Designing multi-actor implementation: A mechanism-based approach

Abstract

The article offers analytical tools for designing multi-actor implementation processes. It does so by proposing a design approach centred on causal mechanisms. Such design strategy requires designers to focus primarily on causal theories explaining why implementers commit overtime to implementing policies. The central proposal is that design procedures should be reversed, i.e. start by reasoning on the causal mechanisms explaining implementers’ behaviour and then go looking for design features. Several advantages of this approach related to designing, reforming, or transferring successful practices are discussed throughout the article. Finally, the article provides six extended examples of such mechanisms in different policy fields: actor’s certification, blame avoidance, earning brownie points, repeated interactions, focusing events and attribution of opportunity or threat.

Keywords

Causal mechanisms implementation learning policy design

Introduction

Policy design was recognised quite early as a primary responsibility of the policy sciences (Dror, 1971). Notwithstanding promising beginnings (Alexander, 1982; Linder and Peters, 1984), however, policy design did not fare well in the academic agenda (Considine et al., 2014; Schneider and Ingram, 1988), and it was almost equated to the study of policy tools.

Recently, there has been a resurgence of scholarly interest into both ‘design thinking’ (Dorst, 2011) and policy design in particular (Howlett, 2014; Howlett et al., 2015). Such contributions have worked on understanding mixes of tools, their complementarity and interaction effects, and how tools should be calibrated to the overall governance arrangement and to existing policy programmes.

In both its past and recent developments, the tools literature helped the progress of design scholarship, but two areas still appear underdeveloped. First, design scholars have focused almost exclusively on the design of substantive policies, i.e. how to formulate programmes and arrange tools aimed at solving policy problems. In this respect, implementation, a fundamental component of policy success, has been left at the margin of the design debate. Second, the focus on tools concentrated scholars’ attention to “design as a noun,” to the detriment of “design as a process,” i.e. the art and craft of individual designers (Considine et al., 2014). Consequently, how to design (i.e., the analytical strategies of designers) has received limited attention. The present article attempts to fill these gaps by discussing the design of multi-actor implementation and proposing an analytical strategy based on causal mechanisms.

Designing implementation was a central concern among top-downers in the implementation literature (Bardach, 1977; Pressman and Wildavsky, 1973), it was part of design scholars’ pleas for ‘framing smarter statutes’ (Ingram and Schneider, 1990), and it was included as a central task for designers in recent contributions (Bobrow, 2006; May, 2012). When approaching a policy problem, in fact, one should certainly design the substantive policy, i.e. arrange a mix of tools aimed at resolving that problem, but this will not be sufficient for the policy’s success. The next section argues that most implementation processes are multi-actor, that those actors have fundamental resources for the success of the policy, and that they are best considered as autonomous players, often able to resist statutes and to pursue their own independent strategies. Such assumptions should not discourage designers, but rather be the initial steps of a realistic approach for designing implementation processes, one that values discretion at the bottom and does not equate implementation to compliance to statutes.

Concerning the ‘how to’ of the design process, the aim is to develop analytical tools for helping designers structure implementation. Designers face a limited ‘design space’ (Howlett and Mukherjee, 2014), and policies and programmes often come out of processes such as log-rolling, conflict, bargaining and so forth. However, policies (or parts of them) are also consciously designed, with policy-makers considering how to reach policy goals more effectively and efficiently. In all such cases, designers will use and need design strategies, i.e. analytical approaches to design (Dorst, 2011).

To take two notable examples, incrementalism (Braybrooke and Lindblom, 1970) is a design strategy cognizant of both information processing and political constraints (Bobrow and Dryzek, 1987), while forward and backward mapping (Elmore, 1985) is one combining top-down and bottom-up implementation tenets. Bobrow (2006) and Weimer (1992) review several other examples, following the assumption that, though policy design is a game, designers need ways to play more effectively (Bobrow, 2006). On the same line, the fourth section presents a design strategy for multi-actor implementation based on causal mechanisms. It consists of focusing on the causal theories explaining why implementers contribute to implementation and designing systems that trigger and support those mechanisms. This is preceded by a short section reviewing the literature on causal mechanisms.

Finally, the article proceeds with the fifth section, where six examples of such mechanisms are proposed through extended examples: actor’s certification, blame avoidance, earning brownie points, repeated interactions, focusing events, and attribution of opportunity or threat. The list is by no means complete, but picks examples from the literature to show the logic and utility of the approach.

The conclusive section discusses directions for future research, such as the prospects of expanding the number of mechanisms and of providing a classification of such mechanisms.

Dilemmas of cooperation in implementation processes

The missing link of implementation was recognised more than 40 years ago, and, since then, the literature in the field has grown continuously (Saetren, 2005). Such literature is vast and varied, but two common assumptions stand out. The first is that the formal adoption of a policy does not guarantee achieving its goals. The second is that in many, if not most, complex policies, successful implementation requires the continuous cooperation of a plurality of public and private actors, provided with their own goals, resources and agendas. In other words, good implementation is not ensured by simple compliance, and actors’ commitment needs to be extended over time. Such complexity is explained further in the following paragraphs.

The first point to note is that implementers are a heterogeneous class, made of both government and non-governmental organisations. Implementation scholars in the bottom-up tradition (Lipsky, 1983) and network theorists (Kickert et al., 1997), for instance, focused their analyses on implementing actors (instead of statutes) in order to account for the prominence of multi-actor interactions. More recently, the debate on governance in the 1990s (Stoker, 1998) pointed to the retreat of governments and the increased participation of several actors from different levels of governments and from the private and voluntary sectors. Typically, in multi-actor implementation networks, those actors will be relevant decision makers within public and private organisations, able to administer organisational resources towards or against policy implementation.

Second, one should not expect cooperative behaviour on the part of implementers. Actors in implementation networks, in fact, will have their own interests and agendas, not necessarily congruent with those of policy implementation. Politicians, bureaucrats and social and private actors participating in such networks will have different goals and preferences, and hence will be sensitive to different kinds of incentives. In this respect, models of policy-makers (e.g., Müller and Strøm, 1999) or bureaucrats (e.g., Niskanen, 1974) may provide examples of the diversity of motives that can characterise implementing actors.

Notwithstanding a possibly high variation of interests, however, one can expect that all actors in multi-actor implementation networks will strive to preserve their autonomy and hence resist external interference or changes in their organisational routines and prerogatives. In this respect, mutual dependency is likely to magnify the inherent political character of actors’ strategies, enhancing their natural inclination to develop a centrifugal trajectory and reassert their autonomy. This inclination can take the form of several ‘implementation games’, based on not cooperating fully and/or of re-interpreting their role in a way more consistent with their own agenda than to policy statutes and successful implementation. In other words, instead of assuming the natural emergence of cooperation, designers should expect a natural degree of resistance or disinterest and work out their designs to contrast such tendencies.

Third, the efficacy of hierarchical directions and market incentives should not be overemphasized. As mentioned above, most social, private, political and bureaucratic actors have autonomous legitimacies and independent resources and are difficult to force towards cooperation through binding contracts or the promise of economic gains. This is true for agencies within government, as well as for implementers outside the public sector. Even when hierarchy may appear to prevail, as in the case of bureaucracies, nominal contributions to programmes, formal applications and tokenism are always available to resistant actors, who can nonetheless appear to be compliant with the statute and the design of the policy (Bardach, 1977).

This is not to say that market and hierarchical incentives will never work or will never be available. Rather, the point is that a wider array of causal theories is at play and worth investigating for explaining cooperative behaviour by implementers. How to investigate such theories and use them for designing policies is the object of the next two sections.

Causal mechanisms: A short review

There is a growing body of literature on mechanisms encompassing several disciplines, such as sociology (Elster, 1989; Hedström and Swedberg, 1998), behavioural economics (Bowles and Gintis, 2011), political science (Mayntz, 2004) and public administration and management (Bardach, 2004; Barzelay, 2007). In political science, the concept became popular thanks to the methodological debate triggered by King et al. (1994). Their predominant quantitative approach made methodologists react by claiming the importance of qualitative studies, as a way to investigate causal mechanisms and limit the risk of spurious relations or uncertain connections (Brady and Collier, 2010).

The concept is now popular and many definitions have been offered (see Gerring, 2010; Mahoney, 2003), all pointing to an in-depth understanding of the causal process by which outcomes are produced. To the use of designers, three points are worth mentioning that help elaborate a ‘mechanistic’ approach to design.

First, mechanisms are causal theories specifying how a certain phenomenon produces an outcome. In its simplicity, this definition entails a specific notion of causation, in which causality is not assumed by observing only the regular connection of phenomena. Scholars studying mechanisms want to face causality upfront, by making hypotheses on causal connections, and searching for empirical tests for such connections.

To take an example, Pawson’s (2002) investigation of Megan’s Law shows how even simple designs make strong implicit assumptions on causal mechanisms. US law prescribes sex offenders to register and make their presence known in the neighbourhood. Common sense suggests that the system increases control for offenders and restricts their freedom of action. Attention to mechanisms would ask why such a connection should hold in the first place by explicitly formulating and testing the theories behind the success of the programme: public disclosure, sanction instigation and offenders’ responses.

Second, mechanisms entail ‘composite’ explanations. Taking a composite notion recalls the lay meaning of mechanisms, which suggests a system made of different parts working together (Fagan, 2012).

In assembling the components of their causal mechanisms, designers may certainly tap on the rich contributions present in the implementation literature. The field has proposed hundreds of independent variables (O’Toole, 1986), that have undergone some attempts at systematisation (e.g., Sabatier, 1986). For the present purpose, implementation theories will entail discussing the causal power of design features (why they should produce the outcome), the response of target implementers (why certain implementers will respond with cooperative behaviour) and the causal power of contextual conditions. Hypotheses on causal mechanisms would be ‘partial theories’ (Winter, 2003), explaining how variables are combined into causal configurations explaining successful implementation.

To take one example, imagine designing a monitoring body for assisting the implementation of a new policy. Referring to the mechanism of “blame avoidance” or “naming and shaming” without further analysis only provides an appealing label for the actions of the monitoring body and the effects of evaluation reports. A satisfactory account of the mechanism should provide a theory explaining the interplay among the actions of the monitoring body (its ability to identify and disclose bad behaviours), those of the public or group of peers (provided or not with sanctioning capacity based on blame), and the response of implementers (who may be shame-sensitive or insensitive).

Such composite character is certainly necessary to avoid tautological or circular explanations, such as that the monitoring body worked in enhancing cooperation thanks to its effective monitoring. More fundamentally, it helps avoid overconfidence in the causal power of design, adjust for changing contextual conditions (e.g., shame-sensitivity of implementers), and hence increase the likelihood of successful transfer.

Third, causal mechanisms need to be formulated at a middle level of abstraction. This qualification is not a purely ontological curiosity on the nature of mechanisms, but responds to a pragmatic aim. At a zero level of abstraction, one may investigate why in a single case the monitoring body helped implementing the policy. Such research may end up as a rich narrative, outlining all details of the implementation process and building a fully satisfying causal account of that case. However, to be useful for designers, the research must climb the ladder of abstraction and build a causal model of the programme (Rose, 1993). This not only means abstracting case materials to more general causal theories, but also “cleaning” the case to understand what is causally necessary and what is contingent to the case (Sayer, 2000). Designing by analogy, in fact, entails abstracting some properties of interest and leave back those that are not (Cross, 2006).

A mechanism-based approach to multi-actor implementation

Although attention to policy tools has been prominent in the literature, the importance of causal mechanisms is not new to policy designers. Schneider and Ingram (1988) proposed a comparative approach to study designs, aimed at dissecting policies and investigating their structure. Among the recurrent structural elements of any design, the authors mentioned the theories explaining why and how tools would produce the desired results. More explicitly, in his studies on lesson drawing, Rose (1993) included the construction of a model of the programme as a preliminary step to learn from practices and transfer lessons. Such a model should not be a description of the design of the programme or a checklist of programme requirements but rather an abstract cause-and-effect model specifying the causal relationships between programme elements and its effects. In a similar vein, Bardach (2004) and Barzelay (2007) have called attention to the problem of the atheoretical study of best practices (Overman and Boyd, 1994). Studying successful cases through best practices has the advantage of moving on from the typical focus on implementation failures (the ‘misery research’ as Meier (1999) defined implementation). However, the practices were usually described in their design elements, without investigating the causal relationships making them work and thereby hindering learning.

The common point in such positions is that an effective design description cannot be obtained by reporting only design features and tools, but must also investigate the causal mechanisms triggered by those features and tools. For instance, it is not sufficient to know that a programme for preventing illegal dumping worked in that city and then investigate details of the design features of such a programme. Instead, one should focus on the actors responsible for the success in implementing the programme (e.g., the waste service provider, the local police, and/or the environmental department in the municipality) and investigate why they responded positively to certain programme features (e.g., a new reporting procedure).

In analysing successful implementation processes, the first step would be to uncover and postulate causal mechanisms explaining the interplay between design, context and implementers. Such an approach will aid designers in three ways related to: the understanding of success in implementation processes, the adjustment of existing designs to new cases, and the search for cases from which to import successful practices.

Concerning the first point, programmes are often made of complex mixes of tools. In the example, a programme aimed at preventing illegal dumping may include control systems on selected sites, public campaigns, education initiatives, community involvement schemes and so forth. Hence, if a drop in illegal dumping is observed, it is not easy to know which parts of the programme had an effect. Instead, having a notion of why the implementing actors (e.g., public officials) responded positively to the programme (e.g., issue salience making citizens press officials) may give clues regarding those design features responsible for the change in behaviour (e.g., public campaigns and community involvement schemes).

By illuminating which design features are causally relevant and why, knowledge of causal mechanisms has important implications for designers. In fact, it spares designers from copying everything from a successful programme (including useless design features), and it limits the risk of disregarding causally relevant features. Further, working on causal mechanisms helps designers understand the effect of design details, discriminate across similar designs and adjust designs to reinforce the desired causal relationship. Similar programmes or design features, in fact, may include minor differences whose causal importance becomes clear only when the causal mechanism is made explicit.

Concerning the second advantage, the same programme often produces different results across space and time. In other words, designed features are not always sufficient for producing the wanted effects, since other causal drivers affect programme implementation. The implementation literature is full of such examples. To take one, when Sabatier (1986) synthesized the top-down and bottom-up approaches to implementation, he mentioned a number of non-statutory variables, such as socio-economic conditions and technology, media attention to the problem, public support, resources of constituency groups, commitment and leadership skills of implementing officials and so forth. Such variables are likely to vary across cases, so that a design that was successful elsewhere would not necessarily work in the target case.

In such cases, designers might not surrender to context dependency. To go back to the example of illegal dumping, if in the target case, dumpsites are remotely located from residences and not accessible by citizens, the effect of public campaigns for raising salience may be poor. However, if designers have an abstract causal model of the successful practice, they can look for functional equivalents (Rose, 1993) that are able to trigger the same mechanisms. In the example, one may find other ways to raise salience for citizens or other ways to pressure public officials. There are many possible configurations for adjusting designs, but the general point is that by focusing on the causal theories producing successful outcomes, designers may have a guide to rework their designs and to redesign not a copy of the exemplar, but a functionally equivalent system.

Finally, by starting with the causal mechanisms instead of the design features, the pool of practices from which to pick ideas for new designs will multiply. Learning from second-hand experiences (Barzelay, 2007) through ‘pinching’ (Schneider and Ingram, 1988), ‘borrowing’ (Bobrow, 2006), or ‘tinkering’ (Weimer, 1992) is a typical strategy for designers. The design process is, in fact, not only a matter of creativity, but also one of information retrieval, selection and adjustment (Alexander, 1982; Lawson, 2006). In this respect, by focusing on the causal mechanisms explaining implementers’ cooperation, the search for exemplars need not be limited to the same (or similar) policy sector. As an example, suggestions for setting up a reputation system for hospitals may well come from analysing star ratings of restaurants, hotel rankings, or other sectors using the same causal theory (Pawson, 2006: 151). Such an advantage is likely to be even more prominent when designing implementation, since implementation processes are likely to vary less than the programme theories informing the design of the substantive policy.

The whole discussion is not to deny that designers never follow causal theories when they approach the implementation of a policy. However, such theories need to be explicit, work as the cornerstones of their designs, and not assume the natural emergence of cooperation.

Illustrating causal mechanisms, or, the beginning of a catalogue

This section provides six examples of causal mechanisms that – sometimes by other labels – can be found in previous literature. These are actor’s certification (Barzelay, 2007; McAdam et al., 2001; Ongaro, 2006), blame avoidance (Hood, 2010), earning brownie points (Coletti, 2013), repeated interactions (Dente and Goria, 2004; Ostrom, 1990) focusing events (Birkland, 1998; Kingdon, 1984), and attribution of opportunity or threat (Barzelay, 2007; McAdam et al., 2001; Ongaro, 2006; Tilly, 2001). In addition, some of them were the object of several research projects (DG IPOL, 2014; ESPON and Politecnico di Milano, 2013; IRS and IGOP, 2011) that – although they did not add up to a unitary research programme – were part of a consistent commitment towards inquiring the use of mechanisms in policy analysis.

As mentioned, the approach has its roots in best practice research (Overman and Boyd, 1994), evidence-based policy (Pawson, 2002) and policy learning (Bardach, 2004; Barzelay, 2007; Rose, 1993). In particular, the last strand of literature explicitly claimed the centrality of causal mechanisms and causal configurations for the analysis and replication of exemplars. This intuition suggested a two-step strategy for (re)design inspired by ‘reverse engineering’ techniques.

Reverse engineering is a technology of ‘reinventing’, i.e. reproducing artefacts by analysing working exemplars (Wang, 2010). Contrary to invention, where new ideas are transformed into designs and then implemented, reverse engineering starts with an already implemented artefact and go backwards to decode its underlying working. In sum, the process requires two main steps: (1) unpack working systems into their constituent elements and causal relations; and (2) design new systems triggering and supporting equivalent causal processes. The six mechanisms presented here provide good examples of such two-step strategy, by presenting case exemplars, abstracting from the case to identify a mechanism, and then providing suggestions for redesign.

Concerning the first step, one has to investigate why in the source-case implementers behave in such a way as to ensure success, by exploring which design and/or contextual features trigger and support such behaviour. To this end, all six mechanisms are presented by abstracting the configuration of causal factors explaining the outcome. Although some labels may suggest the prominence of one component in the configuration (e.g., context features for ‘focusing events’ or design features for ‘earning brownie points’), no example in the list should be thought of as dependent on the causal effect of one variable only. The outcome is in fact dependent on the interaction of multiple factors and all examples attempt to elucidate those multiple components, highlighting background and context conditions, design features and implementers’ reactions.

Turning to the second step, the analyst needs understand if and how the source causal configuration can be transferred to a target case. This step requires a transformational effort, by reproducing not necessarily the same design or context features, but ones that will deploy the same causal power in a possibly different case. While not all mechanisms in the literature might be reproduced, the six mechanisms presented are all suitable to extrapolation, redesign and transfer. In sum, although they show different degrees of transferability (e.g., certification vs. focusing events), they are proposed here because their causal process can be (re)designed in all cases.

Before presenting the six mechanisms, one caveat is in order. The six examples are not exemplifications of classes of mechanisms and do not represent typical categories. Granted, grounding the catalogue in a classification or typology of mechanisms would provide more solid analytical foundations and a more powerful analytical tool. Such search for an organising criterion for the catalogue goes beyond the scope of the present article and is left for further research.¹

Cognisant of this limit, providing a preliminary list is a first step towards increasing designers’ attention to mechanisms. Were such a catalogue increased, designers would have nonetheless an inventory of solutions that would provide a useful add to their analytical toolkit.

Actor’s certification

When certification takes place, implementers adopt cooperative behaviour because the actor mandating implementation receives endorsement by another actor. The certifying actor needs to be considered particularly worthy, thanks to intellectual prestige, charisma, reputation, past achievements, political standing and so forth. Figure 1 provides a diagram outlining the mechanism, which is based on the production and uses of and reactions to endorsement. As depicted, the relevant components include a source of certification (either a contextual or designed feature), the action of certification (a design feature), and the use of certification and sensitivity to it (the reaction of implementers).

Figure 1.

Actor’s certification.

The mechanism was identified by McAdam et al. (2001: 122) to explain different trajectories of mobilization events, showing the importance of, for instance, Cardinal Sin’s endorsement of the protest movement against Marcos. A closer example to our aim is Barzelay’s account of how programme directors of ‘Brazil in Action’ were aided in implementing the programme by the fact that the President of Brazil considered it a priority. The ‘shadow’ of the President’s preferences certified the requests of directors, thereby enhancing cooperation (Barzelay, 2007).

In a research on the development of institutional capacity, the success of the Evaluation Unit in the Italian region of Puglia was attributed to the action of a similar mechanism (ESPON and Politecnico di Milano, 2013). The Unit was set up in 2002, following law 144/99, which mandated that all regions and ministries set up Evaluation Units. The Unit was in charge of evaluating public investments and was located within the regional administration. In Puglia, it was composed of independent experts, coupling both technical expertise and third-party independence. In-depth interviews across the administration revealed that assessments and evaluations of the Unit were considered highly valuable, and were used by project proponents to prove the technical and financial validity of their project. By assuring the quality of the proposal, the Evaluation Unit triggered a mechanism of certification, increasing the resources of proponents, decreasing those of opponents and easing implementation interactions.

Interestingly, most evaluation units across Italian regions were not so successful. Apparently, the success of the Unit in Puglia was due to its capacity to promote its credibility as both an independent body and as the holder of technical knowledge. Less successful Evaluation Units did not have the same credibility or independency, so that their reports and opinions had no certifying powers and were not used a strategic resource by leading implementers.

Cross-regional comparison gives a clear example of how similar designs may provide different results depending on the activation or not of a certain causal mechanisms. Certification is one possibility to support leading implementers. Designers should search for a source of certification in their policy process and build a system of interactions where leading implementers can use such certification to their advantage, by increasing their resources and the likelihood of cooperation by the needed actors.

Blame avoidance

In this case, cooperative behaviour is adopted because defection would be considered inappropriate and would be sanctioned by peers. Many systems of horizontal reputation are characterized by such a mechanism. As already mentioned, for blame avoidance to arise, there is a need for an agreement on good behaviour (a designed or context feature), a disclosure system (a designed feature), and the attribution of blame by peers and blame sensitivity by implementers (the reaction of implementers). Figure 2 provides a diagrammatic representation of the mechanism.

Figure 2.

Blame avoidance.

Melloni (2013) analyses the implementation of the EU Impact Assessment, uncovering such a blame avoidance dynamic. Several design features in the policy contribute to producing the mechanism by enhancing crosschecks and horizontal competition among peers. In particular, an active monitoring body and several procedures enhancing transparency increase the likelihood of public disclosure of bad behaviours. In addition, cross-sectoral steering groups institutionalize peer-review.

Curiously, these features are common delivery systems, but they are not always conducive to success. Their effect in the EU administration has to do with the fact that they impacted a bureaucracy where technical recognition is a relevant point of distinction. On the basis of in-depth interviews, Melloni (2013) reports the importance attached to ‘having something to say’ in cross-sectoral meetings as a way to signal technical credibility.

Activating such a mechanism can be relevant in all cases where substantial compliance can be difficult to obtain, ‘tokenism’ is hard to detect, and collusive behaviour between peers is a likely scenario. Designers would need to build a blame system, understand what implementers are sensitive to, and enhance public disclosure and peer-review. Notice, however, that it is highly likely that context will play an important role in supporting such blame system.

Earning brownie points

When the mechanism is activated, cooperative behaviour is adopted because implementers hope to improve their own position in a parallel or subsequent process. The leading actor can induce cooperation because he/she is the holder of resources that are of interest to implementers for reasons external to the policy. Implementers will play fair and cooperate (even though they could do otherwise) in the hope that the leading actor will treat them fairly in subsequent interactions.

Figure 3 presents the elements triggering and supporting the mechanism. The relevant components include an unequal distribution of resources in two parallel processes (a designed or context feature), symmetrical interests in those processes (a characteristic of the implementers) and the coupling of the two processes (a designed feature).

Figure 3.

Earning brownie points.

This mechanism was suggested by another example of ‘smart regulation policy,’ i.e. the implementation of the Standard Cost Model (SCM) in the Netherlands (Coletti, 2013). Here, the main design feature used to induce cooperation is the coupling of the smart regulation agenda with the budget cycle. The Regulatory Reform Group (RRG), the Unit responsible for monitoring the implementation of the policy, is located within the Ministry of Finance, and the whole process of decreasing regulatory costs runs parallel with the budget process.

Before budget negotiations begin, the RRG monitors all Ministries, and, if discrepancies between reduction targets and achievements are found, the Minister of Finance is informed. A process of negotiation with the RRG begins, and, if no agreement is reached, the non-compliant Minister will need to talk directly to the Minister of Finance when the preparatory negotiations to the budget begin. The Minister of Finance always attends budget negotiations together with the RRG director and the political appointees in charge of ‘smart regulation.’ Not only there is a clear incentive to settle things before being obliged to discuss them with the Minister of Finance, but also, all actors are reminded that the policy is on the front line for the government. Hence, non-compliant Ministers make all efforts to settle problems before budget negotiations, and all Ministers perceive their performances in reducing regulation costs as a trump card for negotiations on the budget.

Notice that this is not a direct exchange, a pure hierarchical direction, or a shadow of hierarchy. In addition, it is a much more generalizable mechanism than ‘blame avoidance,’ since there is no need for implementers to be blame-sensitive. Designers wanting to activate ‘earning brownie points’ will need to connect the policy of interest (easily escapable and not at the core of implementers’ interests) to a parallel process for which implementers would strive to earn brownie points.

Repeated interactions

Here, cooperative behaviour is adopted because implementers learn to value relations, and the costs of defecting become prohibitive. Repeated interactions may have all sorts of good or bad consequences; the paradigmatic example of the former is the solution of the prisoner’s dilemma, and the latter is represented by collusion. Although apparently simple, the mechanism depends on the interplay of several components related to both context features increasing interdependence (i.e., problems’ interrelatedness and the unavailability of exit options) and design features increasing interactions (i.e., shared responsibility and frequency of meetings). These are exemplified in Figure 4.

Figure 4.

Repeated interactions.

A positive example of the effects of repeated interactions is that of Lake Idro in Lombardy (Dente and Goria, 2004). The lake basin was regulated artificially for using water in energy production and irrigation, but it was also an important tourism resource. Until the end of the 1980s, tourism and environmental interests were not included in the governance of the lake, and the farmers and the energy company had negotiated a rule permitting a range of about seven meters of variation for water levels.

Clearly, such a rule was opposed by the tourism industry and by environmentalists. When water concessions to the farmers expired and the rules governing water use had to be readjusted, a change of governance happened with the establishment of a ‘Users Committee,’ including not only the farmers and the energy company but also the lakefront municipalities, the environmentalists and the provincial governments. Instead of agreeing on a fixed rule, the committee decided to leave flexibility in deciding water levels, hence permitting close adjustments to the effects of raining and temperature. The rule was to be rearranged and agreed on periodically by the Users Committee. In summer, which was the most critical period, the Committee met every 10 days to fine tune water levels among users.

Such extreme frequency of interactions, together with problem interrelatedness and mutual dependency, made defection too costly and produced a mutual recognition of users’ conflicting (but interdependent) interests. Not surprisingly, the experimental period of three years for the flexible rule was extended further once expired.

The point for designers would be to build a system of forced interaction, where switching to unilateral action would be too costly and where actors could learn the benefits of cooperation, hence neutralizing temptations to free riding.

Focusing event

When this mechanism is triggered, cooperative behaviour is adopted because certain characteristics of an event (e.g., sudden, uncommon, harmful, and known to the public) push actors to focus on the issue, mobilize and solve the problem with a non-routine rapidity and a high level of cooperation (Birkland, 1998). Although ‘focusing events’ are typically linked to theories of agenda and policy change (Kingdon, 1984), they may have some applications to the case of implementation. In fact, non-standard activities or tasks may be a good terrain for triggering a similar mechanism (think, for instance, to the implementation of infrastructures for mega-events).

The diagrammatic illustration follows in Figure 5. The relevant components include mainly design features that enhance novelty, the non-routine character of the programme, or the importance of solving the collective problem addressed. These features would have a ‘focusing’ effect on implementers, by producing a reaction of enhanced commitment.

Figure 5.

Focusing events.

In-depth interviews within the administration of the Toscana Region in Italy revealed that the implementation of Integrated Urban Projects for Sustainable Development (PIUSS) profited from a similar mechanism (ESPON and Politecnico di Milano, 2013). Toscana was new to European Regional Development Fund (ERDF) projects, and working on them was considered a non-standard activity that attracted the interests and efforts of implementers, producing a higher level of cooperation.

To take another example, although without mentioning such a mechanism explicitly, when Barzelay (2007) describes ‘Brazil in Action,’ he refers repeatedly to the importance of media attention and the perception that the programme was a non-routine activity.

If designers want to activate the mechanism, they should work to increase the focal character of the programme by treating it as non-standard, increasing its visibility, and communicating a sense of necessity (i.e., making the programme a focusing event). This is a dynamic typically found when implementing pilot-projects and, clearly, has the same problems with maintaining the focal character once programme institutionalization comes in.

Attribution of opportunity (or threat)

Cooperative behaviour is adopted because the good implementation of the policy provides benefits to the policy actors. This is, possibly, a simple incentive mechanism, where design features enlarge the opportunity set of implementers, and implementers respond accordingly to the incentive.

The mechanism is portrayed in Figure 6. It is composed of both design and context features that link implementation success to a perception of opportunity by implementers.

Figure 6.

Attribution of opportunity or threat.

McAdam et al. (2001) refer to attribution of opportunity as a basic mechanism for mobilizing previously inert populations. From contentious politics, the mechanism can be imported to the case of local development projects funded by ERDF. In such projects, the engagement of local actors is both a precondition and a goal in itself and hence should be at the core of designers’ interests.

An elementary theory of actors’ engagement would suggest that, once European funds land on a territory, they are seen as an opportunity that will make actors cooperate naturally to use the funds. Stories of implementation failures, local conflicts and non-spent funds discredit such a simple picture, however.

One positive example is given by the case of the URBANA programme in Andalucía (IRS and IGOP, 2011). Here, interviewees reported that competitive calls for approving fundable projects produced unprecedented engagement by local actors. The beginning of the crisis and fear that 2007–2013 was going to be the last funding period for the region increased the sense of opportunity attached to getting the funds. In addition, selection of local projects not at the regional level but rather by the national government limited the importance of consolidated political ties and produced true selectivity instead of dynamics of political distribution. All these factors increased the sense of challenge and the importance attributed to winning the competition, so that municipalities worked hard to increase the quality of their projects.

This is a good example of how common delivery systems, such as, for instance, competitive allocation of funds, may embody complex causal theories, supported by both contextual and design features. This is also a good reminder of how an implicit and common sense causal theory (i.e., equating funds to incentives and opportunities) will not hold empirically. In such cases, the work for designers will be to inquire about what could be considered an opportunity for different classes of implementers and enhance that opportunity through the provision of triggering and supporting design features.

Conclusion: Mechanism-based design as a practice-oriented research programme

The article discusses the importance of designing the implementation process and proposes an approach to design based on causal mechanisms. The examples above have shown how to investigate designs by elaborating causal mechanisms on implementers’ behaviours, looking at a composite array of variables interacting together to trigger and sustain actors’ cooperation. The illustration provides material for discussing the utility, methods and relevance of a research programme on mechanism-based design.

First, a preliminary question regards if – and when – investigating mechanisms is worthwhile. The pragmatic answer is: not necessarily, not always. In particular, the lower the complexity of the programme, the lower the need to uncover mechanisms. As an example, if a municipality increases the costs of using private cars to reach the city centre (by, for instance, introducing a congestion charge), no great in-depth analysis on causal mechanisms is needed to understand subsequent improvements in the use of public transport. In this case, the underlying theory is linear, and implementation is reduced to a choice between alternatives by car users (i.e., spending more or switching to public transport). A theory of compliance for such a policy will boil down to demand elasticity to the price of using private cars, and if a drop in car use does not materialize, it is probably due to miscalculations in assessing elasticity.

In short, if the programme is simple and prior confidence of its underlying causal theory is high, there is no need to investigate causal mechanisms. Notice, however, that not only are many policies complex but also, such complexity is subtle and sometimes hard to detect. Among the examples above, the implementation of two similar tools such as the impact assessment in the EU and the standard cost model in the Netherlands would have been considered “simple,” since they regard individual organizations and are provided with rules for enforcing behaviours. However, the mechanisms explaining their success are not only different in such two similar cases, but also anything but trivial.

Second, designers should elaborate and test hypotheses on causal mechanisms that are non-tautological and practice-oriented. In this effort, choosing the right level of abstraction and taking into account the composite character of mechanisms are fundamental requirements. The case of ‘attribution of opportunity’ described above is a clear example of a possibly correct theory, which, if too abstract, may be of scant use for designers. One would need to qualify what an opportunity means for different classes of implementers (hence, reducing abstraction) and to exemplify, at a minimum, how opportunities are created and why implementers respond to the attribution of opportunity.

Third, research on mechanisms is needed in order to understand and transfer programme success (and not necessarily programme features). Taking ‘earning brownie points’ as an example, it is only by focusing on the causal understanding of the exemplar that one can have a guide to identify relevant drivers for success (e.g., the coupling of two processes and so forth). In addition, only by understanding the theory underlying cooperation in the exemplar (i.e., earning brownie points) can one transfer and adjust that design to other settings. If, for instance, the Ministry of finance in the target case does not have the same power as in the Netherlands, since budgeting is mostly determined by formula funding, designers should find another process to couple to reinforce the implementation of the policy. In this transformative process, designers would come up with a different design that nonetheless will embody the same causal mechanism. The design can therefore very well be formally different but nonetheless causally identical.

Finally, drafting a catalogue of mechanisms may be a useful complement to the toolkit of designers. As typologies and lists of policy tools have helped designers’ reasoning on classes of instruments available for solving policy problems, a catalogue of mechanisms would provide a set of causal theories explaining how to enhance commitment by implementers. As in the examples above, such a catalogue will include causal mechanisms at a middle level of abstraction, describing designed triggers, implementers’ reactions, and the reasons explaining why implementers cooperate. The preliminary “taste” offered in the preceding section highlights that a great variety of theories is potentially available for enhancing cooperation.

A first step in going ahead with a research programme on mechanismic design will therefore be to increase the number of mechanisms included in the catalogue, beyond those listed above. To take one example, ‘bandwagon’ is a recurrent mechanism in the literature that one can imagine as relevant to multi-actor implementation. Furthermore, for each mechanism one could think of listing the different programme features as well as the contextual elements that, in different cases, were able to trigger the mechanism. Finally, as mentioned, one may work towards the definition of classes or typologies of mechanisms. This last avenue is certainly promising, since it would give designers a systematic overview of classes of mechanisms potentially available for affecting behaviours. However, the research is in its early stage, and a preliminary expansion of the list is necessary before meaningful classifications can be built.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Note

References

Alexander

(1982) Design in the decision making process. Policy Sciences 14(3): 279–292.

Bardach

(1977) The Implementation Game: What Happens after a Bill Becomes a Law, Cambridge, MA: MIT Press.

Bardach

(2004) Presidential address – The extrapolation problem: How can we learn from the experience of others? Journal of Policy Analysis and Management 23: 205–220.

Barzelay

(2007) Learning from second-hand experience: Methodology for extrapolation-oriented case research. Governance 20: 521–543.

Birkland

(1998) Focusing events, mobilization, and agenda setting. Journal of Public Policy 18: 53–74.

Bobrow

(2006) Policy design: Ubiquitous, necessary, and difficult. In: Peters

Pierre

(eds) Handbook of Public Policy, Beverly Hills, CA: Sage, pp. 75–96.

Bobrow DB and Dryzek JS (1987) Policy Analysis by Design. Pittsburgh: University of Pittsburgh Press.

Bowles

Gintis

(2011) A Cooperative Species: Human Reciprocity and its Evolution, Princeton, NJ: Princeton University Press.

Brady HE and Collier D (eds) (2010) Rethinking Social Inquiry: Diverse Tools, Shared Standards. Lanham: Rowman & Littlefield Publishers.

10.

Braybrooke

Lindblom

(1970) A strategy of Decision, London: Macmillan Publishing Company.

11.

Coleman

(1990) Foundations of Social Theory, Cambridge: Harvard University Press.

12.

Coletti

(2013) Evidence for Public Policy Design, London: Palgrave Macmillan.

13.

Considine

Alexander

Lewis

(2014) Policy design as craft: Teasing out policy design expertise using a semi-experimental approach. Policy Sciences 47(3): 209–225.

14.

Cross

(2006) Designerly Ways of Knowing, London: Springer.

15.

Dente

Goria

(2004) Competing integration principles in a decentralising state. In: Bressers

Kuks

(eds) Integrated Governance and Water Basin Management, New York: Springer, pp. 189–212.

16.

DG IPOL Directorate-General Internal Policies of the Union (2014) Gender Mainstreaming in Committees and Delegations of the European Parliament. Available at: http://www.europarl.europa.eu/RegData/etudes/etudes/join/2014/493051/IPOL-FEMM_ET(2014)493051_EN.pdf (accessed 22 November 2016).

17.

Dorst

(2011) The core of design thinking and its applications. Design Studies 32(6): 521–532.

18.

Dror

(1971) Design for Policy Sciences, Atlanta, GA: American Elsevier Publishing Company.

19.

Elmore

(1985) Forward and Backward Mapping: Reversible Logic in the Analysis of Public Policy, New York, NY: Springer.

20.

Elster

(1989) Nuts and Bolts for the Social Sciences, Cambridge: Cambridge University Press.

21.

ESPON and Politecnico di Milano (2013) Smart-Ist: Smart Institutions for Territorial Development, Final Report. Available at: http://www.espon.eu/main/Menu_Projects/Menu_TargetedAnalyses/smart-ist.html (accessed 22 November 2016).

22.

Fagan

(2012) The joint account of mechanistic explanation. Philosophy of Science 4(79): 448–472.

23.

Gero JS (1990) Design prototypes: A knowledge representation schema for design. AI Magazine 11(4): 26–36.

24.

Gerring

(2010) Causal mechanisms: Yes, but…. Comparative Political Studies 43: 1499–1526.

25.

Hedström

Swedberg

(1998) Social Mechanisms: An Analytical Approach to Social Theory, Cambridge: Cambridge University Press.

26.

Hood C (2010) The Blame Game: Spin, Bureaucracy, and Self-preservation in Government. Princeton: Princeton University Press.

27.

Howlett

(2014) From the ‘old’ to the ‘new’ policy design: Design thinking beyond markets and collaborative governance. Policy Sciences 47(3): 187–207.

28.

Howlett M and Mukherjee I (2014) Policy design and non-design: Towards a spectrum of policy formulation types. Politics and Governance 2(2): 57–71.

29.

Howlett

Mukherjee

Rayner

(2015) Components of public programs and effective program design. In: Perry

Christensen

(eds) Handbook of Public Administration, San Francisco, CA: Jossey Bass.

30.

Ingram

Schneider

(1990) Improving implementation through framing smarter statutes. Journal of Public Policy 10(1): 67–88.

31.

IRS Istituto per la Ricerca Sociale and IGOP Institut de Govern y Polítiques Públiques (2011) Study on the Contribution of Local Development in Delivering Interventions Co-financed by the European Regional Development Fund (ERDF) in the Periods 2000–06 and 2007–13. Final Report. Available at: http://ec.europa.eu/regional_policy/sources/docgener/evaluation/pdf/eval2007/local_dev_interim.pdf (accessed 22 November 2016).

32.

Kickert

WJM

Klijn

Koppenjan

JFM

(1997) Managing Complex Networks, London: Sage.

33.

King

Keohane

Verba

(1994) Designing Social Inquiry: Scientific Inference in Qualitative Research, Princeton, NJ: Princeton University Press.

34.

Kingdon

(1984) Agendas, Alternatives, and Public Policies, London: Longman Classics.

35.

Lawson

(2006) How Designers Think: The Design Process Demystified, London: Routledge.

36.

Linder

Peters

(1984) From social theory top policy design. Journal of Public Policy 4: 237–249.

37.

Lipsky

(1983) Street-Level Bureaucracy: The Dilemmas of the Individual in Public Service, London: Sage.

38.

McAdam

Tarrow

Tilly

(2001) Dynamics of Contention, Cambridge: Cambridge University Press.

39.

Mahoney

(2003) Tentative answers to questions about causal mechanisms. Annual Meeting of the American Political Science Association. Available at: http://ciece.com.ar/ciece/wp-content/uploads/Mahoney-James-Tentative-Answers-to-Questions-about-Causal-Mechanisms.pdf (accessed 22 November 2016).

40.

May

(2012) Policy design and implementation. In: Guy Peters

Pierre

(eds) The Sage Handbook of Public Administration, London: Sage, pp. 279–291.

41.

Mayntz

(2004) Mechanisms in the analysis of social macro-phenomena. Philosophy of the Social Sciences 34: 237–259.

42.

Meier

(1999) Are we sure lasswell did lt this way? Policy Currents 9(1): 5–8.

43.

Melloni E (2013) Ten years of European impact assessment: How it works, for what and for whom. World Political Science 9(1): 263–290.

44.

Müller

Strøm

(1999) Policy, Office, or Votes?, Cambridge: Cambridge University Press.

45.

Niskanen

(1974) Bureaucracy and Representative Government, Piscataway: Transaction Publishers.

46.

Ongaro

(2006) The dynamics of devolution processes in legalistic countries: Organisational change in the Italian public sector. Public Administration 84(3): 737–770.

47.

Ostrom

(1990) Governing the Commons, Cambridge: Cambridge University Press.

48.

O’Toole

(1986) Policy recommendations for multi-actor implementation. Journal of Public Policy 6(2): 181–210.

49.

Overman

Boyd

(1994) Best practice research and bureaucratic reform. Journal of Public Administration Research and Theory 4: 67–83.

50.

Pawson

(2002) Does Megan’s Law Work?, London: ESRC UK Centre for Evidence Based Policy and Practice.

51.

Pawson

(2006) Evidence-based Policy: A Realist Perspective, London: Sage.

52.

Pressman

Wildavsky

(1973) Implementation, Oakland: University of California Press.

53.

Rose

(1993) Lesson-Drawing in Public Policy, London: Chatham House.

54.

Sabatier

(1986) Top-down and bottom-up approaches to implementation research: A critical analysis and suggested synthesis. Journal of Public Policy 6: 21–48.

55.

Saetren H (2005) Facts and myths about research on public policy implementation: Out-of-fashion, allegedly dead, but still very much alive and relevant. Policy Studies Journal 33(4): 559–582.

56.

Sayer

(2000) Realism and Social Science, London: Sage.

57.

Schneider

Ingram

(1988) Systematically pinching ideas: A comparative approach to policy design. Journal of Public Policy 8(1): 61–80.

58.

Stoker

(1998) Governance as theory: Five propositions. International Social Science Journal 50: 17–28.

59.

Tilly C (2001) Mechanisms in political processes. Annual Review of Political Science 4(1): 21–41.

60.

Wang

(2010) Reverse Engineering, Boca Raton: CRC Press.

61.

Weimer

(1992) Claiming races, broiler contracts, heresthetics, and habits: Ten concepts for policy design. Policy Sciences 25(2): 135–159.

62.

Winter

(2003) Implementation perspectives: Status and reconsideration. In: Peter

Pierre

(eds) The Sage Handbook of Public Administration, Thousand Oaks, CA: Sage, pp. 212–222.