Applying Behavioural Science in the Italian Public Administration

5. BPA for policy implementation

Abstract

This chapter describes how behavioural science can strengthen policy implementation by addressing systematic biases in project planning and delivery. Practices such as pre‑mortems, calibrated forecasting, smart anchors, reference class forecasting, red teaming and stop/go decision points can mitigate over‑optimism, over‑confidence, anchoring and sunk cost thinking. Shifting from ad‑hoc to institutionalised, behaviourally-informed practices can help governments to implement policy on time, within scope and under budget.

Key messages

• Conduct pre-mortems. Encourage the use of pre-mortems – a project planning exercise that imagines the project has already failed and team members must understand why. This technique helps counteract early-project optimism and to identify implementation risks.

• Develop, communicate, and calibrate forecasts. Prompt civil servants to make concrete forecasts, communicate and record them, and then consider their based on new information as it arises. These forecasts should consider type, time, cost, outcomes and challenges of previously delivered projects and be transparent on how these inputs are used to determine the future outlook. This will enable civil servants to develop better, more accurate plans and more realistic appraisals of whether a project will succeed.

• Set smart anchors for developing project plans that are based on forecasts and which nudge project managers to set more realistic expectations of success. These smart anchors could inform project guidance and templates so that planners and managers base their implementation plans on a more realistic assessment of the likelihood a project will experience a cost over-run, timeline delays, staff turnover, and other risks.

• Create Red Teams to stress-test polices and plans. Involve arms-length experts to provide structured dissent and constructive criticism of important decisions. These adversarial collaborators can stress test the assumptions, evidence and quality of the decision making underpinning a course of action, leading to better implementation and delivery.

• Mitigate sunk cost thinking. Break major projects into distinct phases with clear gateway decisions to proceed or not. Set explicit guidance to only consider future costs and benefits, and none of the “sunk” costs already spent.

Why it matters

Public administrations are responsible for effective policy implementation and project management. This includes activities as diverse as publishing a major report, performing a lapsing programme evaluation, hiring new staff, delivering an inter-ministerial meeting, conducting a suite of public consultations, co-ordinating a network meeting or delivering a technical support instrument. In all cases, the project must achieve a set of outcomes by delivering agreed outputs within a defined scope, budget and timeframe.

Implementation is a core responsibility at the intersection of policy, what the government will achieve, and operational delivery. Yet poor implementation endangers the government’s agenda by failing to achieve the policy outcomes and organisational objectives set for public administration. The scale of public activities means that small errors in project management and implementation can compound to create substantial problems and lost public value.

The success and failure of implementation is often behavioural. Successful delivery can be endangered by over-optimism, poor planning and ineffective decision-making. The absence of best-practice project management can mean that civil servants fall into mental traps, ignoring uncomfortable information and doubling down on failing projects. By implementing behaviourally informed decision points, tools and practices, public administrations can mitigate these risks to project management and policy implementation.

Whom it involves

• Central project management offices who monitor progress and facilitate public administrations to deliver projects across the whole of public administration.

• Strategy and implementation units within a public administration who actively support or deliver priority or complex projects.

• Oversight bodies who audit government projects, assess whether the project was managed well, and evaluate whether it followed best practice project management.

• Project managers at all levels of administration.

How to improve implementation

The behaviourally informed implementation practices in this chapter draw on low-to-moderate quality evidence, although the field is growing. There is clear evidence that how civil servants deliver projects and policies can be impacted by over-confidence, anchoring and over-optimism – all are well-established in behavioural economics. However, there are few experimental studies of public administrations who implemented practices to combat these tendencies.

Therefore, the evidence for how these practices combat these tendencies is preliminary. They are based on behavioural guides, case studies and practitioner wisdom as well as qualitative methods and analogies from related fields. These practices are conceptually and theoretically sound, but there is a need for these practices to be evaluated experimentally and for these evaluations to be shared widely to establish their impact in public administration.

Conduct pre-mortems

Many projects experience implementation difficulties due to over-optimism. Psychologist Daniel Kahneman illustrated this when authors were writing a textbook for a Ministry of Education. The authors estimated the textbook would take two years to write, perhaps two and a half years at most. Kahneman asked a pedagogical expert how long other teams had taken to write similar textbooks. They responded that about 40 percent never finished, and those that did finish never completed it in under seven years. The team then began writing the textbook, which took eight years to finish, by which time the Ministry of Education no longer needed the textbook (McKinsey, 2011[1]). This same over-optimism can impact a civil servant’s estimation of any task, such as the time required to write a report, the budget necessary to deliver a programme of work, or the approvals necessary before a decision can be issued.

One solution is that over-optimism may be addressed by conducting a pre-mortem – an exercise during a project planning phase where teams imagine that the project has already ended in failure, and they need to understand why. Participants independently write their reasons for failure and then discuss them as a team (Klein, 2007[2]). A pre-mortem legitimises sharing negative concerns, and its “prospective hindsight” frame can stimulate more creative thinking (Buehler, Messervey and Griffin, 2005[3]; Klein, 2007[2]). Pre-mortems have the potential to mitigate over-optimism by prompting civil servants to identify more substantive reasons why their projects and policies might fail. Civil servants may consider developing pre-mortem templates, include them in project management toolkits and mandate pre-mortem exercises when planning projects to combat over-optimism.

Calibrate forecasts

Civil servants, like most people, can be prone to over-confidence (Tetlock, 2005[4]). A literature review found that when people estimate the likelihood of an event, they experience an “overconfidence gap” of 15 percentage points on average, where an estimated 90% likelihood tends to only happen 75% of the time (Alba and Hutchinson, 2000[5]; Egan, 2025[6]). An online study of over 4 000 participants in five countries found that respondents were, on average, 20 percentage points over-confident in their answers (Egan, Tran and Whitwell-Mak, 2025[7]). Over-confidence is a common phenomenon, and civil servants are not immune. One study of civil servants found they grew more overconfident the more years of experience they had (Liu, Stoutenborough and Vedlitz, 2016[8]).

This matters for policy and project implementation. Civil servants make estimates all the time: how long they need to write a report, a white paper or an audit; how many staff are needed to deliver a project; what will be the impact of a policy; how much money is required in a project budget; and how many stakeholders will participate. Miscalibration jeopardises this implementation: Deliverables may arrive late; staff may be over-worked; policies may backfire; budgets may run out; and stakeholders may not engage. Properly forecasting these outcomes enables civil servants to better manage these risks and their projects.

One approach to mitigate this over-confidence is to implement calibration feedback. Calibration feedback requires participants to make concrete estimates of project outcomes like “a 75% chance the project is delivered on time”, rather than ambiguous statements like “a distinct possibility the project is delivered on time”, so their calibration accuracy can be assessed. Later, participants’ calibration is scored based whether their estimates were over-confident (e.g. their 90% estimates only happened 70% of the time), under-confident (e.g. their 50% estimates happened 70% of the time) or well-calibrated. This feedback enables participants to become better calibrated, leading to more accurate estimates in the long term (Mellers et al., 2015[9]).

Another approach is check for over-confidence or mis-calibration. Civil servants can be encouraged to use cautious language in order to reduce uncertainty. For example, civil servants could be encouraged to say: “this depends on…” or to express conditional thinking such as “if this, then that”. Avoiding asking civil servants “are you sure?” and instead asking “how sure are you?”, invites a more nuanced estimate of what will happen (Egan, 2025[6]). These practices are small but may promote better epistemology in public administration.

Set smart anchors

Anchoring is a tendency to over-rely on one piece of information, typically the first piece of information, when making decisions (Flyvbjerg, 2021[10]). When project managers estimate the time, effort or budget to deliver a project, they may anchor to an initial estimate and under-correct their initial judgement, e.g. “Most projects take 12 months, but this one is complex, so let’s plan for 14 months” when an estimate of 24 months or more might be more appropriate.

Civil servants could mandate setting “smart” anchors in project planning templates, for example, recommending that managers consider, as starting point, the risk that their project’s costs over-run by X%, where X% is the anchor that managers build their estimate upon. A smart anchor would set X% at a more pessimistic anchor that nudges civil servants away from their initial, potentially optimistic estimate. Similar smart anchors can be set for project delays, team members leaving the project, and other risks, nudging project managers away from overly optimistic estimates towards estimates grounded in past practice.

Develop reference class forecasts

Another solution to over-optimism is reference class forecasting – a method of estimating project implementation based on the outcomes of similar past projects (the reference class). The reference class anchors project managers on a more realistic "outside view" of their project and its base-rate of success (Kahneman and Tversky, 1977[11]; Flyvbjerg, 2008[12]). Research suggests that reference class forecasting can mitigate optimism bias in major infrastructure projects if the reference class is calibrated to the project (Baerenbold, 2023[13]). Reference class forecasts may also help project planners delivering smaller-scale projects, but this is an extrapolation from infrastructure studies. There remains a need for reference class forecasts to be evaluated, and for these evaluations to be published.

Some reference classes are hard to define. For example, it is not obvious what the correct reference class was for a novel coronavirus in March 2020: all pandemics, or only certain types of pandemics? Reference classes can be hard to delineate. However, they still offer value for common and clear classes of forecasts, such as project timeframes, outcomes and implementation risks.

In practice, civil servants could require project managers to consider reference class forecasting as an explicit step in their project management guidelines. They could engage researchers to develop high-quality reference class forecasts for common, high-stakes or high-risk classes of projects and decisions. Project Gantt chart templates could prompt users to consult past projects when estimating the time needed for their own project. Template agendas for internal planning meetings could include structured prompts to consider references classes in their discussion of upcoming work.

Organise “Red Teams”

Another solution to over-optimism is to involve outside parties who are less emotionally invested in the project and, therefore, may be less prone to biases (Buehler, Griffin and Ross, 1995[14]; Buehler, Griffin and Ross, 2002[15]). These adversarial collaborators can take the form of Red Teams, a team at arms-length from the project who challenge its assumptions, evidence and decision making. Effective Red Teams do not just offer criticism, but rather provide structured dissent, as described in public sector handbooks (US Government, 2009[16]; UK Ministry of Defence, 2021[17]). These include:

• Key Assumptions Check, in which a Red Team reads a policy, plan or project closely, notes any explicit and implicit assumptions that they see, assesses each assumption’s confidence, contingency and other factors, and reports whether they consider each assumption to be “supported”, “caveated” or “unsupported” (UK Ministry of Defence, 2021[17]).

• Analysis of Competing Hypotheses, in which a Red Team reviews a policy team’s reasoning why they think a scenario is true (e.g. that the wait-list for a service has increased because demand for the service increased); develops a matrix of competing hypotheses (e.g. that the supply of service providers has decreased, that there is a bottleneck in the service front-end, etc.); attempts to systematically disprove each hypothesis by considering contrary evidence; and reports the relative strength of each hypothesis and what evidence is needed to disconfirm any remaining hypotheses (US Government, 2009[16]).

The evidence of Red Teams’ effectiveness is primarily based on case studies (UK Ministry of Defence, 2021[17]). Systematic evidence of its impact across public administration is limited. Red teaming is difficult to scale and the importance of adaptation to different decision-making processes lends itself to more qualitative evaluations. Nonetheless, Red Teams are a promising BPA practice that warrants further study.

Mitigate sunk cost thinking

A challenge in public administration is that projects and policies continue to be implemented even when they no longer serve a useful purpose. Public administrations may hesitate to submit a briefing that recommends ending an ineffective policy or cancelling an unsuccessful project because of sunk cost thinking – the tendency to continue implementing ineffective policy or projects based on money already spent and giving too little weight to future costs and benefits (Miller, 2019[18]). Sunk cost thinking risks leading public administrations to throw good money after bad. One test of sunk cost thinking is to ask when a project is running overbudget: “If we weren’t already delivering this project, would we start now?” If the answer is No, but the project continues nonetheless, that may indicate sunk cost thinking.

Sunk cost thinking is wide ranging. It is most often associated with infrastructure projects, where project managers focus on “what has already been spent rather than what could be saved by taking a difficult decision to say, No, that is it” (House of Commons, 2019[19]). However, sunk cost thinking also impacts activities such as public administrations continuing to procure goods that no longer represent value for money, and continuing public sector communication campaigns that do not have the desired effect (Banuri, Dercon and Gauri, 2017[20]). Sunk cost thinking can impact any project that invests public resources.

Box 5.1. Case study: Sunk cost thinking among civil servants in the UK and World Bank

In a World Bank study, researchers examined the prevalence of sunk cost thinking among development policy professionals from the World Bank (n = 2 053) and the UK’s Department for International Development (n = 825). Participants were asked to review a $500 million biodiversity project that was no longer viable because it was located on the site of a newly announced hydropower dam.

Participants were asked whether to invest the unspent funds. Participants were randomised into groups where the share of funds already spent was, e.g. 30%, 50% or 70%. The level of funds already spent (sunk cost) was irrelevant to the decision to invest the unspent funds because the new development plan meant any additional spending would be of no value.

The researchers found that the higher the sunk cost, the more respondents invested the unspent funds. When 30% of funds had been spent, 40% of respondent chose to invest the unspent funds. When 50% had been spent, the share rose to 43% (p < 0.1). When 70% were spent, it rose to 49% (p < 0.01).

Respondents reported that they thought their colleagues were even more likely to continue funding the project than they were themselves. Environmental experts were particularly susceptible, being 7.3% more likely to continue funding, which suggested that professional identity and domain commitment may exacerbate sunk cost thinking.

Source: (Banuri, Dercon and Gauri, 2017[20]).

One promising mitigation strategy is to explicitly break projects into distinct phases with clear gateway decisions to proceed or not proceed. The helps to “rule out” consideration of past costs because the decision point is explicitly about future costs and benefits. Civil servants could create gateway decisions such as stop / go decision points in projects and policies. These decision points would state what costs and benefits can and cannot be considered whether to stop or continue implementing a policy or project. The UK Government’s National Audit Office reports that the introduction of such decision points improved assurance and control of major projects (UK National Audit Office, 2010[21]). Civil servants may consider developing structured “decision briefings” which omit or separate past costs, before the decision point, from future costs and benefits.

Sunk cost is far from the only reason these projects may continue: project cancellations grab attention, but ongoing and underperforming projects are less noticeable; and public administrations may practice defensive bureaucracy if they’re punished for cancelling projects but face lesser repercussions for delivering projects that no longer serve a purpose (Lorenzoni, 2023[22]). However, where sunk cost thinking may impact policy and project implementation, there is value in public administrations exploring means to address it.

Behaviourally informed insights

• Behavioural practices are not a silver bullet. Public administrations may continue ineffective programs for reasons unrelated to evidence or outcomes, such as “defensive bureaucracy”, where civil servants do not cancel, or recommend cancelling, projects due to the perceived greater risk of cancelling rather than continuing it (even if the project does not deliver public value). Practices like stop / go decisions alone are unlikely to mitigate this problem. Broader changes may be warranted, such as evidence-informed decision tools that justify cancellation and may protect civil servants from the repercussions of cancelling a project (Lorenzoni, 2023[22])

• Embed better decision-making in practice. Civil servants may consider creating a “better decision-making network” across and within public administrations. This network might include members who contribute to the development of reference class forecasts, help develop smart anchors for planning templates, participate in pre-mortems, rotate into and out of Red Teams, and participate in calibration exercises.

For further reading on practical behavioural approaches that can support civil servants and public administrations in strengthening decision making, effective policy implementation and project management, see for example: Biased Policy Professionals (Banuri, Dercon and Gauri, 2017[20]); Performing a Project Premortem (Klein, 2007[2]); Red Teaming Handbook (UK Ministry of Defence, 2021[17]) and Stay calibrated: a practical guide to debiasing decision-making (Egan, 2025[6]).

References

[5] Alba, J. and J. Hutchinson (2000), “Knowledge Calibration: What Consumers Know and What They Think They Know”, Journal of Consumer Research, Vol. 27/2, pp. 123-156, https://doi.org/10.1086/314317.

[13] Baerenbold, R. (2023), “Reducing risks in megaprojects: The potential of reference class forecasting”, Project Leadership and Society, Vol. 4, p. 100103.

[20] 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.

[15] Buehler, R., D. Griffin and M. Ross (2002), “Inside the Planning Fallacy: The Causes and Consequences of Optimistic Time Predictions”, in Heuristics and Biases, Cambridge University Press, https://doi.org/10.1017/cbo9780511808098.016.

[14] Buehler, R., D. Griffin and M. Ross (1995), “It’s About Time: Optimistic Predictions in Work and Love”, European Review of Social Psychology, Vol. 6/1, pp. 1-32, https://doi.org/10.1080/14792779343000112.

[3] Buehler, R., D. Messervey and D. Griffin (2005), “Collaborative planning and prediction: Does group discussion affect optimistic biases in time estimation?”, Organizational Behavior and Human Decision Processes, Vol. 97/1, pp. 47-63, https://www.sciencedirect.com/science/article/pii/S0749597805000348.

[6] Egan, M. (2025), Stay calibrated: A practical guide to debiasing decision-making, https://www.bi.team/publications/stay-calibrated-a-practical-guide-to-debiasing-decision-making/.

[7] Egan, M., C. Tran and J. Whitwell-Mak (2025), Overconfidence is the Norm: Results From a Five-Country Survey, https://media.licdn.com/dms/document/media/v2/D4D1FAQFi9j75CLbuJg/feedshare-document-pdf-analyzed/B4DZp8dAq7HwAY-/0/1763024573846?e=1764201600&v=beta&t=iPjmMfXgCuFa4MOrhvMpY3KZE01C28v46yI5vnhOgMo.

[10] Flyvbjerg, B. (2021), “Top Ten Behavioral Biases in Project Management: An Overview”, Project Management Institute, Vol. 52/6, pp. 531-546, https://doi.org/10.1177/87569728211049046.

[12] Flyvbjerg, B. (2008), “Curbing optimism bias and strategic misrepresentation in planning: Reference class forecasting in practice”, European planning studies, Vol. 16/1, pp. 3-21.

[19] House of Commons (2019), Public Administration and Constitutional Affairs Committee, https://committees.parliament.uk/oralevidence/9545/html/.

[11] Kahneman, D. and A. Tversky (1977), Intuitive prediction: Biases and corrective procedures.

[2] Klein, G. (2007), Harvard Business Review.

[8] Liu, X., J. Stoutenborough and A. Vedlitz (2016), “Bureaucratic expertise, overconfidence, and policy choice”, Governance, Vol. 30/4, pp. 705-725, https://doi.org/10.1111/gove.12257.

[22] Lorenzoni, L. (2023), “Defensive bureaucracy in Italy: An introduction”, Rivista Trimestrale di Scienza dell’Amministrazione, https://doi.org/10.32049/RTSA.2024.1.02.

[1] McKinsey (2011), Beware the ‘inside view’, https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/daniel-kahneman-beware-the-inside-view.

[9] Mellers, B. et al. (2015), “Identifying and Cultivating Superforecasters as a Method of Improving Probabilistic Predictions”, Perspectives on Psychological Science, Vol. 10/3, pp. 267-281, https://doi.org/10.1177/1745691615577794.

[18] Miller, C. (2019), “Sunk Costs and Political Decision Making”, Oxford Research Encyclopedia of Politics, https://doi.org/10.1093/acrefore/9780190228637.013.1022.

[4] Tetlock, P. (2005), Expert Political Judgment: How Good Is It? How Can We Know?, Princeton University Press, https://www.jstor.org/stable/j.ctt1pk86s8.

[17] UK Ministry of Defence (2021), Red Teaming Handbook, https://assets.publishing.service.gov.uk/media/61702155e90e07197867eb93/20210625-Red_Teaming_Handbook.pdf.

[21] UK National Audit Office (2010), Assurance for high risk projects.

[16] US Government (2009), A Tradecraft Primer: Structured Analytic Techniques for Improving Intelligence Analysis, https://www.cia.gov/resources/csi/static/Tradecraft-Primer-apr09.pdf.