Institutional Legacies and “Sticky Layers”: What Happens in Cases of Transformative Policy Change?

Introduction

There has been a strong tendency in analyses of institutional change to focus (explicitly or implicitly) on “punctuated equilibrium” models, whereby long periods of stability are interrupted by sudden exogenous “shocks” which produce transformative change (Béland, 2007; Streeck & Thelen, 2005). This has been described by some scholars as a “conservative bias” within the literature, which needs to be challenged (Streeck & Thelen, 2005). Recent attempts to move beyond the punctuated equilibrium dichotomy have seen an enhanced focus on gradual processes of institutional change (Mahoney & Thelen, 2009, 2010; Streeck & Thelen, 2005). Mahoney and Thelen (2010), for example, have argued that institutions change in subtle and gradual ways over time often from endogenous sources. This emerging body of work argues that we must look more deeply at the endogenous sources of institutional change, and not just endogenous sources of institutional stability, which is a much commented upon short coming of the new institutionalist literature.

In this article, we examine a “critical” case study to explore the transformative–incremental dichotomy. We use the case of the Australian National Disability Insurance Scheme (NDIS) as an “outlier” case of transformative change (Yin, 2014). Launched in 2013 with bipartisan support, the NDIS will provide no-fault insurance cover for Australians who are born with, or acquire, a severe and permanent disability (Australian Productivity Commission, 2011; Bonyhady, 2014; Collings, Dew, & Dowse, 2016). The NDIS is a rare example of an attempt to create comprehensive, system-wide transformative change supported by the community and accompanied by a strong bipartisan political mandate (Thill, 2015). We contend that even in the context of rare transformative change, many features more commonly associated with incremental change are evident. In particular, our analysis explores the phenomena of institutional layering and “institutional stickiness” in attempts to implement this reform. Diverging from the current literature, we find that layering is not merely an unfortunate by-product of previous formal and informal institutions but a tool that is actively sought and used by policy makers to tackle implementation challenges even within a transformative policy context. Hence, the article contributes to the extension of the concept of policy layering in contexts of transformative, non-incremental cross-jurisdictional policy reforms. Consistent with scholars such as Mahoney and Thelen (2010), we argue against the bifurcation of the study of policy dynamics into a stability versus big bang dualism. We do this by revealing that many influential and constraining factors are common across both incremental and transformative reforms. However, we argue that current efforts to develop more nuanced theories of institutional change need to engage with cases of planned transformative change, not just gradual change, which may in time accrue transformative outcomes (Streeck & Thelen, 2005).

Policy Design and Implementation of Major, Non-Incremental Reforms—Clean Slate or an Act of Layering?

Within the important seam of policy studies dedicated to understanding how institutions and policies change over time, attention is being increasingly paid to a set of patterns that produce change in the absence of “big” reform (Béland, 2007, 2010). The ambition is to capture analytically such sequencing patterns that do not display salient, unambiguously important moments or episodes of formal policy change. Conventionally, such patterns have been labeled as “policy continuity.” However, new work has queried casual attributions of stability and instead begun to identify gradual and directional sequences of public policy change under the surface of apparent observed policy continuity, and “emphasize the cumulative impact of small changes, through processes of policy displacement, layering, drift, conversion, and exhaustion” (Wincott, 2013, p. 810).

The concept of policy layering, used to describe the phenomenon where new policy goals are added to, or layered onto, existing policy commitments without removing others, has attracted recent scholarly attention as one type of these incremental but significant policy change (Béland, 2007, 2010; Jacobs & Weaver, 2015; Mahoney & Thelen, 2009; Streeck & Thelen, 2005). The concept has been formed as part of this broader effort to catalog the many different policy dynamics that play out beyond a stability–change dualism. Anchored in the work on historical institutionalism, layering tends to be viewed as the unintended outcome of a series of incremental policy-making interactions and decisions at the margin of already existing policy trajectories (Streeck & Thelen, 2005; Van Der Heijden, 2010, 2011).

When we look across the diverse literatures on institutionalism, a central line of inquiry is to articulate the potentially “uneasy relationship” between policy legacies and policy change (Kay & Daugbjerg, 2015). An inherited policy, which could include previous goals, objectives, or instrument calibrations, can function as an institution by circumscribing the set of options available to current policy makers. Insofar as these institutional constraints are tight, filtering out new policy blueprints and tabula rasa designs, policy decisions of the past weigh heavily on the policy options of the future.

In such terms, incremental policy reform is about changing certain elements of established policy regimes (May, Jones, Beem, Neff-Sharum, & Poague, 2005), and such change carries some risk of producing ineffective instrument mixes or introducing incoherent goals. Howlett and Rayner (2007) propose evaluating such incremental policy changes to inherited policy regimes in terms of the post-change coherence of goals and the consistency of policy means. New mixes of policy goals are judged as coherent insofar as they are related to the same overall policy aims and objectives and may be achieved simultaneously without requiring trade-offs, temporal sequencing, or value balancing. Alternatively, incoherence of policy goals may be understood in terms of contradictions and tensions. The consistency of policy means, or instruments, is considered functionally—the extent to which instruments complement each other in supporting a policy goal. They are inconsistent when they are counterproductive and drive policy in different directions.

The interaction and relations among different parts of policy can make the policy “whole” more than the sum of its parts. The interactions and relationships in a policy system form institutions, generate variety, and produce complexity in the interaction of ideas, interests, and material circumstances. This makes the composition of policy change important: These relations between different parts of policy over time are not simple additive ones; rather they are dynamic and complex, and typically have emergent, self-organizing properties. This is missed by the extant three orders or levels of change introduced by Hall (1993), and developed further in recent work by Cashore and Howlett (2007), Howlett and Cashore (2007), and Howlett and Rayner (2006).

This body of work is a challenge to reductionism in policy studies: the view that every observed policy phenomenon is something intended by a policy actor or the unintended consequence of an individual action. However, by always focusing on the individual agent and how decision making is affected by institutional structure, evolutionary effects in terms of the emergent properties of composite “wholes” are liable to be missed. Institutions are collective: it is groups of agents following a rule that constitutes an institution. Examples of significant interaction effects include tipping points, network effects, combinational effects, bandwagon effects, reinforcement, emergence, and, additionally, we argue, layering. In this article, we argue that the concept of policy layering can be used to structure accounts of policy processes without requiring any assumption of individual agency, that is, that some entity intends to layer policy and acts on that intention.

Furthermore, acknowledging that there are properties of the whole that are not reducible to its constituent elements raises the analytical possibility of meso-level effects. This is where the causal mechanism runs from meso to meso level—from policy to policy—rather than macro to micro—from policy to an individual agent in the policy system. For policy purposes, this means causal mechanisms are operating at a level higher than individual agency. This can be seen in, for example, the many studies of the regulatory state or meta-governance effects (Moran, 2002; for example, Morgan & Dubash, 2013), for example, where a policy aimed at extending the use of market mechanisms to help ration and reallocate water resources in New South Wales created its own policy and governance dynamics which can explain change without requiring an explicit account of agency (Bell & Park, 2006).

We seek to extend the scope of the concept of layering by demonstrating that layering may also be observed in cases of comprehensive, system-wide policy change as well as incremental change, which traditionally fall into the category of “transformative” change. Mahoney and Thelen (2010) outline four models of institutional change: displacement, layering, drift, and conversion. In transformative, system-wide change, we would expect to see displacement—“the removal of existing rules and the introduction of new ones” (Mahoney & Thelen, 2010, p. 15). Displacement is characterized by a fairly abrupt type of change, with a radical shift and breakdown in institutions and replacement of new ones. Mahoney and Thelen (2010) present these four types of change as independent, arguing that in cases of displacement, we are unlikely to see layering, drift, or conversion. We present a case study of layering which challenges these views, revealing that layering exists in cases of non-incremental reforms. We find that even when the policy design is system-wide and transformative in ambition, layering can occur with important consequences for understanding and explaining the trajectory of implementation and, ultimately, the impact on policy outcomes. In our case study, we not only identify layering in the sense proposed by Streeck and Thelen (2005) but also show an important empirical extension of the concept to apply to implementation processes. That is, that layering exists within implementation strategies of phased transitional arrangements. Here, layering and implementation are not mutually exclusive categories—where phases or stages are scaled up. Rather, far from making implementation easier by breaking down big changes into a series of small steps, we show how layering can help apprehend how interim arrangements may get stuck and raise consistency and coherence issues with new, later phases of the scheme layered on top.

The role of agency in layering sequences is poorly understood and articulated (Van Der Heijden, 2011). In particular, it is not clear whether policy layering should be understood as something caused directly by a political strategy used by an influential policy actor, or rather as a policy sequence that is the unintended outcome of other change processes playing out. We argue that it is the latter leaving an important but necessarily separate question of the agency of policy actors—the extent to which they recognize their interests in, and have the capacity to develop appropriate responses to, layering sequences.

As we will explore in this article, the temporality of policy implementation and institutional change is also important. Consistent with this, it has been central to recent discussions of endogenous institutional change (Mahoney & Thelen, 2010). Due to the sheer scale of the NDIS, policy makers committed to a phased process of implementation. This, we will show, has contributed to the emergence of layering and stickiness at key implementation junctures, and is likely to prohibit the maturation of the NDIS into the planned final phase.

Method

This article draws on data from a longitudinal study of the implementation of the NDIS. The study aims to investigate implementation with a particular focus on how governance structures enable and/or constrain policy learning and change (necessary for the successful implementation of complex reforms). The study utilizes a case study research design because it enabled us to investigate these changes in-depth and in their real-life contexts (Yin, 2014), across time. Case study methodology gives us a way of defining cases, rather than a way of analyzing cases or modeling causal relationships.

In many ways, the NDIS is a critical case. Critical cases enable the testing and building of theory (Yin, 2014) because of their unique and sometimes “outlier” nature. As previously noted, the NDIS is a rare case of comprehensive system-wide (or transformative) policy change. The implementation of the NDIS is taking place over 5 years (July 2013-June 2018) through a 3-year trial phase in seven areas across Australia and a 2-year transition phase where trial sites are expanded to cover all eligible individuals. Trialing a new approach is not uncommon; however, in the case of the NDIS, not only are new policy instruments being trialed (an insurance model, individualized care packages, and a greater reliance on quasi-markets), but the scheme encompasses new policy objectives, most importantly, a guarantee of lifetime care and a much greater level of choice and control for eligible participants. Furthermore, because co-evolution is so important in the implementation of the NDIS, none of the trials are testing the specific funding and administrative arrangement that will be used post June 2018 when the scheme is fully implemented. This means the NDIS provides a rare opportunity to examine unusual phenomena within context (Prior, 2016).

Semistructured interviews (N = 26) were conducted with key policy makers in the Commonwealth government charged with the design and implementation of the NDIS. Interviews were recorded and transcribed verbatim. Participants were senior policy makers, charged with major decision making and overview of the NDIS. Snowball sampling was used, which began with the departmental secretary to identify appropriate members of the team responsible for the NDIS. Further details of participants are not provided for reasons of confidentiality.

Themes covered in the interviews included decisions regarding the governance structure of the NDIS, implementation challenges relating to the development of the scheme, the markets and national roll out. Data were analyzed using a thematic approach (Blaikie, 2010). “Like” data were grouped together to form categories and subcategories. These categories were developed into more substantive themes, by linking and drawing connections between initial categories and hypothesizing about consequences and likely explanations for the appearance of certain phenomena (Strauss, 1987). This was done through discussion between the team. In the refining of themes, selective coding was carried out, whereby transcripts were revisited with the explicit intent of finding further linkages and connections between the central issue being explored and other themes.

Staged Implementation and Policy Layering

The temporality of policy implementation and institutional change has been central to recent discussions of endogenous institutional change (Mahoney & Thelen, 2010). Due to the sheer scale of the NDIS—the provision of no-fault insurance cover for an estimated 460,000 Australians—policy makers committed to a phased process of implementation:

Because of the roll out . . . three models are being designed. So you were really using the trial to experiment . . . And then you’ve got this disruptor in the middle where you need a rapid intake process, where you actually need elements of the model designed differently, just for transition. (P02)

So the idea of having three years of trials and then two years of transition, you’ll continue to evolve the model during transition. Once you get more clients in you’ll see different patterns and you’ll evolve and then improve the model. So it will continue to evolve right through to full scheme and then it will just have a continuous improvement cycle even then. (P23)

Although this staged implementation approach was anticipated in the Productivity Commission’s inquiry into the design and establishment of the NDIS (Australian Productivity Commission, 2011, p. 934), policy makers expected that the transition phase would simply be a larger version of the individual trials. However, because each of the trial sites (with the exception of the Australian Capital Territory) focused on a particular type of disability or life stage, a “trial” in the true sense was not possible. Even more significant was the need to enroll large numbers of eligible individuals and establish care packages during the transition phase which required a different approach to that used in the much smaller trials.

I think one of the big changes we’ve seen is a lot of things that were operating in trial were never going to be scalable for transition to full scheme. (P08)

When you’re operating on the trial site . . . they had to get the market operator [for that] cohort, which wasn’t the whole market, and in other trial sites it’s been localized. ACT for example, has been able to, to an extent, test whether the workforce worries are really real . . . Providers have probably had . . . more handholding in the trial than you’d expect them to get in full scheme. (P13)

Those involved in implementation describe the transition phase as a “disrupter” within the implementation process because different goals and values became embedded into organizational processes, and the scope of individuals being integrated into the new system changed from those originally envisaged. As noted earlier, within the transition phase, the major focus of implementation effort is on entering people onto IT systems and drawing up preliminary plans:

during transition, because the ramp up is quite strong—like [going] up the side of a cliff—[there is a] big role in just getting people onto their plans. (P13)

However, the values that get institutionally codified at this stage may prove to be difficult to shift. As one senior official noted, “you also want to make sure that things in transition don’t leave a legacy in full scheme that is a bit hard to unpick” (P02).

We conceptualize this observed disruption as a tense-layering effect. This is at the crux of institutionalized “stickiness” in the multi-stage implementation strategy for the NDIS. Our data reveal that the dilemma of trying to achieve a successful transition stage without establishing legacies with self-reinforcing dynamics at the full and final stage is difficult.

Institutional Legacies and Sticky Layers

A range of drivers were identified that lead to both the development of a disruptive transition phase and potential “stickiness.” The rapid expansion from trial to transition meant that interim arrangements had to be put in place during transition. The NDIS has been implemented at a speed unanticipated by its architects, with each stage of implementation being shortened and the start date of the scheme being brought forward by 1 year. In July 2016, the NDIS moved into the transition phase 1-year early, rapidly expanding trial sites to cover 460,000 citizens. As Patashnik (2008) has argued, implementation is imbued with politics, which push in particular directions that may not always be congruent with original policy goals.

I guess there’s always been a fear that the transition phase is so fast. I mean the thing about NDIS is that the former government started a year earlier than the Productivity Commission said and so it’s always been a bit like building and implementing at the same time . . . At this point in time . . . it’s this tension between wait until you get everything right and then do it, as opposed to jump in and have a go and do the learn, build, build, build. (P16)

The rapid nature of implementation and speeding up of each planned implementation phase has added to the “stickiness” of aspects of the reform. During transition, the NDIS was being implemented before key aspects of its design, such as the regulatory architecture, were finalized (Department of Social Services, 2016).

So we knew we’d need rapid intake: trials 25,000; 25,000 to 30,000 by the end of trial; transition 400,000. So you’re going up by 10 to 12 times every month . . . [There are] things that haven’t been ful[ly] designed and signed off yet around the market, and around quality and safeguards. So [for] some of the elements of the scheme . . . the policy settings hadn’t been finalized. So some of those lessons [which were] about helping to inform the design and build of those policy elements, hadn’t yet been settled. (P02)

As the historical institutionalism literatures suggest, policy makers are relying on co-evolving institutions to take the NDIS from its transition state to full scheme (Streeck & Thelen, 2005), and it is hoped that the necessary institutions will co-evolve in a coordinated fashion (Kay & Daugbjerg, 2015). Of course, in practice, such harmony arising serendipitously is unlikely and a number of factors appear to have limited this desirable co-evolution. In particular, the transition phase is characterized by highly complex, and often constraining, financial arrangements between the Commonwealth and State governments. Prior to the introduction of the NDIS, disability services were provided by the States directly, or indirectly, using block funding. A significant challenge in the transition phase of the scheme was managing the financial arrangements produced by this legacy. Critically, State and Territory finances have dictated the shape of the NDIS and individual care plans during transition.

The design of the transition agreements was very State focused, in that it focused on prioritizing the transition of existing State clients to the NDIS as fast as possible so they could free up [State government funds]. (P16)

Finances and implementation are very linked in the transition period . . . The vast majority of the money that they’re moving into the NDIS is existing [money]. (P18)

While the exact principles differ State by State (Department of Social Services, 2016), the NDIS must use State- and Territory-funded services before “cash” services where individuals are given money to purchase alternatives from the market.

The NDIA has found it, understandably, very difficult to administer using in-kind services because they’ve got reference packages to say that this person’s worth this much. If the state provides supported accommodation as an in-kind service, the NDIA is a price taker, and so they can’t apply their price, which relates back to the modeling of scheme sustainability . . . [The States] need to use the in-kind services above cash, because if they don’t, the in-kind services might disappear . . . But that puts pressure on the cash. I think that is a really huge challenge and something if you’re looking at from an implementation perspective, is very difficult to administer. Now the reason we’ve got in-kind is to help us transition out of the existing contracts, but that can take years. (P09)

This need to use in-kind supports first means that participants of the NDIS are not exercising the type of choice and control that sits at the heart of the scheme:

By using in-kind, you’re also restricting choice and control of the participants, because you’re saying, “well we’ve still got this service that we’re going to have for the next three years, as a contract.” So although we’re telling you, “you can go and get whatever provider you want,” you actually have to use this [provider] first because otherwise we’ll run out of money. (P09)

The use of in-kind funding is already set to reach into the full scheme, past transition phase with some contracts being longer than 5 years.

We have asked the States to get out of in-kind and the Commonwealth, we’re getting out of in-kind over the transition period. So by the time you get to full scheme, most in-kind contracts will have ceased. However there’ll be some in the States where the service is administered by the State government rather than an NGO that they will not be able to get out of. So that will continue. (P09)

In drawing on in-kind contributions, institutional legacies are brought into the new scheme, as existing programs continue to be delivered to clients, underpinned by the values and objectives that existed prior to the NDIS. The risk is that these values and objectives will become embedded within the NDIS as participants’ choices continue to be restricted to what they have known in the past. For example, in discussing the issue of choice for persons with an intellectual or developmental disability, Webber and Cobigo (2014) note that limiting choice to a menu of familiar options limits future choice because individuals have not had the opportunity to be exposed to, or learn about, innovative or unfamiliar options. In other words, NDIS participants may have the opportunity to make choices, but these choices will not be “based on a rich understanding of what might be possible” (Burchardt, Evans, & Holder, 2015, p. 62). Limiting choice to what has been provided in the past will also constrain the development of fully competitive markets, as discussed in the next section.

The Impact of Sticky Layers on Policy Implementation Trajectories

The growth and change required to create effective and efficient disability service markets has been a driving force in the alternate arrangements, or “look,” of the scheme in the transition phase. The NDIS relies on the existence of mature and complex quasi-markets to function. However, quasi-markets are notoriously difficult to create and manage (Considine, 1999; Considine, Lewis, & O’Sullivan, 2011; Needham & Glasby, 2015; Nevile, 2013). In the first instance, there are significant upfront costs associated with provider transition from traditional block funding to individualized funding. As a result, a significant lag period exists between the introduction of the NDIS and the flourishing of a fully functional market. As with the financing arrangements, this has created pressures that lead to clients maintaining existing care arrangements.

To be honest, between the timeframes and then the market responses, what plans look like and what people are able to have provided in those first three to five years would be quite different . . . to what they can get in five to ten years time. And that’s because people do need to come in quickly, it’s new for participants. I think some of the trials are saying a lot of the participants are happy just to transition and pretty much get what they’re [getting] now or a bit more of different providers and that sort of thing. (P16)

Here, policy makers are confronted with a “chicken and egg” problem—to implement without the markets in place, or try to generate new market arrangements and then push toward implementation of care packages. Policy makers hope that

over time as participants become a bit more sophisticated in understanding what the NDIS is and how they can use it, and also as the market then becomes a bit more innovative and responsive, there’ll be much more choice, so the aspirations of the NDIS can be realized. (P16)

However, co-evolution could have been better supported by allowing policy makers to have greater control over the market. Markets and care packages exist in a dialectical relationship (Hegel, 1975). That is, care packages assume markets exist and markets assume care packages exist. Undoubtedly, this is difficult for policy makers navigate and therefore contributing to institutional stickiness and layering.

At present, the role of the government in steering the newly created disability markets remains undecided (Department of Social Services, 2016). A small amount of money has been made available on a competitive basis to help providers transition into a deregulated market, but the main “market lever” has been the provision of information about how the scheme will work (NDIS, 2016). The lack of control and support mechanisms within government means that policy makers do not expect the scheme to reach its vision of a fully deregulated disability service market for at least a decade.

[Full deregulation] was a vision I think of the Productivity Commission, whether that becomes a reality—we all hope it and think it . . . There may be areas where [there are] thin markets—remote regional areas where you’ve got to have a different approach where you can’t be totally deregulated. But it’s definitely the vision for total deregulation. You’d be progressively going there, and I don’t think you’d have full market deregulation for a good 10 years. (P16)

This slow progression to a largely deregulated market leaves the scheme open to being caught in the transition phase, as values are embedded which require on-going regulation. The problems of immature markets are further compounded by participants’ lack of experience in exercising full choice and control. As a historically underresourced area, participants have been happy just keeping their existing service arrangements with, potentially, a slight increase in support. Hence, the problems of immature markets are twofold: Providers are also slow to transition because participants and the scheme are not demanding it of them.

Conclusion

As a relatively rare case of transformative policy change, policy makers implementing the NDIS have had more opportunity than most to break free of institutionalized practices and legacies. At the outset of the trials, there were no foundations on which to layer policy changes as often occurs with incremental reforms.

I don’t think there’s anywhere else in the Commonwealth, past, present or future, that such a complex arrangement, with such complex governance and such complex funding arrangements . . . [has] ever been done before. (P17).

[R]ight from the beginning we were just absolutely flat-out to get the thing implemented . . . we didn’t have an Act in place, we didn’t have actual instruments, so we were all working flat out. We actually had to create an Act from scratch, because there was nothing to model it off, and this is the first thing definitely in Australia, and probably in the world. (P1)

As noted at the outset of this article, Mahoney and Thelen (2010) outline four models of institutional change (displacement, layering, drift, and conversion). The literature argues that in transformative, system-wide change, we would expect to see displacement—“the removal of existing rules and the introduction of new ones” (Mahoney & Thelen, 2010, p. 15). Displacement is characterized by a fairly abrupt type of change, with a radical shift and breakdown in institutions and replacement of new ones. Mahoney and Thelen (2010) present these four types of change as independent—where we have displacement, we will not have layering or drift. However, our case challenges this assumption demonstrating that displacement (or at least attempts at displacement) can in fact be accompanied by layering. Mahoney and Thelen (2010) do note that displacement can occur slowly, and our case shows that this temporal dimension is important as it allows layering effects to come into play. These layering effects are exacerbated when co-evolution is not taking place to the extent that was originally anticipated (and desired) by policy makers.

Thus, because implementing transformative, system-wide change takes time, it is likely that displacement will always be accompanied by some degree of layering. However, the implementation experiences of the NDIS to date suggest that layering is not only a by-product of previous institutional structures but also a tool that is actively sought and used by policy makers to address practical implementation challenges. For example, policy makers used existing State-based service delivery models during the transition phase of the NDIS, which meant prioritizing the use of in-kind service provision over care purchased directly from private providers, because doing so allowed them to process large numbers of eligible participants.

This example raises the question of agency: To what extent do policy actors recognize their interests in and have the capacity to develop appropriate responses to layering sequences? As our interviews reveal, senior bureaucrats charged with implementing the various elements of the NDIS are very much aware of the potential long-term impact of sticky layers:

The NDIS is not an insurance model. It’s become a social welfare scheme. Through the negotiations with States and Territories, it evolved from being the model [proposed by] the Productivity Commission, to a sort of more traditional eligibility assessment social support model. It’s not an income support model, by any means, but it is an eligibility assessment model. (P26)

[I]n some ways I think . . . [a] transition of only two or three years is a good thing because it’s really hard once you’ve done something to pull back. There will need to be some pragmatic approaches in transition, but not locked in. (P16)

Hence, while senior bureaucrats could recognize layering and “stickiness,” they were less certain about ways to prevent or interrupt it. Rather, they hoped that over time, some of these institutional legacies would be reversed or modified, even though the ways in which they as individual agents—or a collective—could achieve this were not clear. For example, senior bureaucrats are very much aware that the long-term success of the scheme is dependent on the development of a mature market for disability services, but, to date, there is little consensus on what kind of governance model will best facilitate market development or management (Department of Social Services, 2016).

There is at this stage [what] I would describe as an immature process sitting under the bilateral agreements [between the Commonwealth and State and Territory governments] for monitoring markets . . . It’s not clear to me what mechanisms will be used, or what information will feed into any kind of assessment that gets made about how the market is going, because the markets are all different in different jurisdictions. (P13)

Stinchcombe (2001) has argued that a well-designed institution ought to have the capacity to correct itself and update the reality of what it seeks to do to regulate change. Our analysis suggests that this is difficult to achieve when reforms are enacted over multiple jurisdictions and levels of government. In such instances, because responsibilities are shared and the focus is placed on inter-governmental reorganizing, there may be limited space for institutional reflexivity. As a result, consequences of earlier decisions are likely to be missed and will become further entrenched as layering effects take hold.

Evaluation of changes to inherited policy regimes often focus on the level of coherence of policy goals and instruments. While the NDIS is an attempt at system-wide reform, policy makers were still forced to use existing institutional structures and practices alongside new policy instruments, to shift from a series of diverse, underfunded, relatively small programs to a national one. This included funding arrangements, as noted previously. Even though the scheme introduced individualized care packages, the transfer from the old system to the new meant in practice participants were offered the same (State supported) services within their packages. The inability to establish new service delivery models, despite the introduction of new funding mechanisms (i.e., individual care packages) means that, while NDIS policy goals have a high level of coherence, there is a much lower level of coherence in relation to policy instruments.

This combination of displacement combined with layering raises the question of the long-term sustainability of the reform. When analyzing the interplay of politics with institutional modes of change, Hacker (2011) and Mahoney and Thelen (2010) suggest that there are weak political veto possibilities in the context of displacement, but strong possibilities in the context of layering. In the case of the NDIS, consistent with this theory, there are very weak veto possibilities due to bipartisan support for the scheme. Despite this, we have still seen the emergence of layering—thereby challenging the dualism set up in the current literature. Patashnik (2008) argues that to overcome the tensions and difficulties of implementation requires continued political attention from the coalition that initially enacted the policy. However, he draws on the historical institutionalism aphorism—that policy begets politics—to describe the importance of changes in policy becoming institutionalized and shifting the political incentive structures in the implementation phase. Policy design is one possible source of stability in the face of novel information and ideas: mechanisms that may all come into play with implementation. For Patashnik (2008), the agreement of policy is only ever the start. Policy is not a one shot game but rather a repeated political game in which the inherent features of political life apply a policy that is made. This point is well supported by the existence of layering within a context of weak political veto in the case of the NDIS.

As a relatively rare transformative, system-wide policy change that continues to enjoy strong political support, those implementing the NDIS have more opportunity than most to break free of layering effects and “sticky” policy legacies. However, the implementation experiences of the NDIS to date suggest that layering is not merely an unfortunate by-product of previous institutional structures, or something that operates merely at the margins of policy. Rather, it can be actively sought and used by policy makers to tackle implementation challenges. At a system level, layering appeared as an emergent property of implementation. Interestingly, even though policy makers could, in some instances, reflexively identify that layering was taking place, they felt unable to redress it. As we noted at the outset of the article, currently there is contestation over whether layering occurs as a result of political strategy, an influential actor, or is an unintended outcome of change processes. The case of the NDIS suggests that layering can take place in the absence of political strategy and goals or strong institutional players. We found that layering can be an emergent property within implementation—sought out as a tool to handle institutional complexity, but uncontrollable once set in motion.