Skip to navigation – Site map

HomeIssues5:1Applying the Multiple Streams Fra...

Applying the Multiple Streams Framework in Westminster systems: A comparative case study of pay-for-performance policymaking in primary health care in England and New Zealand

Verna Smith


The Multiple Streams Framework has been criticised for failing to recognise the strong institutional drivers of policymaking in Westminster-type jurisdictions, thereby limiting its relevance for explaining policymaking in such jurisdictions. There has been much recent scholarship exploring its relevance for such jurisdictions. However, a new method has been developed to analyse the application of this popular Framework to case studies of policymaking episodes, using a set of hypotheses to test the Framework’s predictive power. This provides an opportunity to further address two key questions: the applicability of the Multiple Streams Framework to Westminster systems, and the more general question of the relationship between institutions and the Multiple Streams Framework. The research reported here has applied the new method to two episodes of health policymaking in two centralised Westminster jurisdictions with closely aligned political, policymaking and health systems, England and New Zealand. The process and outcomes of each policymaking episode, and the relevance of the Multiple Streams Framework for explaining them using the new method, are presented. While the hypotheses are found to be valid for the policymaking process and outcomes in the English policymaking episode, this is not the case for the New Zealand episode. The findings show that, there is a need for greater recognition of the strong influence of institutional factors in the Multiple Streams Framework, particularly in the decision-making stages of the policy processes, especially with regard to policymaking in centralised Westminster jurisdictions.

Top of page

Full text

I would like to thank the reviewers for their helpful and detailed comments on this manuscript and Professor Nikolaos Zahariadis for his continued support and encouragement of my research.


1John Kingdon's (2010) Multiple Streams Framework (MSF), as explored by Zahariadis (2007), is one of the most popular multi-theoretical explanations of policy change. It has, however, attracted much critical commentary for its neglect of institutional dynamics in policymaking processes (Mucciaroni, 1992; Zahariadis, 1995; Beland, 2005; Zohlnhöfer et al., 2015; Spohr, 2016; Zohlnhöfer & Rüb, 2016). Scholars have also argued that it is too indeterminate, thus preventing the development of testable hypotheses (Mucciaroni, 1992; Sabatier, 2007). Efforts to increase its empirical impact and use it theoretically are gaining momentum (Jones et al., 2015; Cairney & Jones, 2016), particularly as research has shown that it appears to have become increasingly relevant to policymaking in advanced democracies in recent years including Westminster and European democracies (Herweg et al., 2015; Weible & Sabatier, 2018). Much work has been done to explore these issues of transferability to policymaking outside the United States and the importance of institutional arrangements in policymaking (Zahariadis, 1995; Herweg et al., 2015). This article seeks to contribute to these efforts by examining some of the key elements of the MSF in two Westminster settings involving similar health policy initiatives. The article also explores the relationship between the MSF Framework and the institutions that influence public policymaking in these two Westminster settings.

2Westminster systems are parliamentary democracies characterised by an executive composed of and accountable to members of the legislature, the presence of parliamentary opposition parties, and a ceremonial head of state distinct from the head of government, and often contrasted with the presidential system of the United States. In majoritarian, unitary Westminster systems such as New Zealand, the government can engage in non-incremental policymaking with relative ease in contrast to federal or separation-of-powers systems. Ambiguity in policy preferences, as described by Cohen et al. in the garbage can model at the core of Kingdon’s MS framework (Cohen et al., 1972), is therefore less likely in such systems. To what extent does Kingdon’s MS framework accurately explain this pattern in these systems as well? Ambiguity can, of course, always arise in such systems in times of sudden and unexpected events that require a rapid political response. But what happens when planned and top-down change in Westminster systems produces unpredictable policy responses, and what drives these differences? This article describes and analyses the results of policymaking in two such jurisdictions, which have obtained different outcomes, with a view to explaining these unpredicted results.

3The policymaking stages considered in this article are those relating to agenda setting and decision-making with particular reference to the role of political entrepreneurs in these Westminster systems. A framework developed by Zohlnhöfer, Herweg and Zahariadis (Zohlnhöfer et al., 2022) was used as the basis for analysing the evidence gathered.

4 I will present the findings of this empirical study of two closely matched health policy episodes, the introduction of a national pay-for-performance scheme for general practitioners (GPs), in two similar Westminster countries that had different policy outcomes. The two case studies were analysed using a system study design that was as similar as possible. The research found that the different policy outcomes in the two cases can be attributed to certain features of institutional differences within the health systems of the two countries. In testing the usefulness of the MS framework in these two relatively well-matched systems, the similar findings, through replication, increase the level of confidence that these possible reasons for difference are not due to chance (Smith, 2018, p. 131) and that the MS framework needs to be modified if it is to be able to better explain policymaking in such jurisdictions.

The Multiple Streams Framework: deriving and confirming its hypotheses

5In his classic study into health and transport policymaking in the United States, Agendas, Alternatives and Public Policies (Kingdon, 2010), John Kingdon painted a beguiling picture of non-incremental policymaking as serendipitous and entrepreneurial rather than planned and rational. This theory of policymaking characterised by Zahariadis in 2007 as the Multiple Streams Framework (the MSF), is arguably one of the most popular and widely recognised approaches used by scholars to understand policymaking theory and practice today (Jones et al., 2015).

6To summarise, Kingdon argued that discontinuous and non-incremental policymaking is most likely to occur when three streams in the policymaking arena, namely problems, political factors, and policies are connected at critical junctures, called policy windows, through the efforts of policy entrepreneurs. These efforts converge when policy entrepreneurs place a policy problem and its solution on the agenda of decision-makers. These entrepreneurs must seize an opportunity to act in a brief and perhaps unpredictable policy window, linking problems and policies together and harnessing political will to achieve an authoritative decision to act. It should be noted that Kingdon agreed that these actors could be inside or outside government, in elected or appointed positions, but they are defined as being willing to invest their resources for a future return (Kingdon, 2010, p. 122). My research includes the study of policy entrepreneurs who can be termed political or institutional entrepreneurs to describe those who are politicians who found and championed policy ideas and the civil servants who knew about them and championed them within government, as well as those entrepreneurs outside government who sought to place an idea on a government’s agenda at a particular time. This research finds that both political and institutional entrepreneurs play an active role in these two case studies, but not that of the exogenous policy entrepreneur outside elected or appointed government.

7Kingdon said that three additional conditions are also present: the existence of ambiguity about preferences among participants, fluid participation of people in the policymaking process, and unclear technology for making changes to the current situation (Zahariadis, 2007, p. 27). When there is also time constraint on decision-makers, the chances of rational, measured, incremental policymaking are further reduced, and dramatic, non-incremental change is most likely. Kingdon is curiously silent on the importance of institutional topography in facilitating or frustrating non-incremental policymaking, and this has led to concerns that the framework is too US-centric and overlooks important features of non-incremental policymaking in, for instance, Westminster jurisdictions. 

8Since its first publication in 1984, Kingdon’s MSF has been extensively reviewed and expanded. Zahariadis (2007) has made a major contribution to our understanding of the details of the MSF through his organisation and documentation of the five key structural elements (problems, politics, policies, policy windows and policy entrepreneurs) and sets of sub-elements that are key features or inputs to these elements. For instance, items in the problems stream may be more or less likely to come to the fore through the presence of indicators, focusing events, feedback and/or the burden of other problems.

9The Politics stream is influenced by party ideology, national mood and the balance of interests at any given time. Ideas in the policy stream may be more or less attractive depending on their value acceptability, their technical feasibility, the adequacy of resources to implement them and the degree of integration of networks supporting them.

10The MSF has been criticised for its overly indeterminate nature which has frustrated scholars’ attempts to test its key hypotheses, especially for predicting policymaking in more centralised jurisdictions with fewer veto points (Mucciaroni, 1992; Sabatier, 2007; Zohlnhöfer & Rüb, 2016). These jurisdictions are associated with patterns of planned, top-down processes that achieve non-incremental change. Their political institutions have structural features and decision-making processes which are largely resilient to the activities of individuals or random disruptions which are the primary drivers of Kingdon’s MSF. There has been some scholarship exploring the relevance of the framework to such jurisdictions, particularly those with strong institutional frameworks (Zahariadis, 1995; Herweg et al., 2015). In particular, Zahariadis and Allen in 1995 studied privatisation policy in Britain and Germany and found that “differently structured networks affect policy innovation” with highly integrated networks associated with an emergent to convergent pattern of policy adoption in which innovative policy options can be seen to be “swiftly solidifying into a stew” [rather than a soup] (Zahariadis, 1995, p. 92). Herweg et al. in 2015 then presented theoretical refinements to the MSF that “make it applicable to parliamentary systems and the decision-making stage of the policy process (Herweg et al., 2015, p. 435), as well as “setting out operating definitions of key concepts [which derive] a set of falsifiable hypotheses’’ for the agenda setting and decision making stages of policymaking. Sager has also considered the role of institutional factors in health policymaking using Qualitative Comparative Analysis in a systematic comparison of the adoption of policy programmes in Swiss cantons. This study found that institutional factors need to be included in the garbage can model (Cohen, 1972) that underpins MSF, and that in this study of health policy, organised interests (or networks) were key to the successful implementation of policy programmes (Sager & Rielle, 2013, p. 17). Schlager (2007) has also advised that we should incorporate institutional structures in the Politics stream. 

11However, Zahariadis has recently commented that the MSF still does not make fully explicit the institutional topography, variation in entrepreneurial activity and policy outcomes that help us understand and explain policymaking in highly centralised Westminster jurisdictions (Zahariadis, 2018). Efforts to adapt the MSF to show at the importance of institutions in decision-making have been undertaken by scholars in recent times. For instance, Zohlnhöfer, Herweg and Rüb (Zohlnhöfer et al. , 2016) have advocated the use of MSF in multiple types of jurisdictions, given that policymaking in parliamentary democracies is becoming less centralised and rationally driven. Spohr has attempted to merge elements of a single theoretical institutionalist approach with the MSF to explain change in the UK and Sweden (Spohr, 2016).

12It is important to know how and why policymaking processes work to bring about change or produce different outcomes in these particular types of jurisdictions. Ultimately, this builds knowledge that policymakers can then use to achieve success in their endeavours. Diagnosing the strengths and weaknesses of Kingdon’s MSF for centralised jurisdictions is of pressing importance because if its utility can be dismissed, scholars can move on to develop and test the next theory. If not, it needs to be fully explored, tested and its application to these types of jurisdictions translated into useful advice and guidance for policymakers. As the MSF has grown in popularity (it has been applied to over 20 different policy areas and in over 65 countries), pressure to explore and confirm its explanatory power has increased.

13The MSF has generated a large body of literature comprising individual studies that provide empirical analysis but do not engage with the theoretical framework, according to Cairney and Jones et al. In their view, turning the Framework theory into a detailed theory requires hypotheses that can be tested with multiple cases (Cairney & Jones, 2016).

14Zohlnhöfer et al. (Zohlnhöfer et al., 2022, pp. 27-28) have set out an approach to studying the MSF by testing its key hypotheses and operationalising its key individual concepts in the agenda-setting phase which can also be applied to decision-coupling.

15The research reported in this article addresses the following research questions:

  1. Why have pay-for-performance policies been adopted?

  2. What factors made policy implementation more likely to be successful?

16It does so by testing the hypotheses put forward by Zohlnhöfer et al. These are grouped into two areas: one relating to the MSF as a whole and one relating to the key elements of the framework, addressing the three streams, the policy window and the policy entrepreneurs.

17With regard to the hypotheses for the framework as a whole, it is expected that this research will provide answers to the first research question by exploring whether three conditions are met, namely that a policy window is open, the streams are ready to be coupled and a policy entrepreneur promotes agenda change, thus linking the opportunity for agenda change, the availability of a policy solution and the active promotion of the policy solution to decision-makers. The research will examine whether these three interlinked conditions were present in each case study. The hypotheses for the framework as a whole are broken down as follows:

18Hypotheses for the framework as a whole:

  1. Agenda change is more likely when a policy window opens,

  2. Agenda change is more likely when the streams are ready for coupling

  3. Agenda change is more likely when a policy entrepreneur promotes the agenda change.

19With regard to the hypotheses for the key elements of the framework, the research will explore whether there are factors for successful agenda change and implementation, namely that there is evidence of a number of interlinked drivers for agenda change, including that the framing of a problem is stimulated by negative feedback, and that there is evidence that the actions of the policy entrepreneur support the implementation of the policy change.

20The hypotheses for the key elements of the framework are:

  1. Problem Stream – a problem broker is more likely to be successful in framing a condition as a problem the more an indicator changes to the negative, the more harmful a focusing event is, and the more clearly a government programme is not working as expected.

  2. The policy entrepreneur is more likely to successfully couple the streams during an open policy window.

  3. Successful implementation is more likely when the implementation of the adopted scheme is supported by a policy entrepreneur.

21In the framework used for this research, the dependent variables are agenda change and implementation output.

22The unit of analysis is democracies, the level of analysis is the national level, and consideration is given to MSF adaptation requirements, such as extending the framework to consider policy outcomes.

Methodological approach

23The comparative approach can provide greater methodological assurance that the differences between two or more case studies have not occurred by chance if it can be shown that two systems are similar in all respects, except for one or two variables. We can explore this using Mill’s method of difference where we compare cases in which the phenomenon occurs with otherwise similar cases in which it does not (Mill, 1862; Castles, 1991; Freeman, 2000). According to the logic of the “most similar” strategy adopted for this research, if some important differences are found between these otherwise similar systems, then ‘‘if we can locate some particular features in which otherwise very similar nations differ we are entitled to suggest it is attributable to one of the few other features distinguishing them” (Castles, 1991).

24In the research reported here, the selection of comparative policy cases was important. Next, it was necessary to design research methods and apply them equally to both cases. Finally, a transparent and verifiable comparative analysis had to be undertaken. These steps are described below. 

Selection of cases

25In order to design an effective strategy for the most similar systems research approach, it was necessary to keep as many of the variables associated with the two case studies as constant possible. This meant that the selection of policymaking cases for study was a critical matter for decision in the design of this research.

26The first choice was the two countries to be compared. England and New Zealand are two countries with strong similarities in their institutional structure. Both are majoritarian, unitary and highly centralised Westminster political systems with adversarial features, and both give great autonomy to their executive. New Zealand’s system has been described as a “maverick” following the adoption of proportional representation in 1997, but in both countries political parties develop policy programmes prior to elections and are usually able to implement their manifestos without constraints from the legislature once elected (Patapan et al., 2005).

27The choice of policy arena was the next key decision. England and New Zealand have similar taxpayer-funded national health systems, established in 1948 and 1938 respectively, with similar roles for GPs as gatekeepers to secondary and tertiary health care. In a critical departure from the original intended policy outcome, by 1949 New Zealand had abandoned its preferred contractual arrangements for funding general practice care and adopted a fee-for-service model, allowing GPs to begin charging a small additional sum directly to patients for primary health care consultations. This led to the emergence of a substantial private market within the New Zealand general practice sub-system and over time, reduced the scope for public influence over this sub-system. England has never experienced a similar situation. GP care has remained free for patients and a unique governance and ownership structure has developed for GP services that are fully funded by the state. The relationship between the GP community and the state in England is highly corporatist.

28Our study focused on policymaking between 2001-2007. In both countries, there were similar problems of variation in the quality of primary care, associated with rising health care costs and with a disproportionate impact on deprived communities. There was also considerable pressure from doctors at the time to improve their working conditions and income levels in England.

29The GP interest group or network was to some extent similar in both countries, with a representative medical association for general practitioners. In England, however, the British Medical Association (BMA) had bargaining rights for all GPs in annual contract negotiations with the state. It was a highly integrated network, whereas the New Zealand network was more competitive and much less integrated. Doctors in both countries enjoyed a high level of trust from patients and the wider community and were a powerful force to be reckoned with during their opposition to policy change in the past. They shared a strong preference for professional autonomy and self-regulation, which had been granted to them by both countries.

30The nature of the ideological persuasion of the governing political parties was also an important factor to keep constant. In both countries, centrist social democratic parties had recently been elected or re-elected after a long period out of office, and had made clear manifesto commitments to improving health services in their election campaigns. Their successful election provided a policy window in both countries for widespread structural reform of their health systems. The same idea, pay-for-performance for GPs, was selected by politicians in both countries and intended to incentivise improved levels of best practice. Both countries had experimented with encouraging innovation within their primary care systems at the regional level in previous eras of reform, including with pay-for-performance, but this was the first time that both countries had selected this idea for national implementation. It was also important for the research that the outcome evaluations of both schemes had already been published, so that the impact of the policymaking processes could be assessed and compared.

31To take account of socio-economic circumstances, both countries had a buoyant economy in 2000 and both had committed significant new resources to health services to improve performance. 

Unit of analysis

32In a process of purposeful selection, the unit of analysis for the research was the design and implementation of a national pay-for-performance scheme within general practice in each democratic country over a three-to-five-year period between 2001 and2007. The research considered the process of agenda setting, alternative selection, and implementation in both countries.

33As the analysis shows, while the two systems are similar in most respects, they differ in one crucial element: the institutional framework for the funding and ownership of GP services.

Matching methods in both cases

34Other aspects of the conduct of the research were also aligned. First, a comprehensive review of the grey literature relating to each policymaking episode was undertaken. This included documentation relating to the problem identification, policymaking process and decision-making stages, including cabinet papers and published guidelines for each of the pay-for-performance schemes. In both countries, considerable descriptive and evaluative academic literature had been published following the implementation of the pay-for-performance schemes, and this was also read and summarised. Qualitative interviews were then conducted using the same interview schedule and comparable samples of proximate policymakers in each country. These included the ministers responsible for health policy, the civil servants charged with its development and implementation, representatives of organised general practice engaged in the process, and other participants in the policy design process. Elite interviewing techniques based on Fontana and Frey’s guidelines (Fontana & Frey, 2005) were used in one-hour semi-structured interviews with 12 proximate policymakers in each country, providing rich descriptive evidence of the policymaking process through the eyes of these key participants.

35The same interviewer conducted and transcribed all interviews and completed the write up of the data corpus for both case studies. This researcher also completed the process of returning the transcripts to the interviewees for comment and providing the final write up of the data corpus for participants to comment on.

36Each data corpus was organised using the same approach of presenting a background section followed by sections on who was involved, what was done, how it was done, and implementation. This was followed by an overview of barriers and facilitators of the policymaking process with similar subheadings.

37For each case study, a historical note on the development of the GP component of the National Health Service and the pattern of subsequent reform, derived from the published literature, was described and compared.

38Finally, using Kingdon’s own predictions with respect to conditions for non-incremental change, and Zahariadis’ elements and sub-elements of the MSF, each case study was reviewed for consistency with this policymaking framework. In addition, the case studies were tested for the conditions of preference ambiguity, fluid participation and unclear technology, and the participation of policy entrepreneurs as criteria for non-incremental policy change.

Pattern matching

39The comparative analysis undertaken required a process of “pattern matching” of independent and dependent variables to identify “literal” replication (the same result in each case) or “theoretical replication”, where contrasting results are obtained for predictable reasons, i.e., where the same result does not occur in a second case due to predictably different circumstances (Yin, 2009, p. 140). The pattern matching technique used to demonstrate the results was that of rival explanation (or pattern matching for independent variables). In this approach, alternative reasons for the results are sought and analysis is required to demonstrate which reason most accurately predicts the outcome.

40Explanation building in a study of two cases seeks to contribute to a general explanation that fits both cases, even though the cases themselves differ in their details. This research used an iterative process of explanation building. The study of the English case was undertaken first and revealed the importance of the activities of entrepreneurs, in particular the political and institutional entrepreneurs and their ideological preferences, and the institutional drivers of alternative selection and actual agenda change. This process was repeated for the New Zealand case study, where these phenomena were not similarly observed. One objective of this approach is to entertain plausible or competing explanations and to show how they cannot be supported.

A note on the terminology of policy entrepreneurs

41In addition to the exogenous policy entrepreneur of the original MSF, described as an actor outside government exercising policy agency, two categories are described in this research. There also exists the political entrepreneur - a politician inside the government. Finally, an institutional entrepreneur is one who operates within government.

England: a policy window, political entrepreneurs and the process of agenda change

42In England, there was a close and mutually dependent relationship between the Department of Health and the GP profession, with annual negotiations between the doctors’ union, the British Medical Association (BMA) and the Department of Health (DoH) over the terms of the contract of service signed by every GP in the country. By 2000, however, frustration with high demand and poor conditions in general practice led the profession into open rebellion. Prime Minister Tony Blair recalled that it was in 1998 that the BMA “attacked us for the first time” (Blair, 2010, p. 215). Worse was to come during the Blair government’s re-election campaign. The BMA voted to take strike action unless a substantial pay rise was granted. This was a worrying development for the incumbent political party, which had established the National Health Service in Britain, and a focusing event with considerable electoral importance. As one participant recalled, “there was I think considerable anger about the tactic and the timing of that ballot [with Ministers] railing against the BMA’s interference” (Smith, 2015, p. 101) There was a real risk that failure to resolve this matter quickly could jeopardise the re-election of the Blair government, occurring as it did in the middle of the election campaign.

43One participant interviewed for this research recalls that “the Prime Minister was in the middle of the election campaign when the BMA issued their statement …and [with a small group of others the Prime Minister] talked about it…” The participant recalls that the Prime Minister, his Secretary of State and his health adviser all agreed that although a pay rise was inevitable, it should be conditional on significant improvements in quality being achieved through a pay-for-performance scheme.

44The threat of strike action opened a window in the problem stream regarding GP income adequacy, but it was the political entrepreneurs, the Prime Minister and his recently appointed “modernising” Secretary of State for Health, who decided that pay-for-performance was the best policy solution. Blair spoke in his memoirs of his belief in the need for his government to “introduce systems where the money spent was linked to performance” (Blair, 2010). This was certainly a theme in the NHS Plan published in 2000 in which “partnerships, performance, professions, patient care and prevention” were set out as key deliverables (Secretary of State for Health, 2000). In his authorised biography, Blair is reported as hoping that reforms would include “using incentives to kick-start the modernisation…to increase the quality of health care”(Seldon, 2007). Of the Secretary of State, Blair said that he “was fully sympathetic with the direction of change”. (Smith, 2018, pp. 34-35). In his speech to the Labour Party Conference in 2001, he reiterated that “Without reform, more money and pay won’t succeed” (Seldon, 2007, p. 69)

45Coinciding as it did with the election of a new government, this episode confirms the hypotheses that agenda change is more likely when the policy window is open, the streams are ready to be coupled and a policy entrepreneur, in this case Blair, in collaboration with his Secretary of State, promotes the agenda change.

46Looking more closely at the problem stream, it is clear that the problem broker in this case is the BMA, which has successfully framed the condition of inadequate income as a problem, highlighted the continuing negative changes in indicators, including declining GP recruitment, particularly in low-income areas, and observed increasing unexplained variation in quality standards. The BMA effectively created the damaging focusing event which enabled the problem and policy streams to be coupled and set the scene for agenda change.

47Once negotiations for the design of the pay-for-performance scheme began, there was a rapid acceptance that the bigger the scheme the better. This was followed by a tightly controlled negotiation between a hand-picked team for the Secretary of State for Health and the elected representatives of the BMA general practice profession. The role of the junior minister, John Hutton, who led the negotiations, was key. “The previous government had tried and failed and then imposed an agreement on the BMA. I can understand why…I was very reluctant to get to that point…I wanted consensus…that would stand the test of time …this is a very well developed interest group, the BMA, a very powerful group, so imposed agreement can never survive”(Smith, 2018, p. 41)

48The two teams were dominated by serving GPs. Both the BMA representatives and Departmental teams were trained in principle-based negotiation techniques. These two innovations (a carefully selected Departmental team and the use of principle-based negotiation techniques by that team) gave the negotiations a highly constructive atmosphere that had not often been seen in previous negotiations (Smith, 2021). The Secretary of state’s determination to include in the departmental team GPs who were highly experienced in, and committed to, pay-for-performance forced the process along this route and brushed aside opposition. One member of the government team with considerable experience of a pay-for-performance scheme in his locality felt that he had been personally vetted by the Secretary of State, who invited him to a meeting and strongly challenged him on his attitude to general practice before approving his role on the team (Smith, 2015, p. 91).

49The joint decision to make up to a third of GP income dependent on meeting the new targets resulted in a comprehensive system of 146 quality indicators referred to as the Quality and Outcomes Framework (QOF). To add to the incentives, a newly developed data management system was rapidly developed to allow practitioners to share de-identified patient treatment data from their own practice management systems in order to provide evidence of their performance against targets in real time, to further motivate their efforts.

50Although the scheme was voluntary, GPs rushed to participate and there was almost complete uptake. In the first year, 95% of targets were achieved and GPs received very large increases in remuneration as a result. Morale quickly improved and for a while, recruitment to general practice exceeded recruitment to specialities.

51Early impact studies were mixed, but positive findings included that general practices in poorer areas improved quality faster than those in more affluent areas in the first few years after implementation (McDonald et al., 2010), that reductions in outpatient sensitive admissions as a result of the scheme were demonstrated for chronic conditions such as coronary heart disease, hypertension, congestive heart failure, diabetes and chronic obstructive pulmonary disease) (Dixon et al., 2010), and that the cost-benefit analysis of the scheme was positive. An unanticipated benefit was the creation of a rich national shared database of patient health and treatment information for health research. 

52Politicians were concerned about the large and unbudgeted rewards for performance under the QOF, which required the diversion of funds from other areas of the NHS to GP expenditure. In subsequent years and negotiations, the generosity of the scheme was reduced and the development and withdrawal of indicators was centralised through the National Institute of Clinical Excellence (NICE) and developed more within the negotiating teams. However, the scheme has continued to provide up to 15% of practice income and is widely accepted by GPs who, while calling for change, do not wish to see it removed (NHS England, 2018).

New Zealand: a policy window, policy entrepreneurs and the process of agenda change

53New Zealand’s National Health System was established in 1938, ten years before the establishment of the NHS in England. Following bitter disputes between the BMA and the New Zealand Labour government, in which GPs refused to sign a contract for services with the state, state funding of general practice services began through a fee-for-service subsidy, supplemented by patient co-payments. Over time, the proportion of the cost of GP services covered by state funding declined and the co-payment expected from patients steadily increased. In parallel, private insurers and other state funders were also drawn into the funding framework. There was no formal process of negotiation between the profession and the state to resolve disputes about the level of state funding.

54By 2000, most New Zealanders were paying the full cost of their GP consultations out of pocket, and affordability had become a serious problem. So too were the increasingly alarming gaps in life expectancy between indigenous and poorer New Zealanders and the general population, and the growing health burden of chronic disease. While these indicators were deeply troubling to the newly elected Labour government, they were addressed by the wider programme of health reforms, including improved GP funding, which had been promised in their election manifesto, rather than explicitly through the pay-for-performance programme.

55The newly elected Labour government in 1999 restructured the health system as promised, including requiring all New Zealanders to register with a Primary Health Organisation (PHO) for services and requiring all GPs to be funded by these organisations on a contract and capitation basis rather than on a fee-for-service basis. In this case study, the policy window opened when, as a result of these wider structural reforms, the need arose for a nationally consistent mechanism to incentivise appropriate prescribing and referral to services by GPs. Up to this point, some independent GP organisations had been using pay-for-performance for this purpose, but they would now be replaced by new Primary Health Organisations accountable to the Minister for Health, and their contracts managing prescribing and referrals budgets would be terminated.

56There was no focusing event, no change in indicators to the negative and no electoral salience associated with this issue at that time. No entrepreneur was involved in the coupling of the problem and policy streams. The policy solution, pay-for-performance, came from the GP policy community. The Director General of Health at the time described the process as follows: “The idea came forward that it would be very good to incentivise certain performance measures …grew out of the nexus of communication between the Ministry and the sector… the primary health care sector itself had measures and some of the groupings of general practice had gone quite a long way down the pathway”(Smith, 2015, p. 173).

57Members of this community were recruited to a committee, led by civil servants, convened to discuss how to reinstate controls on prescribing and referrals, given that they were major cost drivers. Partly to replace these lost incentives to improve the quality of prescribing and laboratory referrals, the officials recommended the implementation of a national pay-for-performance scheme to incentivise health actions by GPs to improve the prevention and management of chronic diseases. 

58This is not a classic open policy window, and there were no obvious policy entrepreneurs associated with this policy solution identified in the research.

59GPs did not participate in negotiations over the design of the pay-for-performance scheme, but were minority members of the working group convened to design it. They did not feel listened to in the process, with individual GPs saying “I don’t think a single thing [some participants] said was reflected in the programme such as …peer led, based on feedback and performance data to individuals “(Smith, 2018 p. 99). As a profession, they declined to provide access to their practice data to enable tracking of diagnosis and treatment of the most important and potentially most lucrative health conditions as targets for incentivisation (Smith, 2021). Ultimately, the working group was able to design a small set of 13 indicators based on nationally available data.

60The scheme was managed by the PHOs, rather than directly by the practices. The newly established PHOs had different levels of engagement with practices in their area and developed different criteria for making payments earned under the scheme. Implementation of the scheme in New Zealand was hampered by a lack of timely information to support performance payments, owing to a lack of appropriate practice data, and different approaches to payments by the meso organisation charged with managing the scheme. As a result, the scheme attracted highly variable levels of trust or commitment from GPs and therefore resulted in lower uptake, relatively poorer performance against the targets, and lower additional income (though considerable funds were available and remained unclaimed) (Cranleigh Health, 2012). The lack of real-time data on performance against targets, which was available in England, further discouraged GPs from actively participating in the scheme.

61A table summarising the comparative analysis process used is presented below. In this process, John’s five theories of policy change and variation, namely institutions, networks, rational choice drivers, ideas and socio-economic factors (John, 2012), and the elements of the Zahariadis model of MSF (Zahariadis, 2007), are presented as drivers of policymaking and the two sets of results are compared.

Table 1. The methods framework

Overarching Driver


New Zealand



Majoritarian unitary Westminster

Majoritarian unitary Westminster

Health system-level



GP system-level


Single payer

Single Payer







Quality regulation



Data management





Representative strength



Mandate to negotiate with government funder








Political leaders

Closely involved

Closely involved

Civil servants role

Arms length

Arms length


GP leaders

Highly engaged



Endogenous only

Endogenous only

Other key actors

Mainly GPs

Variety of health professionals



Concern about quality



Concern about inequities of health status




Medical professionalism

Highly valued

Highly valued


Positive previous experiences

Positive previous experiences

Curative vs preventive care

Curative preferred

Curative preferred

Socio-economic circumstances

Level of need



Availability of resources




Party Ideology

Collectivist, committed to NHS

Collectivist, committed to NHS

National Mood

High salience for health service quality

High salience for health service quality

Balance of interests

BMA demanding change

GP interest groups divided about need for change



Evidence-based analysis of health outcomes

Evidence-based analysis of health outcomes

Focusing events

GP threatened to strike



Positive feedback about previous pay-for-performance schemes

Positive feedback about previous pay-for-performance schemes


Commitments to priorities domestic policy change

Commitments to priorities domestic policy change


Value acceptability



Technical feasibility

Feasible with data sharing

Feasible but required data sharing for maximum impact

Resource adequacy


Adequate except for data sharing

Network integration



Policy Window

Coupling logic

Followed GP demand for improved income

Followed need to replace incentives for improvements in prescribing and referrals

Decision style



Institutional context



Policy Entrepreneurs


High for endogenous entrepreneurs

High for endogenous entrepreneurs







Ambiguity of preference?



Fluid participation?



Unclear technology?



Source: the Author

The results

62The table below compares the two outputs and outcomes from the two schemes which resulted from the design process.

Table 2. Outputs and Outcomes from the Two Schemes



New Zealand

Policy output

Voluntary national pay-for-performance scheme

Voluntary national pay-for-performance scheme

146 indicators

13 indicators

Clinical/service domains incentivised

Clinical domains incentivised

25-30 percent of income conditional

2 percent of income conditional

1-yr implementation period

3-yr implementation period

Provision for review of indicators

Provision for review of indicators

Policy outcome

95 percent compliance with targets in first year

81 percent compliance with targets in first year 

Ambulatory sensitive admissions impact on:

  • Coronary heart disease

  • Hypertension

  • Congestive heart failure

  • Diabetes

  • COPD

Ambulatory sensitive admissions impact on:

  • Vaccination-related admissions

Choice of indicators in subsequent years undertaken by National Institute of Clinical Excellence

Choice of indicators in subsequent years undertaken by governance group with larger number of GPs

By 2014, reduction to 10 percent of salary subject to achievement of targets

By 2014, proposal to review scheme and to provide quality building grants to PHOs and direct incentives to practices for achievement of targets

2019 Scheme scaled back but widely supported by GPs

2017 Scheme abandoned

Source: the Author

Summary of findings

63Reviewing the hypotheses in the light of the empirical evidence on the emergence and implementation of the pay-for-performance scheme in England, i.e., the hypothesis for the framework as a whole and the hypotheses for the key elements of the framework, I argue that most of them are validated. There was an open policy window associated with the election campaign and the policy, problem, and political streams were ready for Ministers to couple the idea of pay-for-performance to the need for improved national standards of GP performance. The Prime Minister and the Secretary of State for Health were the leading policy entrepreneurs promoting the policy change. The indicators for GP dissatisfaction with the terms and conditions of their work were demonstrably changing for the worse and the focusing event of their threat to strike was perceived by ministers as deeply damaging. The feedback from the GP survey was clear about the need for change. Both the Prime Minister and the Secretary of State were enthusiastic advocates, indeed policy entrepreneurs within their party, of pay-for-performance and the party itself was perceived as sympathetic to new public management approaches to policy problems. The policy window opened in the problem stream because of the change in GPs level of job satisfaction and this led to the focusing event of the threat to strike, supported by the feedback from the survey of GP attitudes to their work. Ministers saw the situation as a threat to their re-election, and this opened the policy window for the national pay-for-performance scheme in the problem stream overnight. The policy window opened in the political stream due to the election of a new government.

64In contrast, the New Zealand case can be seen as confirming only part of the first hypothesis for the MSF as a whole, namely that a policy solution was at hand. No policy window opened in the classic sense described by Kingdon as one that opens infrequently and does not stay open for long (Kingdon, 2010, p. 166), and no policy entrepreneur promoting the agenda change.

Dependent variables in both case studies

65If we look at the dependent variables, the hypotheses should drive the results in the dependent variables of agenda change and the implementation output.

66The dependent variables are most clearly linked to the hypotheses in the English case study, where the implementation output, namely the largest national pay-for-performance scheme ever implemented, occurred. In both case studies, however, the selection of pay-for-performance was achieved through different mechanisms (policy entrepreneurs in England and an officials’ working party in New Zealand). In England, the agenda change was closely linked to the management of serious political risk and re-election chances, but this was not the case in New Zealand. The implementation outcomes were very different in the two countries – a large-scale scheme in England with 98% of GPs adopting it immediately and with strong initial positive results in reducing outpatient sensitive admissions for targeted conditions, versus a measured and prolonged uptake of a smaller set of indicators in New Zealand, with only one showing reduced outpatient sensitive admissions in due course. In the English case the policy window opened in the problem window, demonstrating all the hypotheses in England.


67These findings provide strong confirmation that the hypotheses relating to agenda-setting, implementation, policy windows and entrepreneurial activity are supported in the English case study. They are less convincing in the New Zealand case study.

68The outcomes of both schemes showed success, although in New Zealand on one indicator, but the magnitude of the results in England was greatly enhanced by the large scale and GP-supported nature of the scheme, which continues with GP support today. The scheme in New Zealand was discontinued in 2017.

69The findings support the work of Zahariadis that network characteristics are relevant to policy diffusion (Zahariadis, 1995), and suggest that this needs to be reflected more strongly in the MSF as a driver of policy change and variation. In particular, the research presented here supports the inclusion of criteria such as “interest group structure” and “forms of governance and ownership of public services” within the Politics Stream as important predictors of the autonomy of governments to manage policy change. They support Zohlnhöfer’s recommendation to acknowledge the role of political entrepreneurs in the Entrepreneurs stream (Zohlnhöfer et al., 2016). They support a proposal for the inclusion of “institutional entrepreneurs” in the Entrepreneurs stream and to recognise “antecedent policies” as a sub-element of the Policy stream. In both countries, institutional entrepreneurs were involved in the promotion and design of the schemes (Smith & Cumming, 2017). Lastly, they support the proposal to add a further element, that of policy “outcome” to the Framework to expand our understanding of policy adoption to include the full public policy cycle.

70Perhaps their most important contribution to the literature, however, is to confirm the importance of patterns of planned, top-down processes in parliamentary jurisdictions, which can achieve non-incremental change and that have political institutions with structural characteristics and decision-making processes that are largely resilient to the activities of individuals or random disturbances. The evidence shows that the primary drivers of Kingdon’s MSF are not always found to drive agenda change and implementation success in these jurisdictions, and that institutional factors may be more important drivers of policymaking and its successful implementation. In particular, the importance of parliamentary parties, the election campaign in parliamentary systems and their ability to create a predictable window of opportunity for policy change is demonstrated in contrast to the dynamics of presidential systems where more personal election campaigns are much more separated from the policy process.


71The research provides further confirmation of the value of the comparative approach in exploring the utility of the MSF and refining it over time. It is only by attempting to apply the MSF that we can determine whether it is fit for purpose in the types of jurisdictions and types of policymaking under study. This has led to further confirmatory evidence of applicability of the MSF to parliamentary systems, as well as the value of modifications to the MSF that better illustrate the policymaking processes found in parliamentary systems. The research has provided a body of evidence to support the contention that Westminster systems of government are capable of purposeful and orderly non-incremental policy change, and that Kingdon’s MSF, which theorises policy formation under conditions of ambiguity, needs to be enhanced to improve its relevance to such jurisdictions. In particular, a more explicit recognition of the importance of institutions such as parliamentary parties in facilitating or impeding policymaking and the role of different types of policy entrepreneurs, including political and institutional entrepreneurs would be useful.

72Finally, this study has reinforced the value of comparative case studies in drawing attention to and further interrogating theoretical frameworks such as the MSF in order to explore their relevance in different jurisdictions and over time as circumstances in the world of policymaking change.

Top of page


Beland, D. (2005). Ideas and Social Policy: An Institutional Perspective. Social Policy and Administration, 39(1), 1-18.

Blair, T. (2010). A Journey. Hutchinson.

Cairney, P., & Jones, M. D. (2016). Kingdon's Multiple Streams Approach: What is the Empirical Impact of this Universal Theory?. Policy Studies Journal, 44(1), 37-58.

Castles, F. G. (1991). Australia Compared. Allen & Unwin.

Cohen, M., March, J., & Olsen, J., (1972). A Garbage Can Model of Organisational Choice. American Science Quarterly, 17(1), 1-25.

Cranleigh Health (2012). PHO Performance Programme Evaluation. Cranleigh Health.

Dixon, A., Khachatryan, A., Wallace, A., Peckham, S., Boyce, T., & Gillam, S. (2010). The Quality and Outcomes Framework (QOF): does it reduce health inequalities? Final Report. National Institute for Health Research.

Fontana, A., & Frey, J. H. (2005). The Interview: From Neutral Stance to Political Involvement. In N. K. Denzin & Y. S. Lincoln (Eds.), The Sage Handbook of Qualitative Research (695-727). Sage Publications Ltd.

Freeman, R. (2000). The Politics of Health in Europe. Manchester University Press.

Herweg, N., Huss, C., & Zohlnhöfer, R. (2015). Straightening the three streams: Theorising extensions of the Multiple Streams Framework. European Journal of Political Research, 54(3), 435-449.

Herweg, N., Zahariadis, N., & Zohlnhöfer, R. (2018). The Multiple Streams Framework: Foundations, Refinements and Empirical Applications. In C. M. Weible & P. A. Sabatier (Eds.), Theories of the Policy Process (pp. 17-53). Westview Press.

John, P. (2012). Analysing Public Policy. Routledge.

Jones, M., Peterson, H., Pierce, J, Herweg, N., Bernal, A., Raney, H. L., & Zahariadis, N., (2015). A River Runs Through It: A Multiple Streams Meta-Review. Policy Studies Journal, 44(1), 13-36.

Kingdon, J. W. (2010). Agendas, Alternatives, and Public Policies, Update Edition, with an Epilogue on Health Care. Longmans.

McDonald, R., Cheraghi-Sohi, S., Tickle, M., Roland, M., Doran,T., Campbell, S., Ashcroft, D., Sanders, C., Harrison, S., Mannion, R., & Milsom, K. (2010). The Impact of Incentives on the Behaviour and Performance of Primary Care Professionals: Report for the National Institute for Health Research Service Delivery and Organisation programme. National Institute for Health Research Service Delivery and Organisation programme.

Mill, J. S. (1862). A System of Logic, Ratiocinative and Inductive: Being a Connected View of the Principles of Evidence, and the Methods of Scientific Investigation. Parker, Son, and Bourn.

Mucciaroni, G. (1992). The Garbage Can Model & the Study of Policy Making: A Critique. Polity, 24(3), 459-482.

NHS England (2018). Report of the Review of the Quality and Outcomes Framework in England. London.

Patapan, H., Wanna, J., & Weller, P. (Eds.) (2005). Westminster Legacies: Democracy and Responsible Government in Asia and the Pacific. University of New South Wales Press Ltd.

Sabatier, P. A. (Ed.) (2007). Theories of the Policy Process. Westview Press.

Sager, F., & Rielle, Y. (2013). Sorting through the garbage can: under what conditions do governments adopt policy programs?. Policy Sciences, 46(1), 1-21.

Schlager, E. (2007). A comparison of Frameworks, Theories, and Models of Policy Process. In P. A. Sabatier (Ed.), Theories of the Policy Process (pp. 293-319). Westview Press.

Secretary of State for Health (2000). The NHS Plan: A Summary. Health. The Stationery Office.

Seldon, A. (2007). Blair Unbound. Pocket Books.

Smith, V. (2015). Pay-for-performance in primary health care: A comparative study of health policymaking in England and New Zealand [Doctoral dissertation]. Victoria University of Wellington, Wellington, New Zealand.

Smith, V. (2018). Bargaining Power: Health Policymaking in England and New Zealand. Palgrave Macmillan.

Smith, V. (2021). Engaging general practitioners in pay-for-performance scheme design. Journal of Health Organisation and Management, 35(3), 249-264.

Smith, V., & Cumming, J. (2017). Implementing pay-for-performance in primary health care: the role of institutional entrepreneurs. Policy and Society, 36(4), 523-538.

Spohr, F. (2016). Explaining Path Dependency and Deviation by Combining Multiple Streams Framework and Historical Institutionalism: A Comparative Analysis of German and Swedish Labor Market Policies. Journal of Comparative Policy Analysis: Research and Practice, 18(3), 257-272.

Weible, C. M., & Sabatier, P. A. (Eds.) (2018). Theories of the Policy Process (4th ed.). Westview Press.

Yin, R. K. (2009). Case Study Research. Sage.

Zahariadis, N. (1995). Ideas, Networks, and Policy Streams: Privatization in Britain and Germany. Policy Studies Review, 14(1-2), 71-98.

Zahariadis, N. (2007). The Multiple Streams Framework. Structure, Limitations, Prospects. In C. M Weible & P. A Sabatier (Eds.), Theories of the Policy Process (pp. 65-92). Westview.

Zahariadis, N. (2018). Foreword. Bargaining Power: Health Policymaking in England and New Zealand. Palgrave Macmillan.

Zohlnhöfer, R., Herweg, N., & Huß, C. (2016). Bringing Formal Political Institutions into the Multiple Streams Framework: An Analytical Proposal for Comparative Policy Analysis. Journal of Comparative Policy Analysis, 18(3), 243-256.

Zohlnhöfer, R., Herweg, N., & Rüb,F. (2015). Theoretically refining the Multiple Streams Framework: An introduction. European Journal of Political Research, 54(3), 412-418.

Zohlnhöfer, R., & Rüb, F. (Eds.) (2016). Decision-Making under Ambiguity and Time Constraints. ECPR Press.

Zohlnhöfer, R., Herweg, N., & Zahariadis, N. (2022). How to conduct a Multiple Streams Study. In C. M. Weible & S. Workman (Eds.), Methods of the Policy Process (pp. 23-51). Taylor and Francis.

Top of page


Electronic reference

Verna Smith, Applying the Multiple Streams Framework in Westminster systems: A comparative case study of pay-for-performance policymaking in primary health care in England and New ZealandInternational Review of Public Policy [Online], 5:1 | 2023, Online since 01 April 2023, connection on 28 September 2023. URL:; DOI:

Top of page

About the author

Verna Smith

Victoria University of Wellington, School of Government, Wellington, New Zealand

Top of page



Creative Commons - Attribution 4.0 International - CC BY 4.0

Top of page
Search OpenEdition Search

You will be redirected to OpenEdition Search