute responsibility ‘down’ to bureaucrats who execute policy, while they also assumed that elected officials confronted with performance data will attribute responsibility to public sector leaders more when performance is low (2017). This thesis thus provides more evidence on responsibility attribution and also examines the negativity bias – although this research does not compare positive and negative information, but rather compares historical and social negative information, with negative social performance information being ‘double’ negative information.

The second contribution is that a new distinction between upward and downward responsibility attribution is introduced. Nielsen and Moynihan assume that responsibility

(9)

attribution predominantly happens in one direction, which is downward. They research how “politicians attribute responsibility to those who actually deliver public services” (Nielsen and Moynihan 2017, p. 269). However, this study researches not only whether responsibility is attributed, but also asks whether municipal politicians attribute responsibility to ‘lower’ bureaucrats when faced with adversity, or whether there is also a possibility that they look ‘up’ to national politics? The latter is a realistic scenario as municipalities are dependent on national politics for their budget: the municipalities receive money from the municipality fund (Gemeentefonds) to function as an organisation and to provide their citizens with a certain standard of services (Rijksbegroting). Hence, the national government influences the services of the municipalities, which could be a reason for municipal councillors to also attribute responsibility upward. A distinction between these two ‘directions’ of responsibility attribution has not been researched in the discipline of (behavioural) public administration, which makes this study innovative to the field of behavioural public administration.

In addition, this study researches if the membership identity of municipal councillors influences whether they attribute responsibility either upward or downward. Therefore, this thesis will explore whether the municipal councillors that are part of the coalition in their municipality attribute responsibility differently than the councillors that are part of the opposition. The expectation is that members of the coalition will feel more liable for

overspending in the municipality as they have had influence on budgetary decisions (VNG) and will thus attribute responsibility more up and down (Jordan & Audia 2012) compared to the opposition, who are expected to sooner attribute responsibility to the municipality itself. As the national government has discussed youth care budgets and the municipalities have been requesting higher funding from the national government, it is also expected that the coalition attributes responsibility upward, even though this goes against Hood’s theory of blaming downward (2011). Members of the opposition are expected to not feel as liable for

(10)

the municipality’s overspending as members of the coalition, as they have not had as much influence on policy and budget as the coalition has had. Therefore, the opposition is expected to attribute more responsibility to the municipality itself compared to the coalition.

The third contribution of this thesis is its experimental design. The use of a survey experiment to explore the reactions of municipal politicians to performance information ensures that this thesis can be categorised in the relatively new field of behavioural public administration (Grimmelikhuijsen 2017). In this field, the theory of public administration is often combined with the methods of psychology, in order to gain more insight not only into how public servants act in certain ways, but also why they do so. These psychological insights can consequently be combined with more practical insights which could potentially alter behaviour (of decision makers) in the public sector. In the survey experiment conducted for this study, actual information on municipalities’ budgets was used and the respondents were actual decision makers in their municipalities.

1.4.2 Practical and societal relevance

Next to a scientific relevance, this thesis has a practical and societal relevance. Firstly, it makes a difference for policy making who politicians attribute responsibility to, and consequently, who they blame. For the government, it is not favourable when municipal politicians ardently express their discontent with the government for several reasons: when municipal councillors are discontent, they seek aid of the government – as the letter of the VNG made clear in this case (Van Zanen and Bruls, 2019). Moreover, it might cause general legitimacy issues when the national government is attributed responsibility for a policy that negatively impacts the people. Additionally, the research in this thesis is relevant for the case of youth care, as it still unfolding. There was a first evaluation of the Jeugdwet in 2018, but after that the VNG came with a letter to bring to attention the severity of the budget shortages and eventually Minister de Jonge promised an increase in budget (De Jonge & Blokhuis,

(11)

2019). In 2020, there are plans to adjust the Jeugdwet by including regional cooperation by law (De Jonge & Dekker, 2020). This means that additional research can be done when the policy is in place longer and adjustments are made.

Finally, this thesis has a societal relevance as its outcomes can illuminate how people can potentially more effectively advocate for better youth care. As Jordan and Audia argue, decision makers might not (just) be looking to improve performance when performance is low (problem-solving mode), but they might also strive to see themselves as decision makers who make the ‘right’ decisions (self-enhancing mode) (2012). Depending on what ‘mode’ decision makers are in, strategies to approach these decision makers to effectively bring about change might differ.

1.5 Research approach and thesis structure

In order to research responsibility attribution by municipal politicians, a survey experiment was conducted among the municipal councillors of 330 municipalities in the Netherlands. In total, the survey was sent to 7819 municipal councillors. Of this group, 1162 councillors responded with a complete response. After having given some background information in chapter 1, chapter 2 will review relevant literature, discuss relevant concepts and will conclude with theoretical hypotheses. Chapter 3 will discuss the methodology used and will operationalise the concepts of the theoretical framework. In addition, the selection of the municipalities will be discussed. Chapter 4 consists of the results of the survey experiment and an evaluation of the hypotheses. The fifth and final chapter will consist of a discussion and conclusion of the research findings, implications, limitations and an evaluation of the validity, and suggestions for future research.

(12)

Chapter 2: Theory

This chapter will focus on relevant concepts and theories for this thesis. Firstly, the dependent variable, responsibility attribution, will be discussed. The concept of responsibility attribution will be explained as well as how the concepts of blame and accountability relate to it. The similarities and differences between these concepts will be explained, eventually bridging them in order to come to one comprehensive definition that will be used for the rest of this thesis. After that, direction (upward and downward) is added to this definition. In paragraph 2.2, the independent variable, negative performance feedback, will be discussed and the various definitions and usages of performance information will be reviewed. The chapter concludes with two hypotheses and an explanation of the moderating variable, the coalition membership, or membership identity, of municipal councillors.

2.1 Responsibility (attribution), Blame, and Accountability

In 1958, Fritz Heider published ‘The Psychology of Interpersonal Relations’, in which he investigated ‘common-sense’ psychology and he introduced the concept of responsibility attribution. Attribution is described as the perception people have when they are trying to understand the causes of events and unravel where the responsibility for the outcomes of these events lies (1958). In many articles, either blame and responsibility are not clearly defined, the difference between them is not clear, or the concepts are used interchangeably. This paragraph will attempt to clarify these concepts and the differences between them.

In the article that this thesis – amongst other things – aims to contribute to, Nielsen and Moynihan use the term ‘responsibility attribution’ when describing a ‘mental’ step that comes before the actual act of blaming (Nielsen and Moynihan 2017). Other authors do not consider responsibility attribution as a step preceding blame, but rather as an integral part of

(13)

blame (Shaver 1985; Hood 2011). According to Christopher Hood, blame is “the act of attributing something that is considered bad or wrong to some person or entity” (Hood 2011 p. 6). He states that therefore there are two elements necessary in order to blame; perceived avoidable harm or loss and perceived responsibility or agency (2011, p. 6). The former points to the severity of the consequences of an actor’s actions: how different the consequences would have been had another action been taken. The latter points out that the harm done is caused by the actions or inactions of an actor, organisation, or system, who people perceive could have influenced the course of action, and thus who could have avoided the harm done (Hood 2011 p. 6). Shaver (1985) used the term ‘blame assignment’ to describe the same process. She states that blame assignment is “the outcome of a process that begins with an event having negative consequences, and it involves judgements about causality, personal responsibility, and possible mitigation” (Shaver 1985, p. 4)

As was also stated in the introduction, “responsibility refers to a decision about liability for sanctions based on a rule” (Hamilton, 1978, p. 316). According to Hamilton, people not only make judgements about what an actor did to cause an event (causality), but they also make judgements about what these actors were supposed to do. People make these judgements based on their personal expectations or morals, or on based on the role that the ‘judged’ actor possesses. These elements comprise role or functional responsibility

(Hamilton, 1978, p. 316).

Kelly Shaver in her book compares a legal definition of responsibility to a moral one. She states although it seems straightforward that an actor would be responsible for the events that he causes, it is not actually that simple and causality and responsibility depend on cultural factors and ‘hidden’ circumstances as well. She describes a doctor’s judgement of someone’s mental state to explain someone’s actions and economic crises as the cause for

(14)

the legal understanding of responsibility would render an actor not responsible, while actor might still be seen as responsible from a moral perspective (Shaver 1985). The definition of responsibility Shaver eventually comes to consists of five elements: 1. A (causal) connection between an actor and an occurrence, 2. “a generally accepted set of moral principles by which that occurrence is considered harmful” (p. 70), 3. the perception that the events at least

partially are caused by human action, 4. the actor chose voluntarily to either do something that caused harm, or chose to not act which caused harm, and 5. an examination of the situation; are there circumstances that might render the actor not responsible?

Shaver and Hamilton include the same elements in their definitions of responsibility, although Shaver has a more elaborate version of what responsibility entails. Both authors state that responsibility depends firstly on a causal connection between the actor and the event, and secondly on a moral notion, which Hamilton describes as role or functional responsibility, and Shaver elaborates on in the second and fourth elements of responsibility. When looking at Hood’s definition of blame, responsibility is one of the two elements that comprises blame. Next to perceived responsibility or agency, Hood includes perceived avoidable harm or loss in his definition of blame. The third, fourth and fifth elements of Shaver’s understanding of responsibility have a similar notion: the third element states that in order for someone to be responsible, the events has to be perceived to at least be partially caused by human action – if there were outside circumstances that were beyond control of the actor, such as a natural disaster, the actor cannot be seen as responsible (1985). The fourth element states that in order for an actor to be responsible, the actor should have willingly acted or refrained from action, and thus resonates with Hood’s understanding of agency (Shaver 1985; Hood 2011). Lastly, the fifth element of Shaver’s definition asks whether there are any factors that did not fit the first four elements, yet still could have exempted the actor from being responsible (1985).

(15)

When examining the wording of the literature on blame and responsibility, it seems that these concepts are often similar. Some authors, however, consider responsibility attribution a part of blame or even as a step that precedes blame. The latter is the case in Nielsen and Moynihan’s article (2017). Consequently, in this study the concept of

responsibility (attribution) will be used to describe a preceding step of blame that connects the actor to the event and perceives that the actor ‘ought to have done’ something. The reason for this is that this reflects both causality and morality. Also, responsibility will not be seen as an integral part of blame but rather a step preceding blame, as blame has consequences for actors and responsibility attribution does not. Responsibility attribution then can be defined as the mental process of attributing responsibility to an actor that is firstly perceived to have been connected to the event that he or she is seen as responsible for and secondly that actor is perceived as if he or she was supposed to have acted.

2.1.2 Adding direction to responsibility attribution

Nielsen and Moynihan (2017) researched how elected officials attribute responsibility to the bureaucrats that execute the policies that the elected officials created. Nielsen and Moynihan did this because they assumed that politicians are more constrained in their options to assign responsibility, contrary to voters, whose attribution of responsibility and blame has been more widely researched (Tilley and Hobolt 2011; Shabad and Slomczynski 2011). This is because voters have the option to assign responsibility and blame to the different political parties, but also to other actors (Nielsen and Moynihan 2017). Politicians, on the other hand, have less options and are expected to primarily attribute responsibility to the bureaucrats who execute their decisions (2017 p. 270). Weaver (1986) describes strategies that politicians follow in order to avoid blame, and one of these strategies is to put the blame on other people. In his book “Organizational learning from performance feedback”, Greve states that managers might attempt to evade responsibility by differentiating performance aspiration levels; when

(16)

performance is low, managers might compare their organisation with worse-performing organisations, which would lower the aspiration level and thus would make the organisation’s own performance seem better in comparison. He calls this a ‘self-enhancement rule’ (2003).

In the context of multi-level governance, the assignment of blame may change from a one directional movement ‘down the line’ (Hood 2011) to a multidirectional concept when researching responsibility attribution by municipal councillors. In the Netherlands, the national government determines much of what the municipal authorities can do: They not only create laws that dictate the responsibilities for each level of government (national, provincial, and municipal), but they also allocate subsidies and budgets for the lower levels of government. Hence, responsibility can not only be attributed to those bureaucrats or

organisations that are executing policy, but there is another group to whom responsibility can be attributed when researching (municipal) politicians. Namely, councillors can attribute responsibility to the decision makers ‘above’ them: certain national politicians or the government in general. Since the national government can influence municipal budgets and responsibilities, municipal councillors can attribute responsibility upward, for decisions that impacted them negatively, to the national government.

In accountability research, upward, downward, and horizontal accountability have been researched already. As these concepts seems similar to the idea upward and downward responsibility attribution, these concepts will be compared. In the literature on organisational behaviour, the concepts of upward and downward accountability have been developed. Ebrahim (2003) describes this in the context of the relationship NGO’s have with their

funders. Accountability measures (such as financial reports) are not only created by funders to track the use of the money that they provide to the NGO’s, but they are also used by the NGO’s to leverage more funds by advertising their programmes and putting the legitimacy of the funders at risk (2003). This creates an (asymmetrical) interdependence. In the case of

(17)

NGO’s, upward accountability thus means that NGO’s are accountable to the people or organisations that fund them: NGO’s have to show that they spend the funding for the intended purposes (Edwards and Hulme 1996). Downward accountability in this case constitutes the accountability that the NGO’s have towards the groups (of people) that they are aiming to help (Ebrahim 2003). Horizontal accountability refers to the accountability of the organisation towards itself; for example to its own employees.

One relationship studied in this thesis, which is between municipalities and the national government (their funder), can be described in a similar fashion. Municipalities receive funding from the national government and create budgets in order to show the national government that they are spending this money well. In turn, the municipalities can voice their opinions on the height of the funding and thus create leverage to ask for more funding from the national government. Creating direction in responsibility attribution differs from the direction in accountability in some ways. With upward accountability, the

municipalities are accountable to the national government which provides funding – i.e. inputs. With upward responsibility attribution, in the case of youth care, this relationship is reversed, as the municipalities state that the national government (the funder) is responsible and should thus be (held) accountable and part of that is when other parties or actors attribute responsibility to the national government. Upward accountability can thus be seen as the opposite from upward responsibility attribution, although the term accountability will not be used further in this thesis.

The concept of responsibility attribution by Nielsen and Moynihan (2017) is thus expanded in this thesis by adding direction. Attributing responsibility upward is when the municipal councillors attribute responsibility to the national government (which provides inputs), and attributing responsibility downward is when the municipal councillors attribute responsibility to bureaucrats or contracted organisations who execute policy.

(18)

2.2 (Negative) performance feedback

Most research on how decision makers react to performance information is focused on performance feedback as a catalyst for change; negative performance feedback is expected to drive decision makers to search for solutions (Nielsen, 2014), in what Jordan and Audia call a problem-solving mode (2011). This thesis however does not assume that municipal politicians are decision makers who solely are problem solvers, but it researches the possibility of

decision makers, in this case municipal councillors, being seekers of self-enhancement. In order to research the possibility of decision makers behaving as self-enhancers, this thesis first needs to establish a framework in which performance feedback itself as well as its uses and effects are defined. This section will attempt both.

2.2.1 Defining (negative) performance feedback

According to Greve (2003), performance feedback emerges from aspiration levels that are set first. These levels are set as a means to reach a goal, in order “to assign performance levels to the success and failure categories favoured by boundedly rational decision makers” (Greve 2003, p. 42). These aspirations levels can either be historical or social.

Historical aspiration levels consist of levels of past performance as a goal for the future, or they are even seen as an indicator for future performance. From an organisation’s past performance a standard can be deducted for how the organisation should perform in the future. Historical information is suited to predict an organisation’s performance, but it does not account well for outside factors: the context in which an organisation operates may have changed significantly in the time that the performance feedback was finalised and hence, the historical aspiration levels might not be realistic anymore. Consequently, historical

information is not always useful. It is useful when inside information on the organisation is more useful than information on the whole sector or comparable organisation, or when information on the sector or other organisations is not available (Greve 2003).

(19)

Social aspiration levels do not solely rely on the past performance of the own organisation, contrary to historical aspiration levels. Rather, social aspiration levels rely on the comparison between the own organisation and organisations that are similar (Greve 2003). A drawback of social comparisons is that it might be difficult to obtain all relevant

information from other (rival) organisations, which can lead to a distorted image of other organisations’ performance, as organisations do not publish all information about themselves and the media do not always present a comprehensive, accurate picture (Greve 2003). Still, the “individual drive for external evaluation is so strong that a decision maker may decide to ignore validity problems with the information at hand” (Greve 2003, p. 46). Social

comparisons have the effect that managers attempt to reach aspiration levels that are higher than the performance levels of comparable organisations, and the aspiration levels are set higher than if managers use historical comparisons (Greve 2003).

Taking the different kinds of aspiration levels into account, performance feedback as used in this thesis can be defined as the process of providing information about how an organisation has performed relative to historical or social aspiration levels. As the historical and social aspiration levels provide a framework for success and failure, they can later be used to evaluate performance and are thus a way of providing performance feedback. The effects of this feedback will be discussed in the next section.

2.2.2 Uses and (unintended) effects of performance feedback

For an organisation, using performance information is a way to achieve results and to be accountable. According to Herbert Simon, performance information is a simple way for citizens to form an opinion about the government, or organisations in general (1955). With the philosophy of new public management rising in the 1980s, public sector organisations felt the pressure to show accountability based on performance information (results accountability).

(20)

Performance information would thus not only help achieve an organisation’s objectives, but it would also create accountability (Speklé and Verbeeten 2009).

These two uses of performance feedback yield various effects. According to Prewett (1987), the numbers that make up performance feedback are never neutral: Performance information, and therefore performance feedback, thus always advance a certain goal. Prewitt states that voters will use ‘retrospective evaluations’ (performance feedback) of policies as a way to determine who to vote for. Politicians who already had their positions in times of crisis will far more likely be rejected than politicians who ruled when performance was good (1987, p. 114-115). When policies would perform well, political credit would be claimed. When performance would be worse than the set goals, however, political blame would be assigned (1987). Consequently, Prewitt states that both positive and negative performance feedback advance the goal of contributing to democratic accountability, as they have impact on voters’ choices (1987). Nielsen and Moynihan (2017) describe that performance information and feedback could lead to responsibility attribution in a similar fashion as Prewitt does. They give two causes: Firstly, performance information creates more certainty, which makes it easier to give judgements. Secondly, politicians might be subjected to blame themselves for performance below standards when performance information is made available to the public.

As Prewitt argues, positive and negative performance feedback each have a corresponding positive or negative effect (1987). More recently, however, scholars have picked up on an asymmetry in the reaction of people towards either positive or negative performance feedback: Multiple scholars, such as Hood (2011) and Olsen (2013) state that people give more attention to negative information. This is called a negativity bias. Olsen states that in research on responsibility attribution by citizens and their views on performance information, a negativity bias is prevalent. However, in more recent research, it was found that this negativity bias could be mitigated by providing different sources of information: the

(21)

citizens were found to be less likely to demonstrate a negativity bias when they had multiple sources of information (Olsen 2013). Hood also describes the negativity bias in regards to the risks that politicians are willing to take:

potential losses are commonly weighted more heavily than equivalent gains, and (as noted in the first epigraph) losers are more likely to notice and act on the basis of their potential or actual losses than gainers are to notice and act on equivalent gains” (Hood 2011, p. 10).

Hood then argues that this induces a ‘risk averse’ attitude in politicians, as they feel more afraid to fail than they will be praised for success. (2011). This negativity bias was also found by Nielsen and Moynihan (2017).

When assuming that people are risk averse even in situations of low performance, this has implications for long-held assumptions in organisational theory. Research of how decision makers react to performance information emerged with the behavioural theory of the firm by Cyert and March in 1963. A main assumption was that decision makers set historical and/or social aspiration levels, and that decision makers adapted their actions depending on whether they achieved these aspiration levels or not. If performance was below their (historical or social) aspiration level, it was thought that decision makers would actively adapt their strategies, while if performance was above the aspiration level, decision makers would decrease their actions (Jordan & Audia 2012). As Greve stated, decision makers would

increase ‘search, change and risk taking’ when performance was below the goal, and decrease search, change and risk taking when performance was above the goal (2003).

Also important for organisational theory were March and Simon (1958), who developed the idea that decision makers are not fully rational. They argued that in fully rational situations, decision makers would solve problems according to the information they have. A problem for this idea is that decision makers are almost never presented with

(22)

complete information. In addition, decision makers will seek alternative actions if their course of action is unsatisfactory. According to March and Simon, “Most human decision-making, whether individual or organizational, is concerned with the discovery and selection of satisfactory alternatives; only in exceptional cases is it concerned with the discovery and selection of optimal alternatives” (1958, p. 140-141). Hence, decision makers’ bounded rationality leads them to ‘satisfice’: the decision makers attempt to merely meet the goals that they have set, instead of rationally maximising performance and thus overachieving

compared to their goals. They will alternate and evaluate their strategy until a strategy has been found that can satisfy the goal (1958). Jordan and Audia elaborate on this idea and state that not only may decision makers satisfy a goal due to bounded rationality, but they might even rationally be opting for a strategy that benefits them personally, rather than the

organisation (2012). One of the reasons for this is self-enhancement: the desire to see oneself in a positive light, even if this does not correspond with actual performance. This process of self-enhancement takes place when decision makers face performance below aspiration levels: when goals are being met, decision makers do not have to distort their self-image in order to see oneself in a positive light. This strategy is only necessary when performance levels are low. Nielsen and Moynihan also state that when elected officials are confronted with performance data, they will attribute responsibility to public sector leaders more when performance is low (2017). However, they also state that municipal politicians attribute responsibility in a downward direction, to the actors or organisations who execute policies (2017). It is unclear whether the public sector leaders are necessarily seen by Nielsen and Moynihan as above or below elected officials. The direction of responsibility attribution is thus still disputed. In addition, decision makers would process positive information about themselves more easily than negative information. Hence, they would take sooner take credit

(23)

for their successes than take the responsibility for their failures, and instead attribute responsibility for failures to outside influences (Jordan & Audia 2012).

Thus, performance feedback can aid in achieving aspirational goals of an organisation as well as creating accountability. Performance feedback use was traditionally thought to either increase problem solving or remaining at a status quo, with different ideas on how well decision makers could evaluate. However, this thesis assumes that decision makers are not only boundedly rational, but they can also intentionally pursue personal goals, rather than pursue the goals of the organisation, and take credit for successes while assigning

responsibility for losses to other actors. Moderating variable

Next to the independent variable of negative performance feedback and the dependent variable of responsibility attribution, this thesis proposes a moderating variable. This moderating variable is membership (identity); is the municipal councillor a member of the coalition or the opposition? This could influence the relationship between performance information and blaming (up or down) as coalition members might attribute responsibility more up and/or down than opposition members. As described earlier in this paragraph, decision makers might look to take credit for their successes, but they might try to evade responsibility for their failures (Jordan & Audia 2012). From this assumption it can be deducted that members of the coalition in the municipal council feel more responsible for failures than opposition members, as the coalition has more influence in the chosen policies. As a consequence of decision makers being self-enhancers (Jordan & Audia 2012), members of the coalition are expected to attribute more responsibility to outside influences than the opposition might do. The literature is still ambiguous about the direction (upward or downward) in which decision makers would attribute more responsibility. Attributing

responsibility in a downward direction has been mostly researched and assumed (Nielsen and Moynihan, 2017; Hood 2011). However, in the case that this study focusses on, the national

(24)

government has (had) much influence on the policy making of municipalities. Therefore, it is expected that coalition members attribute more responsibility to outside influences (upward and downward) compared to the opposition. In addition, it is expected that of this attribution to outside influences by the coalition members, the most responsibility will be attributed in an upward direction.

From the literature discussed in this chapter, two hypotheses were formed.

Hypothesis 1

Negative performance feedback increases decisions makers’ upward and downward responsibility attribution.

Hypothesis 2

2a Municipal councillors who are members of a coalition party attribute more responsibility upward

2b Municipal councillors who are not a member of a coalition party attribute more responsibility to the municipality

Figure 1

Theoretical model of the proposed relationships

Membership Negative performance feedback Responsibility attribution Upward Downward

(25)

Chapter 3: Methodology

3.1 Research design

A survey experiment was conducted in order to research the influence of performance

feedback on municipal politicians. A survey experiment is fitting to research this influence as the idea of decision makers being self-enhancers is quite recent. Hence, not much research on this topic has been done. Moreover, survey experiments are a frequently used tool in the domain of behavioural public administration. One reason to use survey experiments is that it is a way to prove causal relationships, as the experimental setting allows all factors but one to remain the same (Krupnikov & Findley 2018). Simultaneously, the research does not suffer from decreasing generalisability as a survey experiment still allows a large population to be surveyed. For this study, nearly all municipalities in the Netherlands were invited to

participate in the survey.

3.2 Data collection

In order to collect data on the reaction of municipal councillors to performance information, a survey experiment was conducted. To collect the data necessary to disseminate the survey, multiple steps had to be taken.

Firstly, the contact information for the municipalities had to be gathered. The

information on the municipal councillors could be collected per municipality on Overheid.nl. Overheid.nl is a website created by the Ministry of the Interior and Kingdom Relations together with their partners. It functions as a ‘signpost’ to all information and services of all government agencies (Over deze website). The names and corresponding political party were copied and gathered in an Excel file. Their party membership was included (Gemeenten), although this data was eventually not used as this information was not necessary for the analysis of the hypotheses mentioned in chapter 2. Consequently, the email addresses of the

(26)

municipal councillors were collected on the websites of the separate municipalities. If there were discrepancies between the councillors listed on the almanac and on the websites of the municipalities, the municipal websites were considered the most accurate and thus leading. This is because the municipalities can easily adjust their website when there is a change in composition of the city council, while the almanac is only updated by a contact person of the municipality (Help bij Overheidsorganisaties).

The second step in data collection consisted of the gathering of performance

information for the municipalities. The indicator chosen to categorise the municipalities in the survey experiment was budgetary information on youth care that was provided by

municipalities on their websites, compared with their actual spending on youth care services. Some municipalities did not have or did not publish (sufficient) budgetary information, hence they were left out of the experiment.

When both the email addresses and the performance information were gathered, it became clear that 7819 municipal politicians from 330 out of the 355 municipalities would be invited to participate in the survey experiment. Then, the municipalities could be

classified into six categories based on the size of their population. This was done in order to make the comparison to other municipalities more interesting to the councillors, as the Table 1

Municipality size per category

Population Category <20.000 1 20.000-30.000 2 30.000-50.000 3 50.000-100.000 4 100.00-250.000 5 >250.000 6

(27)

performance of a significantly larger or smaller municipality is not of much use to them, as resources (monetary and services) differ among municipalities of different sizes. Possibly, it could also provide a basis on which the data could be analysed (do municipalities have different performance levels depending on their size and consequently, do councillors react differently to this performance information?), however, incorporating the differences in responsibility attribution based on municipality size was beyond the scope of this thesis. From a more practical point of view, this made it easier for councillors to find their municipality in Qualtrics as they only had to choose from a list with municipalities in their own size category, rather than scrolling through a long list with all municipalities.

When all municipalities had a corresponding size classification, the average budget overspending of each category in percentages could be calculated as a score. Subsequently, the performance score of each municipality could be compared to this average score for the category of the municipality. When municipalities stayed within budget (performance ratio <1), they were classified to receive positive historical performance feedback. When

municipalities exceeded their budget but stayed below the average score for their size

category, they were classified as being given negative historical performance feedback. When municipalities exceeded their budget and performed worse than average in their category, they were classified as being given negative social performance feedback.

The third step in the data collection consisted of the answers to the survey experiment itself. In paragraph 3.4, the design of the survey experiment will be discussed.

3.3 Timeframe & prerequisites for inclusion

In the survey experiment, we wanted to research the reactions of municipal councillors to actual performance information about their municipality. This sets a natural time limit and prerequisites for inclusion in the experiment.

(28)

The time is limited as the respondents have to be current members of the city council due to multiple reasons. Practically, only the current members of the council have email addresses listed on the municipal websites. Moreover, the response from former council members would be redundant, as this research focuses on feelings of responsibility towards performance information concerning their municipality; i.e. the council members that were elected are expected to act in a way that benefits their municipality.

As the research was conducted with actual performance information rather than fictional statistics, there were some prerequisites for municipalities to be included in the experiment: the performance information (the information about the municipality’s budget for youth care and their actual spending) from 2018 had to be available. Not every municipality had included a specification of the budget for youth care and thus they were excluded from the experiment. Moreover, there were some municipalities that overspent a lot, which skewed the survey groups. Thus, these were excluded. Another group that was excluded consisted of the newly formed municipalities: In 2019, 27 municipalities were merged from multiple pre-existing municipalities. As these municipalities did not exist in 2018, no performance data was available.

3.4 Design of the survey experiment and operationalisation

The design of the survey experiment consists of twelve groups, divided among two axes: the municipalities were classified into three groups based on their performance and they were either placed into the accountability prime group or a control group. These groups all consisted of two subgroups: a treatment group and a control group. As the accountability prime can be disregarded for the purposes of this thesis, the groups with and the groups

(29)

without the accountability prime can be merged, which leaves the number of groups in the experiment that are relevant to this thesis at six.1

Table 2

Description of the sample groups

Performance score 1 2 3 Information type Positive historical performance feedback Negative historical performance feedback Negative social performance feedback Sample groups Control 1 Treatment 2 Control 3 Treatment 4 Control 5 Treatment 6

The survey itself consisted of multiple types of questions. The survey started with open questions which constituted the accountability prime and a prime control which were not utilised in this thesis, after which the performance feedback or the information for the control group was shown. Following was a sliding scale in which the respondents could attribute responsibility. After that, statements about innovation were presented to the respondents, and they could agree or disagree through a seven-point Likert scale. Next were open questions in which the respondents could leave their own ideas pertaining to innovation in youth care. As these innovation-themed questions were not relevant for this thesis, they will not be discussed or used further. After the questions on innovation, the respondents could ask to receive links to more research and information on youth care that was used in this survey experiment. Following was a check for manipulation concerning the accountability prime at the beginning

1_{The accountability prime was a part of the survey experiment, but as this thesis does not} research accountability and the treatment was randomised among the groups, this thesis does not take the accountability prime into account.

(30)

of the survey. Lastly, some questions about the characteristics of the respondents were asked. The full text of the surveys can be found in Appendix III.

Responsibility attribution

The dependent variable, responsibility attribution, was measured by a sliding scale in which the respondents could attribute responsibility to three different parties. The respondents were asked which parties were to what extent responsible for the financial situation of youth care provision in their municipality, with the three options being the national government, the municipality itself, or the youth care providers. The respondents could slide a scale and attribute points to the parties they thought were responsible, using a slider scale in which the total had to amount to 100. The order in which the three parties were listed was randomised in order to prevent a bias due to the order of the three parties.

Membership

The moderating variable, membership, was operationalised by creating a divide between three groups. In the survey, general information about the municipal councillors was gathered at the end of the survey. One of these questions was whether the municipal councillors were a member of the coalition, opposition or ‘other’ in their municipal council. These options were then denoted as 1 (coalition), 2 (opposition), and 3 (other). The category ‘other’ could consist of members of parties that support the coalition but are not part of the coalition themselves (gedoogpartijen), but as this category was not clearly defined, some variance in the reasons respondents have chosen this option could have occurred. As only three per cent of the respondents chose this option, the choice was made to disregard the participants that had chosen this option.

(31)

The independent variable, performance feedback was operationalised as information that was shown in the survey before the respondents were asked to attribute responsibility. Each of the three performance groups received a different message (treatment) on the performance of their municipality regarding the youth care budget. In performance group one, this information stated that the municipality in question had stayed within the budget that the municipality itself had determined. This information was framed by making part of this information green, and it was framed by stating that remaining within the budget was positive as the remaining money could be spent on the policy domains it was intended for. In

performance group two, the performance information stated that the budget had been

overspent by the municipality in question, which was marked in red and the word ‘negative’ was used to describe this overspending, as it created a situation in which the money that had been overspent could have depleted other policy domains from necessary resources.

Performance group three was also shown information that stated that the municipality in question had exceeded their budget for youth care, marked in red, and this was also described as negative due to the effect it could have on other policy domains. Differently from

performance group two, however, performance group three received additional information that stated their budget shortages were worse than the average budget shortages in

municipalities of comparable size – after which was stated which size (number of inhabitants) this was. This was also marked in red. Finally, the control group received general information on the budgets for youth care in all municipalities. The words positive or negative were not used, neither was the marking of key words in either green or red.

(32)

Chapter 4: Results

4.1 Analytical strategy

In order to analyse the gathered data with the purpose of answering the hypotheses, the descriptive statistics were first calculated on five variables and then analysed by comparing the six sample groups – positive performance, negative historical performance and negative social performance, of which each category consisted of a control and a treatment group – to each other on four control variables in ANOVA tests. Hence, it could be verified whether the six groups in the sample were or were not significantly different from each other. Then, the effects that could be found when answering the hypotheses could more likely be attributed to the proposed causal relationship than to a variance in sample groups, as other factors would be controlled. Also, normality tests were done to see whether the distribution of the

responsibility attribution scores were normal according to the Kolmogorov-Smirnov test. A significance level of .05 was used. The distributions were not normal, with the p-values for the national government, the municipalities, and youth care organisations all being .000. The Q-Q plots did show that the plotted data approached normality. As the sample size is 1162 and the smallest sample group is 190, the central limit theorem states that non-normal distributions are not problematic. The Q-Q plots can be found in Appendix I. As the

hypotheses were formed on the effects of negative performance feedback, the testing of these hypotheses will be on the effects in the sample groups which received negative performance feedback and their counterpart control groups and the groups who had a positive performance were thus disregarded from further testing.

A number of variables had to be altered in order to analyse them. The sample groups were distinguished from each other by creating variables with scores. The sample groups were a combination of a performance score (1 for positive performance, 2 for historical negative performance, or 3 for social negative performance), and a score for either being a control or a

(33)

treatment group (0 or 1) and by combing these scores each group received a group number from one to six (see Chapter 3, Table 2 on p. 30). After this, the descriptive statistics could be calculated per sample group. Similar processes were undertaken for the performance feedback type, coalition membership, and a variable where performance feedback ad coalition

membership were combined, in order to create variables that could be analysed per group. Moreover, the variable ‘Duration’ was created to see how long the municipal councillors had been in place. In the survey, it was asked since what year the councillors had been part of the municipal council. To calculate their longevity as councillors, a variable was created in which the year they had been installed in the council was subtracted from the current year, 2020.

To test the hypotheses, independent samples t-tests were used. This was done as the hypotheses required testing two independent set of samples against each other. The t-tests for hypothesis 1 test a control group against a treatment group for the same performance, and the t-tests for hypothesis 2 tested coalition groups against opposition groups. A significance level of p<0.05 was maintained in all tests. Bar graphs were used to visualise the results and data labels were added for clarity. Due to rounding of the mean attributed points, the points in two case do not total to 100, but to 99 (Figure 3 and Figure 4) and in one case to 101 (Figure 6).

4.2 Results

In total, 1162 complete responses were collected in the survey experiment. This resulted in 556 responses in the control group and 604 responses in the treatment group. The total responses for each experiment group are listed below in Table 3.

As it was not mandatory to fill in the control variables, not all 1162 respondents filled in each control variable. The least filled in variable was ‘Duration’ with 1081 responses (see Table 4 and Appendix II). For the descriptive statistics, five variables were used to create an overview

(34)

Table 3

Respondents per sample group

of the total sample, which is shown in Table 4. After that, the differences between the responsibility attribution scores among the six sample groups were checked on four of the variables that were also used for the descriptive statistics in Table 4: how many years the municipal councillors had been in their position (duration), their age, gender, and how they positioned themselves in the political landscape using a left-right division (orientation). Whether they were part of the coalition or opposition (membership) was used to create an overview, but not in the testing whether the six groups differed significantly from each other. To check the differences between groups, ANOVAs were used. These tests showed that the six different sample groups did not differ significantly from each other based on the analyses of the four variables, as all four p-values considerably transcended 0.05. The results of these ANOVA tests can be found in Table 4. As the ANOVA tests showed no significant

differences between the groups, it can be assumed that the differences in responses stem from the differences in treatment, rather than other confounding factors. As there were no

significant differences between the groups and two groups have to be compared to each other in each test, the hypotheses will be tested through t-tests. As the hypotheses proposed in chapter 2 concern negative performance feedback, sample groups 1 and 2 could be omitted from further analysis. In addition, the respondents who answered ‘other’ on the survey question which asked whether they were part of the coalition, opposition or other were

Group number Performance level Type of feedback Number of responses

1 Positive Control 38

2 Positive Treatment 40

3 Negative historical Control 329

4 Negative historical Treatment 365

5 Negative social Control 190

(35)

Table 4

Descriptive statistics and significance results of a one-way ANOVA between the six groups

Note. * P<.05

disregarded when testing the hypotheses. This was done as this group only makes up three per cent of all respondents, and as no theoretical expectations were formed about this group.

4.2.1 Negative performance feedback and responsibility attribution

When the ANOVA tests had proven that the six performance and treatment groups did not differ from each other on four control variables, the first hypothesis was tested using independent sample t-tests in which the control and treatment groups were compared

according to their performance level. Hence, group 3 and 4 were compared to each other and group 5 and 6 were compared to each other.

For the t-test between the negative historical performance control and treatment

groups, significant results were achieved. The p-values were .000 (national government); .000 Total sample (N = 1162)

Mean Minimum Maximum Column N % Sig.

Duration 6 1 38 .528 Age 55 19 81 .981 Gender male 67.8% female 30.9% private 1.3% Total .935 Membership coalition 54.6% opposition 42.4% not applicable 3.0% Orientation 1 7.1% 2 26.3% 3 43.5% 4 22.2% 5 0.8% Total .606

(36)

(the municipality itself); and .034 (youth care organisations) respectively. Hence, there was a significant difference between the performance feedback and a control message for the respondents whose municipality had a negative historical performance. In Figure 2, the responsibility attribution scores of the control and treatment groups for negative historical performance are visualised. In this bar graph, the national government is attributed a higher score by the treatment group (55 points), who received negative historical performance feedback, than by the control group who received general information (47 points). This indicates that, as stated in hypothesis 1, negative performance feedback indeed increases upward responsibility attribution. For the municipality and the youth care organisations, the opposite is true: The responsibility attribution scores there are lower in the treatment group compared to the control group. This is partly contrary to the expectation that negative

performance feedback would increase downward responsibility, as it would be expected that Figure 2

The effect of negative historical performance feedback on responsibility attribution

47 55 28 23 25 22 0 10 20 30 40 50 60 Control Treatment R esp o n sib ilit y attr ib u tio n s co re Feedback message National government Municipality

(37)

youth care organisations would have been attributed a higher responsibility score. For youth care, the control group gave 25 points, while the treatment group attributed 22 points. That the municipality itself would be attributed a lower score in the treatment group (23 points)

compared to the control group (28 points) was actually expected, as the hypothesised increase in upward and downward responsibility would automatically mean a decrease in attribution to the municipality itself, as the total points that could be attributed (100) remained constant.

Hypothesis 1 stated that negative performance feedback increases upward and downward responsibility attribution, but this analysis has shown that for negative historical performance levels only upward responsibility increases with treatment – thus partly refuting hypothesis 1. Consequently, the indication is that negative historical performance feedback increases upward responsibility attribution.

For the t-test between the negative social performance control and treatment groups, no significant results were achieved. The p-values were .178 (national government); .325 (the municipality itself); and .341 (youth care organisations). This means that there were no significant differences between responsibility attribution of the respondents who received the performance feedback versus the responsibility attribution of the respondents who received the control message for the groups whose municipality had a negative social performance. In Figure 3, the responsibility attribution scores of the control and treatment groups for negative social performance are visualised. Similar to what can be seen in the bar graph of the negative historical performance groups, the national government is attributed a higher score by the treatment group (51 points) than by the control group (48 points). For the municipality and the youth care organisations, the opposite is true: The responsibility attribution scores there are lower in the treatment group (i.e. 24 points for the municipality and 24 points for youth care organisations) compared to the control group (i.e. 26 points for the municipality and 26 points for youth care organisations) – also similar to the negative historical performance groups.

(38)

This indicates that upward responsibility attribution increases due to negative performance feedback, but downward responsibility attribution and attribution to the municipality itself actually decreases. However, whereas the differences between the negative historical performance control and treatment groups were significant, the differences between the control and treatment groups of negative social performance were not.

Hypothesis 1, which stated that negative performance feedback increases upward as well as downward responsibility attribution, can thus not be proven when analysing the results of the negative social performance control and treatment groups.

Figure 3

Effects of negative social performance feedback on responsibility attribution

Consequently, hypothesis 1 was partly confirmed as negative performance feedback did increase the respondents’ upward responsibility attribution. The results for both negative historical and social performance feedback were in the direction that was predicted for the

48 51 26 24 26 24 0 10 20 30 40 50 60 Control Treatment R esp o n sib ilit y attr ib u tio n s co re Feedback message National government Municipality

(39)

national government, but only the results for negative historical performance feedback were significant. For attribution to the municipality itself and to youth care organisations, the results were not as predicted in hypothesis 1, but again the results for negative historical performance feedback were significant, while the results for negative social performance feedback were not. Hypothesis 1 is thus refuted as it is and could be transformed to “Negative performance feedback increases upward responsibility attribution.”

4.2.2 Membership and responsibility attribution

After testing for a relationship between performance feedback and responsibility attribution, the proposed moderator, membership, was tested for a direct relationship with responsibility attribution, before actually testing whether it acted as a moderating variable. For this purpose, an independent samples t-test was carried out to test the relationship between the variable membership and responsibility attribution. The test showed that the results were statistically significant for the national government (p= .010) and the municipality (p= .002), but not for youth care organisations (p= .516). When the relationship between membership and

responsibility attribution was visualised (Figure 4), the graph showed that members of the coalition attribute more responsibility to the national government (52 points) than members of the opposition (48 points). Moreover, members of the coalition attribute less responsibility to the municipality (24 points) than members of the opposition (27 points). Youth care

organisations were attributed equally many points by coalition and opposition members: both groups attributed 24 points to youth care organisations. From these results, it is shown that the proposed moderating variable actually has a direct relationship with the dependent variable, responsibility attribution, and hence acts as an independent variable.