Introduction
First published in 1984, Kingdon’s Agendas, Alternatives, and Public Policies (Agendas hereinafter) is the founding book of the so-called Multiple Streams Approach (MSA). In short, Kingdon postulates the existence of three streams. The Problem Stream explores the factors that make problems gain or lose agenda prominence, such as exogenous events or changes in indicator trends. The Policy Stream explores the development of what Kingdon (2003, pp. 5, 20, respectively) calls policy ‘proposals’ or more simply ‘solutions’, and the factors that make them gain or drop agenda prominence, such as budget constraints and value acceptability. Lastly, the Political Stream explores the political factors – ‘political events’ or ‘political forces’ (2003, pp. 18, 20, respectively) – that make problems and solutions gain or lose agenda prominence, such as elections or swings in public opinion. According to Kingdon, these streams flow independently from each other until their coupling, in combination with the opening of policy windows, leads to agenda change.
Despite its seeming clarity, there is however confusion in the literature about two key aspects of the framework. Firstly, the picture that emerges when trying to ascertain whether Kingdon elaborated a framework for agenda-setting or policy change is rather muddled. According to Zohlnhöfer et al. (2016, p. 249), Kingdon ‘only dealt with agenda setting, not with decision making’. Similarly, Sætren (2016, p. 27, emphasis in original) asserted that Kingdon applied his framework ‘only to the agenda and formulation stages of the policy process’, while Greer (2016, p. 423) contended that he ‘clearly demarcates the agenda-setting process from the decision process’. For others, instead, Kingdon developed a ‘model of policy change’ (Baumgartner, 2012, p. 255) where after the streams are merged ‘the result is major policy change’ (Sabatier, 2007, p. 9). For others still, Agendas is ‘[e]ssentially a book on agenda-setting’ and yet in it Kingdon explains both ‘agenda and policy change’ (Sanjurjo, 2020, pp. 102, 108, respectively), or a book whose ‘original purpose’ was ‘to understand agenda setting’ in which however coupling produces ‘the greatest opportunity for policy change’ (Reardon, 2018, pp. 457, 459, respectively). Yet confusion is not a distinctive trait of the scholarly literature only. Indeed, while some on GoodReads.com (n.d., n.p.) define Agendas as their ‘favorite framework for agenda setting’, most users define the book as a ‘blueprint for how decisions are made’, an illustration of ‘how a policy agenda can enter the enactment phase’, an explanation of the processes ‘that lead to a policy becoming law’.
Secondly, despite policy windows being the ‘signature feature’ of the MSA (Dolan & Blum, 2023, p. 87) and ‘the most popular of all the MSA concepts’ (Jones et al., 2016, p. 24), and despite coupling being ‘the critical element’ (Ackrill et al., 2013, p. 880) that is ‘at the heart’ of the MSA (Herweg et al., 2015, p. 443), ambiguous and conflicting definitions of both terms are used in the literature, sometimes by the same authors. Bache & Reardon (2016, p. 34), for instance, argued that ‘[w]hen policy windows are open, policy entrepreneurs can play a key role in coupling the streams’, suggesting policy windows open before coupling. Saint-Germain & Calamia (1996, p. 65), instead, understood policy windows as ‘the convergence of the three streams’, whereas Lober (1997, p. 13) believed their convergence ‘resulted in a […] window opening’ and so did Howlett et al. (2014, p. 421). Béland & Howlett (2016, p. 222) also argued that the interaction of the three streams ‘produce “windows of opportunity”’, only to conclude that only after ‘a policy window opens […] do the streams cross’. This is not unusual, as policy windows and coupling are often used to signify both a concept and its opposite at once. Weible & Schlager (2016, p. 8, emphasis added) thus argued that ‘[p]olicy windows […] [allow] the streams to be coupled’, only to then state that ‘windows of opportunity emerge through the coupling of streams’ (see also Mucciaroni, 1992, p. 460). This problem continues to permeate the literature, demonstrating its ongoing relevance. To give but one example, among the articles published in the latest available month (January 2025) featured on Google Scholar that cited Agendas and defined policy windows, multiple interpretations can be found at once (e.g., Haby et al., 2025, pp. 6, 7; Oksenberg & Zehavi, 2025, pp. 1, 8).
Inconsistencies aside, while some therefore think policy windows denote the merging of the three streams, for others they open after their convergence instead, and for others still before that. This last position is the one also championed by recent scholarship, which has produced guidelines (DeLeo et al., 2024; Zohlnhöfer et al., 2022; see also Zahariadis et al., 2023) according to which coupling occurs after the opening of policy windows. And yet, among scholars who cited such guidelines ambiguity and inconsistency persist. For instance, for Leon-Espinoza (2022, p. 1384) ‘changes in the problem stream or political stream can couple the streams together, opening “windows of opportunity”’ and yet ‘[w]hen a “window of opportunity” opens, policy entrepreneurs can couple the streams together’. This is far from an isolated case (see e.g., Dzhurova, 2023, pp. 15, 31; Ibrahim et al., 2024, pp. 1, 4, 16; Powell, 2024, pp. 584, 593). In fact, such ambiguity can be found in the above textbooks, too. Wenzelburger & Thurm (2023, p. 52; see also p. 50), for instance, argue that what ‘opens the window’ is either ‘a perceived problem’ or ‘political dynamics’ (thus the factors that lead to the development of the Problem and Political streams) which subsequently leads to coupling, only to show in a stylized version of the framework (2023, p. 56) that coupling takes place before the opening of windows which are also separate events from the streams themselves. This underscores two critical issues. First, scholars are citing yet not following the guidelines that the MSA literature has produced. This may of course be due to their lack of attention, but it highlights nonetheless the existence of alternative definitions which demand further investigation into their potential applicability. After all, MSA handbooks do not discuss any alternative definitions, de facto imposing the view that policy windows can only open before coupling. Second, scholars are not following such guidelines in the belief that their definitions align with Kingdon’s, even when they contradict themselves. For instance, on the very same page Powell (2024, p. 584, emphasis added) notes that ‘Kingdon (2011) argues that an issue reaches the agenda when the policy window opens […] to allow the coupling of three independent streams’ and yet ‘Kingdon (2011) suggests that an issue reaches the agenda when the three independent streams couple to open the policy window’ (see also Ruvalcaba-Gomez et al., 2023, pp. 242, 250). Although this may again be simply the result of scholars’ lack of attention, these quotes suggest that Kingdon himself has offered alternative interpretations of policy windows and coupling that differ from the one being advocated in MSA handbooks, reinforcing the need to investigate whether such alternative definitions indeed exist, and if so if they are applicable.
Through an in-depth analysis of Kingdon’s corpus, this article aims to unravel what causes scholars to have opposite understandings of Kingdon’s research aims and to interpret policy windows and coupling in conflicting ways and yet see such interpretations in line with Kingdon’s. Such questions are linked to a more fundamental question: are scholars misinterpreting Kingdon’s work, or is it Kingdon’s work that lends itself to such misinterpretation? Before beginning, however, it is worth clarifying why this article puts Kingdon’s production centre stage. Albeit reissued with different epilogues, Agendas has remained unaltered since 1984. Kingdon (2011, p. xix; 2003, p. xix, respectively) has indeed kept the main body of his book ‘exactly the same’ since ‘the understandings of the processes presented in these pages seem roughly right’ and his explanations remain ‘accurate’. At first glance, Agendas’ seminal contributions may justify its preservation in its original form. Furthermore, one may regard Kingdon’s decision as irrelevant – after all, the MSA literature has evolved significantly since 1984. However, scholars still rely on Agendas as their main reference text. As Cairney (2018, p. 200) noted, ‘[m]ost readers and users of MSA draw on Kingdon’s (1984) classic metaphor without […] engaging with over 30 years of subsequent research’. In fact, ‘theoretically informed studies, demonstrating a greater appreciation of the literature [… represent] a small proportion of MSA-related activity’ (Cairney & Jones, 2015, p. 50). This becomes remarkably clear when comparing the number of works citing Agendas – an indicator frequently used as evidence of its success (e.g., Baumgartner, 2016; Cairney, 2018; DeLeo et al., 2024; Jones et al., 2016) – vs. Zahariadis (1999) – widely regarded as the most illustrative text on the MSA to have not been authored by Kingdon (e.g., Jones et al., 2016; see also DeLeo et al., 2024). According to Google Scholar, in 2023 alone all editions of Agendas were mentioned in 1,726 texts, whereas all five editions of Zahariadis (1999)1 in 248, with more texts quoting the former between 2022 and 2023 (3,543) than the latter since 1999 (2,432). Furthermore, it is extremely rare to find scholars who engage with more than one text written by Kingdon on the MSA. According again to Google Scholar, as of January 2025 Agendas has been cited in 37,442 texts whereas his first MSA-related publication after that in just 228 (Fig. 1). Not only are scholars thus favouring Agendas over other scholars’ works, but also over Kingdon’s own works. Surprisingly, this is also the case among MSA scholars: all texts but one cited in this article only include Agendas in their bibliography.
This signals limited critical engagement and, indeed, many of those mentioning Agendas do so cursorily (Jones et al., 2016; see also Cairney, 2018; Cairney & Jones, 2015), repeating Kingdon’s language without much questioning (Dolan, 2021; see also Dolan & Blum, 2023). Not only is there thus ‘little effort on the part of scholars to engage with related theoretical explanations [… and] with the full richness of Kingdon’s (1984) exposition’ (Weible & Schlager, 2016, p. 6), but also with Kingdon’s scholarly production on the MSA at large. Yet, Kingdon himself has demonstrated limited critical engagement by repeatedly refusing to ameliorate his work. Whilst some see this as a strength (e.g., Greer, 2016), the issues highlighted so far invite more caution as they suggest their root may lie in Agendas itself. These elements strongly justify the need for a critical analysis that looks back at Agendas and that also scrutinises Kingdon’s wider corpus, so far overlooked, to see how Kingdon addresses some of the key aspects of the MSA and whether the decision not to revise Agendas can be justified, or whether it might be Agendas itself the source of confusion.
To be sure, Kingdon’s lack of clarity and use of ‘highly metaphorical’ language (Sanjurjo, 2020, p. 113) have long been highlighted (e.g., DeLeo et al., 2024; Herweg et al., 2015). Yet, the picture that emerges from the literature is again confusing, with Kingdon being both criticised and praised for his writing, sometimes by the same authors. Zohlnhöfer et al. (2022, p. 28), for instance, noted that ‘many key MSF concepts are not clearly defined in Kingdon’s (1984) landmark book’ yet for Zahariadis (2016, p. 12) the use in Agendas of concepts such as policy windows helped Kingdon make his approach ‘comprehensible’. In fact, Agendas has been praised for ‘the quality of the writing’ (Baumgartner, 2016, p. 60) and argued to be ‘replete with vivid metaphors’ (Howlett et al., 2016, p. 76). This is not necessarily a contradiction and may simply highlight a difference in opinions. However, the criticism raised about Kingdon’s lack of clarity warrants more scrutiny since criticising Kingdon for not having adequately explained his terminology would require examining whether and how he has defined his key concepts in other texts first, which instead remain overlooked. These elements therefore further justify the present work which provides a critical analysis of Kingdon’s corpus and his exposition of some of the framework’s core concepts within it to better understand whether this exposition does raise any issues or not, and if so where and why.
Finally, the responsibility for the issues reviewed so far is often attributed to scholars and their use of the framework, yet as we have seen this may well not be the case. Herweg et al. (2023, p. 41), for instance, assert that it is ‘subsequent literature [… that] has applied the MSF to […] further stages of the policy cycle’ beyond agenda-setting, with many other scholars making similar implicit claims (e.g., DeLeo et al., 2024; Zahariadis et al., 2023). However, the existence of alternative interpretations of Kingdon’s research aims, compounded with the lack of research on this issue, invite more caution in holding scholars responsible and require further investigation. Similarly, Jones et al. (2016, p. 30) note that it is ‘MSA analysts’ who, despite using ‘the same vocabulary […] do not all share the same definition of concepts’. However, the underlying assumption that Kingdon uses the same definitions across his texts also lacks supporting evidence and demands further investigation. All in all, a critical scrutiny of Kingdon’s corpus is crucial to bring clarity over the meanings and aims of key aspects of the MSA. Besides contributing to strengthening the framework’s theoretical foundations, this analysis will help ascertain whether the root cause of confusion lies within the literature or Kingdon’s works. Only then will we be able to ascribe responsibilities and identify the best way forward. In light of Agendas’ recent 40th anniversary, its widespread use, and Kingdon’s repeated refusals to ameliorate it, these questions are not only timely but also pivotal for any refinement of the framework for, to use Kingdon’s language, only when we know where a problem lies that solutions can be proposed.
At the very beginning of Agendas, Kingdon (2003, pp. 2, 3) states that policymaking includes ‘at least (1) the setting of the agenda, (2) the specification of alternatives […], (3) an authoritative choice among those specified alternatives, as in a legislative vote or a presidential decision, and (4) the implementation of the decision’. Right after that, he (2003, pp. 3, 2, 8, respectively) clarifies thus that ‘[t]his study concentrates on the first two processes’ – what he also calls ‘predecision public policy processes’ – intentionally excluding what ‘goes beyond the agenda-setting phase’. Or, as he (2003, pp. 196, 2, respectively) repeatedly claimed, ‘[t]he book is not about how […] authoritative figures make their final decisions’ or ‘how issues are authoritatively decided’, but rather ‘how they came to be issues in the first place’. Indeed, for Kingdon (2003, p. 4; see also p. 230) agenda-setting and decision-making are ‘affected by somewhat different processes’. Or, as he stressed in other texts, too, ‘agenda-setting and alternative specification are not the same as final choices’ (1993a, p. 40; see also 1989, p. 282; 2002, p. 98). Rather than focusing on decision-making and policy change, Kingdon’s (2003, p. 235) interview guide thus contains questions that investigate what interviewees are ‘paying attention to’, or what issues are ‘on the front bumer [sic]’ or ‘being serious [sic] considered’.2
However, one can already tell from its very first pages that Agendas is hardly going to be about agenda change when Kingdon (2003, pp. 14, 8, respectively) claims that ‘this is a book on agenda setting rather than enactment’ but then asks himself ‘[w]hat conditions would increase the chances of enactment?’. Or when he (2003, p. 15, emphasis in original) describes his chosen examples as cases ‘of policy changes’ stating to be interested in understanding ‘why [such] changes occur’. In fact, Kingdon (2003, p. 5) himself said that he and his assistants picked case studies ‘covering many policy changes’ – and not agenda changes – which is why examples of the former abound: he explains how academia and public opinion contributed to ‘policy change’ in transportation deregulation (2003, p. 55; see also pp. 65, 125, 128, 156); how regarding Medicare and Medicaid ‘[f]raud and abuse legislation was passed’ (2003, p. 107); how ‘substantial changes in […] regulatory procedures’ of clinical laboratories were made (2003, p. 109); what contributed to ‘the passage of a waterway user charge’ (2003, p. 111; see also pp. 190, 191); how the proposal for an earmarked trust fund ‘found its way into enactment’ (2003, p. 126; see also p. 179); the ‘passage of Medicare’ (2003, p. 128); how ideology sometimes ‘affects public policy outcomes’ (2003, p. 135); how ‘[t]he Senate passed’ a renal dialysis program (2003, p. 136); ‘jurisdictional disputes’ and how they ‘diminished the chances for enactment’ (2003, p. 156) – to mention but a few (see also e.g., pp. 48, 52-54, 102, 103, 153-158, 177, 192, 193, 203).
This issue is not confined to Agendas. For instance, Kingdon (2001, pp. 331, 335, respectively) introduces his ‘model of agenda-setting’ yet talks repeatedly about ‘policy change’, giving the example of aviation deregulation and how the work of specialists helped it ‘come about’ and how deregulation eventually ‘passed’ leading to policy change in other sectors as well. Similarly, Kingdon (1994, pp. 215, 225, respectively; see also 1993b, p. 82) writes that he is interested only in ‘how issues come to be issues in the first place, not […] how they get decided’, yet explores the role and ‘content of ideas in authoritative decisions’. Or he states that ‘[w]e’re talking, here, not about how issues get decided’, and yet describes the policy window seized by Lyndon Johnson as one in which ‘a tremendous amount of major legislation was passed’ (1993a, pp. 40, 44 respectively; see also, 1994, p. 216; 2002, p. 100).
It is therefore untrue that Kingdon elaborated a theory of agenda-setting: Kingdon did aim to do that but his remained a mere statement of intent. This is not to say that Kingdon ignored agenda-setting and the specification of alternatives entirely – after all, the examples of policy change he discussed did involve both to some degree. However, agenda change is not the central focus of his analysis. That, after all, is why Kingdon mentioned ‘policy change’ and ‘enactment’ more than ‘agenda change’ throughout Agendas3; why the MSA was initially referred to as the policy streams approach (e.g., Sabatier, 1991) and not the agenda streams approach; why Kingdon spoke of policy windows and never of agenda windows; of policy entrepreneurs and never of agenda entrepreneurs4; and why, ultimately, his book is titled Agendas, Alternatives, and Public Policies, for Kingdon is not just interested in how alternatives get on the agenda – he is mainly interested in how they turn into actual public policies.
Yet, the problem is not only methodological (in that Kingdon studied primarily policy change by asking questions to his participants about agenda change) or one of internal coherence (in that Kingdon did not follow his own research aims), but also and above all theoretical for Kingdon repeatedly equated agenda with policy change. Policy windows are therefore ‘an opportunity […] to push attention’ and simultaneously ‘opportunities for action’, whereby action denotes ‘legislative enactment or other authoritative decision’ (2003, pp. 165, 166, 167, respectively). Their opening causes ‘the agenda to change’ and simultaneously ‘the major changes in public policy result from the appearance of these opportunities’(2003, pp. 173, 166, respectively, emphasis added; see also 1994, p. 228). In fact, Kingdon explicitly identified the confluence of streams, the defining feature of his alleged agenda-setting model, as a necessary precondition for policy change5 – contradicting his own dismissal of such hard-and-fast rules6 –, repeatedly equating agenda with policy change7 contradicting in so doing his claim that agenda change may as well as may not lead to policy change.8 This explains why, in a confusing yet for that very reason truthful summary of his book, Kingdon (1993b, p. 77, emphasis added) wrote: ‘[i]n agenda setting, Kingdon (1984) portrays policy change as the result of changes in separate streams’.
The above discussion leaves us hanging in a limbo. In theory, policy windows and coupling are supposed to explain agenda change. In practice, they are used to explain policy change, which in turn is both equated to and differentiated from agenda change at once. Trying to understand the meaning of policy windows and coupling is thus challenging from the very outset. The task is further complicated by Kingdon’s lack of clarity regarding the meaning of both concepts. In The Policy Window, and Joining the Streams chapter of Agendas,Kingdon (2003, p. 166) defines policy windows as ‘opportunities for action’ in which the three streams ‘come together and are coupled’ like when ‘target planets are in proper alignment’. Based on this definition, however, it is unclear whether policy windows open before or after coupling – and whether thus they are two different events – or whether both describe in reality the same phenomenon. This confusion is exacerbated by the term coupling. In theory, this means combining two things or ideas, and indeed Kingdon used it sometimes in this sense (see below). Yet, as the quote above demonstrates, Kingdon also used it to describe the merging of three streams.
Besides the vagueness of Kingdon’s given definition of policy windows and the inaccurate use of the term coupling, an initial review of his texts reveals six different interpretations of both concepts (Fig. 2): 1) coupling and policy windows both mean the merging of two streams; 2) coupling and policy windows both mean the merging of three streams; 3) policy windows (not denoting the merging of three streams) open after coupling (merging of two streams); 4) policy windows (merging of three streams) open after coupling (merging of two streams); 5) policy windows open after coupling (merging of three streams); 6) policy windows (merging of two streams) open before coupling (merging of three streams).
At first glance, the last option seems the most valid. After all, for Kingdon (2003, p. 194) ‘coupling is most likely when a policy window […] is open’, meaning that they are different events and that the opening of policy windows precedes coupling, the former denoting the merging of two streams and the latter that of three streams.9 However, this interpretation goes against other interpretations of coupling that Kingdon has offered (see below) and the definition of policy windows that we have just seen. Above all, it is inconsistent with Kingdon’s thinking since policy windows create an opportunity for policy change10 (hence the term policy windows). They are ‘opportunities for passage’ (2003, p. 161), an ‘opportunity for actual adoption’ (2002, p. 101; 1993a, p. 43) or simply an opportunity to make a choice (hence the alternative ‘choice opportunity’ – 2001, p. 332). Yet, the merging of two streams does not create such an opportunity. Indeed, what drives change for Kingdon is the convergence of all streams. As he (2003, p. 195) argued repeatedly, ‘the probability of an item rising on a decision agenda is dramatically increased if all three elements […] are coupled in a single package’.11 Or, as he confirmed in the preface to the second edition of Agendas, ‘big policy changes occur when the streams join’ (2003, xix; see also 2001, p. 332). After all, he conceived his as a three-stream model, not as a two-stream one.
A closer look at the Coupling sub-chapter (Kingdon, 2003, p. 172) supports option 1 instead. In here, Kingdon interpreted coupling as the merging of the Policy Stream with either the Problem or the Political Stream. ‘[P]roposals […] suddenly […] become elevated on the governmental agenda’, Kingdon wrote, ‘because they can be seen as solutions to a pressing problem or because politicians find their sponsorship expedient. National health insurance, for example, has been discussed constantly […]. But the proposal rises on the agenda when the political stream […] opens a window’ (2003, pp. 172-173, emphasis added). Therefore, a problem does not need to be on the agenda for the policy window to open (or for coupling to occur) since what is only needed is the merging of two streams; and coupling and the opening of policy windows denote the same event, meaning they could be used as synonyms. However, this contradicts the interpretations of coupling that we saw above, as well as the use that Kingdon made of both terms in The Policy Window chapter in which he described coupling and policy windows as two different events, also dedicating each of them separate sub-chapters. Additionally, this would again be inconsistent with Kingdon’s thinking since policy windows are opportunities for change the likelihood of which is “dramatically increased” when all streams have merged, not when only two have so. Moreover, this would leave us with no clue about what happens after all streams have merged and the factors that might or not then lead to policy change (which, as we saw previously, is Kingdon’s main preoccupation).
What about options 3 and 5, i.e. the possibility that policy windows may open after coupling, with policy windows not denoting the merging of three streams? This is what Kingdon alluded to when he (2003, p. 183) wrote that ‘coupling [‘of solution, problem, and political momentum’ …] does not take place only when a window opens [… but also] before’. In this case, too, both options are unrealistic. To understand why, one needs only to look at the events that for Kingdon determine the opening of policy windows in the Problem or Political streams, which are the same ones that determine the development of the same two streams: exogenous events, changes of administration, swings in public opinion, etc. (e.g., 2003, p. 174). Thus, it would not be possible for the same factors to lead to the development of a stream and the opening of a window after all streams have merged at two different times. First, because by the time the window opens a problem might not be a problem or on the agenda anymore, or a solution might no longer be available or accepted, or the government might have been overthrown. In fact, none of the streams may be there anymore. Second, because for the same factor to open a window at t2 after a time-consuming process of merging with other streams at t1 would be incredibly demanding. As Kingdon (2003, p. 104) recognised, ‘[i]t takes time, efforts, mobilization of many actors, and the expenditure of political resources to keep an item prominent on the agenda’ – and it is only right to imagine that keeping a problem and a solution prominent will be even harder. To be sure, different factors from the same stream could open a window after all streams have merged. A change of administration, for instance, could take place when all streams have merged already and when such change of administration is not the agenda driver of the Political Stream. However, in this case, too, by the time such change takes place, any of or all the streams might have disappeared, not to mention that as Kingdon repeatedly argued it would be incredibly resource-intensive to keep all items prominent and that time is not only key, but also extremely limited.
In option 4, windows also open after coupling but, in this case, they denote the merging of three streams. This is possibly what Kingdon meant when he said that a policy windows is like a ‘space launch window’ (2001, p. 332) that occurs when ‘target planets are in proper alignment’ (2003, p. 166). This interpretation, however, goes against every other definition of policy windows and coupling he provided. It also contradicts his core arguments, for if policy entrepreneurs ‘constantly hook these [three] streams together’ then windows cannot ‘open infrequently’ (2003, pp. 183, 166, respectively), unless of course they denote a separate event which would however bring us back to option 5.
Finally, option 2. If coupling and policy windows meant the merging of three streams as Kingdon suggests in several instances (see above), this interpretation, too, would contradict Kingdon’s other definitions of both terms and make one of them superfluous, contradicting also his claim that they are separate events and that windows are scarce. Besides, it would leave us again with no clue about the factors that from agenda change may then lead to policy change.
Yet to make things more complicated, there is a further interpretation of coupling offered by Kingdon (2003) at the end of Agendas which as Dolan (2021, pp. 183, 169, respectively; see also Dolan & Blum, 2023) noted has been ‘overlooked’ by scholars who have ‘gloss[ed] over the coupling process by repeating Kingdon’s metaphorical language’. Here, Kingdon explains that coupling takes place twice. First, any two streams are coupled in what he calls ‘partial coupling’ (2003, p. 204). These streams and the remaining third stream are then joined in a ‘complete linkage’ (2003, p. 204), accomplishing what we may call full coupling. It is this full coupling that ‘dramatically’ increases the chances of bringing about change, contrarily instead to partial couplings which are ‘less likely to [make items] rise on decision agendas’ (2003, p. 202).
Three further definitions therefore appear to be possible (Fig. 3): 7) the opening of policy windows corresponds to partial coupling; 8) the opening of policy windows corresponds to full coupling; 9) policy windows open after partial and full coupling.
Option 7 is to be rejected similarly to option 6 for it would imply that the partial coupling of streams and the opening of policy windows denote the same event, which would make one of these terms superfluous and be again inconsistent with Kingdon’s thinking since not only are they thought to be separate events but policy windows are also meant to create an opportunity for policy change, which only full coupling is capable of.
Option 8 resembles the scenario just described, in that if the opening of policy windows corresponds to full coupling, then one of these terms is superfluous, contradicting Kingdon’s claim that they are separate events. In fact, one could simply use the term coupling to denote the merging of two streams and policy windows to denote the merging of such streams with the remaining third one. Yet this would resemble in full option 4.
Finally, option 9 is also to be rejected since policy windows would open after the merging of all streams, hence bringing us back to options 3 and 5.
Yet Kingdon made matters even more complicated, offering in his texts two further interpretations of policy windows and coupling (Fig. 4). According to Kingdon (1994, p. 220; see also 2003, p. 229) streams can ‘get linked in various ways before the critical time of open windows and coupling’. This allows for a further option (no. 10) which resembles in full options 3 and 5, except that this time Kingdon does not call the confluence of streams coupling but conceives this as a separate event, meaning the three streams need to be joined once and then again after a window has opened. This scenario not only further exacerbates confusion and seems impracticable due to time and resource constraints (see above), but it also raises key questions that Kingdon has left unanswered (e.g., how is one to differentiate between coupling and “normal” stream linkage? Just like there are “partial” and “full” couplings, are there also pre-coupling partial and full linkages?).
Lastly, Kingdon (2003, p. 203) argued that ‘[w]indows are opened by events in either the problems or political streams’. For instance, explaining the rise of health care in the 1980s in the United States, Kingdon (2002, p. 103) argued that ‘[t]he problem stream […] opened a window’. Similarly, Kingdon (2003, pp. 168, 169, respectively) argued that a ‘new administration [… represents] an open policy window’, or that ‘turnover [… and a] change in congressional membership […] all open windows’, or that ‘[a]n airplane crash […] opens a window’. There is, therefore, a final alternative definition (no. 11) to consider: policy windows open in one stream, with coupling denoting the merging of two or three streams (Fig. 4). This option is also to be rejected. First, if policy windows open in either the Problem or the Political stream, then the term ‘policy windows’ is superfluous since their opening would merely denote the development of said streams. Indeed, the factors that for Kingdon open windows in these streams are the exact same factors that lead to their development. As noted previously, one could argue that the factors opening a window in the Problem and Political streams are different than the ones that lead to their development. However, setting aside the fact that for Kingdon (2003, pp. 168-169) they are the same factors (e.g., ‘a change of administration […], a shift in national mood, […] [a]n airplane crash’), one would also need to explain why these factors lead in some cases to the opening of a window and in others only to the development of a stream, which neither Kingdon nor any scholar has to the best of my knowledge ever done.
Technicalities aside, this interpretation of policy windows is inconsistent with Kingdon’s thinking since as previously noted (see footnote no. 9) if the development of a stream was enough for the opening of a window, then there would be windows all the time contrarily to Kingdon’s (2003, pp. 166, 184, respectively) claim that they ‘open infrequently’, hence their ‘scarcity’. This definition is also inconsistent because as we have seen Kingdon (2003, p. 166; 2001, p. 332, respectively) believes that policy changes ‘result from the appearance of these opportunities’, therefore policy windows must either denote the merging of three streams or occur right after or else the development of a stream would be enough to bring about change which instead ‘take[s] place when all three of the streams join’. Finally, if policy windows open in the Problem or Political streams, which is why there are only ‘problems windows’ and ‘political windows’ (2003, p. 194), then what are policy windows exactly? Are they just a synonym for both problems and political windows, which in turn simply denote the development of the Problem and Political streams? To be sure, one could argue that policy windows are called so because they help a policy (solution) rise on the agenda. However, this contradicts Kingdon’s (2003, p. 203, emphasis added) view for he believes that policy windows are ‘an opportunity for advocates to push their pet solutions or to push attention to their special problems’ so policy windows may not be about policy at all. In fact, as Kingdon (2003, p. 170) wrote, ‘the window sometimes closes because there is no available alternative’. In short, not only do we not know after reviewing Kingdon’s corpus what a window is, but also what policy windows are.
As if that were not enough, scholars have added another layer of confusion by crediting Kingdon (e.g., Howlett et al., 2016; Reardon, 2018) for having also authored the term “windows of opportunity”. However, Kingdon (1993a, 1993b, 1994, 2001, 2002, 2003) never used such an expression.12 Besides being yet another demonstration of the cursory use that is made of MSA terminology, this raises the question of what exactly these windows are. On the one hand, the word ‘window’ suggests that they are a synonym for policy windows, which in turn is just another expression for problem and policy windows, which in turn simply denote the development of the Problem and Political streams. On the other hand, the word ‘opportunity’ suggests that these windows create a “choice opportunity” therefore implying that the term is a synonym for “full coupling” instead, as only this can “dramatically” increase the chances for change. Such a cursory use of the term not only increases ambiguity as one can see, but it also increases inconsistency. Indeed, most of those using the terms policy windows and windows of opportunity interchangeably believe policy windows open in one stream (option 11). In such a case, however, policy windows cannot denote “windows of opportunity” but, at most, an opportunityfor an opportunity, for only “full coupling” creates an opportunity for change. Alternatively, one could argue that a policy window opens in one stream whereas a “window of opportunity” opens after the merging of all three streams. However, this does not add anything new to the debate as it leads us back to options 11, 2 and 4 already discussed.
Conclusion
Albeit widely used, there is still confusion in the literature about two key aspects of Kingdon’s so-called Multiple Streams Approach. Firstly, whilst for some Kingdon developed an agenda-setting framework, for others he developed a theory of policy change, and for others still a theory of both. Secondly, despite policy windows and coupling being the most crucial and popular concepts of the framework, ambiguous and conflicting definitions of both terms are used, sometimes by the same authors. Further, even though scholars have established guidelines which advocate a specific interpretation of their meaning, those citing such guidelines employ different definitions in their work. Sometimes, they do so in the belief that these align with Kingdon’s, suggesting that Kingdon himself has offered interpretations that differ from the ones advocated in current guidelines. Through an in-depth analysis of Kingdon’s works, most of which were previously overlooked, this article demonstrated that these issues have their origins in such a corpus. Regarding the aim and scope of the framework, Kingdon does claim to be interested in the study of agenda change, building his interview guide accordingly. However, his remained a mere statement of intent as he primarily attempted to explain policy change instead, also repeatedly equating one with the other despite recognising that they are affected by different processes. Regarding the meaning of policy windows and coupling, Kingdon provides a vast number of ambiguous definitions of both concepts in his texts. This article identified no fewer than eleven of them, all of which were rebutted as inconsistent, unhelpful or unrealistic.
These issues emerge from across Kingdon’s corpus demonstrating that they are structural weaknesses rather than minor, circumscribed oversights, as a review of Agendas’ latest edition quickly reveals, too. As we have seen, Kingdon has kept its main body unaltered claiming it necessitates no revisions – a point that he suggested again in a 2016 interview in which he admitted that ‘every once in a while I pick it [Agendas] up and read parts of it and go – that’s really good!’ (2016, n.p.). He has, however, expanded its back matter with new case studies. Yet rather than an opportunity for theoretical advancement, this has resulted in him repeating the same mistakes discussed in this article, thus further validating its very conclusions. Indeed, in the Epilogue of Agendas’ updated second edition, in which Kingdon (2011, pp. 235, 242, 243, 240, 231, 239, 241, 242, respectively, emphasis added) compares health care reform in the Clinton and Obama administrations, he gives conflicting definitions of policy windows (e.g., first claiming that ‘the recession opened a window’ and then that ‘the confluence of all three streams […] open[ed] a window’), differentiating also agenda change from policy change whilst simultaneously equating one with the other (claiming that ‘final passage […] lies somewhat beyond this book’s subject’ yet focusing his entire Epilogue on passage of ‘the final legislation’, hence the choice of health care reform, his main research question – ‘[w]hy has it been so difficult to get fundamental health care reform passed?’ –, the repeated references to policy change and how e.g., ‘the Democrats [were able] to pass the bill’ although ‘opposition complicated the passage’, and his definition of policy windows as opportunities ‘during which there is a real chance that major policy change can ensue’).
Although this article, like previous studies (e.g., Cairney & Jones, 2015), concludes that greater clarity and precision are sorely needed, it does so for different reasons. By showing that confusion in the literature originates in Kingdon’s corpus, the present work disproves claims that the lack of shared understandings of key MSA concepts (e.g., Jones et al., 2016) and the application of the MSA beyond agenda-setting (e.g., Herweg et al., 2023) are attributable to scholars and their use of the framework. Similarly, although Kingdon’s lack of clarity has long been noted, such claims are often made without unpacking and comparing the various definitions Kingdon has proposed or examining whether and how he has defined his key concepts in texts other than Agendas, which remain overlooked. This article addresses that gap and substantiates those claims by analysing Kingdon’s language in detail. This confirms the importance of critically reviewing Kingdon’s texts for only when we know where a problem lies that solutions can be proposed and responsibilities ascribed. That said, scholars do lack a shared understanding of key MSA concepts and they are responsible for having glossed over the meaning of key elements of the MSA (sometimes even crediting Kingdon for authoring concepts that he never used), as well as for not having adequately scrutinised Kingdon’s texts (suffice it to remind here that Agendas is the only work on the MSA written by Kingdon cited by all papers but one mentioned in this article).
Yet to the extent that Kingdon and scholars share responsibility for causing the problem, they also share responsibility for solving it. This article offers three recommendations to strengthen the framework’s theoretical foundations. As far as Kingdon is concerned, it is evident that clarifications are needed about the intended meaning of policy windows and coupling, as well as the framework’s objectives. Increasing clarity and consistency is crucial if the framework is to theoretically advance, for should Kingdon continue to refuse to ameliorate his work, Agendas’ widespread use will continue to perpetuate its flaws, magnifying them. Scholars’ limited critical engagement plays its role here of course, but Kingdon could do much to fix the problem at its very source. Such issues do not necessarily need to be clarified within Agendas itself. In fact, one could argue that the best place for such a disquisition could be Theories of the Policy Process, which Kingdon previously declined to contribute to (see Sabatier, 2007). An “idea whose time has come” at last?
Secondly, different definitions of policy windows and coupling should be considered, compared, and critically discussed. Recent MSA scholarship (e.g., DeLeo et al., 2024; Zohlnhöfer et al., 2022; see also Zahariadis et al., 2023) has produced guidelines which de facto impose the view that policy windows open in one stream and occur before coupling, without discussing any alternative definitions. This view, however, is not only one amongst many in Kingdon’s texts, but even more internally inconsistent than its alternatives. In fact, this article has shown that there are at least eleven definitions in Kingdon’s corpus and none of them is internally consistent. This is a critical issue that extends well beyond Kingdon affecting the whole MSA literature, not least because scholars cite Kingdon’s definitions without much questioning (Dolan, 2021; see also Dolan & Blum, 2023). To be sure, ambiguity, while it undermines conceptual clarity and generates confusion, can also be beneficial insofar as it allows for empirical flexibility and enables wider applicability. However, this is no longer a matter of ambiguity, but rather one of semantic dispersion and theoretical impoverishment. Key elements of the framework are used in so many, often inconsistent ways, that we no longer know what exactly is being referred to: one could go as far as saying that there are multiple Multiple Streams Approaches, and yet none at the same time. This compromises the development and testability of a theory, and a theory. This is not to say that existing definitions need to be discarded. However, choosing one over another needs to be justified, and the ramifications that each definition comes with need to be carefully evaluated. For instance, if policy windows are thought to open in one stream, then as previously noted one needs to explain why factors pertaining to that stream lead in some cases to the opening of a window and in others only to the development of a stream. Similarly, if coupling is thought to occur before and after the opening of policy windows, then one needs to explain how one is to differentiate between coupling and “normal” stream linkage. Moreover, one needs to be open to the idea that some concepts may be unnecessary if they describe the same phenomenon, and if conflicting definitions are found to work, it becomes crucial to understand what makes their coexistence possible and what theoretical implications that entails.
Lastly, appropriate methodological choices should be made based on the aims which the framework is being used for. Some scholars have suggested that ‘[t]here is […] nothing in the MS[A] framework itself that restricts its application to some policy stages at the expense of others’ (Sætren, 2016, p. 27). However, different rules govern different stages of the policymaking process (as ironically Kingdon himself noted), which explains why scholars using the MSA often find a discrepancy between the level of agenda and policy change achieved (e.g., Bache & Reardon, 2016). Applying the MSA beyond agenda-setting therefore requires making different methodological choices for asking people what issues they pay attention to as Kingdon did (see 2003, p. 235), is different than asking them how policies changed or why they supported or rejected policychange: I may pay attention to the health benefits of exercise, but that does not make of me a runner.
Far from introducing ‘comprehensible’ concepts (Zahariadis, 2016, p. 12) and being a book so good for its ‘coherence’ (Greer, 2016, p. 418) as well as ‘the quality of the writing’ (Baumgartner, 2016, p. 60), this article has shown that Kingdon’s Agendas abounds in contradictions and ambiguities and suffers from critical methodological and theoretical shortcomings. The fact, thus, that it has remained ‘almost unrevised’ for forty years is a ‘rare fate’ (Greer, 2016, p. 429) – but this is a weakness, not a strength, and a critical one. This article has shown why and proposed several solutions to bring clarity and strengthen the frameworks’ theoretical underpinnings, opening in so doing a “window” that is now left to its users and analysts to seize.
