How ideas matter in public policy: a review of concepts, mechanisms, and methods

The recent ideational turn in political science and public administration implies that ideas matter. Ideas are an essential explanatory concept for understanding policy changes and decision-making processes. The aim of the paper is to specify how ideas matter as a variable in public policy research, providing students and scholars of public policy with a stock take of the current state-of-the-art literature on ideas in political science and public administration. The paper first identifies three approaches to ideas as a variable in the policy process. It then discusses where ideas come from and the dynamics and drivers of ideational change to shed light on the ideational mechanisms underpinning policy processes. Furthermore, it taps into different research methods that can be used to study ideas. Finally, the paper concludes with five lessons for future research endeavours on the study of ideas in public policy.


Taking stock of how ideas matter
Public policy studies and political science have experienced an 'ideational turn' in recent years. Ideational explanations have allowed public policy scholars to clarify the content of policy changes and how such changes come about (Béland & Cox, 2011;Parsons, 2007;Schmidt, 2008). Key research programmes and frameworks in policy sciences, such as Sabatier's Advocacy Coalition Framework, Kingdon's multiple streams approach, and Hall's work on policy paradigms and social learning, have spurred an academic debate on the role of ideas in public policy and political science (Hall, 1993;Kingdon, 1984;Sabatier, 1988). A growing number of studies suggests that ideas are an important variable shaping public policymaking processes (Baumgartner, 2014;Béland & Cox, 2011;Berman, 2013;Blyth, 2013;Schmidt, 2008). Arguing that 'ideas matter' has become like preaching to the choir: ideational factors are now widely considered an essential variable in the analysis of various political and policymaking phenomena (Kamkhaji & Radaelli, 2019;Mehta, 2011).
However, ideational scholarship is scattered across subdisciplines, and a comprehensive overview of relevant research questions, findings, and methods is presently lacking. Thus, the analysis of the effects of ideas on public policy is hampered by problems that arise from the imprecise specification of what ideas are, where they come from, when and how they change, and how to study them (Berman, 2013;Carstensen, 2011a;Kamkhaji & Radaelli, 2019;Meyer & Strickmann, 2011;Parsons, 2007;Schmidt, 2016). Thus far, ideational scholars have provided us with classifications of different types of ideas (e.g. different levels of generality, ranging from specific policy programmes to overarching cultures) as well as different ways of thinking about ideas (e.g. ranging from positivist to constructivist approaches, stressing ideas as causes of or constitutive of public policy) (cf. Hall, 1993). A result of the concept being used in a variety of subdisciplines is that it has been stretched to include a number of phenomena and concepts, making it close to impossible to understand what is ideational and what is not (Kamkhaji & Radaelli, 2019). This makes it difficult to understand how ideas matter.
Hence, the aim of this paper is not to reconstruct a succinct overview of the development of ideas in political and policy science. Others have done excellent jobs reviewing the ideational turn in political and policy science (for overviews, see Béland &Cox, 2011 andParsons, 2007). Instead, the aim is to provide students and scholars of public policy interested in the concepts of ideas and ideational change with a stock take of the current state of the art to help them navigate the broad interdisciplinary field of ideational scholarship and identify challenges for future ideational scholarship.
To do this, I reviewed articles on ideas and ideational change published over the last 25 years, guided by the PRISMA method (Petticrew & Roberts, 2006). This research approach allows the researcher to obtain a comprehensive, structured overview of a certain concept or study domain. Using PRISMA for this paper, I obtained an overview of the current literature regarding ideas and ideational change that spans different disciplines. The initial corpus of 71 articles resulting from the analysis was examined thoroughly, and complemented with additional literature. 1 In sum, most reviewed studies were published by scholars in Anglo-Saxon universities, focusing on the role of ideas in public policy at the national level (specifically in the domains of economic, foreign, or climate policy), and both theoretical and empirical in nature (see appendix for full overview of quantitative data about the corpus). A majority of the articles does not specify research design or methodology, data collection techniques or modes of analysis. Furthermore, scanning the corpus suggests that current ideational scholarship is scattered throughout the political and policy science landscape (see Figure 1 in appendix) with little to no cross-fertilisation between journals rooted in different subdisciplines (e.g. political economy and political psychology). This leads to the expectation that there are competing answers regarding four key questions policy researchers should be able to answer when they choose to use ideas as a variable in their research: What is an idea? Where do ideas come from? What are the dynamics and drivers of ideational change? And what methodology can I use to study ideas? Answering these questions will help policy researchers shed light on the pivotal question of how ideas matter in public policy.
The conclusion charts five lessons and objectives for scholars of ideas in public policy research: (1) taking micro-level cognitive dynamics seriously; (2) theorising about relationships between different types of ideas; (3) specifying the conditions, mechanisms, and sequences involved in processes of ideational change; (4) expanding their methodological toolbox with innovative methodologies to measure ideas; and (5) examining the ways in which ideas matter in different policy domains and settings.

Specifying the concept of ideas in public policy
Policy researchers have come to accept that ideas should be taken seriously as a variable in explanations of public policy outcomes (Cairney, 2019;Mehta, 2011). The corpus used suggests that the relationship between ideas and policy outcomes takes many forms and depends on the precise specification of what ideas are. Concepts associated with ideas are wide-spread, bringing in a 'conceptual minefield' (Berman, 2013;Kamkhaji & Radaelli, 2019;F. van Esch & Snellens, 2019). Broadly, ideas can be defined as "beliefs held by individuals or adopted by institutions that influence their actions and attitudes" (Béland & Cox, 2011, p. 6). Furthermore, scholars seem to assume that what constitutes an idea ranges from specific, programmatic ideas to more general philosophies (Kingdon, 1984;Schmidt, 2008), or from diagnostic beliefs to principled beliefs (cf. Jervis 2006; Van Esch and Snellens 2019).
Simultaneously, the Béland and Cox definition displays the complex and differentiated nature of the concept as a word and emphasises that there are different approaches to determining what ideas are. For instance, stressing ideas as beliefs held by individuals is ontologically different from viewing ideas as embedded entities in institutions. Depending on the chosen perspective, the effect of ideas on policy outcomes may differ, and different mechanisms may cause these effects.
Herein, I unpack Béland and Cox's (2011) definition by reviewing the literature. 2 The review leads to three broadly different conceptual approaches that scholars working with ideas use across subdisciplines: ideas as sense-making heuristics that guide people's actions, ideas as strategic tools that actors use to craft political discourse, and ideas as institutional frameworks that have an effect on their own and maintain some order throughout the actions of individuals, groups, and society (cf. Blyth, 2001). These approaches vary on their ontological positions on structure and agency and on their epistemological positions on positivism (explaining) and interpretivism (understanding) (Hollis, 1994). Here, in a simplified manner, I address the core positions of these different approaches.

Ideas as heuristics
First, ideas can be conceptualised as heuristics or subjective beliefs. 3 Heuristics are cognitive short-cuts that help people make sense of the complex world around them. Ideas are internalised in people's hearts and mind. People depend on such pre-existing ideas to understand the situations they are in (cf. Jones, 2017).
Scholars term ideas as beliefs, cognitive short-cuts, mental aides, or heuristic devices (Brummer, 2016;Burdein, Lodge & Taber, 2006;Foyle, 1997;Radaelli & O'Connor, 2009;Renshon, 2009). A belief system can be defined as a more or less integrated set of beliefs about the human physical and social environment. "An individual's perception, in turn, is filtered through clusters of beliefs or 'cognitive maps' of different parts of his social and physical environment" (Holsti, 1976, p. 20). Thus, ideas act as the filter through which information is viewed and judged (Fielding, Head, Laffan, Western & Hoegh-Guldberg, 2012;Jacobs, 2009).
Beliefs are subjective as they refer to an individual's understanding of cause-and-effect relationships or normative assumptions about what is good or bad (Jervis, 2006). Beliefs do not provide accurate or objective assertions about the world but 'coloured lenses' through which individuals make sense of the world. This suggests that beliefs limit the capacities of policy actors to review policy debates holistically and constrain the available policy alternatives they consider when making policy decisions.
This approach to ideas originated in foreign policy analysis and political psychology (Jervis, 2006). The fundamental research question asked to understand policy processes, is how to understand what those who make public policy believe. Alexander George (1969( , in Larson, 1994 was among the first to develop this approach in political psychology. He proposed that policy actors had a certain 'operational code' consisting of philosophical and instrumental beliefs, setting their parameters for action. This 'opcode' could be constructed through analysis of the verbal behaviour of policy actors. The pioneer of what recently has started to be referred to as behavioural public administration, Herbert Simon (1947( , in Mintrom, 2015, noted early on that any theory of public administration should consider that policy actors are 'boundedly rational'. Therefore, scholars need to grasp how policy actors process information and form views about their decisions. Simon's work found its way into what have become classic policy studies, such as Lindblom's incrementalism, Kingdon's multiple streams, Baumgartner and Jones' punctuated equilibrium, and Sabatier et al.'s Advocacy Coalition Framework. According to Jones (2017), the 'cognitive approach' focuses on the influence of policy actors' beliefs on policy outcomes, for example by studying the effect of public opinion on the type of policy outcome or the effect of policy-makers' beliefs on policy choices (cf. Yee, 1996). Despite early work denoting that policy actors' beliefs should be taken seriously, many key theories in public policy literature continued to focus on the 'system' as their level of analysis than the individual decision-maker per se. Furthermore, scholars of public administration in part neglected theories and methods from (political) psychology to study these micro-level processes (Grimmelikhuijsen, Jilke, Olsen, & Tummers, 2017). In recent years, scholars have aimed for cross-fertilisation between public policy theories and (political) psychology theories and methods, with an increasing number of studies on behavioural public administration and the work on micro-foundations in policy learning literature (Dunlop & Radaelli, 2017;Moyson, Scholten, & Weible, 2017).
In sum, the cognitive approach departs from the micro-level of individual beliefs to explain macro-level phenomena in policymaking. Actors' beliefs constrain or enable them to influence the policymaking process (e.g. agenda-setting, limiting decision-making alternatives, facilitating groupthink). Contemporary questions for this approach centre on how strongly and under what conditions policy actors hold on to which beliefs and examine what causal mechanisms instigate belief changes (cf. Van Esch and Snellens, 2019). In the analysis of study findings in the corpus (see Table 1), both exogenous factors such as crisis, as well as endogenous factors such as the personal disposition of actors' presumably influence the stability of beliefs, leading to competing hypotheses on mechanisms of ideational change (Golec de Zavala & Van Bergh, 2007;Van Esch & Swinkels, 2015). Ontologically, scholars in this approach argue that policy action can be primarily explained by examining an individual's 'internal computer'. Epistemologically, this scholarship seeks causal explanations for how such beliefs come about.

Ideas as strategic tools
Alternatively, scholars take a linguistic, or discursive, approach to ideas and their potential effects on public policy and institutional change. 4 From this perspective, ideas influence public policy when viewed as the content of discourse (Schmidt, 2008). The role of policy actors is to 'do things with words.' Ideas reflect policy actors' normative orientation towards the context in which they operate, and these determine the behaviour they display in that context (Hay, 2011, p. 67;Blyth, 2001). Scholars of the 'argumentative turn' conceptualise public policy as a social construct (Schon & Rein, 1995). This requires policy analysts to focus on how policy actors perceive, understand, and frame policy issues, and how they make sense of ideas and imbue them with meaning in the policy process (Hajer & Wagenaar, 2003).
Despite different orientations within this approach, scholars take argumentation to be essential to grasp how actors perceive the world and interact with their counterparts. Much of the recent interpretive or argumentative work on ideas and public policy centres around discursive institutionalism (Schmidt, 2008). Here, ideas are the substantive content conveyed through the interactive process of discourse. Discourse allows ideas to be generated and communicated in institutional settings (Schmidt, 2011). Change in these institutional settings arises through "dynamic processes through which agents use not only their 'background ideational abilities'... to create and maintain not only their institutions but also their 'foreground discursive abilities' ... to communicate and deliberate about taking action collectively to change (or maintain) those institutions" (Schmidt, 2011: 685). In effect, ideas can be seen as tools purposefully deployed by policy actors to shape the meaning of discourse.
Conceptualising ideas as strategic tools emphasises an active role for actors to consciously work with ideas. Actors engage with ideas, adjust them, and challenge existing ideas through the use of political discourse. Through discursive practice, actors can build coalitions, shape political agendas, navigate the political arena (Morrison, 2016), and effectively induce policy change. This approach is used in different theories of the policy process, for example in the Narrative Policy Framework and the Social Construction Framework (Sabatier & Weible, 2014). These treat ideas as narrative strategies and social constructions of (groups of) actors.
Policy actors use ideas as rhetorical weapons and armoury in political struggles (McNamara, 1999). If an individual in a group reconfigures the substantive content of an idea he or she holds, the dominant discourse does not necessarily change (yet). Through repeated interaction certain ideas can institutionalise -providing power 'in' ideas (Carstensen & Schmidt, 2016).
For example, Schmidt and Thatcher (2014) show the resilience of neoliberal ideas in Europe throughout multiple crises as an effect of the weak substantive content of alternative ideas and the effectiveness of neoliberal ideational entrepreneurs' in political discourse. However, if most people within a policy domain become convinced of the content of alternative ideas, the dominant discourse becomes vulnerable to change and can even disappear. Policy actors potentially have power 'through' ideas, understood as the capacity of actors to persuade other actors to accept and adopt their views through the use of ideational elements (Carstensen & Schmidt, 2016).
Conceptualising ideas as strategic tools in discourse furthermore links to notions of ideational entrepreneurship or craftsmanship (Bratberg, 2011;Schonhardt-Bailey, 2005). Without strong ideational entrepreneurs, ideas cannot gain prominence in groups or networks. Through positional power and rhetorical skills, entrepreneurs create, represent, promote, and embed group ideas. The key rationale here is that ideas about policy problems or solutions are not a pre-established heuristic device, but are actively constructed through discourse by policy actors (Hajer, 2013;Mehta, 2011;Schon & Rein, 1995). Hence, this approach is situated at the interpretive side of the epistemological spectrum as opposed to the positivist perspective of the cognitive approach.
This ideational scholarship thus zooms in on the meso-level in policymaking; i.e. how policy actors construct the meaning of public policy through social interaction in groups or in society. Actors are embedded in society but have agency to act and initiate change (Hollis, 1994). Contemporary debate in the ideas-as-strategic-tools approach concerns the extent to which actors are capable of bringing about ideational change as well as the conditions under which they can bring about such change. In sum, when do policy actors gain power 'through' ideas and instigate joint ideational shifts that reconfigure a pre-existing discourse?

Ideas as institutional frameworks
Third, ideas can be conceptualised as institutional frameworks. 5 They can be understood as intersubjective understandings embedded in institutions or societies. This approach differs from the heuristics approach, as the causal logic works in reverse. Rather than working their way out from inside people's hearts and minds, ideas work their way into people's hearts and minds from the outside and either hamper or enable them to act. Ideas function as societal norms or paradigms, giving people a sense of direction in uncertain times (Blyth, 2001). Rooted in historical institutionalism, ideas provide policymakers and politicians with "interpretive frameworks that specify the nature of the problems that policy-makers are meant to be addressing" (Hall, 1993, p. 279).
Ideas can thus give 'organised existence' to a public policy domain (Bell, 2012), representing shared "systems of thought consisting of a series of interconnected claims and assumptions about how something functions" (Baker & Underhill, 2015, p. 381). Ontologically, ideas are sit-uated at the level of structure. No included study in the review takes a structural-explanatory perspective on ideas. Instead, a structural-interpretive perspective is prevalent. From this perspective, ideas are sets of meaning telling actors how 'to do' social life. Actors are perceived as 'followers' of these sets of meaning (Hollis, 1994). These sets of meaning are open to contestation, but in public policy, the range of potential interpretations is not infinite. Instead, contestation is shaped by path dependency mechanisms (Pierson, 2000) and hegemonic discourses (Blyth, 2013). For example, Matthijs (2016) argued that the ideational construct of ordoliberalism in Germany shaped German policymakers' responses to the Euro crisis. Carstensen and Matthijs (2018) show how the precepts of neoliberalism survived both the global financial crisis and the Euro crisis.
This type of ideational scholarship zooms in on macro-level phenomena in policymaking; i.e. how interpretive policy paradigms govern policy domains. Contemporary questions in this approach are how coherent such ideational frameworks truly are, how change occurs over time, and how such institutional frameworks affect policy debates. Scholars often depart from posthoc analysis of uncertain and complex policy processes. Hence, some scholars render this a problematic approach to studying the impact of ideas on public policy as it fails to account for the cognitive or discursive processes playing a role throughout the policy process (cf. Kamkhaji & Radaelli, 2019).

Mixed approaches
Many ideational studies use hybrid conceptual approaches combining elements of the different approaches discussed here.
This blurs the boundaries between ontological and epistemological positions. 6 For example, studies perceive ideas both as 'worldviews' and shared mental constructs that are used strategically by actors to achieve their goals (cf. Carstensen & Schmidt, 2016;Helgadottir, 2016;Kuisma, 2013;Parsons, 2016;Saurugger, 2013). This is for good reasons. A one-dimensional focus on ideas as institutional frameworks will tend to leave out a theory of individual action. In contrast, focusing on individual decision-makers only may overemphasise their self-described motivations and exclude the constraints of the social structures under which they are operating (Jacobs, 2015). The 'un-boxability' of these approaches is both a strength and a weakness. The strength of such hybrid approaches is their ability to identify many valuable dimensions of how ideas matter. However, what remains under-theorised is how these different types of ideas relate to one another as a result of competing ontological positions and epistemological views. These imprecise specifications make it easy for other scholars to critique the ideational approach in public policy. Specifying the relationship or interaction between different approaches to a concept allows us to rethink and elaborate on what aspects of the ideational spectrum we study, ultimately spurring our understanding of the potential causal effects of ideas on public policy.
To conclude, ideas can be understood from a cognitive, strategic-discursive, or institutional perspective. Differentiating between these three approaches helps us to answer fundamental questions about where they stand on ontology and epistemology (Hollis, 1994). Cognitive and strategic approaches locate ideas with individuals, whereas the institutional approach locates ideas in structure. A cognitive approach entails a more positivist epistemology, examining causal explanations about how beliefs come about or how they affect policy. Strategic and institutional approaches stand on the interpretive side of the epistemological spectrum, focusing on how social structures are formed through or structure interpretation and meaning making of policy actors in public policy. Departing from different answers to the question of what ideas are, the question of where ideas come from and how they change is a next step of inquiry into how ideas matter.

Where do ideas in policy processes come from, and what makes them change?
As Kingdon (1984, p. 72) stated, "ideas can come from anywhere" and also "from a plethora of sources". The question of where ideas come from links to the carriers of ideas in policy processes. As argued above, different ontological positions determine 'who' carries ideas. Different epistemological positions further specify where to look for them. First, from a cognitive approach, ideas, come from 'microfoundations'. Second, a strategic-discursive approach shifts our focus to factors and mechanisms explaining the distribution and transfer of ideas between individuals or between individuals and groups (micro-to micro-or micro-to mesolevel). Alternatively, ideas can be seen as epiphenomenal frameworks (Kamkhaji & Radaelli, 2019). This shifts the focus to factors such as time, context, and events to explain how taken-for-granted ideas in a policy domain may be altered.
How ideas form and how they change, in short, draws our attention to both institutional and contextual (exogenous) factors as well as actor-specific and entrepreneurial (endogenous) factors (see Table 1 for a full overview). Relating different types of factors to who 'carries' ideas leads to the identification of different mechanisms of ideational change. For example, from an agency-endogenous perspective, identified mechanisms of ideational change include: (instrumental and social) learning, input, persuasion, puzzling (Van Esch & Snellens, 2019). These mechanisms all refer to processes in which policy actors try to make sense of new or alternative ideas in different ways. From a structural-endogenous perspective, mechanisms such as socialisation, imitation or identification describe how policy actors adapt to ideas in a social context. From an exogenous-agency perspective, mechanisms of ideational change are more 'political' and include coercion, powering, or turn-over. Alternatively, from a structural-exogenous perspective, ideational change occurs through dispersion. Ideas then spread among human beings more or less the same way as germs or viruses do (cf. Dunlop & Radaelli, 2017;Moyson et al., 2017 and Van Esch & Snellens, 2019 for a more in-depth discussion of alternative mechanisms of ideational change). In the discussion below, I discuss the three most dominant mechanisms identified in our initial corpus of ideational studies related to ideational change: learning, persuasion and socialisation. 7 As studies in our corpus often identified endogenous factors as triggers for change, while identifying exogenous factors as a scoping condition for change, it makes sense that these three mechanisms are most prevalent in the reviewed corpus.

Ideas come from microfoundations
Microfoundations refers to understanding public policy as the aggregate of policy actors' behaviour. If ideas affecting policymaking originate in the cognition of individual policy actors, then a logical second step is to examine what factors influence cognition (e.g. openness to information, education, emotion, experience) as well as when and how these factors contribute to changes in cognition (Brummer, 2016;Kamkhaji & Radaelli, 2017). The corpus included studies into the effects of personal disposition on both the content and flexibility of policy actors' ideas in policy processes (see Table 1 for a complete overview of expectations and findings of the included studies). For example, studies on the effect of cynicism, values, level of education, anxiety, and arousal have demonstrated that these factors affect the core meaning of the beliefs of individual actors in policy processes (cf. Bolsen, Druckman & Cook, 2015;Brewer & Steenbergen, 2002;Fielding et al., 2012). These studies discovered a positive relation between these factors and their influence on the beliefs of policy actors. For example, Fielding et al. (2012) showed how personal disposition (e.g. level of education) is positively associated with politicians' belief in climate change and this guides their response to policy issues (see Table 1).
Other studies have examined the effect of personal disposition on the stability or flexibility of beliefs as a proxy for their receptivity to new, alternative policy ideas (see Table 1), for example the effects of traits, expertise, partisanship, and emotion on belief stability. For instance, Brummer (2016) suggested a relationship between a political leader's openness to information and their reflective stance towards new policy ideas. As a result, openness to information could decrease the likeliness of leaders to become involved in policy fiascos. Meanwhile, expertise is seen as a constraining factor for belief change (cf. Larson, 1994). Emotions, such as anxiety, could prefigure the beliefs of elites (Renshon, Lee & Tingley, 2015;Widmaier, 2010). These factors could make policy actors less prone to belief changes, ultimately affecting policy stability.
The oft-used causal mechanism through which an individual actor is believed to change their cognitive ideas is instrumental learning (Van Esch & Snellens, 2019). Learning in this sense refers to the 'updating of beliefs' (Dunlop & Radaelli, 2013, p. 599). Through learning, actors can stick to, reinforce, or revise their ideas about public policy (Trein & Vagionaki, 2020, p. 8) According to the literature, learning is more likely to occur under scoping conditions of crisis, external challenges, discrepant events, and failures. Scoping conditions in this sense refers to the context in which a particular mechanism is theorised as likely to be activated (Beach & Pedersen, 2013). Cognitive-ideational scholars often choose such scoping conditions as the empirical research environment in which to study if and when learning occurs. The need to respond to scoping conditions, which generate uncertainty about existing beliefs and the policies based on them, may set people on the path of learning. However, depending on the personal disposition of actors, such challenging circumstances could also produce impulses to adhere to and defend existing ideas (Van Esch, 2014). From a cognitive perspective, it is, therefore, questionable whether political actors are as prone to ideational change as people tend to think. People are often rigid and resistant to new ideas, even under crisis conditions (Moyson, 2017).
In conclusion, the policy process is permeated by the beliefs of individual policy actors and their beliefs are conditioned by cognitive processes. The personal disposition of policy actors determines their receptivity to new, alternative ideas in response to scoping conditions (e.g. crises, exogenous shocks). As such, only when learning occurs and actors update their beliefs in response to new circumstances can we find an effect on policy change.

Ideas come from interaction and entrepreneurship between policy actors
Scholars taking a discursive approach depart from the assumption that ideas are tools for policy actors to transfer and distribute in groups or networks. Questions of ideational change do not centre so much on the stringency or flexibility of ideas, but on questions of reconfiguration or convergence and divergence of ideas. As such, studying the effect of ideas on public policy concerns the ways in which ideas are transmitted in policy processes. Ideational research in this stream takes us to the study of policy entrepreneurs and their capacities to transmit ideas in networks, advocacy groups, and epistemic communities. The concept of policy entrepreneurs was first established by Kingdon (1984) and refers to policy actors capable of linking policy problems to policy ideas to promote policy decisions. Yet, a major challenge ahead lies in the ways in which such communicative interactions of policy entrepreneurs truly function as an explanation of policy processes.
Scholars surmise that the positions of and power divisions between policy actors within existing policy networks, the availability of new ideas, and their rhetorical skills to promote them determine which actors become key entrepreneurs in the transfer and distribution of ideas (Moschella, 2015) (see Table 1). Policy entrepreneurs need considerable resources, legitimacy, and a feasible alternative to the institutional status quo. Organisational or institutional settings play a role in determining who is 'waiting in the wings' and has the resources and positions of authority to become a policy entrepreneur (see Table 1 -institutional factors). For example, policy experts in a policy subsystem with a high degree of autonomy are more often in a position to introduce radical new ideas into the policy process (Baker, 2015). Helgadóttir (2016) argued that influential Italian economic policy ideas could be advanced and diffused throughout the EU as a result of putting people in the right positions.
Once policy entrepreneurs are in place, they can become the 'central defecting actors', creating windows of opportunity for policy change. To do so, they need to persuade other actors that their old ideas are wrong and instigate a 'joint belief shift' (Culpepper, 2005, p. 176). Whether the entrepreneurs are successful in bringing about such shifts depends on the different strategies they deploy (see Table 1). Strategies range from effective problem framing to team building, assembling new evidence, and collaborating with advocacy coalitions or policy networks to use and expand network connectedness (Petridou & Mintrom, 2020). If entrepreneurs have rhetorical skills, they may succeed in making certain ideas more or less attractive (Béland & Cox, 2016). For instance, the French Mont Pélerin society meetings were used by policy entrepreneurs to transmit the idea of neoliberalism into the public realm (Schmidt & Thatcher, 2014). Successful 'idea carriers' are often political and administrative elites that have authority on the basis of their expertise, experience, position, and skills.
The causal mechanism to explain how policy entrepreneurs can evoke ideational change is persuasion. Mobilising language leads people to adopt or adapt to the new or alternative ideas that policy entrepreneurs present to them (Béland, 2016;Kuisma, 2013). Persuasion may result in a consensus (convergence of ideas) among different actors participating in policy debates (Baker, 2015;Bell & Hindmoor, 2015). This mechanism arguably plays a role in highly politicised contexts (with much ideological disparity) wherein multiple, equally powerful actors try to move each other by argument (Flibbert, 2006).
The assumption that ideas originate in the interactions between policy actors has two analytical consequences. First, if change results from interaction, it implies a more active and influential role of policy actors in the policy process (Trein & Vagionaki, 2020). Second, it implies that they are not necessarily coherent entities but composed of different elements that policy actors piece together. Through bricolage, policy actors piece together different elements of meaning to construct a 'web of meaning' (Carstensen, 2011b). Seen from this perspective, ideas are flexible, open, fluid, and always subject to contestation (Crawford, 2016).
In conclusion, the literature on the transfer and diffusion of ideas stresses the importance of studying who becomes a policy entrepreneur and the discursive strategies they use to persuade others of (alternative) ideas towards public policy. Institutional and structural factors predetermine which actors could potentially play a role in changing ideas in policy processes, whereas the successful uptake of new ideas by others is the result of the discursive strategies these actors deploy.

Ideas come from processes of institutionalisation
Taking a structural perspective of ideas, ideas seemingly 'float freely' and cannot be attributed to specific individuals per se. Policy actors are mere followers of social structures and choices they make follow from given conditions in their environment (Parsons, 2007). Alterations in these stable, institutional ideas are often observed through critical analysis, highlighting the retrospective distinction between policy paradigms such as Keynesianism and Monetarism (Hall, 1993). Paradigmatic ideas, such as neoliberalism, may be quite resilient and coherent (cf. Schmidt & Thatcher, 2014). The timeframe to study such processes of change is lengthier: explaining the emergence of such institutional frameworks requires careful process-tracing, often analysing the development of a policy domain over a long time-span, carefully reconstructing the history of such frameworks (Jacobs, 2015;Mehta, 2011).
It is appealing to believe that factors sweeping ideas aside are contextual, for example through emerging external challenges like a crisis (Cairney, 2019) (see Table 1). Schmidt (2016) showed how the contextual pressure of the Eurozone crisis made German policymakers question their taken-for-granted ordoliberal ideas for the first time since these ideas had been first formed in the 1950s. Baker and Underhill (2015) argued that the global financial crisis opened a window of opportunity for actors to push new economic ideas and instigate policy reforms, countering prevailing post-Keynesian ideas in place since the 1990s. As such, one could argue that exogenous shocks can serve as an explanatory concept for deep ideational change. In effect, it disregards both the cognitive and discursive ideational processes that occur in between (Kamkhaji & Radaelli, 2019).
A more fine-grained observation renders the process of change to be 'layered'. Here, external challenges do not provide a full explanation of why some ideas come into being instead of others. Acknowledging that new ideas first develop in the cognition of individual actors, and are subsequently communicated through discourse, allows for an analysis of how more actors come to a 'buy-in' of new ideas that may replace a prevailing interpretive paradigm. This provides a more gradual explanation of ideational change. Through individual learning and collective interaction, the majority of policy actors may ultimately be socialised into a new interpretive paradigmatic framework.
Socialisation is the mechanism through which actors or groups become a part of institutional practices, rules, or norms. It occurs as a result of observing or simulating existing practices (e.g. Bell, 2012) and creates stable policy communities that could potentially become infected with new ideas. Saurugger (2013, p. 894) stated that, "Socialization occurs when norms, worldviews, collective understandings are internalized, and subsequently are codified by a group of actors". Adapting to ideas through socialisation is characterised in the literature as 'mimesis', 'imitation', or 'internalisation' (Becker & Hendriks, 2008).
In sum, once ideas are institutionalised, they are believed to sit at the 'deepest level of generality' (Schmidt, 2016, p. 320). They arrive there as the result of a process preceded by individual learning and collective interaction, wherein the historical and political context of a policy domain and its actors influence how ideas institutionalise over time. The extent to which the historical and political context determine institutionalisation through time vis-à-vis the involvement of active, strategic entrepreneurs influencing this process is a key question in the literature. Pinning down the exact moments when transitioning actors become aware of their background ideas, start learning, and subsequently start using them strategically in political discourse is a less empirically assessed topic (Molthof, 2016).
To conclude, ideas may 'come from anywhere', but I have provided clues about where to look for them and what range of factors and mechanisms influence ideational change. For reference, Table 1 shows a more complete overview of specific hypotheses that ideational scholars in the reviewed literature have both formulated and examined to understand the different types of ideas and ideational change. In addition to the general discussion above, this overview could help scholars formulate expectations and hypotheses for future explorations of ideational change, aiding them in their examination of 'how' specific ideas matter in policymaking.

Positive
Organisational strength led to coherent environmental belief system in NL and GE (Rohrschneider, 1993).  (Morcol, 2001). Nelson (2014: 312): 'graduate training in economics "is a transformative experience for doctoral students that creates strong professional identities."' 2. Helgadottir (2016): power of academic 'Bocconi' network on promoting idea of expansionary austerity in economic policy-making sphere.

Expertise
Ideational stringency Proposition Larson (1994): suggests experts have more constrained belief systems. Butler et al. (1995): suggest relation expertise and lower attitude change due to ability to counterargue and engage in critical thinking.
Highly educated more belief in climate change (Fielding et al., 2012).

Strategies and skills
Strong rhetorical skills (crafting and promoting ideas)

Ideational reconfiguration
Positive Abilities of policy entrepreneurs to craft, promote and work with the idea (rhetorical skill) make certain ideas more or less attractive (coalition magnets) (Béland & Cox, 2016

Methods to study ideas
Adding to the theoretical challenges concerning ideas and ideational change, ideational scholarship also harbours methodological challenges. One key challenge is how ideas can be measured. For example, much ideational scholarship builds on hypothesised assumptions of how cognitive ideas influence public policy but does not necessarily provide methods to do so (Kamkhaji & Radaelli, 2019). Hence, in some aspects, the field of ideational studies remains methodologically underdeveloped.
So how can ideas be empirically established and what data sources could be used to study them? In the corpus, over 40 percent of the articles do not explicate the methodology they used to measure ideas. Those that do showcase methodological pluralism. Table 2 provides a brief overview of different methodologies used in the corpus depending on different conceptualisations of ideas, substituted by examples for each. Of the articles addressing methodology, most use a comparative or single case study research design. The majority relies on document analyses to derive ideas from, either speeches or policy documents. These data are often analysed through qualitative content analysis. Information about the nature of ideas is relatively scarce. Analysing the web of causal assertions an individual makes on a selected topic.
Van Esch (2014): analyses the ideas of political leaders and central bankers in the Eurozone crisis to understand how the crisis affected the economic and policy ideas of these leaders.

Experiments
Measurement of beliefs under controlled conditions to discover hypothesised effects. Renshon et al (2015): survey experiment where respondents watched different videos (relaxed, neutral and anxious), and skin conductance tests measured emotional responses to questions about political beliefs.

Computational modelling
Representations of cognitive structures and the processes that link these structures to decision Taber (1992): Use POLI, a model to understand shared belief systems to analyse U.S. Foreign Policy at the level between individuals and organisations.

Surveys
Analysis of self-reported beliefs or ideas about selected topics using questionnaires Brewer and Steenbergen (2002): survey of American citizens' show reliance on beliefs about human nature to reason about international affairs. These beliefs help citizens to make sense of topics that are relatively unknown to them.

Interviews
Structured, semistructured, or inductive conversations, targeted to uncover self-reported assumptions, beliefs and ideas of individuals. Radaelli and O'Connor (2009) use INTUNE project data to uncover shared governance belief systems in different types of EU committees.
Text analysis (qualitative content analysis) Interpretation of the ways in which individuals, organisations or members groups make sense of who they are Foyle (1997)  Case studies Investigating contemporary phenomena indepth within real-life context. Becker and Hendriks (2008): influence of the Central Planning Bureau in sustaining wage restraint paradigm.

Content analysis
Interpretation of the ways in which individuals, organisations or members groups make sense of who they are Bratberg's (2013): speech analysis of Blair and Chirac in Iraq war to observe how their actions were guided by pre-existing 'guiding principles'

Ideas as strategic tools
Discourse analysis Analysis of the substantive content of ideas, and the interactive processes of distributing these ideas through discourse.

Schmidt (2016): French Mont
Pélerin Society capacity to construct and disseminate 'coordinative discourse' of neoliberal thinking through their annual meetings.

Process tracing
Tracing causal chains between two variables using rich data (interviews, documents) Helgadottir (2016): process-tracing of network of Italian economists that erected and maintained the idea of 'expansionary austerity'.
Biographical analysis Analysis of biographical data (educational background, country, age) to measure the proportion/presence of certain ideas in specified groups.
Nelson (2014): database of biographical data (educational background, work experience) on 2000 policy officials working with the IMF to show proportion of neoliberal ideas in the organisation.
Narrative analysis / Text analysis Methods aimed to identify the '...repertoires, or shared patterns of interpretation, active processes of reasoning that draw attention to the form as well as content of argumentation and can be linked to broader social and political structures and processes.' (Finlayson 2004: 539).
Alceste software to analyses thematic classes in rhetoric of speeches by Bush and Kerry (Schonhardt-Bailey, 2005).

Source: The Author
Cognitive-ideational studies employ methods such as operational code analyses, experiments, surveys, or cognitive mapping, aiming to assess the cognitive processes that lay bare or influence an actor's or actors' beliefs about certain subjects (Table 2). Research questions focus on the causal links between contextual factors, personal disposition, and cognitive beliefs of policy actors and how these may affect policymaking.
Scholars approaching ideas as strategic tools employ methods that uncover either 1) the structural composition of networks, 2) the presence of a certain discourse in an organisation or in rhetoric, or 3) the propagators (i.e. policy entrepreneurs) and followers of a certain idea. Methodologies include process tracing, discourse analysis, or biographical analysis. These methods aid scholars to examine compositions of committees, research units, policy/epistemic communities, advocacy coalitions, and think tanks, as well as their analyses of how dominant narratives and discourses can be traced in these groups (see Campbell & Pedersen, 2015;Radaelli & O'Connor, 2009).
Institutional analysts of ideas use methods to explain the longitudinal processes of how ideas gain or lose acceptance in policy communities or society. These can be categorised under the heading of 'critical policy analysis', for example story-telling or historical analyses. These studies often contributed to theory-building in ideational scholarship.
While the aforementioned methodologies can be used to study ideas, not all are designed to do so. This obfuscates empirical research in dissecting how ideas can be separated from other influential factors in the policy process. Of the reviewed methods, only cognitive mapping and operational code analysis were originally designed to study ideas. Hence, ideational scholarship would benefit from methodological innovation and rigorous application of such and other methodologies. Process-tracing, for example, addresses some of the challenges regarding the study of ideational causation (Jacobs, 2015;Molthof, 2016). It is a useful method to study the development of a policy domain over a long timespan. Discourse Network Analysis (DNA) provides insights into both the composition of actors in a discursive network, as well as the content of the ideas that they exchange in it (Leifeld, 2016), combining actor-centred and content-centred approaches. Comparative Cognitive Mapping (CCM) is useful to systematically measure ideas of actors or groups, allowing scholars to trace different forms of ideas, establish their strength, and compare ideas over time (Van Esch & Snellens, 2019). Furthermore, Qualitative Comparative Analysis (QCA) is useful to assess configurational hypotheses of different factors and mechanisms of ideational change. Nonetheless, these methods have their limitations as well. They may be labour-intensive, or face challenges regarding validity and reliability of the analysis. As general advice, scholars need to consider the relationship between their level of analysis and the most appropriate research method carefully.

Conclusion
This paper demonstrated that ideational scholarship in public policy evolved along several distinct lines of inquiry. These lines range from micro-level cognitive ideational analysis, mesolevel discursive ideational analysis and macro-level institutional ideational analysis. It provides students and scholars of public policy with an approach to organise the existing literature and a language to explicate choices when choosing to work with ideational variables in public policy research. In effect, it enables us to provide answers to the question of how ideas matter. I also identified several blind spots in the existing scholarship. Here, I articulate what new directions are needed to address these.
First, although many scholars acknowledge that cognition plays a role in ideational scholarship, the role of cognition of individual actors in public policy is currently understudied. To better understand the percolation of ideas in policy, these micro-level cognitive dynamics should be brought back into the focus where Simon once put them (Kamkhaji & Radaelli, 2019).
Second, scholars need to theorise and examine the relationships between the three functions ideas have: as heuristics, strategies or frameworks. For example, to what extent are policy and political entrepreneurs successful in shaping policy agendas when they advance wholly novel, frame-breaking ideas, as opposed to pre-existing but repackaged ideas that fit more within existing institutional discourses (Béland & Waddan, 2015)? Examining such questions requires rigorous concept formation and testing. Early efforts to do so include Van Esch et al. (2018) and Swinkels (2020), who studied how Keynesian and ordoliberal paradigmatic ideas are reflected as beliefs in political leaders' speeches, using CCM to determine if such individually held paradigmatic beliefs changed over time. Likewise, Molthof (2016) applied process tracing, showing how political "actors strategically tailor their discursive use of ideas according to the political context and the policy position to be justified" (Molthof, 2016, p. 204). This enabled him to pin down the exact moment when actors transition from being 'mere slaves' of institutionalised ideas to using them strategically in political discourse and, ultimately, to reshape policies.
Third, this paper suggests that ideational change could be considered a 'layered' process; therefore ideational scholarship needs studies that specify the mechanisms and sequences involved. An example of such an approach is Rinscheid et al.'s (2020) model that shows how and when different factors and mechanisms align to trigger a 'joint belief shift' among (a coalition of) policy actors during an exogenous shock (scoping condition), which then enables institutional change. The broad overview of triggers and constraints presented in this review (see Table 1) encourages public policy researchers to test a configuration of variables to explain the development of ideas as well as the relationship between ideas and policy changes. Swinkels' (2020) qualitative comparative analysis of the belief changes of key leaders in the Eurozone crisis is an example of a study that assesses the configurational hypotheses of individual belief changes during a crisis.
Fourth, scholars of ideas can benefit from expanding their 'methodological toolbox' to measure ideas. Policy science scholars have recently started using experiments, seen as a useful addition to the methodological toolbox for scholars doing cognitive-ideational research in a controlled setting (Grimmelikhuijsen et al., 2017). Diary studies and 'political ethnography' (Kamkhaji & Radaelli, 2019) are potentially useful methods to study the dynamics of beliefs and arguments in policy arenas (Bevir & Rhodes, 2015). Likewise, by using social network analysis methodology, Flickenschild and Afonso (2019)  Fifth, comparative studies of ideational change across different political systems, layers of government, and policy domains expose the limitations of the grand theories of radical policy change and the importance of middle-range, contingent theorising about ideational change (Baker, 2013). Future studies should therefore examine the multiple ways in which ideational change occurs in different policy settings (Radaelli & O'Connor, 2009;Saurugger, 2013). Different settings may determine when a particular mechanism for change is more or less likely to be activated. For example, under similar conditions of a crisis such as COVID-19, ideational change at the level of national governments may be triggered by different mechanisms than ideational change at the EU level. As states have modes of centralisation, ideational change may be triggered by exogenously motivated mechanisms, such as powering or coercion, as central actors can weigh in their interests. Alternatively, in the EU, where decision-making authority and resources in times of crisis are dispersed, such centralised decision-making structures often take time to develop. Hence, persuasion or socialisation may be more prevalent mechanisms in these settings.
Finally, this paper attempted to systematically track how different kinds of ideas, mechanisms, and factors involved in ideational change matter in public policy. One limitation of this review may be that it focused on a narrow search scope related to the concept of ideas (cf. beliefs, paradigms, worldview, ideas), whereas the literature on 'ideational elements' (cf. Berman, 2013) includes other concepts such as memes, narratives and frames. Future literature reviews may benefit from including such a wide range of search terms to allow for a broader perspective on the concept of ideas. administration) and shows how these are interconnected. The Journal of European Public Policy is revealed to be the key node in the network, being most densely connected to other journals. Figure  1 also shows there is virtually no interconnectedness between articles and journals that take a more agentic approach (e.g., in political psychology) towards ideas and those that take a more structural perspectives (e.g., in political economy) on ideas in politics and policy.

Figure 1. Bibliographic interconnectedness of journals in this review
Of the remaining 599 articles, 52 abstracts and 15 full-texts were not available for screening. Eventually, 532 abstracts were screened and coded on inclusion criteria as 67 did not include an abstract or abstract was not accessible. The inclusion criteria were 1) a reference in the abstract to the concept of ideas (if possible explicit reference to ideas as a dependent variable), 2) a reference in the abstract to researching ideas (methodology), 3) a reference in the abstract to ideational change, and 4) a reference in the abstract to conditions that may cause ideational change.
A total set of 78 articles published in 27 different journals matched one or more of the inclusion criteria. These 78 articles were subsequently coded for full-text eligibility using coding criteria that focused on the key questions of the study: conceptualization of ideas, carriers and methods to research ideas, theories of ideational change and drivers of ideational change (see Table 1 for a full description of the codes used in this review). These four questions were derived from articles and book chapters discussing the challenges for ideational scholarship (cf. Daigneault, 2014bDaigneault, , 2014aMehta, 2011).

Figure 2. Growth of articles per year
Ultimately, 71 articles were selected for full-text analysis. The remaining 7 were excluded because they ultimately did not focus around the key concept of the review (n=3), were not presented in article format (n=1) or full-texts were not readable (n=4). After several conference workshops, relevant books, dissertations, and 'grey' literature were added. The chief objective was to capture the key debates and differences across different (sub)disciplines, not to compile an exhaustive corpus. Figure 1 depicts the 'ideational turn' in the growing number of articles published over time.
The remaining 71 articles were also coded for background factors, such as focus area, policy area, and article type. Furthermore, the institutional affiliation of the first author and the country of the institutional affiliation were coded. After the coding process, the first results were published at two different conferences. These sessions led to reviewing ten other books and publications that were used as complementary readings. The full selection process of articles is shown in figure 1.
Synthesizing the results of the articles in a systematic literature review can be done by either a quantitative meta-analysis or qualitative narrative analysis. Given the paper's aim of presenting a substantive review of current conceptualizations and theories of ideas, hypotheses and findings, I chose the narrative approach (Petticrew & Roberts, 2006). 9 A quantitative scoping of the background factors that were coded for all articles is presented below.
A systematic literature review on a fuzzy concept in the field of social sciences is a complex endeavour (Daigneault et al. 2012). Therefore, I want to stress that this study knows limitations. First, as the searches only include peer-reviewed published articles in journal, a pitfall is that newly published articles, conference proceedings, dissertations, books, or 'grey literature', were not included in the searches. Second, as this is a single-authored study, the articles were coded by one coder only which has the potential to lead to flawed coding. The researcher conducting the review did not have sufficient resources to hire research assistants to double the work. Third, defining the selection criteria (inclusion and exclusion of articles) for a review is a complex matter, and decisions in this stage can have consequences for the corpus as scholars can, for example, be easily overwhelmed by the sheer number of studies to screen. After deliberation with an advisory team, the search terms were defined in narrow rather than broad terms (focusing on ideas, beliefs, paradigms, instead of the breadth of ideational elements such as norms, frames, narratives, images). This may have excluded relevant articles, but as searching with these 'narrow' terms already led to >500 articles to screen, I believe that to keep such a review manageable for a sole author, these search terms have been sufficient to capture the state of the art in the current debate, spanning different subdisciplines. Future reviews could definitely benefit from expanding the scope of search terms. Taking these limitations into account, the results from the SLR provided a core corpus that served as a point of departure for the paper, yet additional references were included after conferences, over time, through feedback, and through reviews.  Notes or excerpts from original articles about the concept of change. Aim here was to see how scholars write about change, possibly distinguish between the speed of change (fast/slow), the process (revolutionary or evolutionary), and the direction of change (reinforcement, conversion, stasis, etc.) (n=3), Queens University Belfast (n=3), University of Queensland (n=3), and the University of Saskatchewan (n=4) appear three or more times in the set. These first findings do suggest that ideational scholarship may be dominated by scholars based in English speaking countries and institutions.

Focus area and topics
Of the articles published in English-speaking countries (n=55), the majority focuses on the national level (36.4%). Specifically, the domains under investigation in these articles range from migration policy, economic policy, US leadership, a country's foreign policy, or climate change. Fewer articles published in English-speaking countries focus on the European level (14.5%). Of these, the articles focus on Eurozone governance, financial regulation, economic integration, or EMU. Another 14.5% of the studies is coded having comparative focus, focusing on the topics of environmental policy, political economy, welfare state reform, tax policy or war studies/foreign policy. 10.9% of the articles by scholars based in English-speaking institutions focus on the international level, for example on the International Monetary Fund (IMF), foreign policy, or international political economy. The remaining 23.6% of these articles are either theoretical or methodological in nature and as such do not focus on one specific focus area or policy domain.
Scholars based in Non-English, European research institutions, do not focus more on the EU level. Two studies focus on EU welfare state reform and European integration respectively. Another five articles zoom in on the comparative national, or international level and topics under study here are foreign policy (n=2), welfare state reform (n=1), wage restraint (n=1) and international monetary finance (n=1). Again, a fair share of the articles here (n=5) is more theoretical in nature.
For all studies combined, we can conclude that most studies in our full set (n=71) either focus on the national level (n=27; 38.0%) or do not have a specific focus (n=18; 25.4%). Relatively few are focused on comparison, international relations or EU studies (n=9, 12.7%;n=7, 9.9%;n=10, 14.0%). The locus areas (subdisciplines, policy domains) are quite diverse and specific. Figure 2 presents an overview of all studies per locus area. Most articles in our set focus on the realm of economic of financial policy, or foreign policy (31%), another 27% is do not focus on a specific topic as they are more theoretical in nature.

Journals
In the main text, we have showed how articles were connected and distributed over our search results (n=532). More specifically, for the included articles in our set, the journal distribution is presented in figure 3. We can observe that the Journal of European Public Policy (JEPP), the British Journal of Politics and International Relations (BJPIR) and the Journal of Political Psychology (PP) are most frequent in our set. Figure 2. Number of articles per journal.

Article type and methodology
We have elaborated extensively on the different, specific types of methodologies and research practices in the main text of this article. However, all articles were also coded with regard to article type. First, articles coded as empirical contain studies that focus more on a rich description or explanation of empirical data, and make more limited use of theoretical concepts. Of the articles included in our set, 40.8% were coded as empirical. Second, 14.1% of the articles were coded as purely theoretical. This means that these have an explicit theoretical focus. Furthermore, 36.6% were coded as theoretical, supported by empirical evidence. These studies often have an explicit theoretical grounding, and focus on the presentation of a theoretical argument. To make the case for the theoretical argument, these studies use empirical data to illustrate the relevance of the theories they are describing. For these studies, it is often difficult to establish the methodological quality of the studies and processes of data collection. In sum, over 50% of the articles were more theoretical than empirical in nature. A further 4.2% of the article were about methodological innovations or explanations, and another 4.2% were introductions or commentaries for special issues/sections in journals.