This chapter describes how behavioural science can inform policy analysis and design, reinforcing the quality and consistency of government decision‑making. There’s a risk that civil servants, like all persons, may be prone to patterns like confirmation seeking, framing and messenger effects, as well as decision “noise”. These risks can be mitigated through structured processes such as presenting evidence in a structured way, providing decision checklists, and sense-checking policy with citizens. International case studies show how these behaviourally informed tools can strengthen evidence-informed policy, reduce error, and support better policy outcomes.
Applying Behavioural Science in the Italian Public Administration
A Blueprint for Best Practices
Abstract
Create systems and processes that help avoid confirmation bias. Civil servants are likely to prefer and act on evidence that confirms their prior beliefs, which can bias the advice they give to decision makers. It is important to build steps in the policy process where assumptions are checked, and contradicting evidence is explicitly considered.
Look for and mitigate the messenger effect. Be mindful of how the source of communication of policy evidence can affect its perceived quality and how that may result in good evidence being disregarded or poor evidence being unfairly considered.
Build on existing evidence by tailoring it to local context. Challenges civil servants’ assumptions and address gaps in the literature by engaging with citizen stakeholders when designing policy.
Frame risk carefully to mitigate framing effects. Adopt guidelines requiring briefing documents and options papers to consider using a neutral frame when presenting the estimated impacts of different policy options. Where framing is unavoidable, develop guidance on how gain and loss framing can shape risk-preferences and align them with the administration’s priorities.
Develop checklists and tools to support consistent, high-quality decision making. Audit key decision points, such as policy briefings, recommendations and project stop/go points, and identify important considerations that, if overlooked, would result in worse outcomes. Develop checklists that guide administrations to consider all relevant factors and to assess the quality of their reasoning.
Civil servants develop policy proposals by gathering evidence and stakeholder input, by weighing options, and briefing governments on the courses of action available to them. At all points of policy design, civil servants make complex decisions to draw on the best existing evidence, generate new evidence, develop policy options and brief ministers so they can make an informed decision. However, they have only a finite amount of time and attention to process these trade-offs so they rationally employ behavioural strategies, rules of thumb and heuristics to manage this complexity (Simon, 1990[1]; Blanco, 2022[2]) While these strategies work well in many contexts, they can lead to decisions which, in some settings, are not aligned with long-term interests (Tversky and Kahneman, 1973[3]). Research has shown that civil servants are susceptible to predictable, psychological errors in their decision making. However, these errors can be mitigated by making behaviourally informed changes to the context in which civil servants operate.
Centres of government policy units responsible for policy design, development and providing advice to ministers. These teams might be embedded in central co-ordinating ministries or agencies. Their work shapes the quality of problem analysis and policy design across public administrations.
Policy teams in line ministries who are responsible for portfolio areas in their administration, e.g. policy teams responsible for primary care, mental health, or communicable disease in ministries of health, etc. These teams conduct the problem analysis and policy design which is reviewed by top management and provided to government in the form of briefings and options papers.
Policy evaluation, analysis and research teams who specialise in evidence-informed policy and methods to synthesise research that informs policy design. These teams might conduct original research to inform policy, articulate the uncertainties and gaps in existing research, or train policy teams in evidence-informed problem analysis and policy design.
Civil servants can adopt practices during policy analysis and design to support decision-making and evidence-informed policy. The overall evidence base is moderate-to-high quality. Behavioural science has improved policy design across a range of policy and public administration contexts, see Behavioural Government (Hallsworth et al., 2018[4]), Biased Policy Professionals (Banuri, Dercon and Gauri, 2017[5]), work by the Victorian Department of Premier & Cabinet (Australia), Applying behavioural insights to organisations (OECD, 2018[6]) and Behavioural insight and regulatory governance (Drummond, Shephard and Trnka, 2021[7]).
Practices in this chapter can support decision making. However, not all policies have been evaluated in a wide range of settings (e.g. they may have been tested in regulatory agencies and public hospitals but not in ministries). The evidence for specific practices in specific settings can be low-to-moderate. More research is needed into how these practices translate into diverse administrative settings.
These behavioural practices do not start from scratch; they enhance existing policy design and analysis. These practices reinforce the OECD’s Recommendation of the Council on Open Government to actively engage stakeholders in all phases of the policy-cycle and service design and delivery [OECD/LEGAL/0438], and as operationalised OECD Guidelines for Citizen Participation Processes (OECD, 2022[8]). In policy design, behavioural science is a complement, not a replacement.
Use behavioural science in policy design
Behavioural public administration seeks to embed evidence-informed approaches across all levels of government and all stages of the policy cycle. Doing so ensures that policies incorporate the best available behavioural science, supporting more impactful and cost-effective policies. The behavioural literature that civil servants can apply is vast. To organise this knowledge and make it tractable, civil servants can draw upon behavioural frameworks that summarise and organise this literature. A common starting point is the OCED’s ABCD framework (OECD, 2019[9]). This framework guides users to change behaviour by considering the following principles and questions:
Attention. People’s attention is limited but it can be guided towards behaviour change. What information seizes people’s attention? Is the targeted decision point well-timed? Is there a default that happens if the person is inattentive?
Beliefs. People rely on mental shortcuts that can lead to erroneous beliefs, although these beliefs can be corrected. What are people’s pre-existing beliefs? Are people over-optimistic?
Choice. People’s behaviour is influenced by the framing, social and situational context of their choices. How are choices framed? What choices are more desirable in that context?
Determination. People’s motivation can be harnessed to change behaviour. Do people see how the behaviour benefits them? What points of friction might demotivate people?
ABCD has been applied, for example, to understand the drivers of regulatory behaviour (Drummond, Shephard and Trnka, 2021[7]). Civil servants may draw on many different frameworks alongside ABCD. There is COM-B: Capability, Opportunity and Motivation combine to produce Behaviour (Michie, Stralen and West, 2011[10]). There is EAST, which seeks to make behaviour Easy, Attractive (or, Attention-grabbing), Social and Timely (BIT, 2024[11]). These frameworks connect broad behavioural principles with specific changes to the decision-making context (Drummond, Shephard and Trnka, 2021[7]). They are a valuable tool in civil servants’ toolkit to design better, more effective policies.
Civil servants often design legislation with a focus on legal clarity and administrative feasibility, while underestimating how citizens will experience and respond to new rules. In practice, laws can create unintended frictions if they fail to account for the everyday realities, capacities, and behavioural patterns of those affected.
To address this, Germany has developed a structured approach to integrate behavioural and citizen-centred thinking into the legislative process. The methodology, known as the Citizen Check, supports lawmakers in considering how new legal provisions will unfold in citizens’ lives before they are adopted.
The Citizen Check comprises five practical tools: the Addressee Profile, Insight Studies, Citizen Timeline, User Journey/User Blueprint, and the Rule Mapping Tool. These tools prompt legislators to identify the primary target groups, assess behavioural barriers and motivators, and evaluate whether citizens have the capacity to act as intended. For example, the Addressee Profile encourages reflection on who is affected, in what numbers, and what contextual factors may shape their behaviour. This helps identify blind spots, prioritise action, and inform more inclusive lawmaking.
One of the core tools, the Citizen Timeline, helps legislators visualise how a draft law might interact with a citizen’s daily life. It highlights potential points of confusion, burden, or disengagement, helping ensure that legislation is not only legally sound but also behaviourally realistic.
To support implementation, the tools will be embedded into a new digital platform, designed to make Citizen Check a regular part of the federal law-making process. Early pilot testing with federal ministries showed strong interest in the approach, with feedback highlighting both its practical value and the need for support in applying it. To address this, a peer-based support model has been introduced, with ministry staff trained to guide colleagues in using the tools, even without formal behavioural science expertise. By embedding behavioural and citizen-centred reflection into the early stages of lawmaking, this approach not only improves the usability of legislation but also contributes to institutionalising a culture that draws on evidence and behavioural insight to shape more realistic, inclusive, and effective laws.
Source: Based on information provided by the Federal Chancellery of Germany.
Mitigate confirmation seeking
Civil servants process information based on the best available evidence. However, studies by the World Bank and the Victorian Government in Australia suggest civil servants may be susceptible to confirmation seeking when they assess evidence. For example, civil servants may ignore or overlook evidence that contradicts their views when designing policy. Civil servants may consider contradictory information but hold it to a higher standard than confirmatory information, or they may process ambiguous information in a way that disproportionately favours their existing views (Hallsworth et al., 2018[4]). Studies by the World Bank and the Victorian Government in Australia suggest civil servants may be susceptible to confirmation seeking when they assess evidence.
Researchers assessed the extent to which civil servants displayed confirmation seeking at the World Bank (n = 2,053) and the UK’s Department for International Development (n = 825). In the study, participants were asked to interpret a study’s results and describe whether the study supported the effectiveness of an intervention. For half the participants, the study was about an ideologically neutral topic – whether a medical cream improved or worsened a skin rash. For the other half, the study was ideologically charged – whether raising the minimum wage is associated with higher unemployment for low-income workers. Within each topic (cream vs minimum wage) half of participants saw results that supported the intervention, and the other half saw the same results flipped to not support it.
The results found that participants showed strong signs of confirmation seeking. In the neutral skin cream condition, 65% of all participants correctly interpreted the data. In the minimum wage conditions, this figure dropped to 45%. Participants’ interpretation of the minimum wage evidence was strongly correlated with their prior beliefs. Participants who preferred a more equal distribution of income were more likely to say that the evidence supported the minimum wage when it did not. Participants who did not prefer more equal income were more likely to say the evidence did not support the minimum wage when it did. These results suggest that how civil servants interpret evidence may be susceptible to confirmation seeking, especially when they hold strong prior views.
Source: (Banuri, Dercon and Gauri, 2017[5]).
The risk of confirmation seeking is in the process of developing policy, not in assessing its outcomes. Oftentimes the evidence for or against a policy clearly favours one position over another. In such contexts, there are two promising practices that might mitigate confirmation seeking in policy analysis.
Simplify evidence to mitigate confirmation seeking
Simplifying complex evidence can reduce the scope for confirmation seeking to affect the interpretation of results. Civil servants who assess evidence and report findings from evidence may benefit from guidelines on how to present results more impartially. For example, frequency tables and graphs could state the results as a number and a proportion. Guidelines could recommend that percentage results state the baseline, the relative change and %-point change, for example “This policy was associated with a 50% increase in the outcome of interest (from 10% to 15%, or 5%-points)” Reviewers may benefit from a checklist for assessing graphs and tables to ensure they follow best practice methods of communicating results.
In a 2019 study, the Behavioural Insights Unit of the Victorian Department of Premier & Cabinet in Australia assessed civil servants’ susceptibility to confirmation seeking. The researchers asked participants how much they supported mandatory bicycle helmets on a 10-point scale (1 = strongly against, 10 = strongly in favour). Participants then read about a study assessing the impact of mandatory bicycle helmets on health outcomes. The fictitious study analysed areas that introduced a mandatory helmet law vs those that had not. Half of participants were shown results supporting the intervention, and the other half were shown the opposite results which did not support the intervention. Participants were asked if the study’s results showed that the intervention was associated with more or fewer healthy people.
Participants showed strong signs of confirmation seeking. Participants' interpretation of the study was strongly correlated with their prior beliefs. The more the results agreed with participants’ views, the more likely they were to interpret the evidence correctly, consistent with the World Bank’s study of confirmation seeking. The researchers then tested the impact of simplification on confirmation seeking. Participants were shown the same results but with results expressed as percentages rather than absolute numbers.
The simplification dramatically reduced confirmation seeking. Participants were significantly more likely to accurately describe the study’s results when it was described in percentage terms.
These results suggest that civil servants’ interpretation of evidence may be susceptible to confirmation seeking, especially when they hold strong prior views. However, public administrations can mitigate confirmation seeking by simplifying and disambiguating evidence so that the results are clearer.
% of correct responses
Source: (Nudgestock, 2021[12]).
Use “consider the opposite” framing
Another technique to mitigate confirmation seeking is to “Consider the Opposite”, which is for civil servants to ask themselves “What are some reasons that my initial judgment might be wrong?” (Hallsworth et al., 2018[4]; Larrick, 2004[13]). This approach is especially useful for mitigating the first forms of confirmation seeking, where civil servants exclude information from consideration, because it expands the range of evidence they consider. Civil servants can adopt “Consider the Opposite” framing in how they conduct policy research, and reviewers can also adopt this as part of the evaluation process.
Mitigate decision “noise”
Behavioural science traditionally seeks to reduce unjustified and systematic decision-making error. However, recent research has begun to also address noise – unjustified and non-systematic decision-making error (McKinsey, 2021[14]). Noise in administrative decision making represents random, inconsistent variability in decisions that ought to be uniform and consistent.
Noise can affect policy design and briefing. It is not limited to policy design, for example, it also impacts policy implementation, evaluation and hiring decisions. In the context of policy design, noise can affect:
Whether to include or exclude a study in a review of the policy evidence. One analyst may include the study, while another would exclude it, even when there are inclusion criteria that establish consistent rules for what sorts of studies public administrations are expected to draw upon.
Whether to include or exclude a policy option in an options paper. One manager may include an option while another may exclude it, even when the government has set clear expectations about what options include and exclude in options papers.
Whether to submit a draft briefing to the next level of approval or hold it back for further review.
Case conferences
Public administrations can minimise decision noise by implementing case conferences – a process whereby different people review the same information independently, share their decisions with each other, explain any differences in their decisions and then reach a consensus decision (Kahneman, Sibony and Sunstein, 2021[15]). The consensus decision can be reached by various means, including discussion and agreement, or a simple algorithm such as taking the average or median score (Hastie and Kameda, 2005[16]). These algorithms reduce unjustified variation in decisions, leading to more consistent outcomes.
Case conferences only work if they aggregate independent judgements because the differences in independent decisions cancel out unjustified variation. If participants discuss the information together before they reach their own decision, their decision is no longer independent. Case conferences are a promising decision tool in public administration, but more research is needed to evaluate them in practice.
Checklists
Public administrations can also reduce decision noise with behaviourally informed checklists. Checklists can improve the consistency of decision making and improve outcomes where it’s critical to ensure certain steps are not overlooked. Checklists have been shown to: guide Regulatory Impact Assessments in the UK (UK Comptroller and Auditor General, 2001[17]); conduct risk-based audits in Italy (OECD, 2021[18]); assess qualitative research in public health decisions in Canada (Yost et al., 2014[19]); improve aviation safety (Clay-Williams and Colligan, 2015[20]); and reduce surgical errors (World Health Organization, 2009[21]; Haynes et al., 2009[22]). When checklists were used at six Dutch hospitals, mortality rates almost halved from 1.5% to 0.8% (de Vries et al., 2010[23]).
Effective checklists use behavioural science to promote consistent decision making. High-quality checklists help decision-makers make use of their limited cognitive bandwidth, often summarised as the brain’s capacity to actively process 7 ± 2 pieces of information at once (Miller, 1956[24]). A checklist helps decision makers offload information from working memory to the checklist. Effective checklists are collaborative – they enable one person to check off the work of another, which leads to higher compliance than one person checking their own work (Clay-Williams and Colligan, 2015[20]).
Civil servants looking to develop checklists to support policy analysis and design can start with a checklist developed by Daniel Kahneman and colleagues to assess the quality of advice (Kahneman, Lovallo and Sibony, 2011[25]). The checklist comprises twelve questions that each address a decision-making risk:
Is there any reason to suspect motivated errors, or errors driven by the self-interest of the recommending team? (Self-interested or motivated reasoning)
Have the people making the recommendation fallen in love with it? (Affect heuristic)
Were there dissenting opinions within the recommending team? (Groupthink)
Could the diagnosis of the situation be overly influenced by salient analogies? (Saliency)
Have credible alternatives been considered? (Confirmation seeking)
If you had to make this decision again in a year, what information would you want, and can you get more of it now? (Recency and availability heuristic)
Do you know where the numbers came from? (Anchoring)
Can you see a halo effect? (Halo effect)
Are the people making the recommendation overly attached to past decisions? (Endowment and sunk cost thinking)
Is the base case overly optimistic? (Optimism bias, overconfidence, planning fallacy and competitor neglect)
Is the worst case bad enough? (Optimism bias and planning fallacy)
Is the recommending team overly cautious? (Loss aversion and risk aversion)
Civil servants may benefit from behaviourally informed checklists to support policy design and other high-stakes decisions. There is a great deal of research developing behaviourally informed checklists, and many public administrations use checklists in their policy workflow. However, relatively little research has evaluated the real-world impact of behaviourally informed checklists in public administration and policy design specifically. More research is needed to evaluate the impact of checklists in this context to understand the heterogenous circumstances where they add the most value.
Mitigate the messenger effect
Civil servants often assess evidence from multiple sources. However, it is mentally taxing to answer the difficult question “is this evidence correct?”. There is a risk that civil servants substitute this hard question with a simpler one, “is the messenger credible?”. This over-reliance on assessing the messenger is the messenger effect, where people evaluate the credibility of evidence based on irrelevant attributes of the source of the information rather than the merits of the evidence.
The scope of the messenger effect is sometimes misunderstood. Evaluating evidence based on its messenger is sound when the messenger's expertise, track record, or access to information genuinely affect the quality of the evidence they provide. For example, it is sensible to respond differently to research published by a high-performing policy team versus an industry-funded lobby group.
The challenge of the messenger effect is that it assesses evidence based on irrelevant aspects of the messenger. For example, in a mock jury study where participants were asked to assess the evidence provided in a witness statement, half of participants saw the witness make a statement themselves, while the other half saw the same statement made by a surrogate who read the statement on the witness’ behalf. Many of the jurors’ assessments of the statement were affected by the perceived credibility of the surrogate, even though the surrogate had no relation to the content of the statement (Kassin, 1983[26]). Those who were swayed by these irrelevant attributes of the surrogate messenger demonstrated the messenger effect.
In a 2019 study, the Behavioural Insights Unit of the Victorian Department of Premier & Cabinet in Australia assessed whether public servants were prone to the messenger effect when they assessed the advice of external consultants relative to internal policy units.
The researchers ran a field experiment with nearly 1 100 public servants in Victoria where participants read a fictitious business case for a temporary housing service. The business case was deliberately mediocre, containing spelling errors, cost estimates that did not add up and did not follow the government business case template.
The respondents were randomised into one of two conditions: one where the business case was said to be prepared by an “internal business unit” and another where it was prepared by an “external consultant”. The business cases were otherwise identical, and no more information was given about the authors. Participants were then asked whether they recommended the proposal be funded or not.
The researchers found evidence of a messenger effect in the study. When the business case was said to be authored by an internal business unit, just over 50% of participants recommended it be funded. When the case was said to be authored by an external consultant, the likelihood that participants recommended it be funded fell over 10%-points to just under 40%. This decrease was a statistically significant messenger effect that disadvantaged the external consultant. The researchers reasoned that this difference may have been because participants held the external consultants to a higher standard than an internal business unit, so the external consultants’ errors were penalised more.
Source: (BIT, 2021[27]).
There is little direct evidence of the messenger effect within public administration outside of these rare studies. Those studies that exist tend to not evaluate solutions to address them. Therefore, public administrations looking to mitigate the messenger effect may wish to explore mitigation strategies during funding processes by identifying opportunities to “blind” the reviewer from the name of the author of the business case – building on the findings of the Australian case study in box 4.4. This can make it easier for decision makers to decide funding and policies based on the strength of evidence presented, not the nature of the people presenting the evidence. For example, civil servants may update their guidelines to require that evidence reviews, funding submissions and meeting attachments be anonymised for reviewers by removing names and branding.
Tailor policy to local context
Civil servants seeking to change behaviour can benefit immensely from understanding the context in which it takes place. Human behaviour is driven by situational, social, cultural, economic and personal factors. Civil servants cannot understand these drivers and barriers based on published research alone. This is a particular challenge for civil servants, who may be prone to the illusion of similarity, where they over-estimate how much citizens share their preferences and circumstances; and the mind projection fallacy, where people assume their subjective interpretation of the world describes objectively what it is like for other people (Jaynes, 1990[28]; Hallsworth et al., 2018[4]). Civil servants may also experience the Curse of Knowledge, which leads experts to overestimate how familiar non-experts are with their subject matter. Civil servants who specialise in areas such as tax, health and roads, for example, are likely to over-estimate how much the average citizen knows about these areas.
Civil servants can overcome these limitations by conducting exploratory research. Early in policy development, exploratory research seeks to understand “What is the context shaping target behaviours?” and “Why people behave as they do” (OECD, 2019[9]). Administrative communications benefit from understanding the context that communication passes through to reach citizens (OECD, 2021[29]). When trying to shift citizens’ behaviour, seek first to understand its context.
Exploratory research is valuable because it engages citizens. Research methods from behavioural science, such as interviews, focus groups and fieldwork, are essential for understanding the qualitative experiences and reasons why people may behave the way they do. Quantitative methods, like surveys and data analysis, are well-suited to understanding the broad patterns of behaviour, how common it may be and how many people are affected (BIT, 2022[30]). There is value in civil servants selecting the right method for the question they seek to understand.
One example of this citizen engagement comes from the Netherlands. Recognising that behavioural barriers often limit the effectiveness of well-intentioned policies, the Dutch government has formally integrated behavioural research in its policymaking through a “capacity to act” standard. This case illustrates how behavioural science can be embedded at multiple stages of the policy cycle to sense-check that its interventions will be feasible once deployed in the real world.
Civil servants often overestimate the behavioural effects of policies and underestimate the difficulties people face in complying with rules and accessing services—particularly when it is assumed that simply providing information and (financial) incentives will lead to the desired behaviours. In practice, however, even when people know what they ought to do, they may act differently due to emotional, social, or contextual barriers. These might include psychological distress, major life events, the cumulative burden of interacting with multiple government agencies, or other psychosocial factors.
To address this, the Netherlands adopted a behavioural approach known as “capacity to act”, introduced by the Netherlands Scientific Council for Government Policy in 2017 and now formalised as a quality standard in Dutch policymaking. The goal is to ensure that policy and legislation more accurately reflect the real-world circumstances in which people are expected to operate and provide adequate measures to help people to be able to act.
The quality standard is operationalised through the Capacity to Act test (“doenvermogentoets”), which prompts civil servants to analyse how feasible the proposed policy is in the context of people’s daily lives. This can involve mapping citizen journeys, consulting with target groups and frontline staff, and, where possible, piloting measures to evaluate their behavioural impact. In doing so, it encourages civil servants to look beyond knowledge and technical capability, and to systematically consider the behavioural and situational factors that may limit citizens’ ability to respond to policy as intended.
The Netherlands integrates this behavioural lens through the Beleidskompas (“Policy Compass”), a comprehensive framework for Regulatory Impact Assessments (RIA) that all ministries apply when proposing, revising, or evaluating policies and regulations. By embedding the Capacity to Act test early in the policy cycle, civil servants create more feasible and effective policies tailored to citizens’ real-life contexts. In addition, the Unit for Judicial Affairs and Better Regulation Policy within the Ministry of Justice and Security is responsible for scrutinising compliance with the RIA framework and has made capacity to act one of its main priorities for scrutiny of regulations. The Advisory Division of the Council of State; the final independent advisory body in the preparatory phase before a legislative proposal is presented to Parliament; has also added capacity to act to its Assessment Framework. As a result, advice is now based not only on legal quality and enforceability (such as IT systems), but also on a behavioural lens.
Making capacity to act a quality standard for policy development is an example of how governments can embed the behavioral lens in their policy development process. This has the potential to make policies and regulations more realistic by accounting for how citizens actually behave, rather than assuming how they behave. Thereby increasing the effectiveness and efficiency of new policies in reaching policy goals.
Source: Based on information provided by the Behavioural Insights Network Netherlands (BIN NL).
Engaging with citizens enables better policy. Exploratory research ensures that policies have a clear-eyed view of the barriers faced by citizens, enabling civil servants to select the most appropriate solution for the problem. This research also ensures that interventions address citizens’ most difficult or common challenges, ensuring that policy is designed with outliers in mind and aligned with principles of human-centred design (OECD, 2024[31]). These methods help ensure that policy is based not on civil servants’ preconceptions but on the real experience of citizens, ensuring they feel heard and providing them with a platform to constructively shape policy.
Frame risk mindfully
Civil servants’ briefings, options papers and proposals are a key input to political authorities’ policy decisions. These policy options involve trade-offs between expected costs, benefits and risks. How civil servants present these options is based on their expertise, the government’s priorities and risk preferences, and the best available evidence. An under-appreciated factor shaping how this advice is presented is its framing. Research has shown that people’s choices are sensitive to whether risks and outcomes are framed in terms of loss, e.g. 1 000 people will die but 2 000 won’t die, or in terms of gains, e.g. 2 000 people will live but 1 000 will not (Tversky and Kahneman, 1981[32]). Even when the underlying outcomes are identical, people’s decisions can be influenced by the framing.
Framing can also influence decision makers’ choices. Risk aversion is the tendency to prefer an option where the outcome is certain rather than uncertain. For example, risk-averse persons may prefer an option to earn EUR 10 for certain, over an option with a 50% chance to earn EUR 20 but a 50% chance to earn nothing. Risk seeking is the opposite – a preference for riskier options over more certain ones. Research suggests that presenting options in a loss-frame promotes risk-seeking while the same options in the gain-frame promote risk aversion (Tversky and Kahneman, 1981[32]). This is reflected in the tendency for gamblers who are winning to prefer conservative bets that consolidate their gains, but gamblers who are losing prefer risky bets to recover their losses.
Civil servants may be naturally risk-averse. A study of Italian professionals found that individuals in employments with fixed incomes (e.g. teachers and civil servants) are likely to exhibit higher risk aversion that individuals whose employments entail a variable income stream (Mauro and Musumeci, 2011[33]). However, research suggests that civil servants’ risk preferences are nonetheless influenced by framing. A World Bank study of framing and risk among civil servants found that framing exerted a significant impact on decision making. Similar effects have been replicated across meta-analyses in non-government settings (Kühberger, 1998[34]; Steiger and Kühberger, 2018[35]).
A World Bank study evaluated how civil servants’ preference for risk changed under gain and loss frames. The study was conducted with development policy professionals from the World Bank (n = 2,053) and the UK’s Department for International Development (n = 825). Participants were shown a hypothetical scenario where they had to choose between two potential treatments for a new virus that was expected to infect 12 000 people. Participants were randomly assigned to either a gain frame, which expressed their policy options in terms of lives saved, and the other half a loss frame, which expressed their policy options in terms of lives lost. Participants were presented with two policy options: A risk-averse Treatment A which would save / lose a certain number of lives, and a risk-seeking Treatment B which had a 1/3 chance of saving all lives / preventing all deaths but a 2/3 chance it would not. Despite the different framing, both the risk-averse and risk-seeking policy responses had identical expected outcomes.
The framing was associated with a significant change in risk preference. When presented with the gain frame, nearly 22% of respondents chose the risk-seeking Treatment B. However, when presented with the loss frame, the share who chose the risk-seeking Treatment B grew to 65%, a 43%-point increase (p < 0.01).
Source: (Banuri, Dercon and Gauri, 2017[5]).
Civil servants can mitigate the effects of framing on decision making under uncertainty. Civil servants may wish to:
Neutralise framing by presenting the gain and loss frames together. Civil servants can implement explicit guidelines that suggest briefing documents and options papers adopt a neutral or joint gain/loss frame in how they present policy options. For example, when briefing on the economic impacts of a policy, civil servants might say “We estimate that under Option A, 1 000 jobs will be lost but 2 000 jobs will be created”, which combines both the gain frame (jobs created) and loss frame (jobs lost) simultaneously, neutralising the framing effect. Civil servants can consider embedding this technique in guidelines for reviewers to check the framing of policy options in briefing documents and require reviewers to confirm if they approve the framing in the final version.
Use framing deliberately. There may also be circumstances where civil servants may not wish to eliminate the framing effect and wish use it deliberately because some areas of public administration may wish to encourage risk-seeking or risk-averse behaviours. For example, government innovation teams may wish to encourage civil servants to be less risk-averse in terms of trying new innovations, and more willing to “fail fast” in safe, sandboxed environments. Here, civil servants may benefit from using a frame that encourages or discourages risk-averse behaviour.
Public administrations may wish to audit how they currently frame risk in policy documents, such as briefings and options papers, to assess where framing may impact decision making. Where a neutral frame would be appropriate, civil servants could create checks and instructions how to combine the gain and loss frames. Where decision making may be too risk-averse or too risk-seeking, they could consider whether employing a deliberate frame would be appropriate to counteract this tendency. Civil servants could develop guidelines for reviewers to check the framing of policy options in briefing documents and require the reviewer to approve the use of joint, neutral or deliberate framing in the document.
Create decision tools that help, not hinder, the user. Civil servants may be wary of processes that appear to challenge their expertise or autonomy. Framing behavioural tools as supports for professional judgement, rather than constraints, can help overcome this barrier.
Monitor for and prevent superficial compliance. Tools must be tailored to context, kept concise, and regularly updated. Overly complex or generic tools might be ignored. For example, an effective checklist is as short as possible. Civil Servants should evaluate, iterate and design these decision tools with users to ensure they continue to add value.
For further reading on how civil servants can apply practical behavioural approaches throughout policy analysis and design to strengthen decision making and support evidence informed policymaking, see for example: Applying Behavioural Insights to Organisations: Global Case Studies (OECD, 2018[6]); Behavioural Government (Hallsworth et al., 2018[4]); Behavioural Insight and Regulatory Governance: Opportunities and Challenges (Drummond, Shephard and Trnka, 2021[7]); and Biased Policy Professionals (Banuri, Dercon and Gauri, 2017[5]).
References
[5] Banuri, S., S. Dercon and V. Gauri (2017), Biased policy professionals, https://openknowledge.worldbank.org/server/api/core/bitstreams/306d3735-0e22-5b10-8f87-32bc3ba5bdbe/content.
[11] BIT (2024), EAST Framework: four simple ways to apply behavioural insights (revised and updated edition).
[30] BIT (2022), Target, Explore, Solution, Trial, Scale. An introduction to running simple behavioural insights projects.
[27] BIT (2021), International approaches to applying BI: Part Two - Asia Pacific, https://www.youtube.com/watch?v=pVjHKnJSZe0.
[2] Blanco, F. (2022), Cognitive Bias, Springer, Cham, https://link.springer.com/rwe/10.1007/978-3-319-55065-7_1244.
[20] Clay-Williams, R. and L. Colligan (2015), “Back to basics: checklists in aviation and healthcare”, BMJ Quality & Safety, Vol. 24/7, pp. 428-431.
[23] de Vries, E. et al. (2010), “Effect of a Comprehensive Surgical Safety System on Patient Outcomes”, The New England Journal of Medicine, Vol. 363/20, pp. 1928-1937.
[7] Drummond, J., D. Shephard and D. Trnka (2021), “Behavioural insight and regulatory governance: Opportunities and challenges”, OECD Regulatory Policy Working Papers, No. 16, OECD Publishing, Paris, https://doi.org/10.1787/ee46b4af-en.
[4] Hallsworth, M. et al. (2018), Behavioural government. Using Behavioural Science to Improve How Governments Make Decisions, The Behavioural Insights Team.
[16] Hastie, R. and T. Kameda (2005), “The Robust Beauty of Majority Rules in Group Decisions.”, Psychological Review, Vol. 112/2, pp. 494-508, https://doi.org/10.1037/0033-295x.112.2.494.
[22] Haynes, A. et al. (2009), “A surgical safety checklist to reduce morbidity and mortality in a global population”, New England Journal of Medicine, Vol. 360/5, pp. 491-499.
[28] Jaynes, E. (1990), Probability Theory as Logic, Springer.
[25] Kahneman, D., D. Lovallo and O. Sibony (2011), “Before you make that big decision”, Harvard Business Review, Vol. 89/6, pp. 50-60.
[15] Kahneman, D., O. Sibony and C. Sunstein (2021), Noise: A flaw in human judgment, Little, Brown Spark.
[26] Kassin, S. (1983), “Deposition testimony and the surrogate witness: Evidence for a“ messenger effect” in persuasion”, Personality and Social Psychology Bulletin, Vol. 9/2, pp. 281-288.
[34] Kühberger, A. (1998), “The influence of framing on risky decisions: A meta-analysis.”, Organizational Behavior and Human Decision Processes, Vol. 75/1, pp. 23-55.
[13] Larrick, R. (2004), Debiasing, Blackwell Publishing.
[33] Mauro, C. and R. Musumeci (2011), “Linking risk aversion and type of employment”, The Journal of Socio-Economics, Vol. 40/5, pp. 490-495.
[14] McKinsey (2021), Sounding the alarm on system noise, https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/sounding-the-alarm-on-system-noise#/.
[10] Michie, S., M. Stralen and R. West (2011), “The behaviour change wheel: A new method for characterising and designing behaviour change interventions”, Implementation Science, Vol. 6/1, https://doi.org/10.1186/1748-5908-6-42.
[24] Miller, G. (1956), “The magical number seven plus or minus two: some limits on our capacity for processing information”, Psychol Rev, Vol. 2/81, pp. 81-97.
[12] Nudgestock (2021), Chiara Varazzani - On the Pace of Behavioural Science in Government | Nudgestock 2020.
[31] OECD (2024), Global Trends in Government Innovation 2024: Fostering Human-Centred Public Services, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/c1bc19c3-en.
[8] OECD (2022), OECD Guidelines for Citizen Participation Processes, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/f765caf6-en.
[18] OECD (2021), OECD Regulatory Policy Outlook 2021, https://doi.org/10.1787/38b0fdb1-en.
[29] OECD (2021), OECD Report on Public Communication: The Global Context and the Way Forward, OECD Publishing, Paris, https://doi.org/10.1787/22f8031c-en.
[9] OECD (2019), Tools and Ethics for Applied Behavioural Insights: The BASIC Toolkit, OECD Publishing, https://doi.org/10.1787/9ea76a8f-en.
[6] OECD (2018), “Applying Behavioural Insights to Organisations: Global Case Studies”, OECD Regional Development Papers, No. 76, OECD Publishing, Paris, https://doi.org/10.1787/50a1919d-en.
[1] Simon, H. (1990), Bounded rationality, Palgrave Macmillan, London, https://doi.org/10.1007/978-1-349-20568-4_5.
[35] Steiger, A. and A. Kühberger (2018), “A meta-analytic rep-appraisal of the framing effect”, Zeitschrift für Psychologie, Vol. 226/1, pp. 45-55.
[32] Tversky, A. and D. Kahneman (1981), “The framing of decisions and the psychology of choice”, Science, Vol. 211/4481, pp. 453-458.
[3] Tversky, A. and D. Kahneman (1973), On the psychology of prediction, Psychological Review, https://psycnet.apa.org/doi/10.1037/h0034747.
[17] UK Comptroller and Auditor General (2001), Better Regulation: Making Good Use of Regulatory Impact Assessments.
[21] World Health Organization (2009), WHO Surgical Safety Checklist, https://www.who.int/teams/integrated-health-services/patient-safety/research/safe-surgery/tool-and-resources (accessed on July 2025).
[19] Yost, J. et al. (2014), “Tools to support evidence-informed public health decision making”, BMC Public Health, Vol. 14/1, p. 728.