Since the beginning of reforms in the late 1970s, and the move away from a state-dominated and planned economy, China has been transformed from a poor country, in which incomes of most of the population stood below the global median, to a middle-income country, raising hundreds of millions out of poverty in the process . Latterly, it has actively developed social protection systems, including rural and urban health financing schemes, pensions for rural and urban populations, and others. China’s approach to reform has been government-led and unorthodox. The speed of development, and the complexity and interconnectedness of reforms have led many to talk of a specific ‘Chinese model’ of development . Experimental policy making and policy innovation are part of the policy toolbox, though most analyses have discussed experimental policy making in economic reform and similar areas, rather than health or social policy. This section argues that China’s health system development has relied substantially on a range of approaches to policy making, implementation, and policy learning that can be broadly classified as experimental. The discussion that follows is not exhaustive: health reforms have been underway for more than 20 years, and reforms are ongoing both in the health system itself, and in the larger institutional-bureaucratic system. It is also not to say that Chinese approaches are optimally efficient, or that they are necessarily effective (or are used well), but this section argues that China’s approaches (China displays not one form of experimental policy making, but a range of approaches that fall under this umbrella) provide real-world examples of the use of experimental reform management techniques in reforming complex systems, and that these deserve attention from researchers and policy makers concerned with scale up of health interventions.
Complexity in China’s health reforms
As argued above, health systems should be analysed in their social, economic and institutional context. Seen in this way, Chinese health system reforms show a high degree of complexity. While this is hard to capture overall, at least the following points are highly salient:
The health ministry sits at the apex of a hierarchical bureaucratic system and makes policies that guide the actions of multiple tiers of lower level government, but health policy exists within a broad ecology of policy making involving many institutional actors and interests.Footnote 6 Policy making responsibility is distributed, and many policies with direct relevance for health agencies, healthcare providers, etc., are the responsibility of other parts of government. Actions of a range of other government and institutional actors may have an impact on those of health agencies. Central government coordination groups are often set up at times of intense reform activity to coordinate ministry interests and link mandates.
Central government policy must be sufficiently broad to be implementable in jurisdictions (principally cities and rural counties, henceforth: counties)Footnote 7 that vary across many parameters, including levels of economic development, financial and human resources, population health, geographical, environmental, socio-cultural and spatial factors. Equally, counties’ bureaucratic and institutional structures differ, reflecting naturally occurring variation in the political economy, and the legacy of past restructuring processes (both within and outside the health system) that have been implemented differently, or to different degrees, in different places. The importance of non-state actors (including civil society and commercial interests) differs by place.
The number of counties and the range of parameters across which variation may occur create great information asymmetry for central ministries and even provincial governments, and data able to inform intervention are often unavailable, or are limited, before the launch of a given reform. The results of interventions across a range of counties with different conditions and starting points are hard to foresee and the possibility of divergent outcomes in different counties, or over time, is high. The interactions between reforms in the health system and other social, administrative, economic, and environmental systems are harder to predict, especially given ongoing and rapid change in almost all such systems, an increasing amount of which falls outside the direct influence of government, as the society and economy are liberalised. Rapid change in all these areas creates challenges for predicting the effects of any given intervention.
Since the early 2000s, China has embarked on a process of rebuilding a functioning health system, following the partial dismantling of its pre-reform system, increasing marketisation of many health services, and worsening access and equity [27, 28]. Since the early 2000s, waves of reforms have focused on a wide range of areas, including launching of rural and urban health insurance programmes to underpin population health seeking, increasing government health spending and, under the aegis of the 2009 new health reform package, expanding insurance coverage, establishing a national essential medicines system, improving primary care and improving availability of public health services, and piloting public hospital reforms . The above changes have been accompanied by vast transformations of China’s economy, infrastructure, population, and society.
As above, the starting point of much scale up literature is an intervention that is thought to merit application at greater scale, and this literature increasingly points to the importance of context and an understanding of complexity in ensuring successful roll out. On any possible interpretation, building a functioning health system in China requires dealing with complex, high speed and intersecting reforms across a wide geographic area in which the starting points of implementing counties are hard to know, and in which the results of reforms are hard to accurately predict. The following sections discuss Chinese thinking about experimental policy development, and a range of experimental management techniques visible in China’s health reforms.
Central control and local discretion
In China, local experimentation as a tool for policy development dates to before the establishment of the People’s Republic of China (PRC) in 1949, and was pioneered by Chinese communist party (CCP) cadres as a way to deal with a range of priority issues, including land reform and organising agricultural production. For the CCP, this was a pragmatic approach to reforms under conditions of scarcity of resources and expertise and in an environment characterised by extreme variation in conditions. Approaches trialled in the 1920s and 30s were progressively codified as party/government working methods and as an identifiable administrative vocabulary of reform management . Such pragmatic approaches to reform management have existed in tension with more command-oriented approaches over the history of the PRC, but have been frequently been stressed in China’s ‘reform’ era, starting in the 1970s, through use of experimental zones and experimental policies. Such deliberate experimentalism coexists with increasing discretion within the government system through decentralisation of administrative authority and creating incentives for local governments to pursue reforms (such as in economic development) that have system-wide benefits . A number of factors underlie Chinese thinking on this, not least the difficulty of adopting one-size-fits-all solutions in a country of continental size and encompassing very great variation. Inevitably, the range of dynamics created by policy approaches that include direct experimentation as well as decentralisation mean that a range of patterns of learning exist, from (in some cases) directed scale up of relatively managed experiments to more organic dissemination of scattergun innovation processes tied to decentralisation .
In many policy areas, central government leaves substantial space to sub-national governments within an overall policy framework and direction set by central government, creating a ‘paradox’ of central control and sub-national space for discretion and initiative: while central government controls the policy agenda, the personnel system, much resource allocation, and approval for large projects, sub-national units have discretion and space for local initiative, which can support policy innovation . This is frequently rationalised with reference to common understandings of the ‘spirit’ of policy:
“China is a unitary polity. Sub-national government must obey central government, and must carry out reform under the unified arrangement and direction of the centre. [However] central institutions and policy give local government a lot of space for exercising initiative. Acting according to the requirements of the centre, and under the unified leadership of the centre, doesn’t mean blindly or mechanically acting according to the instructions of superior levels [of government]. Rather, each place can […] in line with their local conditions, experiment boldly according to the spirit of central [policy]” .
This framing of local government flexible implementation and innovation within the overall spirit of central policy pre-dates the establishment of the PRC. The paradigmatic statement of this comes from Mao Zedong:
“legislative powers are all vested in the central authorities. But, provided that the policies of the central authorities are not violated, the local authorities may work out rules, regulations and measures in the light of their specific conditions and the needs of their work […] We want both unity and particularity. [I]t is imperative to have a strong and unified central leadership and unified planning and discipline throughout the country […] At the same time, it is essential to bring the initiative of the local authorities into full play and let each locality enjoy the particularity suited to its local conditions” .
This framing is tied to policy discourses which emphasise the importance of ‘implementing according to local conditions’ (yin di zhi yi), and the impossibility of adopting ‘one size fits all’ (yi dao qie) policy that can be summarised as a “deep-seated one-size-does-not-fit-all pragmatism” . Central leaders and policy makers frequently express rhetorical commitment to sub-national policy adaptation, experimentation and innovation,Footnote 8 underpinned by a policy discourse that allows signalling of emergent practices. This discourse of local particularity is shared by many actors within the policy community, including sub-national officials, think tank researchers and academics,Footnote 9 and one aspect of China’s reform process has been to create norms of tolerance of variant sub-national policy practices that can allow initiative and risk taking by local governments for systemic benefit .
What does experimentation look like?
Multiple forms of activity should be considered under the umbrella of ‘experimentation’ in Chinese policy development. According to Heilmann [37, 38], Chinese central government supports managed pilots for the exploration of novel policy options intended to support central decision making, but other analyses have shown widespread policy ‘tinkering’, in which sub-national governments carry out broad low-level experimentation on a range of issues under loose policy frameworks . In most cases, experimentation is not controlled piloting, but distributed problem solving under overarching national policy frameworks, often signalled by terms such as ‘innovation’ [40, 41]. This fits closely with the typology of approaches to intervening in complex systems to promote the emergence of desirable order discussed in Fostering emergence in complex systems and Table 1. Experimental policy making of some form exists in many policy areas, including health programming [36, 42,43,44].
This section develops an indicative typology of forms of experimental policy making, implementation and innovation. This is a heuristic device, whose aim is to indicate the breadth of activity that should be considered under this rubric, to show commonalities with the analysis, above, of programming under conditions of complexity, and to raise the question of the relevance of such approaches for researchers and policy makers concerned with scale up of health interventions (Table 2).
A brief overview of development of China’s rural health insurance scheme, the New Cooperative Medical Scheme (NCMS), illustrates a number of these aspects. A section on Important features and limitations, below, extracts distinctive features from this narrative and frames them in the context of the analysis provided in the section on fostering emergence in complex systems.
Following the decline and collapse of China’s pre-reform rural health insurance, a number of pilots were run in the 1990s and onwards, both managed pilots, involving a substantial degree of expert design or oversight [Type I] , and pilots in which sub-national jurisdictions were granted a degree of autonomy to decide their own reforms [Type III] . Pilots often had support from Chinese and foreign researchers and international agencies, aiming to provide a model for a new insurance scheme. While lessons were learned about possible modes of structuring a new rural insurance scheme, and system dynamics, these failed to have substantial policy impact until the opening of a policy window in the early 2000s . Lessons from these early pilots were, to an extent, taken into account in establishing a national ‘pilot’ scheme [36, 48].
The national pilot scheme initially gave around 300 local governments a range of implementation choices [49, 50] in scheme design [Type II], though much sub-national management was unscripted trial and error within the overall parameters of the scheme [Type III]. Incomplete dismantling of the previous rural health insurance programme from the 1980s onwards meant that some counties retained insurance schemes [Type IV], while variation in economic, institutional and political economic factors provided varying starting points for reforms in different locations [Type IV].
This range of experimental processes produced a wide range of approaches and practices of varying degrees of local and supra-local usefulness, as well as some approaches deemed illegitimate. Efforts were made to learn from this scattergun implementation experience, to disseminate lessons from it in reports and guidelines [51, 52], and to use it in codification of policy. In many areas, sub-national governments (counties and provinces) made concrete policy which was then learnt from and/or codified in provincial or national regulations, in a kind of policy ‘crowd sourcing’ .
Even following roll out, much scheme management remained under-institutionalised , and sub-national governments engaged in much learning by doing [Type III], figuring out how to manage a complex insurance scheme, and to develop suitable management arrangements that are contextually appropriate as they went along, often with little support. Such local experimentation / innovation underpins ongoing system reform and resilience, is often widely reported, and may be the focus of debate within the policy community, but may or may not have much policy impact or be codified .
Faced with specific management problems, national and provincial policy makers, research institutes and international agencies commissioned or carried out managed pilots [Type I] or made use of experimental policy frameworks [Type II] in an attempt to develop policy models with broad applicability, for example in developing functioning provider payment systems [55, 56]. Some such models were propagated widely, but policy impact is often limited and/or hard to show.
Sub-national governments carried out a range of reforms within the overall parameters of the scheme, whether through proactive policy entrepreneurship or through pressure to implement the scheme [Type III], exploring different scheme management problems, including fundraising approaches, cost control measures, payment reforms, linking rural and urban schemes, extending the scheme to rural-to-urban migrants, and similar. This scattergun approach produces a wide range of results, of varying degrees of usefulness. National, city and provincial policy makers, as well as academics and researchers, made use of a range of methods (site visits, meetings, evaluations, etc.) to identify and propagate potentially useful practices – a form of screening for, and learning from, positive deviance – and there is often substantial debate over reforms and specific practices, though propagation is often limited.
Important features (and limitations)
The above is a sketch of a highly complex reform process run across almost three thousand implementing counties over the course of approximately two decades, from small scale managed pilots preceding national policy backing, to the launch of a loosely-articulated piloting process in the early 2000s, nationwide roll out, and ongoing adaptive management, problem solving, learning by doing, and increasing codification over time, including linking of the rural health insurance scheme with urban insurance schemes .Footnote 10 The case presented here illustrates very many, if not most of, the approaches to managing for the emergence of order in complex systems described from the literature in Fostering emergence in complex systems. The following features are striking:
Central government controls the overall policy agenda and direction of reform, sets implementation targets for local governments (frontline implementers of the national scheme), and has oversight of scheme financing. Coexisting with this control, many implementation parameters are only broadly defined, and sub-national governments act as problem solvers and change managers, rather than straightforward implementers of policy. Central government encourages and tolerates a wide range of policy practices. Variation between implementing counties is substantial, and a wide range of policy practices exists at any one time.
Sub-national reforms frequently ‘die’ or are discontinued, in their place of origin, whether because they are deemed unsuccessful, or because of changes in local leadership, incentives or conditions . On the other hand, some sub-national reforms show patterns of iterative learning and deepening .
While variation creates substantial information asymmetry and difficulties for learning, multiple patterns of policy learning exist, from ‘vertical’ codification of policy practices deemed useful in provincial or national policy, through to relatively organic spread through a range of ‘informational infrastructures’ , including meetings, trainings, organised site visits, exchange of documentation, etc. Reforms are accompanied by substantial information flows, through policies, ministry bulletins, media, etc. Policy champions and trusted intermediaries, such as government research institutes, may be important in promoting certain sub-national practices.
Variation across jurisdictions can occur along multiple dimension; scale up inevitably encounters multiple contexts and requires ongoing policy adaptation. Codification of diverse policy practices often occurs late in the policy cycle, maintaining space for sub-national flexibility and one-size-doesn’t-fit-all scale up. Even where codification occurs, reforms are often ‘under-institutionalised’ .
There are many criticisms of such approaches to policy development . This is not an optimally efficient approach to policy development: many pilots fail to have substantial impact, while many policy innovations fall by the wayside or fail to be propagated . Specific policy innovations are rarely ‘best practice’, but the system can produce ‘appropriate’ solutions that show contextual fit, and underpin ongoing system adaptation . Overall, there is space for policy learning to take place and for system adaptation over time.
As discussed above, the above dynamics are visible in a range of reform areas, in health and elsewhere, and a substantial body of Chinese analysis is devoted to profiling potentially useful or important sub-national reforms.