Providing feedback to the policy cycle

PROVIDING FEEDBACK TO

THE POLICY CYCLE

Using system dynamics to gain insight into the complexity

of the public policy process

Words: 29169

Supervisor: Prof. E.A.J.A Rouwette Second examiner: Dr. F.A. Bekius Business Analysis and Modelling Business Administration

Nijmegen School of Management, Radboud University

Mathijs Ambaum

1

Preface

In front of you lies the master thesis: “Providing feedback to the policy cycle. Using system dynamics to gain insight into the complexity of the public policy process”. This research is the last hurdle of the

Business Administration master: Business Analysis and Modelling. The thesis also represents my last moments of being a student. I enjoyed my time at Radboud University, but I am ready for the next step. This thesis project has been conducted in a very odd time. The COVID-19 crisis forced me to study, work, exercise and relax all on the same 14m2_{. On the one hand, this posed challenges to keeping a}

healthy study-work-life-balances. On the other hand, the intelligent lock-down provided also the chance to really lock-in with my thesis. This lock-in was necessary as I, again, chose a rather difficult topic by examining the full policy process. The literature was so diverse and so interrelated that initially my model looked like spaghetti. Although the model is still quite complex and definitely not the easiest model to understand, I am satisfied with the end product. I also believe that the model could definitely have some value in analysing the policy process.

I could not have done this process alone. First, I want to thank my girlfriend, Simone, for providing mental support and being my best sparring partner. You really helped me move on when I was stuck every now and then. Secondly, I would like to thank my thesis supervisor Etiënne Rouwette for providing feedback on my progress and stimulating me to find my own solutions to the issues I encountered. Further, I would like to thank my family and my roommates for their mental support and providing the necessary social distractions in this otherwise lonely process. Last, I would like to thank all who read my thesis to help me with the last check.

2

Abstract

For a long time, the policy cycle has been used as the model to provide insight into the public policy process, but at the same time it has been unable to provide a valid representation of the public policy process due to oversimplification of reality. There has yet to emerge a model that does provide insight into the full policy process and embraces its complexity. This study aims to fill this theoretical void by using system dynamics to model the recent applications of public policy process theories. A systematic review of contemporary literature was converged into a single model that displays how feedback, strange attractors, emergence and initial movers cause the public policy process to be a complex system. The model shows that the public policy process is a continuous process of decision-makers feeling pressured to address certain issues and attempting to resolve them. This process is dependent on the capacity of a decision-maker to address issues and enforce policies, support from policy actors for either maintaining the status quo or adopt new policies, and the fit of an idea within current rules and norms. The model contains 25 feedback loops as a first sign of complexity. There is a crucial role for policy evaluations, events and conflict stability to disrupt the equilibrium, i.e. the status quo, and reveal new issues. Because of the omnipresence of feedbacks, the policy process is highly dependent on its own history and is able to drive its own behaviour.

3

Content

Abstract ... 2

Figures and tables ... 5

1. Introduction ... 6

1.1 Research aim and question ... 7

1.2 Scientific Relevance ... 8

1.3 Societal Relevance ... 9

1.5 Reading Guide ... 9

2. Theoretical Framework ... 10

2.1 Public policy process ... 10

2.2 Theorising the public policy process ... 11

2.3 Complexity theory ... 13

2.4 Public Policy in a complex context ... 15

2.5 Conclusion ... 16

3. Methodological framework ... 17

3.1 Research Strategy ... 17

3.2 Systematic Literature Review ... 17

3.3 Modelling complexity ... 18

3.3.1 System dynamics ... 19

3.4 Data collection ... 20

3.5 Data Analysis ... 22

3.6 Validity and reliability ... 24

3.6.1 Internal Validity ... 24 3.6.2 Model validity ... 24 3.6.3 External validity ... 25 3.6.4 Reliability ... 26 3.7 Research ethics ... 26 4. Results ... 27 4.1 Research designs ... 27 4.1.1 Research aim ... 27

4.1.2 Research design and sample ... 29

4.1.3 Conclusion ... 31

4.2 Concepts ... 32

4.2.1 Actors making choices ... 32

4.2.2 Institutions ... 33

4.2.3 Networks and subsystems ... 33

4.2.4 Ideas and beliefs ... 33

4.2.5 Context ... 34

4.2.6 Events ... 34

4 4.4 Dynamics ... 35 4.4.1 Feedback ... 35 4.4.2 Initial mover ... 41 4.4.3 Emergence ... 41 4.4.4 Strange attractor ... 42

4.3 Public policy process model ... 42

5. Analysis ... 45

5.1 Case description ... 45

5.2 Case application ... 46

5.2.1 Basic education ... 46

5.2.2 Second phase ... 47

5.2.3 Preparatory middle‑level vocational education ... 48

5.3 Complexity in the 1990 education innovations ... 49

5.4 Building confidence: model assumptions and limitations ... 50

5.4.1 Model assumptions ... 50

5.4.2 Model limitations ... 50

6. Conclusion ... 52

6.1 Model of the public policy process ... 52

6.2 Discussion ... 53

6.2.1 Implications research ... 53

6.2.2 Limitations ... 54

6.2.3 Recommendations for future research ... 55

Literature ... 57

Appendix 1. Sample ... 65

Empirical journal articles ... 65

Non-empirical journal articles ... 68

Appendix 2. Study profiles ... 70

Appendix 3. Process of analysis ... 81

Procedure ... 81

Concept tree ... 81

Dynamic identification ... 95

Dynamics source ... 112

Excluded articles ... 118

Appendix 4. Model structure ... 120

Concept definitions ... 120

5

Figures and tables

Figure 1.1 The policy cycle p.6

Figure 2.1 Elements of a policy process theory p.13

Figure 2.2 Population as an illustration of feedback p.14

Table 3.1 Most impactful journals 2010-2018 p.21

Table 3.2 Sensitizing concepts p. 23

Table 4.1 Themes among scholarship p.28

Table 4.2 Research design and research level scholarship p.30

Figure 4.1 Geographical area scholarship p.31

Table 4.3 Overview model concepts p.35

Figure 4.2 Feedback loops actors making decisions p.36

Figure 4.3 Feedback loop institutions p.37

Figure 4.4 Feedback loops networks and subsystems p.38

Figure 4.5 Feedback loops role of knowledge p.39

Figure 4.6 Feedback loop role of ideas p.39

Figure 4.7 Feedback loops idea environment p.40

Figure 4.8 Feedback loops issue context p.40

Figure 4.9 Feedback loop issue familiarity p.41

Figure 4.10 Feedback loop issue salience p.41

6

1. Introduction

Governments make decisions that influence all of society. Sometimes it is clear why certain issues were enacted upon, but most of the time we could raise the question why these decisions were made (Cairney, 2012a). For instance, governments could not ignore the financial crisis of 2008 nor the COVID-19 pandemic in 2020. But why did not every government respond in the same manner to these issues and why did not every government act at the same time? In other instances, governments do not act upon an issue at all. Why do some issues get attention, while seemingly equally important issues do not get the same attention? How can it be that some policies radically break with the past, while policies in other areas seem to evolve so slowly that change is noticed only years later? Public policy research is the study of how and why these government decisions come about (Howlett, Ramesh and Perl, 2009; Weible, 2014).

For a long time, the policy cycle has been used to explain the development of public policy (Jann and Wegrich, 2007). The policy cycle is used as both a prescriptive and descriptive model to organise and systemise the public policy process (Jann and Wegrich, 2007; Howlett et al., 2009). It is used as a prescriptive model because it depicts an ideal sequence of stages that government should go through when developing and implementing policy. It is also used as a descriptive model to describe actual policy processes. Although there are many different versions (see Jann and Wegrich, 2007 for an overview), the most commonly used policy cycle contains the stages of agenda-setting, policy formulation, decision making, policy implementation and policy evaluation (Jann and Wegrich, 2007) (Figure 1.1). The main merit of the model is that it is an easy and understandable representation of the policy process (Weible, 2014; Howlett et al., 2009; Jann and Wegrich, 2007). Moreover, the stages could be separated, allowing for theorising within each stage (Howlett et al., 2009).

The use of this model has come across several critiques over the years. The first shortcoming of the policy cycle is that it oversimplifies reality (Cairney, 2013; Johnson, 2015). The model has a chronological order which is not representative for the way actual policies are developed (Jann and Wegrich, 2007) and would thus not be suitable to describe the policy process. Secondly, the cycle lacks causal theory and testable hypotheses according to Sabatier and Jenkins-Smith (1993). This would also make the model less useful to apply to real-world situations. Lastly, the unit of government to which the cycle should be applied remains implicit, or even unknown (Howlett et al., 2009). These shortcomings have led the model to be abandoned or merely seen as a categorisation of policy stages (Howlett et al., 2009).

Aside from the inherent flaws of the policy cycle itself, there is another cause for the abandonment of the model. A chronological and linear model does not fit with contemporary society anymore (Astill and

Figure 1.1 Policy Cycle Agenda setting Policy formulation Decision making Policy Implementation Policy Evaluation

7

Cairney, 2015). Contemporary society is a complex system (Johnson, 2015). A complex system is made of “elements [that] interact with each other to produce outcomes that cannot simply be attributed to individual parts of a system” (Cairney and Geyer, 2015, p.2). The notion of policy development as a linear process with separable stages is a misleading description of how policy is actually developed

(Cairney, Heikkila and Wood, 2018). Contemporary public policy development is characterised by a

multitude of actors in and outside of government that are dependent on each other but also influence each other to arrive at a certain decision (Cairney et al., 2018).

One way to adjust policy process theory for the complexity of contemporary society is to make use of complexity theory (Cairney and Geyer, 2015). Complexity theory advocates shifting the focus of analysis to the behaviour of a system as a whole instead of the sum of its separate parts (Cairney and Geyer, 2015). For the policy process this means that the development of certain stages in policy cannot be considered without considering the other stages too, because of their interrelatedness. Complexity theory is founded on two main elements. First, feedback processes drive the behaviour of the system (Cairney and Geyer, 2017). Through a string of interactions different elements of a system have either a balancing or reinforcing effect on each other. Second, these feedback processes cause different elements in the system to react to each other. A string of reactions results in a certain behaviour of the system. In complexity theory, systems are thus path-dependent (Cairney and Geyer, 2015). The challenge, however, is to incorporate complexity theory into existing knowledge of public policy processes.

The challenge presented here is twofold. First, there is the aim to incorporate complexity theory into current public policy theory to unravel the black box of causality. This should result in a detailed description of the interactions between elements of a certain system in order to explain the behaviour of the system as a whole (Johnson, 2015). A pitfall in complexity theory is getting lost in the details. Yet in order to advance policy process research, complexity should be made understandable (Johnson, 2015). Therefore, modelling approaches are used to visualise this unravelling of the black box. A model is a simplified version of reality that shows only the relevant aspects of a system (Edmonds and Gershenson, 2015). The question remains what is relevant to include in such a model of the public policy process. Second, existing knowledge of the policy process is rather diverse (Howlett, 2019; Cairney, 2013). Theories that have emerged as replacement, adjustment or as complementary to the policy cycle each use a different scope, level of analysis and draw on different assumptions (Cairney and Heikkila, 2014). Although there is some overlap between these theories, one could still pose the question how does public policy actually develop? A comprehensive and useful theory has yet to emerge in this field. The challenge is to synthesise these public policy theories in order to be able to integrate them with complexity theory in a model.

1.1 Research aim and question

This thesis confronts these challenges and aims to develop a model that describes contemporary public policy development and its complexities in order to provide researchers, students of public policies and policymakers with a more accurate description of the process. The research objective is thus to gain theoretical knowledge on the different elements of the public policy process, in order to construct a model that provides insight into the complexity of the public policy process. This study fills the theoretical gap in public policy research on the lack of a valid model of the public policy process. This aim requires a two-step approach. First, the researcher has to identify how public policy actually develops, synthesise the concepts and relations necessary to explain the development of public policy and translate this into a model. Second, the researcher has to explain how complexity is present in this process. Consequently, the research question of this thesis is:

8

What concepts and relations used in contemporary journal articles explain the complexity of the public policy process?

Five sub-questions serve as the stepping stones to answer the research question: 1. What are, according to literature, the characteristics of a public policy process? 2. What are the characteristics of a complex system?

3. How does complexity theory relate to the public policy process according to literature? 4. What are the concepts and relations in the public policy process?

5. What features of complexity are present in the public policy process?

1.2 Scientific Relevance

This research contributes to the scientific field in four ways. First of all, the scientific field addressing the public policy process is splintered (Howlett et al., 2016; Cairney and Heikkila, 2014). There is a tension between keeping theory simple and comprehensive and including all details of the actual process. Keeping theory simple and comprehensible allows to cover the full public policy process, like the policy cycle, but fails to be a realistic depiction of the policy process due to oversimplification (Weible, 2014). This thesis adds to the scientific field by finding a new balance between comprehensibility and comprehensiveness for the depiction and understanding of the public policy process.

Secondly, including all complexities often leads to theoretical frameworks that only explain a part of the process. One of the reasons why complexity and the full public policy process have not been unified yet might be that previous public policy research has not included complexity explicitly in their research design. Complexity has been an outcome of research rather than used a methodology to better understand the output. Hence, complexity is not new to public policy research, but has to be set in a new daylight (Cairney and Geyer, 2017). This research uses complexity as a new lens to consider the full public policy process as a means to advance public policy research.

Third, according to Cairney (2013), the next step in public policy research is to acknowledge methodological pluralism and to compare different approaches to the same problem. Albeit out of necessity, this thesis will use multiple methodological approaches to answer the research question. Complex systems are difficult to explain comprehensively. Therefore, several tools have been developed to make those systems more insightful (Astill and Cairney, 2015). Among these tools are system dynamics and agent-based modelling. Each of these approaches has its own way of structuring complexities into comprehensible frameworks (Edmonds and Gershenson, 2015). This research will identify which of these complexity tools is best suited for explaining the development of public policy. In order to do so, this thesis will make use of a systematic review to organise and synthesise the work of previous research. Acknowledging complexity does not require developing new theories, rather it should make use of the insights that have been developed already (Cairney and Geyer, 2015). This thesis can be seen as one of the first steps in the direction of using methodological pluralism to provide insight into complexity based on previous findings.

The last contribution to literature is also rooted in the notion of pouring “old wine in new bottles” (Cairney and Geyer, 2015, p.3). Much research has been conducted on public policy (Cairney and Heikkila, 2014; Weible, 2014; Howlett et al., 2016). It has turned out to be challenging to either synthesise, complement or contradict all these theories in order to come to one singular framework (Cairney, 2013). This thesis does not claim to overcome the issues with all of these approaches and provide the one true framework on public policy development. Rather, this thesis adopts an interpretative scientific philosophy, meaning the context determines how one should interpret the theory or model.

9

This thesis takes complexity as the context of the public policy process and will identify how a comprehensive framework can be synthesised in this particular context. This approach will advance the scientific field of public policy by connecting the various theories in a context of complexity.

1.3 Societal Relevance

Apart from a contribution to the academic field, this study is also highly relevant for society for three reasons. First of all, this study remains close to the actual policy process. The data for this study are empirical studies in the field of policy research. So even though no primary data is used, this research uses empirically tested theories and hypotheses. The framework that will be built is thus based on an analysis of the actual process instead of normative and prescriptive works. Therefore, this thesis will lead to more insight into the actual complexities of the public policy process. By knowing that the public policy process develops in a much more complex context than the policy cycle depicts, a much more realistic idea could be adopted by education, politics and others who talk about developing public policy. Secondly, planning and evaluation of public policy could be done much more accurately. As said before, the public policy process is surrounded by much more complexity than the policy cycle indicates (Cairney, 2013; Weible, 2014). When using the policy cycle as a structure for planning this process (Jann and Wegrich, 2007) these complexities could be overlooked and the actual process could differ from the planned process. Due to cognitive limitations, people focus on parts rather than the full system (Vennix, 1996). Consequently, people do not recognise their influence on the entire system. Hence, when planning or evaluating policy issues people tend to focus on the results that are close to their actions (Vennix, 1996). A depiction of the public policy process as a complex system could show the interrelations between all elements of the process and indicate how a small part of the system could make a huge difference for the outcome. A framework that maps complexities could thus help planners make a more realistic planning and a more accurate evaluation.

Thirdly, by combining theories and modelling the complexities of the public policy process, education on the topic could be adjusted. This would lead to realistic expectations, rather than knowing an ideal and normative model (Howlett et al., 2009). By not acknowledging the complexities of public policymaking, people tend to consider the policy process as a rational controllable process based on evidence (Jann and Wegrich, 2007) Yet, only a small part of the total field of policymaking can adhere to those assumptions (Geyer, 2012). The conceptualisation that will be developed in this thesis will encompass a holistic view on the public policy process as well as acknowledging the complexities of the process. This will make sure that those who will educate themselves using this model will have a holistic realistic rather than a holistic idealistic picture of the public policy process.

1.5 Reading Guide

This thesis is structured as follows. The second chapter provides the theoretical framework. The theoretical framework consists of an elaboration on what a public policy process contains and how complexity relates to this process. The third chapter consists of the methodological considerations for this research. The methods used are a systematic review and system dynamics to construct a model that provides insight into the complexities of the public policy process. The fourth chapter covers this model and how the concepts and relations were elicited from the sample. The fifth chapter aims to build confidence in the model by applying the model to a case and a reflection of the model. The last chapter provides the conclusion and discussion of this research. It summarises the findings, provides an answer to the research question and provides suggestions for future research regarding this topic.

10

2. Theoretical Framework

This chapter serves as the theoretical framework. In the first section, I identify what a public policy process is. The second section describes the elements that constitute a public policy process. The building blocks of theories on public policy are explained in order to provide an overview of how one can theorise the public policy process The third section explains what a complex theory is and what its features are. The last section shows how complexity theory is present in the elements of a public policy process theory.

2.1 Public policy process

Public policy is a debated concept (Howlett et al., 2009; Cairney, 2012a). Many definitions that have been used in the past have stumbled upon criticism. One of the first and most well-known definitions of public policy is Dye’s (1972, in Howlett et al., 2009, p.4) definition: “public policy is anything that the government decides to do or decides not to do”. This definition is a solid start, but also highlights three issues for defining public policy. First of all, is government the only actor involved in the public policy process? Second, is public policy always a conscious decision? Third, is a decision to not do anything also a policy (Howlett et al., 2009; Cairney, 2012a)? One can easily imagine that these questions can raise wide-ranging discussions. Therefore, the aim here is not to provide a precise definition of what falls within and outside the boundaries of public policy. Rather the aim is to come to a working definition; a definition that is sufficiently clear to get the main idea of what a public policy is, but some grey areas might remain. The three issues presented above will guide this query to a working definition. First, unless the subject of interest is an anarchy, there must always be a government involved with public policy. Governments are the institutions that have the unique authority to make decisions on behalf and ideally in the interest of the citizens of their territory (Howlett et al., 2009). Yet, the question remains whether governments are the sole actor involved in making decisions, or if there are more actors involved. According to Sabatier (2007) a usual public policy process involves many different actors assisting the government. He refers to interest groups, different levels of government, researchers and journalists, among others. Based on this insight, I deliberatively choose to leave the decisionmaker open for interpretation in my definition. A public policy is therefore constituted by, or involves, a government institution.

Second, the issue whether public policy should be a conscious decision is best illustrated by an example. Imagine a government that aims to reduce emissions from cars by increasing taxes on fuel. Government plans for people to start using public transport, but instead more people move to urban areas to live closer to their jobs and consequently need less fuel to get from their home to their job. The question is whether the government can claim the result of these movers as a public policy. This question requires us to make a distinction between outputs and outcomes, which already relates to public policy being a process. Outputs are direct intended effects of a certain action, while outcomes are all effects of a certain action combined (Cairney, 2012a). Jenkins (1978, in Howlett et al., 2009) proposes to define public policymaking as a selection of a goal and a certain means to reach that goal. This implies that the public policy is a conscious decision. In order to put boundaries around what counts as public policy and what does not, this is taken as a part of the definition of public policy.

This also solves the third issue with Dye’s (1972) definition of public policy. If a public policy is a selection of a goal and the means to reach it, this could also imply that the means to reach the goal is to maintain the status quo. These deliberations combined make public policy a decision by a government institution or involving a government institution concerning the means to achieve a goal set by a government institution or involving a government institution.

11

A public policy process suggests that there is a system that translates these goals and means into actual policies which can be implemented and evaluated (Birkland, 2005). This requires thinking of the public policy process in terms of input, output and throughput. Input of the model is any translation of factors that the public policy system can react to, that is the goal is set on what to achieve (Birkland, 2005). The output of the public policy process is the various means that are set in place to achieve the goal. The conversion of inputs into outputs happens in the black box of the process. A black box is a metaphor for unknown processes and often, like in this thesis, the subject of interest of research. The effects of the means on achieving the goal can, but does not have to, be a factor for new input in the policy process (Birkland, 2005). The public policy process is thus a chain of reactions between setting and achieving goals through a translation of these goals into means by or involving at least one government institution.

2.2 Theorising the public policy process

This thesis aims to develop a model to explain the dynamics of the public policy process. In order to construct a model of the full public policy process one should build a framework which includes the main theoretical features of the public policy process (Schlager, 2007). A framework displays and organises the elements that must be included in order to build a theoretical model. It provides a metatheoretical language that applies to most, if not all, cases, while a model is much specific on relations and assumptions (Ostrom, 2007). Theory on public policy should describe and explain decision-making by actors, the role of institutions, networks and subsystems, ideas and beliefs, context and events(Weible, 2014; Cairney and Heikila, 2014). Each of these six elements is discussed in the following paragraphs.

The first element is ‘actors making choices’ (Cairney and Heikila, 2014). From the definition of the public policy process in the previous section it is clear that public policy is a choice for a goal and a means to achieve those goals (Howlett et al., 2009). This poses two questions for a theory: who makes these choices and how does an actor make a choice? As was stated previously, there are many actors involved in making decisions for public policy (Sabatier, 2007). Sometimes this could lead to up to a thousand actors involved. Therefore policy theories only identify the key actors, or group key actors under a common denominator (Weible, 2014). Depending on the level of analysis, the actors in the public policy process can be characterised as individuals, collectives or structures (Howlett et al., 2009). Individual decision-makers might seem clear without any further elaboration, collectives and structures deserve more detail. Collectives are organised interests consisting of multiple individuals, but acting as one voice. Structures consist of institutions that shape the society we live in (Howlett et al., 2009). Government is an institution, but also churches and firms count as institutions that shape society. The second element of actors making choices, is the way this happens. One of the most widely accepted theories on making choices is the rational choice theory (Cairney, 2012). Rational choice is the premise that a decision-maker seeks to act according to his or her preferences in order to gain as much satisfaction as possible. In the definition of public policy this would mean that the goal chosen is according to the beliefs of the decision-maker and that the means really help to achieve that goal. The satisfaction that is gained from making the decisions is dependent on the context of the choice. Rational choice is acknowledged as a normative decision-making strategy (Howlett et al., 2009; Cairney, 2012a). In reality, decision-makers use heuristics in combination with rational choice conceptions to come to their choices (Tversky and Kahneman, 1986).

These actors operate within certain sets of rules and norms, which forms the second element in theory on public policy processes: institutions (Cairney and Heikila, 2014). Institutions are not only possible decision-makers, they are also “the rules, norms, practices, and relationships that influence individual and collective behaviour” (Cairney and Heikila, 2014, p.364). Institutions thus function as the boundaries within which actors can legally operate in the public policy process. Yet institutions are also

12

more than that. Institutions shape behaviour (Cairney, 2012a). There are different ideas on how institutions shape behaviour. Some say that institutions are sets of rules formed in the past that influence behaviour now. Institutions are hard to dismiss and only critical events create entirely new institutions. This view on the influence of institutions is called historical institutionalism (Cairney, 2012a). Another variant is rational choice institutionalism. In this view, institutions provide the context for how individuals make choices. For instance, they might create incentives to cooperate, reduce transaction costs and make sure through rules that there is stability in society (Cairney, 2012a). Normative institutionalism represents the rules a society has regarding what is considered to be appropriate behaviour. Lastly, there is constructivist institutionalism which differs from the other institutionalist approaches in that it does not explain why actors behave in a certain way, but rather explains how actors approach their context (Cairney, 2012a). This perspective states that institutions are shaped by shared frames and that these shared frames form the context in which actors make decisions. If one does not agree with a certain frame, he/she can break with this frame through discussion and other forms of communication.

Third, public policy processes happen in networks or subsystems (Cairney and Heikkila, 2014). Numerous times it has been mentioned that there are multiple decision-makers involved in the public policy process. Networks and subsystems represent the relationships between the decision-makers and the ones who seek to achieve a certain goal (Cairney and Heikkila, 2014). These could be interest groups that seek to achieve some goal through public policy, for which they need the government, or different layers of government who operate in different jurisdictions and need each other’s help to get access to certain means. There is a difference between subsystems and networks. Subsystems represent the whole issue area where actors, collectives and institutions are found. Networks are the groups of actors who actually are engaged in the public policy process (Howlett et al., 2009). Networks can have few or many players. In a network there are different combinations of actors possible. There could be multiple government institutions involved, or combined with civil society, or combinations with market institutions.

Fourth, public policy is based on ideas (Cairney and Heikila, 2014). The goal that a policy is aimed at comes from an idea or belief that an issue should be dealt with in a certain way. Ideas are beliefs, thoughts and opinions (Cairney, 2012a). Ideas help to understand a problem and propose solutions on how the issues can be solved. Ideas can shape the way society thinks and thus also become institutionalised and constrain which ideas can be brought into the policy process (Cairney, 2012a). It is thus important to know how ideas get picked up in the public policy process. A widely accepted theory on how ideas are picked up is the multiple streams model of Kingdon (1984). He proposes that a window of opportunity for an idea to get picked up arises when problems, policies and politics are aligned. Policies are solutions and politics refers to the realm of the day. If a problem occurs and a solution is present which fits at least to some extent with the realm of the day, a window of opportunity opens and new ideas can get picked up.

The fifth element of a public policy theory is the policy context (Cairney and Heikila, 2014). The context refers to the environment of the public policy process and how it could influence decision making (Cairney and Heikila, 2014). The public policy context is a broad category. Rose (1990) sees the context as the inheritance of public policy. Closely related to historical institutionalism, he argues that contemporary governments inherit the legacy of previous governments, that is laws, rules, institutions and programs. New governments have to deal with what their predecessors have left them with. Aside from policy inheritance, two meta-institutions also matter in shaping the political-economic context of the public policy process. In most countries, capitalism and democracy are institutionalised, but in each country it works slightly different (Howlett et al., 2009). This has its consequences for how public policy

13

decisions can be made. Other factors that are mentioned as context setters are infrastructure and culture of a state (Weible, 2014).

The last element of a theory on public policy is the role of events (Cairney and Heikkila, 2014). Events are both anticipated and unanticipated incidents (Weible, 2014). Events are often sudden and are critical in the sense that they have an impact on the public policy process (Birkland, 1998). Elections are a typical example of anticipated events, while natural disasters are a fine example of events that are more difficult to anticipate. Events can but do not necessarily open the earlier mentioned windows of opportunity. An example is when there is a new government formed after elections, there is room for change in what kind of public policies will be developed. In order for events to have a real influence on the public policy process, policy entrepreneurs need to grasp the moment to connect the event to known ideas and propose how things could change for the better (Cairney, 2012a).

The development of a policy process theory requires to consider all these elements together. Even though these meta concepts are fluid and overlap with the interpretation of some other elements, it is only together that they can explain the full policy process (Cairney and Heikkila, 2014). Only explaining the rules and norms and how actors make choices within those boundaries does not explain the full process. The choices are connected to relationships with other policy actors and what ideas are considered appropriate. This all happens within a certain context which cannot be ignored. This is only a very brief argument for why it is necessary to consider these elements together if we want to develop a model of the full policy process. Figure 2.1 displays the six elements of a policy process theory.

2.3 Complexity theory

Complexity theory is a scientific approach that studies systems to explain outcomes instead of investigating each element separately (Cairney and Geyer, 2015). A system is a set of interrelated parts that operate together to produce a certain outcome (Jervis, 1997). Complex behaviour is caused by the different elements being interdependent, interrelated and interacting (Mitleton-Kelly, 2003). Consequently, if one element of a system changes it causes a cascade of actions and reactions through

Policy process theory Actors making choices Institutions Networks and subsystems Ideas and beliefs Context Events Figure 2.1

14

the entire system (Jervis, 1997; Mitleton-Kelly, 2003). A system is thus more than the sum of its parts (Cairney, 2012b; Mitleton-Kelly, 2003). Complex systems exhibit four characteristics (Cairney, 2012a): (1) feedback processes cause dynamic behaviour, (2) systems are path-dependent, (3) a system’s behaviour is emergent and (4) systems may contain strange attractors. Each of these characteristics will be elaborated on in this section.

The first feature of complex systems is the presence of feedback processes. “Feedback is a process in which action and information in turn affect each other” (Vennix, 1996, p.31). Feedback thus creates a closed loop of actions and reactions that flows back to the element that initially started the behaviour. A system with feedback loops is thus capable of driving its own behaviour. Feedback processes can be either positive or negative. A positive feedback means that elements A and B move in the same direction, while negative feedback causes the elements to move in opposite directions (Vennix, 1996). An often-used example to illustrate the dynamics of feedback loops is the population model (Figure 2.2). This model shows that when there are more births than deaths, population will grow. And as a consequence, births will increase as more people will give birth. This is a reinforcing effect leading to an exponential growth of the population. However, there is also the negative effect of deaths. If more people live, more people will eventually die. If this happens, population will decrease and fewer people will pass away. What also becomes apparent from this simple model, is that births and deaths do have an influence on each other. Through an increase of population, births have an influence on the number of people that will pass away. Vice versa, if more people pass away, the lower the population that can give birth. In this way the system is never stable and will behave in a dynamic fashion. Also, because these feedback processes are closed the system will drive its own behaviour (Sterman, 2000).

The second feature of a complex system is that this string of reactions requires an initial mover (Cairney, 2012b). Actions in the past are the information that will be used for reactions in the future. Complex systems are thus path dependent (Cairney, 2012b; Sterman, 2000). A system does not reset every time a new choice has to be made, instead it builds on the conditions of the past. The population model serves as an example again. The size of the population in the initial state influences how births and deaths will develop over time. For instance, if population size is a thousand people both procreation and deaths are very likely. Yet, if the population size is only one, then births are not possible as it requires two people to procreate. The conditions of a system at a certain point in time determine the options for development of the system in the future (Cairney and Geyer, 2017; Sterman, 2000).

Third, complex systems are characterised by emergence (Cairney, 2012b; Sterman, 2000). Emergence refers to “behaviour that evolves from the interaction between elements at the local level rather than central direction” (Cairney, 2012b, p.348). It is thus a process of self-organisation (Mitleton-Kelly, 2003; Sterman, 2000). Small changes in the system create this string of reactions which generates the larger patterns of the whole system (Sterman, 2000). It is for this very reason that single elements of a system can influence the outcome but not control the entire system’s behaviour (Johnson, 2015). An example of the difference between the action and interaction of individual parts and the behaviour of

Births

₊

R

+

Population

-

B

₊

Deaths

Figure 2.2

15

the whole system is a traffic jam (Bonabeau, 2002). Even though all cars are individually moving forward, the traffic jam – the sum of all individual elements – increases in the opposite direction. The last feature of complex systems is that they may contain strange attractors (Cairney, 2012a). Complex systems are associated with irregular and unpredictable behaviour, yet often they do show extended regularities in their behaviour (Cairney, 2012b). Through their connectedness systems find a state of balance or repeating behaviour, this is called an equilibrium (Sterman, 2000). A strange attractor is an event or trigger that disrupts this equilibrium and changes the behaviour of a system radically (Cairney and Geyer, 2017; Cairney, 2012a; Mitleton-Kelly, 2003). The COVID-19 pandemic can, for example, be understood as a strange attractor. Before the COVID-19 crisis, people shook hands to greet each other. The more people shake hands, the more appropriate it seems. Over time, this has become an institutionalised habit. Yet, because of COVID-19, it is thought of as inappropriate because of a fear of infection. Corona is thus a strange attractor as it is an event that causes a system to move away from what has been perceived as normal for a long time.

2.4 Public Policy in a complex context

In this section I apply the different elements of complexity theory to the six elements of a public policy theory as defined in section 2.2. For the first element of complexity, the system is more than the sum of its parts, is it hard to pinpoint how exactly this applies to the public policy process. Cognitive limitations prevent people often from seeing the whole system and they thus merely consider its parts (Vennix, 1996). Here too it is difficult to arrive at a conclusive answer on how this applies to the public policy process. That is why the structure of this system is also central to this study. An indication that this feature of complexity also applies to the public policy process, is that there is not one central element in the whole process (Cairney, 2012b). Rather the process is a combination of actors, institutions, networks, ideas, context and events (Cairney, 2012a).

The second element consists of negative and positive feedback which each is composed of a string of actions and reactions caused by the interconnectivity of different elements in a system (Vennix, 1996). This feature is omnipresent in the six elements of public policy theory. Actors make decisions based on ideas, an idea may thus lead to a certain decision. These ideas are shaped by institutions, the boundaries around how we think. Context and events might open up windows of opportunity for actors to make decisions or to change the institutions that we agreed upon. It might become clear that all these factors influence each other and create a string of actions and reactions that together shape the system’s behaviour (Cairney, 2012b; Eppel, 2017).

Third, public policy does not change overnight, usually public policy develops incrementally (Lindblom, 1959). Sometimes it might take years for a policy to change, or something radical has to happen. This is an indication that public policy can be seen as upheaval to a trajectory subject to a certain path dependence through institutions as a norm setter (Cairney, 2012a). Initial conditions do matter and it requires an initial mover, someone with radical ideas, or an event that shifts the focus of society completely, to change a course of action (Birkland, 2005). With this element of complexity, the context of the public policy at stake plays a critical role. A stable environment will create stable system behaviour, yet an unstable environment will cause a public policy to be less predictable (Eppel, 2017). Fourth, emergence is a feature of public policy processes. Although institutions have a top-down function in influencing behaviour, these norms are usually not imposed top-down. Rather the norms of institutions emerged through what people have found appropriate or not. This strongly relates to the previous comments on path dependency and an initial mover. Apart from this, policy processes are characterised by networks and subsystems. Networks and subsystems represent interactions between decision-makers and stakeholders that jointly create a policy (Howlett et al., 2009). Path dependence

16

and the role of networks show how emergence is a characteristic of public policy development in a complex system.

Fifth, strange attractors are also omnipresent in public policy development. In between the lines, one could have gotten hints on that from the previous paragraphs. For instance, the initial mover idea and how it relates to public policy also clearly is connected to the ideas of a decision-maker. The initial mover has the ability to change a system after a long time of stability (Cairney and Geyer, 2017). A change in ideas or a change in what is perceived as appropriate has the potential to disrupt the public policy process. Apart from this, decision-makers have the ability to maintain the status quo just like institutions have a tendency to keep things the way they are. Yet, an unstable environment or a gradually changing context might disrupt this and cause a public policy to change drastically.

2.5 Conclusion

Concluding, in order to develop a model that describes the public policy process one has to know what constitutes a sound theory on the public policy process and how complexity manifests itself. The public policy process was defined as a chain of reactions between setting and achieving goals through a translation of these goals into means by or involving at least one government institution. A metatheory on theory building in public policy process identified six elements as the core features of a theory in the field. A holistic theory for the public policy process one should give an idea of actors making decisions; both who and how they do that. There should be a notion of the role of institutions and how ideas are developed and what role they play in the process. Also, governments are not alone in policymaking, there are subsystems and networks that are involved and they operate in a certain policy context. Lastly, there are also events that are either regular or irregular but can change the proceedings in a public policy process.

In the public policy process there are also features of complexity present. The features of complexity are that the system is more than the sum of its parts due to feedback processes, initial movers, emergence and strange attractors. The public policy process has many interdependent connections where an initial mover can set a string of actions and reactions in motion. It includes feedback processes which lead to the emergence of a distinct behaviour of the process and could be disrupted by strange attractors. The highly abstract elements of a public policy process could all be linked to each other through these features of complexity.

17

3. Methodological framework

In this thesis, I make use of methodological pluralism. Two different methods are used to answer the research question: What concepts and relations used in contemporary journal articles explain the

complexity of the public policy process? This question has two components, which are each analysed

best by a different method. On the one hand, a method of inquiry and data collection is necessary to identify the concepts and relations that represent the elements of a public policy process theory. A systematic review is used to identify these elements. On the other hand, the relationships between these concepts need to be characterised and interpreted in terms of a complex system. A system dynamics approach is used to accomplish this task. Why these methods were chosen and how they were applied in this research will be elaborated on below. Subsequently, a discussion on data collection, operationalisation, reliability and validity follows.

3.1 Research Strategy

The objective of this research is to develop a model that describes and explains the dynamics of the public policy process. In its very essence, this research uses an abductive approach as the model should be capable of providing testable hypotheses. Abductive research is research aimed at producing new hypotheses and theories based on evidence (Timmermans and Tavory, 2012). Unlike deductive approaches, abduction actively seeks to build hypotheses. The difference with inductive reasoning is that induction is based on primary observations while abduction is based on previous evidence (Timmermans and Tavory, 2012). The abductive approach suits the research aim for two reasons. First, this thesis aims to build a model that describes and explains the contemporary public policy process. Neither current models nor theories explain the full process or strike a useful balance between description and explanation. An abductive approach allows to build on the evidence those theories generated while still constructing a model that provides both description and explanation of the full policy process (Schwaninger and Hamann, 2005). Second, even though the process must be an accurate description of actual policy development, this thesis still aims to build a generalisable model. An abductive approach builds on previous evidence, which allows using a broader range of units of observation as opposed to an inductive approach (Timmermans and Tavory, 2012).

This thesis will synthesise previous evidence into a single model that represents the contemporary public policy process. It will do so by means of two methods. First, a systematic review will help gathering the evidence and to elicit the important elements of the policy process. Second, utilizing system dynamics a model will be constructed that shows the relationships between the concepts of recent scholarship and will reveal the complexity of the policy process.

3.2 Systematic Review

A systematic review is “an integration of evidence about an effect or intervention involving the summary of findings from a defined set of relevant and usable primary sources” (Yardley, 2020, p.187). These primary sources are previous literature studies, but the systematic review goes beyond an ordinary literature review as the systematic review has an explicit method to execute this review (Harden and Thomas, 2015). By explicitly stating the actions of each step there is a bigger chance to uncover bias (Harden and Thomas, 2015). The different steps of the systematic review will be elaborated on in a later section but are divided in a planning, conducting and reporting stage (Yardley, 2020).

A systematic review can be done both qualitatively and quantitatively. Quantitative systematic reviews combine the results of multiple studies statistically to elicit generalised principles (Grant and Booth, 2009). This is also called a meta-analysis. In order to conduct this analysis, the content and the fashion of these studies should be rather homogenous so the results could be integrated (Harden and Thomas, 2015). Qualitative systematic reviews have a more interpretative character. The findings depend more

18

on the reviewer’s judgement (Harden and Thomas, 2015). This provides the reviewer with the chance to synthesise a heterogeneous group of literature (Grant and Booth, 2009). It requires a more detailed description of the synthesis process, but it might lead to a new narrative and a more meaningful synthesis than “isolated comments from local questionnaires or surveys” (Grant and Booth, 2009, p.100). The qualitative systematic review is a suitable method for this research because of two reasons. First of all, a systematic review fits with the aim of this research. The aim is to develop a model of the public policy process that reflects the contemporary complexities and practices of this process. A systematic review captures these elements. This method is aimed at collecting a large body of literature that reflects a certain phenomenon (Grant and Booth, 2009). This thesis is not only aimed at synthesising the literature itself, that would be a literature review, but it is also aimed at processing the empirical evidence for these theories. A systematic analysis is aimed at synthesising the implications and findings of theory (Mertens, 2018).

Second, the qualitative approach to synthesising the findings is better suited to heterogeneous bodies of literature. Public policy research is characterised by case studies (Weible, 2014). The chance of finding heterogeneous studies is thus substantial and a qualitative approach to analysing these findings is preferred (Petticrew and Roberts, 2006). A qualitative study provides a narrative of a certain field of study (Petticrew and Roberts, 2006) and can be used to develop new theories or hypotheses (Harden and Thomas, 2015). This study aims to develop a model of the public policy process that accommodates the shortcomings of the traditional policy cycle. One of the flaws of this framework was that it lacked testable hypotheses and theory (Sabatier and Jenkins-Smith, 1993). By organising the methodology of this study towards the development of new theoretical insights, this thesis can include the testable hypothesis in the conceptualisation.

There are different interpretations of which steps exactly must be included in a qualitative systematic review (Yardley, 2020; Petticrew and Roberts, 2006; Tight, 2019). Ultimately, they all come down to approximately the same procedure: a three-stage procedure with several sub-steps. The first stage is to organise the study (Petticrew and Roberts, 2006). That means that the research problem and scope are clearly defined in order to identify the sample (Tight, 2019). This will be done according to principles of the systematic review and will be further elaborated on in section 3.4. The next stage is the within-study analysis (Petticrew and Roberts, 2006). In this stage the findings of each within-study are described and the quality of the studies is assessed (Petticrew and Roberts, 2006). The last stage is the development of cross-study synthesis. Here the outcomes of the different studies are interpreted and integrated (Petticrew and Roberts, 2006; Tight, 2019; Yardley, 2020). In section 3.5 the data synthesis will be explained and operationalised.

3.3 Modelling complexity

A critical phase in this thesis is to model the relationships between the identified elements of the public policy process. There are multiple methods and tools to make complex systems comprehensible. These methods can be separated into system dynamics type models and agent-based models (Boccara, 2004). “To explore the complexity in a complex system, system dynamics (SD) and agent-based modeling (ABM) are the two most commonly used approaches” (Ding et al., 2018, p.2). System dynamics (SD) type models are either quantitative or qualitative models that describe the behaviour of a system as a whole by focusing on feedback processes between the elements of the system (Vennix, 1996). Agent-based modelling (ABM) focuses on individual decision making entities that collectively determine the

19

behaviour of a system (Bonabeau, 2002). The main difference is that with SD structure drives behaviour whereas ABM operates from the assumption that individual behaviour drives group behaviour. When modelling complex systems there are two tensions that the researcher needs to consider when choosing one modelling method over the other. These are the tensions between simplicity and validity, and between formality and generality (Edmonds and Gershenson, 2015). Simplicity refers to the comprehensibility of a model. Validity refers to the extent to which the model accurately represents the real world. Generally, the more accurately a model resembles reality, the less simple it is. There must be a balance between the two. A too simple model would create to the same problems as the policy cycle stumbled upon, while a too accurate model would be unusable for potential users without the explicit knowledge on how to interpret models of complexity. Formality refers to the specificity of a model to a particular case or situation. This would allow a model to be more valid without losing much of its simplicity because users are familiar with the case. Generality refers to the opposite of formality. High generality of a model means that it is applicable to many situations and cases. Although all aspects are important, the goal of this study is more oriented towards a general model as it should be a model of the public policy process rather than a public policy process. The tension between validity and simplicity is more problematic. Yet, this thesis aims to provide an accurate description of the process of public policy development as a reaction to the simplistic policy cycle, hence it is clear that the modelling approach should be able to accurately reflect the policy process.

Given this prioritisation of generality over formality and validity over simplicity, one could start comparing ABM and SD modelling approaches with regard to their suitability. SD is a more suitable approach for three reasons. First, even though both modelling approaches are aimed at simulation modelling, SD also provides the option of qualitative modelling (Bonabeau, 2002; Sterman, 2000; Vennix, 1996). This allows a model to be less formal and more abstract as there is no need for empirical data to simulate the system’s behaviour. Secondly, SD is an approach that makes use of aggregated entities. When describing a full process, this helicopter view might be more useful than modelling each individual decision-maker. Moreover, ABM is notorious for developing highly complex models which would pose a challenge to keeping the model sufficiently simple (Edmonds and Gersherson, 2015). The helicopter perspective of SD is better able to keep the balance between simplicity and validity. Thirdly, this thesis aims to describe the process and its complexity. SD offers an approach to model structures that drive behaviour. This allows us to consider how different elements in the structure of the system influence each other and jointly produce a certain outcome. Considering the elements of a public policy theory, several parts of that structure were already identified: institutions, context, events and subsystems. Only one element is decision making, something that ABM excels at. Yet, this thesis considers the public policy process more as a structure rather than a decision making procedure. This means that it is about the relationship between all factors of the system instead of a sole focus on how agents behave within a certain set of rules (Bonabeau, 2002; Sterman, 2000).

3.3.1 System dynamics

SD is a method to model the structure of complex issues in an attempt to explain its behaviour (Sterman, 2000; Vennix, 1996). The core element in SD to explain complex behaviour is feedback (Vennix, 1996; Sterman, 2000). Further, it uses closed boundaries and the discrepancy between desired and perceived state as its core principles for building models. It thus perceives a system as a self-maintaining structure. There are three types of modelling approaches in SD: causal loop diagrams, qualitative stock and flow models, and quantitative stock and flow models (Vennix, 1996). Causal loop diagramming (CLD) is a tool to present a system’s structure of feedback processes (Vennix, 1996). Some SD practitioners claim that CLDs are not real SD modelling techniques, but merely stepping stones for stock and flow models

20

(SFD) (Forrester, 1994). The advantage of CLDs is their simplicity and generality. SFDs are more specific and often case-specific, but they allow for simulations and thus a simulation of dynamic behaviour over time. An SFD displays aside from feedback processes also accumulations in stocks (Vennix, 1996). They are thus more informative than CLDs, yet developing them is more time and resource consuming (Veldhuis et al., 2015).

In this study, a CLD is sufficient for satisfying the research aim and answering the research question for three reasons. First of all, the aim and question are focused on obtaining insight into the role of the elements of the public policy process in order to be able to explain the system’s behaviour. A CLD is capable of displaying the structure of the policy process and its complexity in two manners. One is that complexities in the structure are identified and shown via feedback loops. The second element is the use of polarity to provide more insight into how the different variables influence each other (Sterman, 2000). The combination of these two features of CLDs allows for a qualitative depiction and assessment of the public policy process.

Secondly, modelling the policy process requires a balance between simplicity, generality, formality and validity (Edmonds and Gershenson, 2015). Although all four features should be satisfied, the goal of this thesis is aimed slightly more towards generality and validity. It should be a generally applicable model and it should provide insight into the complexity of the policy process. It is thus more important to have a general model than a formal model, and more important for the model to be valid than that it is simple. Because of the qualitative nature of a CLD, there is more room for interpretation and more room for abstract terminology. This allows me to satisfy the two prioritised features of the model. This does not mean that the model will become too detailed in order for it to be as valid as possible, as generality requires also simplicity to some extent. In the same vein, a valid model requires the model to lean towards a certain formality, because a too abstract model might not be a sufficiently valid representation of reality. Hence, there is a focus on generality and validity, while still aiming to keep the balance with formality and simplicity.

Third, a CLD is more capable of providing a narrative that can be interpreted by different users in their own particular cases. A quantitative SFD would require more modelling knowledge to adjust the numbers in such a way that the model tells their story. A qualitative SFD also includes stocks and specified units of analysis, which would already make the model much more steering in the interpretation of the model. A CLD is thus more versatile in its construction and use than an SFD. Even though this research is theoretical in nature, it still needs to appealing for practical use. “What policymakers want, above all, is validity, with generality […] and simplicity” (Edmonds and Gershenson, 2015, p.210-211). So a CLD would satisfy the theoretical goal of providing insight into the complexities of the public policy process, while at the same time being more suitable to be a useful model to policymakers.

3.4 Data collection

The data for this research is formed by articles published in the most impactful journals that relate to the public policy process. The reason to use this as data is twofold. First of all, this thesis aims to contribute to the development of policy process theory not by adding a piece to the puzzle but by putting existing pieces together. If this research would have been based on primary empirical data, I would have added another piece to the puzzle, which might have fitted within the frame but at the same time would make the puzzle that attempts to explain the public policy process more complicated as the number of pieces would increase. By using articles as data I try to find a shared language between them and find the pieces among the large scholarship that already exist and put them together to help to complete the puzzle. Secondly, this thesis uses journal articles as its data to construct a CLD as opposed to combining grand

21

theories1 _{because the assumptions of those grand theories might be incompatible (Cairney and Heikkila,}

2014). Grand theories such as the advocacy coalition framework, multiple streams theory and punctuated equilibrium have proven their value and are still used to explain at least a part of the policy process, but due to mutually incompatible assumptions, they have failed to explain the full policy process (Weible, 2014). Therefore to transcend these incompatibilities the concepts in the articles are used instead of the full grand theories that lie at the basis of them.

The Journal Citation Report of the Web of Science search engine was used to determine the journals in the field of the public policy process that have the highest impact. As public policy process theory is a typical research field of public administration, I searched only in that particular category. To determine which journals have the highest impacts, the Journal Impact Factor (JIF) is used (Web of Science Training, 2017). The JIF “provides a ratio of citations to a journal in a given year to the citable items in the prior two years” (Annual Reviews, n.d.). In order to collect the most recent applications of theories, the journals were assessed for the last ten years. That means the JIF’s between 2018 (last reported date) and 2010 were taken into account. The selected journals had on average the highest JIF’s between 2010 and 2018. The selected journals are found in table 3.1.

Journal _{Number of articles found}

Climate policy 5

Governance- an international journal of policy administration and institutions 5

Journal of European public policy 6

Journal of policy analysis and management 1

Journal of Public administration research and theory 0

Policies studies journal 17

Policy sciences 15

Public administration 2

Public administration review 4

Regulation and Governance 3

Total 58

The next step to collect the sample of articles that are used to analyse the dynamics of the policy process is to set up a search string. Via Web of Science the following phrases and nouns were used divided by the Boolean operator OR to look for keywords in the title, abstract, author keywords and keywords plus (Web of Science, n.d.): policy development, policy cycle, policy research, policy dynamics, policymaking, policy-making, policymaking, or policy formation. Using the Boolean operator AND, these keywords were combined with the topic “policy process”. This way the articles that were not concerned with the policy process were eliminated. This search gave a total of 947 articles between 1945

1 _{Grand theories are widely accepted, generalisable theories. In the field of public administration these are the}

punctuated equilibrium theory, multiple streams model and the advocacy coalition framework. For an overview, see Cairney and Heikkila (2014).

Table 3.1