1.1 Behavioural Insights as a Policy ‘Toolkit’

Behavioural insights (BI) are becoming increasingly omnipresent in the public policy and administration landscapes of many countries around the world. Since the publication of Reference Thaler and SunsteinThaler and Sunstein’s (2009) seminal text on the use of psychological insights to influence behaviour and achieve policy goals, several national governments have been experiencing a new behavioural turn, while others have built on already-existing frameworks for incorporating policy tools of moral suasion into existing regulatory toolkits. Evidence abounds, in both cases, of the adoption of BI for policy-making in some of the most important aspects of public life, such as health care, education, the environment, transport and security.

Behavioural insights in this regard are broadly defined as the collection of approaches for designing public policy, which incorporates evidence about human psychology into policy instrument formulation (Reference Hallsworth and KirkmanHallsworth & Kirkman, 2020). As such, BI can be situated at the micro-level of policy design wherein policy instrument settings and calibrations are adjusted based on behavioural evidence (e.g. modifying the frequency and form of tax reminders), and also at the operational level of policy programmes (e.g. designing a public health programme on healthy eating) (Reference RuggeriRuggeri, 2018). The BI approach recognises that people are not entirely rational agents, and their decision-making process is subject to various cognitive biases and heuristics. Behavioural insight-based policy interventions are usually implemented by altering the ‘choice architecture’, or the way in which policy choices are presented to the citizens whose behaviour the government is looking to change. Nudges, for example, are designed to direct citizens to a particular decision through the use of defaults, where the desired decision is the default one. However, BI is not limited to nudging designs and may take the form of commitment devices or alternative framing to encourage specific action or non-action (Reference Gopalan and PirogGopalan & Pirog, 2017). More cognitive forms of nudges, or ‘thinks’, have also been widely employed wherein the targeted behaviour change is meant to be long-term, relying on individuals’ own deliberation and thought processes (Reference John, Cotterill and MoseleyJohn et al., 2020). The growing BI ‘toolkit’ in policy design thus spans a range of responses that, in turn, necessitate specific capabilities, procedures and institutional frameworks to be in place for their design.

However, a comparative discussion of the procedural considerations of BI tool design is currently lacking, with a few notable exceptions (Reference LunnLunn, 2012; OECD, 2017; Reference Whitehead, Jones and LilleyWhitehead et al., 2017). Some forays into the administrative structure of BI tools have been made recently, and while this literature remains mostly scattered, they point to several patterns of BI tool design that can contribute to broad policy instrument and policy formulation studies, for example, the governance resources of organisation (both substantive and procedural) that these tools rely on. The organisational configuration of BI policy design in countries like the United Kingdom, for instance, manifests itself through dedicated BI units located at the heart of the government. Other governments rely on a more networked approach with mobile teams of BI experts across ministries and agencies. Finally, some countries utilise a more flexible, ad-hoc format of BI input, where behavioural insights are used for specific projects and initiatives, and experts (both internal and external) are involved on a contractual basis (Reference Afif, Islan, Calvo-Gonzalez and DaltonAfif et al., 2019; Reference FeitsmaFeitsma, 2018, Reference Feitsma2019; Reference Mukherjee and GiestMukherjee & Giest, 2020). The empirical evidence on such patterns, while disjointed, suggests that there is no single favoured approach to designing BI tools, or to integrating them into existing policy toolkits.

1.2 The Evolution of Behavioural Insights (BI) within the Policy Design and Formulation Literatures

Behavioural insight intervention within the policy design literature is a vibrant yet under-theorised area of discourse, even though it has been explored in other research domains pertaining to policy, such as policy mechanisms (Reference Capano and HowlettCapano & Howlett, 2020); epistemic communities (Simons & Schniedermann, 2021; Reference WeibleWeible, 2018); policy transfer (Reference Ball and FeitsmaBall & Feitsma, 2020); and comparative studies (Reference Jones, Pykett and WhiteheadJones et al., 2013; Whitehead et al., 2019), among others. The proliferation of BI policy tools has been especially prominent, for example, during the pandemic era, as the global movement to scale up community behaviours quickly and critically for supporting public health has led to unique interactions between novel BI interventions and existing or more traditional health policy instruments. Other sectors, such as environmental and climate policy, have also seen a marked increase in the use of BI mechanisms to enhance or supplement regulatory or incentive-based policy instruments (Reference Carlsson, Gravert, Johansson-Stenman and KurzCarlsson et al., 2021; Reference Gravert and ShreedharGravert & Shreedhar, 2022).

Studies of these developments in the area of BI policy research, beyond their application, have also burgeoned over the last decade. Most of this scholarship retains a clear emphasis on the substantive content and contributions of BI interventions, and their place and role in the multi-instrument ‘toolkits’ that governments design during times of relative stability, as well as during times of crisis. There is agreement in the policy sciences literature that policy design can be more or less systematic in attempting to match policy ends and means in a logical fashion. In all cases, formulators make assumptions about the capabilities of policy tools and their ability to achieve predictable policy outcomes (Reference Capano, Capano and HowlettCapano, 2020; Reference Capano and EngeliCapano & Engeli, 2022; Reference Howlett, Ramesh and CapanoHowlett et al., 2020; Reference Howlett and LeongHowlett & Leong, 2022). Usually, the choice of instruments is based on either ideological premises or the empirically informed perception of how different categories of instruments can engage and modify the behaviour of policy targets (Reference Schneider and IngramSchneider & Ingram, 1990). The unique capabilities of different classes of policy instruments and their resulting ability to achieve policy goals involve a consideration of how sets of actors behave, implicitly or explicitly, through the deployment of tools to deliver specific kinds of responses on the part of their intended targets. This is the case even though policy-makers may not always be familiar with a causal way of thinking (at least, they may not always be aware that they activate certain behaviours when implementing their chosen policy design).

The discussion regarding the need to study this underlying behavioural logic of design choice has been recently invigorated in the policy sciences (Reference Capano and HowlettCapano & Howlett, 2020). More active engagement with the literature on behavioural policy tools, their formulation and deployment, through the perspective of traditional policy instrument studies, can help to address this emerging research agenda and to move beyond any deterministic stance regarding how particular behaviours can arise only by deploying particular policy instruments.

1.3 A Framework for BI Policy Design

The modern discourse on behavioural public policy presents four main theses related to the evolution of BI tools. Firstly, and most encompassing, is that behavioural policy evidence strongly suggests that there is more than one class of BI policy interventions that infuse behavioural notions (dominated by ideas of automatic individual actions guided by mental heuristics, or ‘nudges’) with more varied cognitive considerations (deliberate actions or decision strategies that can be more community-driven) (Reference John, Cotterill and MoseleyJohn et al., 2020).

Secondly, the venue of action for nudges is specific and often minor in scope in terms of the modification or recalibration that they propose as a low-cost policy tool (Reference Marteau, Ogilvie and RolandMarteau et al., 2011; Reference Thaler and SunsteinThaler & Sunstein, 2009). The micro-perspective employed in the design of nudges, while scalable, has also attracted criticism in terms of lacking public autonomy or consent (Reference Bovens, Grüne-Yanoff and HanssonBovens, 2009). The procedural as well as substantive content of nudges tends to be limited to easily testable activities with randomised controlled trials (RCTs), public message framing and minor one-off changes in public actions, rather than sustained long-term behaviour change.

Thirdly, as the emphasis on a simple behavioural nudge further expands to emphasise agency, deliberation and autonomy, it goes beyond its ‘one-size-fits-all’ implementation and tool choices centred on information-based cues and public messaging mechanisms. A resulting, more ‘think’-oriented research agenda is emerging in response, explicitly incorporating human cognition and the ability for policy targets to exercise deliberation through the design of BI operations, thus diversifying them beyond simple, micro-level readjustments or tweaks to address rationality biases (Reference Hertwig and Grüne-YanoffHertwig & Grüne-Yanoff, 2017; Reference John, Cotterill and MoseleyJohn et al., 2020; Reference Krueger and FunderKrueger & Funder, 2004). Recent discussions on the design of nudge ‘plus’ tools reflect this major development within the field (Reference John and StokerJohn & Stoker, 2019). While nudges are designed to work towards quick, almost automatically elicited socially desirable behaviours, the contemporary work of behavioural policy scholars indicates a consensus that some element of self-awareness, internal deliberation and participatory engagement leads to more effective implementation and long-term viability of this category of tools (Reference Banerjee and JohnBanerjee & John, 2021; Reference Richardson and JohnRichardson & John, 2021).

In doing so, simpler assumptions about policy target decisions made inherent in nudges can be refined and augmented by making them more nudge plus, and eventually even becoming boosts, or behavioural interventions that capture more cognitive and motivational competences. For example, the framing of public announcements can help to guide policy targets to demand and accept health statistics that are based on absolute numbers (e.g. decline in the number of COVID-19 cases who are vaccinated individuals) instead of relative figures (e.g. vaccine A is x per cent more effective than vaccine B). Such efforts to ‘boost’ domain general competences are able to target human cognition within the environment (e.g. community norms) (Reference Hertwig and Grüne-YanoffHertwig & Grüne-Yanoff, 2017; Reference Richardson and JohnRichardson & John, 2021). Boosts are also being used to address cognitive challenges in a digital world, whereby incorrect public health information and countering fake news and disinformation have become a significant part of COVID-19 response toolkits. In other words, a successful boost relies on improving the decision-making tools available to policy targets to better deliberate and more effectively discern pertinent information (Reference Grüne-Yanoff and HertwigGrüne-Yanoff & Herwig, 2016).

This distinction reiterates that while nudges are used to guide target population behaviour by creating micro-level implementation conditions that trigger an action, more ‘think’-oriented boosts change behaviour by cultivating more long-term and sustained competences to make better decisions. As such, there is growing agreement within the behavioural public policy community that boosting and nudging can readily complement each other by combining meso–micro-level design considerations (Reference Richardson and JohnRichardson & John, 2021). The actual blend between them depends on the policy design goals for which they can feasibly become the means. For example, boosts are considered to be better aligned with policy designers’ goals centred on producing generalised, sustained behavioural change (such as social distancing), rather than a more specific, temporary action (such as giving up a seat on a bus), wherein the latter could be simply classified as a ‘nudge’ (Reference Banerjee and JohnBanerjee & John, 2021; Reference Grüne-Yanoff and HertwigGrüne-Yanoff & Hertwig, 2016; Reference JohnJohn, 2013).

The design of ‘think’-based BI means, with their more ambitious policy goal of enhanced public cognition, also encompasses a strong normative emphasis on greater public participation. Peter Reference John, Cotterill and MoseleyJohn and colleagues (2020), who have been at the forefront of the analysis of this category of BI devices, continuously emphasise that these tools procedurally entail more governance resources and are markedly ‘harder to do’ than nudges. Increased incorporation of public deliberation, autonomy and participation, which are fundamental to the design of such tools, eventually will augment the task of formulating mostly information-based interventions towards more organisation-based developments. Deliberative tools, participatory approaches and even the co-production literature that highlights broader public management implications have been empirically examined and discussed in the behavioural public policy literature as being an extension of the ‘think’ family of policy interventions (Reference Durose and RichardsonDurose & Richardson, 2016; Reference John, Cotterill and MoseleyJohn et al., 2020). The role of the state in designing these interventions is found to go beyond being the supplier of customised public information (as is the case with nudges), to becoming the creator of ‘new institutional spaces to support citizen-led investigation’ (Reference JohnJohn, 2013, 19), thus demanding more and different governance resources. In other words, for the ‘think’ design strategy, ‘the policy-maker needs to be open minded and willing to act as an organizer of citizen-driven investigation’ (20).

While more work is needed to conceptually better clarify the ‘think’ family of tools, there are broad patterns to their design that distinguish them from the more primal and simpler notion of ‘nudges’ (Figure 1). Firstly, despite sharing a similar starting point (in espousing notions of bounded rationality), ‘thinks’ propose to overcome the limits to rational behaviour through greater deliberation, more dialogue, increased participation and creation of enhanced conditions for collaboration. This agenda, which implies a significant commitment by and cost on the part of the state, is distinct in scope (more meso-level) from the design of simpler nudges that work alongside existing policies or regulations, looking to behaviourally tweak or enhance these tools at the micro-level of on-the-ground tool implementation. Secondly, as nudges go more ‘along the grain’ of existing tools, they supply recalibration mechanisms that are more low-cost and information-centred. By contrast, as more synergetic, cognitively sustainable behavioural outcomes are sought with the input of an increasing diversity of stakeholders, the ‘think’ family of behavioural tools requires more governance resources and more prominently organisational-oriented changes.

Figure 1 A continuum for BI policy design

Several important next steps are needed in the policy sciences in terms of generalising about the BI family of concepts within existing efforts to classify policy instruments (Reference Einfeld and BlomkampCapano & Engeli, 2021). Especially pertinent are discussions about the kind of new hypotheses that need to be drawn about how BI means interact with more ‘traditional’ policy tools such as regulations and incentives, and especially how they interact within existing policy toolkits. The field of policy studies is at a point in theorisation that needs a focus on the outcomes of effective policies and an evaluation of the unintended consequences and side-effects of BI, which are outcome-driven. This is discussed at length in Reference HallsworthHallsworth’s (2023) manifesto for applying behavioural science, in addition to several important actionable steps aimed at addressing the issues in the scope, methods and values of BI. Still, the focus of this Element is procedural rather than evaluative, so it operates on the assumption that BI instruments are adopted and implemented in a bid to increase the effectiveness of governments. In addition, some of these issues, for example policy ‘over-design’ (Reference MaorMaor, 2020), are already discussed in the policy literature.

Furthermore, what happens differently during public crisis situations? Comparative work on this question picked up in the midst of the pandemic (e.g. Reference Bavel, Baicker, Boggio and WillerBavel et al., 2020) and continues to highlight the need to address the formulation of behavioural tools in times of crisis. Reference Hartley and JarvisHartley and Jarvis (2020), for example, in their exposition of the pandemic response in Hong Kong, call for a more nuanced exploration of how public behaviours and norm-driven motivations are constructed and proliferate during times of crisis, and what this means for community-level capacities in the face of pandemics. To tie together these important empirical insights with efforts at theorisation, this Element presents one possible ‘next step’ in the procedural understanding of how BI, as a policy innovation, is blended into more traditional policy design choices.

1.4 Organisation of This Element

As alluded to earlier, several important gaps are emerging in the current scholarship on BI in policy-making, which indicates the salience of the key question explored in this Element: What are the key capacity and procedural prerequisites for the integration of BI into the design of policy tools? The main objective of this Element is to address the pronounced lack of consolidated understanding about what is required of governments in terms of capacity and established rules of operation to effectively design behavioural insights and integrate them in multi-tool policies. This fragmentation in the literature builds on questions of whether BI tools represent a separate category or subcategory of instruments, or components of larger policy instruments (Reference Einfeld and BlomkampCapano & Engeli, 2021), but essentially goes further and calls for a discussion on what capabilities and procedural endowments are needed in order to integrate them into policy.

In the following two sections (Sections 2 and 3), the Element develops a discussion of a novel theoretical framework combining the academic literature on procedural policy tools and innovation adoption studies, focusing on the capacity and institutional prerequisites of BI. In Section 4, we present three illustrative case studies with respect to their experience of essential procedural endowments facilitating the effective use of BI. Namely, we introduce and discuss the broad developments in BI policy design in Australia, Singapore and the Netherlands, and their experience in creating an administrative and institutional environment conducive to effective BI utilisation. In the concluding section, we discuss the practical implications of the procedural prerequisites for BI integration in public policy-making by motivating future research questions.

2.1 BI in Information and Nodality Policy Tools

Reference Hood and MargettsHood and Margetts (2007, 6) define this category of policy tools as pertaining to the centrality of the government in social and information networks: ‘nodality equips the government with a strategic position from which to dispense information and likewise enables the government to draw in information for no other reason than that it is a centre or clearing house. The limiting factor is credibility, and the ‘coin’ – how government spends this resource – is messages sent and received’. Instruments in this category include government tools for information collection and release; advice and exhortation; advertising and the formation of inquiry commissions and various modes of e-government (Reference HoodHood, 2007; Reference Howlett, Ramesh and PerlHowlett et al., 2009). The behavioural mechanisms that are activated by this suite of instruments are grounded in information dissemination, norm-driven citizen participation and interaction, enhancing public image and knowledge sharing (Reference Carter and BélangerCarter & Belanger, 2005; Reference DawesDawes, 1996) and increasing public accountability for social behaviours (Reference Andersen and DawesAndersen & Dawes, 1991; Reference Gil-Garcia and Martinez-MoyanoGil-Garcia & Martinez-Moyano, 2007).

Nodality-based instruments can also work to establish and ingrain prevailing norms and ideas about appropriate interactions, participation and coordination. That is, the norm-informed trajectory of such policy instruments can precede self-maximising calculations on the part of policy targets, and also reflect the dominance of a particular intervention logic that directs and constrains policy choice (Reference BernsteinBernstein, 2001; Reference Bernstein and CashoreBernstein & Cashore, 2012; Reference Cashore, Bernstein and HumphreysCashore et al., 2019). Furthermore, with the advent of information technology and unprecedented technical advancements, such tools of ‘nodality’, or those that depend on the government to communicate with policy targets, collect and disseminate information, establish cognitive frames of appropriateness, sway civic actions or persuade members of society to comply with public directives, can be seen to be increasingly crucial to the success of all other categories of policy tools.

Furthermore, the evolution of networks of information sharing has brought information as a policy resource to many actors beyond governments. This scenario is in contrast to Hood’s original, earlier conceptualisation of this class of instruments, whereby nodality was considered to be an endowment that the government possesses ‘by virtue of being government’ (Reference HoodHood, 2007). This is because, from a mechanistic perspective, information tools are better adapted to activate ‘first-order’ behaviours such as citizen engagement and interaction and information diffusion, while for second-order behaviours and cognitive commitments made through public participation in policy, nodality appears to activate coordination as well as the redistribution of power.

2.2 BI in Authority-Based and Regulatory Policy Tools

Instruments of regulation mainly exercise government control and authority, and they therefore closely reflect the use of mechanisms related to coercion. Popular styles of regulation-based policy instruments include command-and-control systems, laws and mandates, and forms of delegated and (self-)regulation (Reference SalamonSalamon, 2000). Most types of regulation and authority instruments award ‘the legitimate legal or official power to command or prohibit’ societal actions to the government (Reference Margetts, Hood, Peters and ZittounMargetts & Hood, 2016). Economic regulations, such as those governing the prices and volumes of agricultural resources or regulating pollution and other environmental offences, are formulated to control particular aspects of a market economy (including the entry or exit of firms), influence the return on investments and correct perceived inequities in economic relationships (Reference SalamonSalamon, 2000). The progression of such tools strongly reflects the ‘rules’ pathway identified by Reference Bernstein and CashoreBernstein and Cashore (2000, Reference Bernstein and Cashore2012). Reference Bernstein and CashoreBernstein and Cashore (2000) use the example of regulatory mandates and agreements, which demand a focus on not just how and how well rules and regulations are followed but also the influence of such rules on problem definitions and recasting of policy goals in line with policy target compliance. As discussed previously, behavioural mechanisms activated based on the government’s exercise of coercive powers and regulatory alignment lead to second-tier mechanisms related to monitoring compliance, political legitimation, blame-avoidance and isomorphic dynamics. One example of the importance of BI for regulatory action is the multiple regulations set up by the European Commission, such as data protection, regulation of privacy, consumer protection and business conduct, among others (see Reference Alemanno and SibonyAlemanno & Sibony, 2015).

In the case of self-regulation or voluntary agreements, sometimes also known as ‘non-state market driven’ (NSMD) governance, a different set of behavioural mechanisms can encourage voluntary certification schemes whereby industry conglomerations develop and certify standards of good business practice (such as the sustainability certification of ecologically detrimental agricultural products and practices) (Reference Auld, Balboa, Bernstein, Delmas and YoungAuld et al., 2009). This class of instruments usually activates behavioural mechanisms such as compliance, normative isomorphism and participation. The second set of such mechanisms set off by these practices concerns building legitimacy and gaining political acceptance as an authority-based instrument. These mechanisms are required because the success of modes of self-regulation is heavily dependent on how effectively they are able to develop contextually grounded institutions and endure as politically feasible rule-making entities (Reference OstromOstrom, 2000). Often, these legitimacy-seeking mechanisms also depend on how well self-regulation systems ‘create and maintain appropriate structures of property rights, rule governing contracts, liability rules’ (Reference Brondizio, Ostrom and YoungBrondizio et al., 2009, 39).

2.3 BI in Treasure-Based or Financial Policy Tools

The functioning of this class of policy instruments relies heavily on the allocation of public wealth and financial resources, as well as on the ways in which the government raises and distributes funds. Mechanisms based on utility calculations and the weighing of incentives versus disincentives are most closely linked with this family of policy instruments. The deployment of incentives or disincentives can unleash causal processes or behavioural responses akin to those identified in the ‘markets’ pathway by Reference Bernstein and CashoreBernstein and Cashore (2000, Reference Bernstein and Cashore2012). The use of such instruments encompasses the entire gamut of financial tools at the disposal of the government, such as subsidies in the form of grants, tax incentives and loans (to reinforce desirable economic activities), as well as taxes and user charges (to raise government funds or penalise socially undesirable behaviour) (Reference Howlett, Ramesh and PerlHowlett et al., 2009). Government expenditures of the former sort, especially grants, are offered to support ‘some end worthy in itself, sometimes as a form of recognition, reward or encouragement’ (Reference PalPal, 1992, 152). Tax incentives or the reduction or deferral of payments owed to the state are another form of subsidy (Reference MitnickMitnick, 1980). Subsidies as a class of instruments make specific use of incentivising as the main mechanism of their deployment.

By contrast, taxes and user charges provide financial deterrents to curb some aspect of society or modify the consumption behaviour of an economy. Taxes can be deployed for a wide array of public goals, and as such, the primary behavioural mechanisms that are activated by taxes are disincentives because they increase the cost of a particular action or behaviour, especially the consumption (e.g. sugar tax) or production patterns of private market goods. Support for innovation as a mechanism may follow to uphold the efficient production of such goods and services. Similarly, when taxes are employed to compensate for public ills, such as environmental taxes to curb carbon emissions, they may be followed by innovation in pursuit of more efficient practices and competition among existing industry players. From the perspective of norms, through the use of such taxes, governments are able to activate a reorientation of public values and linked priorities between broad goals such as environmental protection and economic development.

2.4 Organisation-Based Governance Resources for BI Design: A Research Gap?

The research agenda of behavioural public policy (BPP) and behavioural public administration (BPA) has also tended to focus on more micro-level individual behavioural processes, with comparatively less attention to broader political, social and, at the policy level, organisational considerations of the procedural aspects of BI design (Reference Ewert, Loer and ThomannEwert et al., 2021; Reference MoynihanMoynihan, 2018). As a result, a prevailing criticism faced by the BI policy scholarship is that most analytical approaches to studying BI in the policy process remain descriptive with a limited scope when it comes to accommodating different policy-making contexts as well as discussing how the instrumental, organisational and macro-political considerations of behavioural policy formulation are related to each other, even when taking a comparative view (Reference Battaglio, Belardinelli, Bellé and CantarelliBattaglio et al., 2019; Reference Feitsma and WhiteheadFeitsma & Whitehead, 2022). In response to these observations, recent discourse on the ‘what next’ of BI policy studies has highlighted that in investigating the organisational implications of BI policy, ‘they should produce concrete practical insights for the management and performance of and decision-making in public organizations, the success and failure of network collaboration and contribute to the diagnosis, evaluation and development of policy design and its assumptions’ (Reference Ewert, Loer and ThomannEwert et al., 2021, 5).

In other words, while the literature remains steadfastly focused on the public outcomes and effects of BPP, scholars have cautioned that in perceiving nudges and minor behavioural calibrations as more ‘simple’ than traditional policy instruments, their organisational design and administration may also be misconstrued as being simple. Relevant experience from the environmental sector, for example, indicates that while nudges have been reportedly treated as ‘easy’ and comparatively ‘low-cost’, their design has necessitated extensive research on human behaviour, in particular multilevel contexts such as sustainability (Reference Mont, Lehner and HeiskanenMont et al., 2014). That significant effort to generate and allocate resources, capabilities and knowledge is often needed to access, review and incorporate this research from behavioural economics and psychology into nudge policy design may not be fully reflected in studies surrounding their implementation. Further, integrating this evidence in different policy domains (such as health, environment and public spending) is often a separate endeavour from using it for adjusting existing policy instruments (such as regulations versus tax incentives) (Reference Giest and MukherjeeGiest & Mukherjee, 2018).

Additionally, the vast diversity of behavioural interventions, as indicated in Section 1, has warranted a proliferation of methodological innovations that help policy-makers go beyond just looking at behaviours that defy economic rationality. Modern behavioural public policy is developing as a multidisciplinary and multi-methodological concept that derives insights from ‘anthropology, geography and sociology as well as the application of qualitative methods’ (Reference Ewert and LoerEwert & Loer, 2021, 25). This awareness further suggests that while certain methods (such as experimental social science research, and most prominently RCTs) have dominated the behavioural policy formulation arena, it is likely not to remain limited to these approaches as the field advances. For example, research on incorporating behavioural insights into policy design burgeoned in response to COVID-19 as governments sought to improve the outcomes of regulatory-based tools (lockdown-related prohibitions), safe-distancing measures and curbing the spread of disinformation, to name a few (Reference Debnath and BardhanDebnath & Bardhan, 2020). The application of RCTs in refining BI design, in this situation, necessitated mobilisation beyond labs and government research units to quickly translate ‘improved’ designs into practice, thereby heavily relying on methods such as qualitative research, participatory action research and network analysis to inform policy design (Reference Abaluck, Kwong and StyczynskiAbaluck et al., 2022; Reference van Roekel, Giurge, Schott and Tummersvan Roekel et al., 2023).

Focusing on these developments, this section organises what is known about government’s institutional and organisational considerations for behavioural public policy by outlining the procedural resources and instruments, policy capacities and subject expertise necessary for their effective design as public policy.

This discussion is keenly aware of other factors that play a role in BI adoption and implementation, such as the role of epistemic communities (Simons and Schniedermann 2021; Reference WeibleWeible 2018), transnational knowledge brokers (Reference FeitsmaFeitsma, 2019) and the shift towards RCT in policy, and in evaluation of BI tools specifically (Reference Banerjee, Alsan and BrezaBanerjee et al., 2020; Reference de Souza Leão and Eyalde Souza Leão & Eyal, 2019). There is an acknowledgement of the extent to which policy translation played a role in BI adoption in the cases used in this Element. However, the primary scope of this Element is firmly on addressing the gap of the organisational aspects of adoption and implementation of BI at the national level. The research gap in the interplay between national-level organisational resources and transnational communities is important and should be addressed in future research.

2.5 Procedural Resources and Instruments

Unlike substantive instruments, which have frequently been the topic of analysis in the policy sciences, procedural instruments have enjoyed much less academic investigation, and this is also starkly evident in the study of behavioural policy. Procedural instruments are understood as those resources and actions that go towards supporting the administrative and managerial processes taking place within government for the design of policy. As such, their study reflects the scholarly overlap between public administration and the policy sciences and includes an exploration of government’s own instruments, created to support its internal design and policy formulation activities. These instruments include network management mechanisms and public participation, activities linked to the delivery of organisational services, such as the formation of advisory committees to regulatory agencies, and policy processes more generally, such as the creation of public information commissions, government data portals and repositories and judicial review processes (Reference HowlettHowlett, 2018). Such instruments are designed internally within the government to support the creation of substantive policy instruments – such as taxes, subsidies or public awareness campaigns – that directly impact public and societal behaviour.

Unlike substantive tools that directly affect the production, consumption and distribution activities of public goods and services, procedural instruments have been envisioned as instead focusing on affecting governments’ own internal actions related to the policy-making process. For example, network management and governing the behaviours and interactions of policy-making agents is one significant function of procedural tools (Reference Klijn, Koppenjan and TermeerKlijn et al., 1995). Beyond their administrative function, procedural elements can also shape the substantive policy decisions that follow from government’s process-oriented actions. Examples include government forming an advisory committee of select citizens or experts to help timely deliberations on sensitive issues such as nuclear energy regulation, or the creation of freedom-of-information or access-to-information laws facilitating citizen access to government legislation and records. Additionally, reorganising an administration’s own internal structure can impact the effective formulation of policies, as occurs when new regulating agencies are created by merging personnel from energy and environmental ministries, forcing the two to adopt a new collaborative operating arrangement.

Organisational changes made to initially set up or further broaden BI-centred policy design within government may be as radical or distinct as setting up of a new nudge unit or appointing new personnel, or more incremental as the commissioning of one-off pilots or experiments within existing policy initiatives. In the first instance, the popularity of nudges and other behavioural interventions have been seen to propagate the creation of specialised policy advisory bodies or behavioural insight units to consolidate expertise on public behaviour and citizen choice (Reference HalpernHalpern, 2015b). Often referred to as behavioural insights teams (BITs), per their origins in the United Kingdom and the United States, these units have emerged as the main originators of policy formulation activities explicitly based on understanding the ‘choice architectures’ of public policy targets. As expressed by Reference Thaler and SunsteinThaler and Sunstein (2009), such experts take on the role of ‘choice architects’ who have ‘the responsibility for organizing the context in which people make decisions … there is no such thing as a “neutral” design’ (1). As such, in contemporary policy formulation situations, BITs have thus been instrumental as activators of policy designs that are specifically aimed at bringing about modest yet measurable modifications in the actions of the public towards socially and environmentally desirable behaviours through the use of ‘nudges’ and other behavioural cues. Often supplementing the use of traditional instruments that make use of government resources of coercion, law enforcement or fiscal endowments, BITs are tasked with designing mechanisms able to stimulate nearly subconscious compliance with government directives (Reference HowlettHowlett, 2018). For example, BITs have been widely successful in helping to reduce energy consumption by working with public utilities to display own and peer electricity-use patterns for households (Reference Lehner, Mont and HeiskanenLehner et al., 2016; Reference Rafiq, Salim and NielsenRafiq et al., 2016).

In other situations, whereby the diffusion of behavioural insights research and policy design has been more variable, behavioural experts or consultants may negotiate institutional logics within the government in a more incremental fashion, especially in ‘backstage’ governance contexts that have previously rejected strong rationalist ideas prior to engaging with BI (Reference FeitsmaFeitsma, 2018). For example, Reference FeitsmaFeitsma’s (2018) critical ethnography of behavioural experts at work for policy-making (in the example of the Netherlands, which is covered in this Element) has depicted four main ‘balancing acts’ on the part of experts that have organisational bases, namely, ‘knowledge brokering, focusing on outputs, analytical satisficing and horizontalizing the hierarchy of evidence’ (160), which determine the negotiation of the institutional logics of their work within the government. Analytical capacities and expertise are further discussed in the remainder of this section.

2.6 What Next?

How, then, can we explore the ways in which experience with behavioural public policy as an innovation is changing and in turn being changed by governments’ own work? The general discourse strongly indicates that the use of BI in policy represents a coming-of-age example of policy innovation and is more than just a passing policy fad (Reference Battaglio, Belardinelli, Bellé and CantarelliBattaglio Jr. et al., 2019; Reference StraßheimStraßheim, 2020), especially considering the rise of artificial intelligence (AI) and its potential to augment behavioural science to develop targeted nudges based on large swaths of data. The rise of the digital BI tools, G2C (government-to-citizen) governance, regulation of the web and so forth make this discussion even more timely and pertinent. While it is still not clear to what extent BI will penetrate policy-making and existing frameworks of policy instrument choice in the long term, or whether it will even eventually be recognised as a policy tool on its own rather than an extension of the existing instruments (Reference Hill and VaroneHill & Varone, 2016; Reference Kosters and Van der HeijdenKosters & Van der Heijden, 2015), BI remains at the foreseeable forefront of policy design ambitions in many countries around the world (Reference Afif, Islan, Calvo-Gonzalez and DaltonAfif et al., 2019).

With this in mind, there are still important, yet presently unanswered questions to be asked about the necessary procedural and administrative endowments of governments for the successful and appropriate integration of BI, as a policy innovation, into the existing policy design framework. Behavioural insight’s institutionalisation as a governance innovation is discussed in the following section.

3.1 Stages of Innovation Adoption

The existing frameworks of the innovation adoption process vary from the two-stage (Reference Gopalakrishnan and DamanpourGopalkrishnan & Damanpour, 1997) to more detailed models that cover three (Reference Damanpour, Chiu and WischnevskyDamanpour et al., 2009; Reference Hameed, Counsell and SwiftHameed et al., 2012), five (Reference Klein and SorraKlein & Sorra, 1996), six (Reference Frambach and SchillewaertFrambach & Schillewaert, 2002) or more stages (Reference Kwon, Zmud, Boland and HirschheimKwon & Zmud, 1987). Some authors conceptualised this process as linear (Reference Howieson, Lawley and SelenHowieson et al., 2014), while others argued for a dynamic and recursive approach characterised by feedback and feed-forward loops (Reference Adams, Bessant and PhelpsAdams et al., 2006). However, the most frequently used approach to the innovation adoption process is the ‘unitary sequence model’, divided into three more general phases: initiation, adoption decision and implementation, consistent with Reference LewinLewin’s (1952) model of the change process. In this section, we rely on this conceptualisation of the innovation adoption process and describe these three stages in more detail.

3.2 Determinants of Innovation Adoption

Beyond exploring what stages comprise the innovation adoption process, scholarly attention has been paid to factors that contribute to innovation adoption. Since our primary interest lies in the institutionalisation of BI, we are particularly keen on understanding what organisational antecedents serve as an important impetus for innovation adoption. This is crucial for the theoretical development of the requisite institutional arrangements for effective BI utilisation in government policy. One of the efforts that discusses organisational and top management characteristics that aid innovation adoption can be seen in Pichlak (2015), which proposes two factors: organisational structure and resources.

In this section, we discuss how neatly institutionalisation of BI in terms of procedural instruments, three types of capacities – analytical, political, and managerial – and knowledge and expertise can be fitted into the framework of innovation adoption determinants at every stage of adoption. Pichlak’s (2015) effort, used here as a guiding framework, provides a useful heuristic as to what is required at every innovation adoption stage for the innovation to succeed. We attempt to synthesise insights from the innovation adoption literature with those from the public policy and public administration literature in order to produce a theoretical approach, which we further develop throughout the remainder of this Element.

3.3 Institutionalisation of BI through the Lens of Innovation Adoption

Based on our discussion so far, what emerges is a more systematic way of thinking about what is needed within governments to establish BI as a policy innovation. Three important conversations or common bridges exist within the policy design and policy innovation literatures in discussing the governance ‘pre-conditions’ that are needed in place for BI policy-making to develop over time.

• The first precondition has to do with procedural tools or organisational decisions.

• The second precondition has to do with the capacities and resources endowments for innovation adoption.

• The third precondition has to do with the role of specialists, subject-matter experts and managers both within government departments as well as external contracts/partnerships.

Figure 2 is a summary of the overlap in the conversations in policy design and policy innovation scholarship. Policy design literature offers various perspectives on the pre-conditions or capacities needed for policy-making. These, however, do share several features such as administrative capacity, sound political management, operational readiness in terms of resources, information, legal underpinnings, among others (Brenton et al., 2022; Reference Howlett and RameshHowlett & Ramesh, 2016; Reference Mukherjee, Coban and BaliMukherjee et al., 2021; Reference Saguin, Ramesh and HowlettSaguin et al., 2018; Reference Wu, Ramesh and HowlettWu et al., 2015).

Figure 2 Conceptual overlaps in policy design and policy innovation literature

Policy innovation literature focuses heavily on organisational structure (Pichlak, 2015) and dynamic capabilities as essential for innovation adoption (Reference Eisenhardt and MartinEisenhardt & Martin, 2000; Reference TeeceTeece et al., 1992, Reference Teece, Pisano and Shuen1997). Further, knowledge integration is given much attention as an important prerequisite for innovative solutions to take root (Reference Okhuysen and EisenhardtOkhuysen & Eisenhardt, 2002). The pattern of overlap is emerging when we look at these two scholarship streams together – analytical capacity or expertise; organisational/operational capacity or resources, legal arrangements, organisational scaffolding; and the intersection of organisational and political capacity or the ability to secure and sustain political support (Howlett & Ramesh, 2015; Reference Wu, Ramesh and HowlettWu et al., 2015).

Based on this pattern of overlap, we propose a theoretical framework for the institutionalisation of BI through three stages of innovation adoption, presented in Figure 3. Three types of capacities are needed for effective introduction and institutionalisation of BI in policy – analytical, organisational and political. Organisational and analytical capacities are more crucial at the initiation and decision adoption stages, while some organisational and political capacities are indispensable at the implementation stage.

Figure 3 Theoretical framework

4.1 Methodology: Approach to Case Selection

In this section, we turn to three country cases that represent three distinct examples of a BI implementation journey with a focus on the three stages of innovation adoption and the level of policy design. The three country cases are Australia, the Netherlands and Singapore. All three consistently feature in conversations on BI policy adoption (World Bank, 2014) as examples par excellence. Each successful in their own right, these three examples are instrumental for our case for a differentiated approach to institutionalisation of BI depending on the level of policy design and the stage of innovation adoption.

Our selection of these three cases is driven by the methodological approaches discussed at length by Reference Seawright and GerringSeawright and Gerring (2008). All three cases are influential examples of BI application that can be characterised by high levels of success and are often discussed in the context of effective BI implementation (Reference Ball and HeadBall & Head, 2021; Reference DetenberDetenber, 2021; Reference Feitsma, Schillemans, Straßheim and BeckFeitsma & Schillemans, 2019). In addition, despite similar levels of success, these cases were selected using the ‘most different’ criterion (Reference Seawright and GerringSeawright & Gerring, 2008) due to the differences in the level of BI intervention at which they were adopted. We rely on a range of scholarly sources, as well as government reports and media coverage.

We expressly omit routinisation as a stage of innovation adoption when presenting the three country cases. Routinisation, or ‘automaticity in behavior, typically including unintentionality, uncontrollability, lack of awareness, and efficiency’ (Reference Ohly, Sonnentag and PluntkeOhly et al., 2006, 258), is an important stage in the innovation adoption process; however, due to its association with established routines and a certain degree of ‘calcification’ of processes, it might be too early to apply this lens to BI adoption journeys in the country cases included in this Element.

4.2 Australia: The Structured Collaborative Approach

Even before Thaler and Sunstein’s conceptualisation of ‘nudge’ became well known, the ideas underpinning behavioural decision-making were influential in some of the economic and regulatory agencies of the Australian government, as well as in the health and human services agencies, since at least 2007. For example, the Australian Department of Finance had a long-standing interest in ‘better practice’ regulation and voluntary codes of compliance, while the Australian Tax Office (ATO) had explored behavioural issues concerning tax evasion and tax compliance (Reference Jones, Head and FergusonJones et al., 2021).

One of the early adopters of BI insights in policy, Australia was largely influenced by the success of the UK’s Behavioural Insights Team (BIT), which prompted its government at the state level to forge links with BIT and start experimenting with micro-level instrument calibrations. This evolved into solutions for more complex policy problems, culminating in the establishment of a behavioural policy unit at the central government level with an emphasis on identifying low-cost measures to influence behavioural change.

The projects undertaken by the central government BI unit, Behavioural Economics Team Australia (BETA), and similar units in other government departments (see Box 2), reveal a broad cross-section of activities, including strengthening the resilience of students, dealing with cyber security, improving census participation, the workplace experience of apprentices, improving support for drought-affected farmers, improving consumer engagement with electricity retailers and managing unconscious bias in public sector recruitment practices (DPM&C Australia, 2023).

At the state level, the range of projects implemented with the help of BI is also extensive, from childhood obesity, outpatient appointments, and applications for jobs in rural areas, to participation in justice programmes, improving public sector diversity, improving the success of return to work programmes and boosting productivity (NSW Behavioural Insights Unit, 2020).

4.2.4 Lessons Learnt from the Australian Case

Figure 4 illustrates how stages of innovation adoption interact with the policy framework in Australia. As we have discussed in this case, BI adoption in Australia is focused heavily on application to policy-level BI, including national programmes tackling wicked problems such as health, tax compliance, employment policy, social policy interventions, market competition and so forth.

Figure 4 Policy design continuum and stages of innovation adoption in Australia

Most of the work undertaken for BI application has sought to apply experimental methods, such as randomised control trials (RCT), which is in line with the skillsets of the staff in most BI units across the government departments. This is also consistent with the influence of the UK BIT, which has conducted over 400 RCTs, trials and projects, meaning that RCTs are seen by Australian policy practitioners as most capable of providing robust evidence for policy innovation and assessing the relative efficacy of options.

We use the example of wage subsidies project by DoJB from Box 1 to illustrate the approach to BI adoption in Australia at the policy level. At the initiation stage, co-design of the intervention takes place with relevant stakeholders. This is followed by a launch of the RCT or trial at the decision adoption stage. The results of the trial are available publicly to secure support and ensure transparency of the proposed intervention. Feedback is received at this stage that is instrumental for further refining the intervention. Finally, at the implementation stage the results are carefully calibrated in line with the feedback and policy changes are proposed for implementation.

In terms of capacities, the commitment to analytical capacity, as expressed in BETA’s efforts to recruit experts, build thought communities, share expertise between departments and provide tools for training, was especially evident at the initiation and decision adoption stages (Reference Ball, Hiscox and OliverBall et al., 2017; Reference Ball and HeadBall & Head, 2021). At this time, importance was placed on creating an impetus for a more centralised approach to BI after the success at the state level. This helped bring the issues of funding and securing the federal government’s support for a centralised unit to the fore later, at the implementation stage. The Behavioural Economics Team Australia enjoyed relatively quick development into an influential government unit with the requisite financial resources due to the ability of the interested government agencies to build analytical and organisational capacities at earlier stages of policy adoption.

4.3 The Netherlands: The Networked, Bottom-Up Approach

As with many governments around the world, the Dutch state has been influenced by behavioural insights for several decades (Reference RoseRose, 1998). However, policy-makers started taking a focused interest in BI for policy only around 2008. This growing interest resulted in a symposium proposed by the Scientific Council for Government Policy (Wetenschappelijke Raad voor het Regeringsbeleid (WRR)). The symposium was aimed at exploring the potential uses of BI in policy-making. The WRR symposium was an attempt to follow in the footsteps of other country governments, such as the United Kingdom and the United States in their efforts to engage with the behavioural sciences. In the case of the Netherlands, however, the uptake of BI in public policy was met with some academic, bureaucratic and ethical resistance that prevented early discussions from evolving into tangible steps.

Nevertheless, these early discussions culminated in the creation in 2012 of the first BI unit in the Dutch government. This small unit was formed in the Ministry of Infrastructure and Environment, initially with only two staff members. At the same time, some efforts were undertaken to form a more coordinated engagement effort to utilise BI in public policy-making. The impetus for this development can be attributed to three main reasons. First was the much lauded effort of the UK’s BIT (Reference HalpernHalpern, 2015b).⁴ Secondly, 2012 saw the emergence after the election of a Liberal and Labour coalition government headed by Mark Rutte. The coalition, founded upon the principles of liberalism but also forged in a time of austerity, found the low cost and libertarian nature of BI appealing. Thirdly, this time also saw the emergence of an opportune political environment for policy experimentation.

The eventual institutionalisation of BI for policy-making in the Netherlands was preceded by extensive discussions of how this could be achieved in the most effective way. The Ministry of Economic Affairs lobbied for the formation of a central BIT that would mirror the structure of the UK’s BIT. This, however, raised several concerns. Firstly, it was not immediately clear which government department would host such a unit. This is because the structure of Dutch government departments is non-hierarchical, where no department has direct control over others. Second, an argument was put forward that a single BI unit would become an easy target for the BI doubters, giving them an easy argument that BI are excessively centralised.

It was out of this discussion that the Behavioural Insights Network Netherlands (BIN NL) was eventually formed. The network has representatives from all ministries and facilitates knowledge sharing about the application of BI to policy-making and its implementation, supervision and communication.

More recently, the Information Council (Voorlichtingsraad), which formulates the joint communication policy of the central government for the Prime Minister and the ministries, started a government-wide trial behaviour lab for communication in 2017. As of 2018, the cabinet continues to advocate for the strengthening of BI at the departmental level and for BIN NL to continue working to bring departments together. Behavioural insights have also been incorporated into the government Integral Assessment Framework for Policy and Regulations, published by the Ministry of Justice and Security, to guide policy-makers on instruments and guidelines to formulate policies and regulations.

The types of projects implemented by BIN NL are wide-ranging; however, the emphasis is on instruments that have small- to medium-level effect, such as communication tweaks and incorporation of behavioural considerations in analysis or policy-making, through to the widespread implementation of policies based on experimentation utilising BI. Examples include specific messages and reminders, advertisements, field experiments and financial incentives, among others. Box 3 is an example of such measures that focus on overcoming the limits to rational behaviour through regulation, understanding context and barriers to certain behaviours, ‘budging’ beliefs and attitudes and presenting opportunities for testing desirable attitudes and behaviours. Thus, the level of engagement with BI in the Netherlands can be characterised as mostly ‘meso’.

4.3.4 Lessons Learnt from the Dutch Case

Recent analyses of the emergence of BI in the Netherlands indicate that there has been an ongoing process of contestation and politicisation of behavioural expertise (see https://journals.sagepub.com/doi/10.1177/2399654419867711; Reference FeitsmaFeitsma, 2018). The contestation of expertise has in part been a product of the inevitable decisions that must be made in relation to which forms of expertise, with which particular behavioural insights, should be engaged with. It has also been a result of the process of working out which forms of behavioural expertise are actually most relevant to the policy-making process itself (Reference FeitsmaFeitsma, 2018b). Furthermore, the institutionalisation of expertise is not about bringing scientists into government, as it is governmental officials actively filtering scientific insights to meet their own needs (Reference FeitsmaFeitsma, 2018b). It is thus clear that the emergence of BPP in the Netherlands has not so much involved the hardwiring of behavioural science expertise into government but rather the training and repurposing of existing civil servants in new policy-making skills (Reference Feitsma, Schillemans, Straßheim and BeckFeitsma & Schillemans, 2019). In addition, the Dutch example is illustrative of a bottom-up approach predicated on limited resources and weak links to policy-makers and institutional actors, which translated to uncertain support from key government actors and the constant need of BI experts to defend their legitimacy until the gains became clearer, as reflected in the growing global recognition of BI tools and a rise in prominence of epistemic communities and knowledge brokers (Reference Feitsma, Schillemans, Straßheim and BeckFeitsma & Schillemans, 2019).

Figure 5 Illustrates how stages of innovation adoption interact with the policy framework in the Netherlands. Behavioural insight adoption in the Netherlands places emphasis on meso-level instruments, or ‘thinks’ such as developing an understanding of what drives human behaviour through RCTs, tweaking communication instruments, or ‘shoves’ aimed at changing behaviour by removing barriers and creating incentives (Reference Afif, Islan, Calvo-Gonzalez and DaltonAfif et al., 2019).

Figure 5 Policy design continuum and stages of innovation adoption in the Netherlands

We use the example of the ‘Optimising Use’ project from Box 3 to illustrate the approach to BI adoption in the Netherlands at the meso level. At the initiation stage, research is carried out that aims at identifying barriers to and enablers of the desirable user behaviour, done through surveys, interviews and analysis of the existing user data, among others approaches. This is followed by communicating with interested stakeholders and engaging in co-production with non-government actors at the decision adoption stage to secure additional resources for implementation and enhance legitimacy and participation. The solutions are aimed at budging existing beliefs that lead to unwanted behaviours. The resulting ideas are sent to a trial stage to ensure the workability of the proposed intervention. At the implementation stage, a clear and detailed communication campaign is used as the main driver.

The Netherlands example is also a good illustration of the need for analytical, organisational and political capacities to be built at different stages of policy adoption. Analytical capacity or expertise was a major focus of the government departments, who emphasised drawing on BI expert knowledge across departments to be shared widely through BIN NL. Organisational arrangements and political support for a more formal approach to using BI in policy came later, at the implementation stage, when a BI lab was proposed for implementation in 2017 to cement the use of BI in policy in a more centralised fashion.

4.4 Singapore: A Networked but Centrally Sponsored Approach

Singapore is another example of an early adopter of BI in policy with a high rate of success and recognition. From its earliest days as a nation, Singapore has enacted many measures to try to get its citizens to behave ‘properly’. Among the early policies set out were heterogenous public housing estates (i.e., state-mandated integration), mother tongue language requirements in schools and moving food vendors off the streets into regulated spaces. These social developments, coupled with economic progress, allowed Lee to claim that Singapore had gone from a third world to a first world country in the span of a generation (Reference LeeLee, 2012).

However, the true BI revolution was catalysed by the success of the BIT team in the United Kingdom, much like in Australia. Unlike the United Kingdom, the application of behavioural sciences in Singapore’s public policy did not begin with a big bang driven from the centre. Instead, it was a ground-up movement, with various agencies exploring and experimenting with small-scale projects. The first teams started out as ‘skunk works’ (Reference SoonSoon, 2017), learning the techniques while scrounging for willing partners prepared to give these new approaches a try. Quick wins were needed to gain confidence and win support from senior management.

Today, with over 250 members in a community of practice across fifty public agencies, the use of BI in Singapore public policy has evolved from an initial fascination with how cognitive biases challenge the traditional way of designing policies, to a more sophisticated framework of testing and accumulating insights on behavioural interventions. This shift has also shown that the use of BI is more than improving the last mile experience of citizens – it has the potential to fundamentally challenge the way we think about government policies and programmes.

The domains in which the Singapore government successfully implemented BI are numerous – from finance, health, and public utilities to environment, transport and, public communications, among others (see Box 4).

4.4.4 Lessons Learnt from the Singaporean Case

A positive outlook towards adoption of BI is common among policy-makers and public administrators. They recognise, however, that there are ethical issues and practical concerns when using BI, as well as limits to nudging. Hence, the civil servants regard the approach not as a panacea but as a resource to be thoughtfully applied. As the Deputy Secretary (Development) in the Ministry of Manpower put it, ‘BI is not a silver bullet’ (Reference KokKok, n.d., 110). Behavioural insights are implemented in Singapore through a variety of approaches, including policy design, experimentation and partnerships with private sector organisations.

Figure 6 illustrates how stages of innovation adoption interact with the policy framework in Singapore. Behavioural insight adoption in Singapore places emphasis on micro-level interventions, both those aimed at encouraging mindful choice and those tackling attentional shifts (Reference DetenberDetenber, 2021).

Figure 6 Policy design continuum and stages of innovation adoption in Singapore

We use the example of MOM’s Work Pass Division example from Box 4 to illustrate the approach to BI adoption for micro-level interventions in Singapore. At the initiation stage, research is carried out aimed at identifying the source of the problem and potential solutions through surveys, interviews and analysis of user data, among others. This is followed by rigorous experimentation to test the effectiveness of behavioural interventions before implementing them on a larger scale. This approach has helped to identify effective interventions and avoid wasting resources on ineffective ones. The solutions are deployed and regularly monitored for effectiveness, making implementation and improvement an iterative process.

Analytical capacity is at the heart of Singapore’s approach to staffing its government. Singapore is world-famous for offering its public servants competitive salaries. Hiring experts with extensive knowledge of BI and collaborating with academia was the priority at the initiation and decision adoption stages, as well as setting the stage for a networked approach to BI with departments making their own decisions on how to approach the organisational and resource endowment of BI in policy. At the implementation stage, government agencies in Singapore make sure to secure public and political support prior to rolling out the initiative.

4.5 Additional Considerations

This section aimed to analyse the experience of the three frontrunners in BI institutionalisation with application of BI at three different levels of policy instruments. While this Element assumes a neutral, apolitical approach in describing the cases in order to highlight the procedural factors and, specifically, the capacities and resources needed to successfully adopt and implement BI, it would be remiss to not at least discuss some important contextual considerations that contribute to BI implementation. After all, Reference Thaler and SunsteinThaler and Sunstein (2009) note that ‘there is no such thing as a “neutral” design’ (40). Following are some of the important political factors that underscore the essential conversation governments need to be having around adopting BI, beyond capacities and resources.

Firstly, two out of three cases described are from settings that can be depicted as varying shades of paternalistic, which is hardly surprising given that Reference Thaler and SunsteinThaler and Sunstein (2009) describe this brand of behavioural approach as ‘libertarian paternalism’. More discussion is needed on how BI would fare in settings that are on the other end of the scale and find any restrictions of freedom of choice unsavoury, and especially what this would mean for the type of capacities needed for adoption and implementation.

Secondly, out of the perceived limits placed on citizens’ autonomy and liberty justifiably come ethical concerns with BI application. Attitudes of the public have to be well understood and factored in. Seen by some commentators as manipulation or even coercion (Reference Hausman and WelchHausman & Welch, 2010), either BI needs to be repackaged to be more palatable to the public or their benefits need to be articulated along with potential side-effects and how the government is planning to mitigate them. The issue of information asymmetry and availability needs to be solved before BI is used as a policy solution. Whether this would need to be factored into the conversation on capacities merits a separate discussion.

Thirdly, to avoid selective reading of the application of BI, policy-makers need to have a clear idea of the impact of BI application on not just individual behaviour but also legal institutions and mechanisms (Reference Lepenies and MałeckaLepenies & Malecka, 2015). Nudges are non-normative because they are not legally prescriptive regarding how people should behave. Hence comes the debate on the most effective way to reconcile the application of BI with potential questions around their legitimacy.

Finally, additional contextual variables to be considered include administrative contexts, policy transfer culture, advisory systems and regulatory traditions. This Element is the first effort in trying to understand the procedural tools needed for effective BI adoption and implementation and could be the first step towards an informed discussion in this regard.

5.1 Organisational Adaptations

Through the empirical exposition presented in this Element, what is clearly evident is that the position of a dedicated behavioural insights unit within a government ministry or agency can have a significant impact on how behavioural insights are formulated as well as implemented. What is also evident, especially in the aftermath of the global COVID-19 pandemic, is that this impact can be diverse depending on its manifestation related to the content of policy tools functioning across different sectors and focal areas. For example, according to the World Health Organization (WHO), behavioural units or behavioural research units that were set up in Europe in response to COVID-19 performed a variety of tasks ranging from scaling up cultural behavioural and media activities across agencies relevant to health and welfare (e.g. Cultural, Behavioural and Media Insights Centre [CUBE], Government of Finland), to more specifically and only dealing with national COVID-19 responses and their behavioural implications (e.g. Corona Behavior Unit, Netherlands). For both of these examples the actual tasks taken on by the BI unit in delivering BI projects may be very similar, with the difference lying only in the sector and the scale. Distilling the cases studied in this Element, Figure 7 shows the four main phases undertaken in the design and delivery of a typical BI project. As such, the capacities and capabilities housed within a BI unit need to reflect (1) skills related to problem framing and prioritising policy goals; (2) research and monitoring skills related to analysis of individual behaviour social norms as well as historical trajectories of policy design; (3) technical skills related to methods of designing pilots, randomised controlled trials and intervention design; (4) evaluation skills related to impact evaluation and monitoring of evidence that is being generated about the process outcomes and impacts of proposed behavioural interventions and any unexpected repercussions, and (5) political and stakeholder management skills to scale up projects with proof of concept and do so successfully through political and social buy-in. These capacities further need to coordinate action in order to plan and secure resources for the perpetuity of activities generated by the unit.

Figure 7 Policy capacities and the institutionalisation of BI policy design

In doing so, the presence of a behavioural unit can permeate different phases of the policy-making process, and its flexibility or ‘nimbleness’ in being able to adapt to the various phases of policy-making can become a significant indicator of its tenure within the government.

So, how does this flexibility translate to strategic organisational decisions that must be taken when instating a behavioural unit within the government? Beyond project and programme design, behavioural units have been shown to have a critical role in delivering policy advice. Most often this advice has to do with building not just a behavioural tool in isolation from other existing policy tools but also one that is meant to complement and enhance the work of existing policy toolkits. As mentioned before, the literature on policy design and behavioural insights finds that nudges in the most general sense are rarely stand-alone instruments being implemented; instead, they tend to be situated in a larger implementation context that includes existing regulations as well as potentially counteracting behavioural interventions deployed by private stakeholders (Reference Ekhardt, Wieding, Mathis and TorEkhardt & Wieding, 2016). Therefore, in providing behavioural policy advice for existing policy portfolios, the behavioural unit’s role goes beyond analytical and technical capabilities for programme design, to one that approaches the role of knowledge brokering and traversing the science–policy divide.

5.2 Remaining Puzzles

An alignment with context represents the cornerstone of effective policy design. While the consideration of the influence of social and community norms is treated as foundational through different implementation contexts, specific attention to behaviourally informed instrument components in policy design is a relatively new phenomenon. Many policy instruments have behavioural assumptions as part of their foundation in order ‘to get people to do things they might not otherwise do or enable people to do things that they might not have done otherwise’ (Reference Schneider and IngramSchneider & Ingram, 1990, 513). Such considerations become especially pertinent, as evinced through the outcomes of a global public health crisis such as COVID-19.

As reiterated in this Element, despite their seeming embeddedness, the conceptual treatment of behavioural tools within the scholarship of policy instruments has remained piecemeal even though behavioural insights have been seen to inspire a wide variety of policy responses. These responses have been defined differently in the literature depending on whether researchers take on the narrower view of nudging or the wider scope of ‘thinks’ in BI, which galvanise more sustained and deliberative participation in the policy process. As surmised by scholars navigating behavioural public policy and policy instrument studies, ‘the spectrum of behavioral policy instruments is large, including efforts of political and administrative simplification to reduce the cognitive burden on citizens, education programmes for decisions under the conditions of risk and insecurity, techniques of social norms marketing, and behaviorally informed regulation’ (Reference Straßheim and BeckStrassheim & Beck, 2019, 3).

As the examples presented in this Element indicate, there is a need to organise the sheer diversity of behavioural means in terms of their place in the policy process. Some work towards this end has begun in line with distinguishing between policy instrument components that are ‘behaviourally-tested (i.e. initiatives based on an ad-hoc test, or scaled out after an initial experiment), or behaviourally-informed (i.e. initiatives designed explicitly on previously existing behavioural evidence), or behaviourally-aligned (initiatives that, at least a posteriori, can be found to be in line with behavioural evidence)’ (Reference Ciriolo, Lourenço, Almeida, Straßheim and BeckCiriolo et al., 2019, 6). Nudges, in this reasoning, fall into the last category of behaviourally aligned initiatives as these represent instrument calibrations by simplifying and framing of information; making changes to the physical environment in which information is presented; manipulating the default policy option; and the use of social norms, which need not always emanate from dedicated experiments or randomised controlled trials but may rather constitute a ‘what works’ principle on the ground during policy implementation.

That is, behavioural means working in policy toolkits dedicated to tackling public crises need not only appear early on in evidence-based design processes. The examples of BI modifications or interventions shown in this Element are most prominent at the level of calibrations made to instruments during their on-the-ground implementation, at sub-jurisdictional levels. At the outset, such calibrations are shown in this context to be mainly information-based and utilise governance resources of nodality in support of regulatory measures issued at the state level.

With more long-term motivations for policy design and the inclusion of more sustained behavioural implications in managing a variety of sectors, governance resources beyond information-based tools become necessary. The principle described earlier for nudges engenders a trade-off for policy design between supporting quick reactions in the short term and embracing the complexity of human understanding and cognition in the long term. To induce long-term behavioural reflections, there is a need to design more ‘thought provoking’ nudges that also uphold public and individual agency (Reference JohnJohn, 2018; Reference StraßheimStrassheim, 2021). Over time, this deliberate turn of the behavioural policy revolution can require more organisational decisions to be made in order to scale up policy responses with changing governance contexts. For example, in a public health crisis this means being able to deploy BI resources for health monitoring and testing, ramping up social distancing efforts, and recognising the need to quickly augment and mobilise health resources during disease spikes. At the same time and within the same crisis, as shown in the global experience with COVID-19, more organisation and incentive-based decisions can concurrently be taken to sustain behavioural change, therefore reflecting more ‘think’ style BI design choices that favour deliberation and collaboration.

All in all, a rapidly growing array of opportunities exists with which to hypothesise about the evolving role of BI means in the policy design process (at times of crisis and otherwise). Firstly, is there a temporal nature to how BI means are designed initially as calibrations (micro-level) but lead to instances of collaboration and co-creation through more multi-instrument policy mixes and programmes (meso-level) over time? Is there an implication of causality between the design of procedural elements (e.g., establishing analytical behavioural insights units, project-specific taskforces and commissions) and substantive elements (e.g., adjustments made to vaccination messaging) of particular policy instrument mixes in policy sectors that have become most aligned with behavioural research? Secondly, are there ‘orders’ to which combination of governance resources are necessitated by the design of BI policy mechanisms, and do these relate to the suggested temporal trajectory of their diversity? Thirdly, how do these concerns of temporality and governance resources in BI tool design reflect different policy capacities that must be activated at the operational, programme and, eventually, system-wide levels of policy design? Finally, past and potential limitations of BI require a critical discussion and an adjusted outlook for their application. This is especially relevant in the light of recent criticisms levelled at BI in several instances around the world: from publication bias, failure to reach the necessary scale and flawed assumptions to neglect of social context and ethical concerns, among others (Reference HallsworthHallsworth, 2023). These limitations warrant a deeper collaboration between various actors and across policy communities to ensure that BI means are implemented as part of a suite of policies without placing too much confidence in any one measure.

More completely, is the rise of behavioural policy tools spurring an equal shift in the work of policy design?