Applying Behavioural Science in the Italian Public Administration

7. BPA for people management

Abstract

This chapter describes how behavioural science can enhance public sector people management. There’s a risk that civil servants, like all persons, may be prone to patterns like in-group favouritism, the halo effect, and anchoring on early information. Behavioural practices – such as skill-based assessments, joint evaluation, and better defaults for flexible work – can mitigate these risks, improving fairness and effectiveness. Behavioural practices can also improve how public administrations develop their workforce by basing their people management and training on the science of learning. Applying these behavioural techniques to people management can help civil servants to recruit the best people for the job, perform to a high standard, and continually develop in their role.

Key messages

• Change the default on flexible working. Many workers prefer flexible work, and evidence suggests that flexible work can enhance productivity. However, the default in many public administrations is that flexible work must be requested. Switching the default to advertise flexible working when it is available can help fill vacancies and increase workforce productivity.

• Apply techniques to improve hiring decisions, such as crowd-sourcing multiple viewpoints, using joint evaluation across written responses, randomising the order that assessors review responses, and anonymising written tasks. These techniques are promising but under-researched and would benefit from closer evaluation as they are implemented.

• Encourage applicants to do their best and reapply if unsuccessful. Under-represented groups may be at risk of attrition during recruitment and may be less likely to reapply if unsuccessful. Small, behaviourally informed messages of encouragement sent to all candidates can have a disproportionate and positive effect on the chances that candidates from under-represented groups persist with their application and apply for future roles if unsuccessful.

• Prioritise effective training courses. Some professional development courses are not as impactful as they could be because they do not employ effective learning techniques. Professional training tends to be more effective when it uses spaced repetition, behavioural modelling, and test-based learning. Help participants apply their learning with training tailored to their work and creating post-training implementation plans. Evaluate the impact of training to ensure it represents value for money.

• Foster an environment of continuous learning. Use performance evaluations as an opportunity for learning and motivation to stretch oneself. Foster communities of practice. Embed informal and ongoing project learnings through after-action reviews and debriefs.

Why it matters

Public administrations are called on to recruit, onboard, develop, promote and offboard their staff. Public administrations can adopt proactive strategies that tailor human resource management, including recruitment, to promote the systematic attraction, identification, development, retention, and deployment of talent within the organisation. Introducing behavioural science alongside these strategies can make public sector management more effective. The civil servants who manage staff are, like all people, prone to heuristics and mental shortcuts. For example, how managers assess the civil servants’ performance may be influenced by anchoring on their prior performance, or by a halo effect where managers’ assessments in one area are unjustifiably correlated with ratings in unrelated areas. Public administrations would benefit from efforts to mitigate these biases. However, as with many decision-making techniques, merely raising awareness of the risk is unlikely to be effective on its own.

Box 7.1. Case study: Mitigating noise in public sector workers’ performance evaluations in Italy

Researchers conducted a survey experiment with nursing personnel at a local health authority in Italy to improve staff management practices. As part of a series of RCTs, the researchers tested interventions to remove the impact of anchoring and the halo effect in workers’ performance evaluation.

To assess anchoring, the researchers asked participants to evaluate the performance over the last year of “your subordinate [who] has reached satisfactory results”. Participants were randomised into a high anchor group who were told the subordinate’s rating last year was 91/100 and half were randomised to a low anchor where the subordinate’s rating last year was 51/100. Participants were then randomised to receive or not receive instructions that “Remember that last year's performance score should not affect the score you are about to assign.” Results indicated that evaluations were anchored on past year’s results. The average rating was 13.8 points higher (out of 100) in the high anchor than the low anchor condition (p < .01). The prompt to ignore past performance was not associated with significant change in anchoring.

To assess the halo effect, the researchers asked participants to evaluate a subordinate separately on technical skills and interpersonal skills. All participants were told that the subordinate had good interpersonal skills, but half were told that they had poor technical skills and the other half that they also had good technical skills. Participants were then randomised to receive or not receive instructions to “Remember that the score on one dimension should not influence the score on the other dimension.” Results indicated that there was a halo effect – subordinates with poor technical skills scored lower on interpersonal skills by 3.33 points (out of 100) than when they had excellent technical skills (p < .05). Instructions to ignore the halo effect were not associated with a change in rating. Both results are consistent with other studies that merely telling people to ignore their intuition is often insufficient to change behaviour. Civil servants need to explore other, more structured ways to counteract these tendencies in people management.

Source: (Cantarelli, Belle and Belardinelli, 2020[1]).

Whom it involves

• Civil service workforce commissions who advise and support public administrations on the governance, management and development or their workforce, as well as support whole-of-government workforce priorities.

• Schools of government or administration who design, deliver, and evaluate training for civil servants.

• Human resources teams who are responsible for recruitment, performance management, and learning and development. This includes teams who set HR policy centrally and the organisational teams that must implement it. These teams often operationalise recruitment and what training public administrations may access.

• Line managers who coach, performance manage and support the learning of their staff. Line managers often commission or approve spending on L&D activities for their staff and play an important role in directing this funding at more effective forms of training.

How to improve recruitment

Public administrations have explored many opportunities to apply behavioural science to recruitment, including how they advertise openings, guide candidates through the process, as well as deciding which candidates to interview and, eventually, hire. These practices are based on mixed quality evidence. Some practices have high-quality evidence that they improve recruitment, other practices are based on very low-quality evidence and warrant scrutiny. Some techniques show a consistent and positive impact, others a mixed effect, alongside others that are promising and deserve further study.

Box 7.2. Case study: Increasing applications to the civil service entrance exam in Greece

The Hellenic Ministry of Interior, in partnership with the Supreme Council for Civil Personnel Selection (ASEP), the OECD and the European Commission’s SG REFORM, sought to address a persistent challenge: how to strengthen the attractiveness of the Greek civil service as an employer, especially among younger and IT/STEM-qualified candidates. A particular challenge is attracting a larger number of candidates, especially more qualified candidates, to take the civil service entrance examination. This examination determines whether candidates are eligible for public hiring.

A large-scale online behavioural experiment was run to increase registrations for the civil service entrance examination. The study was done on LinkedIn in May 2025, targeting the general Greek population aged 21-40, reaching a mix of IT/STEM and non-IT/STEM graduates, and totalling 42 982 participants. The study tested four different recruitment messages:

• Civic Duty: Emphasising societal contribution and the honour of public service.

• Career Stability: Highlighting job security and long-term benefits of a public sector career.

• Personal Growth: Focusing on skill development and career advancement.

• Control: A neutral ad to serve as a baseline comparison.

Each participant saw only one ad, ensuring there was no spillover between messages. The campaigns were run independently for each message and target group, with budgets and exposure balanced across arms.

The team found that Civic Duty was the most effective message, increasing unique click-throughs by 47% compared to the Control group (from 2.52% to 3.70%, p <. 05). This effect was consistent across genders, and across IT/STEM vs non-IT/STEM graduates. Career Stability performed comparably to the Control group, suggesting that career stability may be valued but does not seem to drive additional interest. Personal Growth was the least effective, underperforming the Control message (1.40% down from 2.52%).

The study suggested that public sector recruitment campaigns can benefit from placing civic duty at the forefront, presenting public service as a career that contributes to democratic institutions and social cohesion. Messages based on personal growth were less effective and may work better when explicitly linked to serving society or tackling complex challenges. There is also value in future campaigns testing and refining messages further. The online environment is always changing. Messages that attract candidates one year may be less effective in future years as the candidate pool and their priorities change.

Source: Based on information provided by the Hellenic Ministry of Interior.

Nudge hiring managers to advertise flexible working options

Public administrations are increasingly called on to attract applicants of various ages, backgrounds and skillsets to meet the needs of the public sector workforce. How hiring managers advertise, promote and fill vacancies plays a critical role in meeting these workforce needs (OECD, 2023[2]). These efforts can be enhanced through behavioural science. For example, public administrations may consider nudging public sector hiring managers to provide flexible working options and to promote their availability when they advertise positions. Surveys suggest that just over 90% of workers would prefer to work flexibly at least in some capacity (Timewise, 2019[3]; Timewise, 2023[4]). However, from 2020 only 9.3% of workers across the European Union worked from home at least once per week on average (Eurostat, 2024[5]). Promoting the availability of flexible work in the public sector, when it is available, benefits public administrations by encouraging more applicants to apply and it benefits applicants by offering them flexibility that suits their circumstances.

Public administrations can take an active role in nudging hiring managers to promote flexible work in their job listings. For example, public administrations may implement a pop-up on public sector hiring portals when new openings are created that nudges hiring managers to state the availability of flexible work in the job listing. In a study with the job site Indeed.com, researchers tested the effect of a such a pop-up when hiring managers created new job listings on the site. If the hiring manager had not selected any flexible working options in the listing, the site prompted them to include flexible working options, if any were available. The researchers found that the pop-up increased the share of listings that mentioned flexible work by 20% (from 34.5% to 41.5%, p < 0.001) and flexible work listings received 30% more applications than those that didn’t (from 23 per listing to 30 per listing, p < 0.05) (Londakova et al., 2021[6]). Public administrations may wish to explore similar prompts that nudge public sector hiring managers to advertise flexible work, and advertise it more prominently, in their job listings.

Anonymising CVs may, in isolation, have mixed or undesired effects

Public administrations may consider anonymising candidate’s demographic information, like age, gender and ethnic background, during the application. Anonymisation mitigates in-group favouritism, a tendency for assessors to favour candidates who share superficial characteristics with themselves. Anonymisation is usually limited to the screening of applicants because once they attend an interview their characteristics can often be inferred. Public administrations that seek to anonymise applicant screening may wish to:

• Consider how in-group favouritism can impact the entire recruitment process, not just the screen-in. Without addressing unfair assessment across the entire recruitment process, anonymisation of CVs alone may simply delay the effect of in-group favouritism. Public administrations open to anonymisation in recruitment may wish to combine it with other techniques across the entire recruitment process, see section below.

• Prioritise scalable, low-cost anonymisation methods. Public administrations that have anonymised CVs by manually redacting these PDF and Word documents found the process to be overly burdensome. A more scalable solution is to not use applicants’ CVs during the screen-in and instead have applicants enter their details in a standardised, electronic form which automatically anonymises each candidate.

In practice, anonymisation in government hiring may not consistently increase the screen-in rate of any one group or groups of disadvantaged applicants. Evidence to date has shown mixed results. For example:

• In Finland, researchers ran a difference-in-difference pilot of anonymous recruitment for certain roles within the City of Helsinki from February to December 2020 with over 74 800 applications. The researchers found that applicants with foreign-sounding names were 5%-points more likely to be interviewed when their application was anonymised. Compared to standard recruitment, job listings that advertised that applications would be anonymous also attracted more women applicants but did not attract more minority applicants (Kanninen, Kiviholma and Virkola, 2023[7]).

• In Canada, the Treasury Board Secretariat and the Public Service Commission of Canada ran an RCT to test how anonymisation affected screening decisions in public sector hiring in 2017. The study involved 2 226 applications to 17 federal government organisations. The researchers found that visible minority candidates were neither advantaged nor disadvantaged by anonymous review, whereas screen-in rates for all other candidates declined under anonymous review (Government of Canada, 2018[8]).

• In France, researchers ran an RCT to test how partially anonymous CVs influenced recruitment outcomes across nearly 1 000 branches of the French Public Employment Service (PES) in 2010. CVs in the treatment group were partially anonymised by erasing the CV topline: name, address, gender, nationality, photo, age, marital status, and children. More complex anonymisation was not possible, which left intact text that may have implied these characteristics, such as gender-specific terms, alma mater, and languages spoken. The researchers found that women with partially-anonymised CVs were slightly more likely to be interviewed than those with normal CVs, however, applicants from minority backgrounds or disadvantaged areas were less likely to be called for interviews (Behaghel, Crépon and Barbanchon, 2012[9]).

• In the Netherlands, researchers ran a field experiment in a large city government from 2006 to 2007, testing anonymous application procedures across seven departments with over 1 200 applicants. They found that anonymisation equalised the chance that candidates with foreign-sounding names were interviewed compared to candidates with Dutch-sounding names. However, candidates with foreign-sounding names were no more or less likely to be hired whether their application was anonymised or not (Bøg and Kranendonk, 2011[10]).

• In Sweden, researchers ran a difference-in-difference analysis from 2004 to 2006 of over 3 500 applications to the local government of Gothenburg, where some districts implemented anonymous recruitment procedures while others did not. Anonymous applications were associated with an increase in the chance that female and ethnic minority candidates progressed to the interview stage. However, anonymisation was only associated with an increase in the chance that women candidates received job offers – there was no change in job offers for minority candidates (Åslund and Skans, 2012[11]).

Public administrations don’t just need to consider whether to implement anonymisation, but also how to do so. Redacting applicants’ CVs by hand is time intensive. Despite the positive effect of anonymisation in Sweden, administrators expressed displeasure with the administrative burden of manually redacting CVs (Åslund and Skans, 2012[11]). Manually redacting CVs also runs the risk that not all demographic information can be removed consistently. A more promising approach is to replace or complement CV screening with standardised electronic forms. Standardised electronic forms are a scalable way to collect applicants’ demographics and can be programmed to only reveal relevant demographic data at the appropriate point in the application, with no apparent negative effects on applicant’s willingness to apply (IZA, 2024[12]). A German study found that standardised application forms that masked demographic data were more practical than having civil servants redact applicants’ CVs (Krause, Rinne and Zimmermann, 2012[13]). Public administrations that consider anonymising job applications may find it is worth the upfront investment in developing standard forms to avoid the sludge and inconsistency of redacting CVs manually.

Assess candidates through structured tasks and interviews to improve decision making

Anonymisation alone may not eliminate bias in recruitment because there are many points where bias can occur. Even when a CV is anonymised, reviewers may spend as little as ten seconds reviewing it, overlooking important work experience or fixating on trivial details (Ladders, 2018[14]). Anonymous CVs can invoke a halo effect, for example, if the candidate has worked in a prestigious ministry or administration. Later, interviews may also fail to identify the best candidate, especially if the interview is one-on-one and the interviewer is free to pose or adapt questions as they wish (Whillans and Polzer, 2021[15]).

One promising, behaviourally informed alternative is to replace CV screening at the start of an application with written tasks based on the work the candidate would perform in the role. For example, the initial screen-in may be based on the candidate drafting an email they might send to a public administration’s stakeholders, how they would give difficult feedback to a colleague, or describe how they would approach a research task. Assessors than grade each written response and use these scores to decide who to offer an interview. These written tasks can be enhanced through behavioural techniques, including:

• Situational judgement tests with competency-based scoring. Situational judgement tests present applicants with a situation they would face in the role and ask them how they would respond. Research suggests that situational judgment tests are moderately correlated with candidates’ abilities and years of experience (Cabrera and Nguyen, 2001[16]). Competency-based scoring provides a rubric to assess responses based on what a poor, mediocre and good response to the situation would be, based on a consensus among subject matter experts (Whetzel and McDaniel, 2009[17]). Competency-based rubrics reduce noise – unjustified variation in scores – by tying assessors’ scores more clearly to how candidates are expected to perform in the role, supporting greater consistency and fairness.

• Case conferences. Case conferences involve having multiple assessors score each written response independently before reaching a consensus score. Despite using a scoring rubric, an individual assessor’s scores may be noisy if they are distracted, tired or respond to the written answer idiosyncratically (Kahneman, Sibony and Sunstein, 2021[18]). Therefore, as discussed in earlier chapters, a better approach is a “case conference” approach where multiple assessors reach a decision independently, share their independent scores and apply a decision rule, such as the average or median of their scores, to derive a less noisy consensus score of the candidate’s response (Hastie and Kameda, 2005[19]). This can support greater consistency and accuracy in grading written application tasks.

• Joint evaluation, by assessing all responses to a given question together. For example, assessing all responses to question one together, then all responses to question two, then all responses to question three, etc. This is distinct from separate evaluation which assess all the responses from candidate one in order, then assessing all responses from candidate two, then all from candidate three, etc. (Bohnet, Geen and Bazerman, 2016[20]). Joint evaluation prevents assessors from demonstrating the halo effect where the score on the candidate’s first question influences the assessor’s score on later, unrelated questions.

• Randomisation of the order that each candidate’s responses are assessed within joint evaluation. This means that a candidate’s response to a given question might be the first response assessed by the first assessor, the final response assessed by the second assessor, and in the middle by a third assessor. Randomisation may also occur between joint evaluation, where one assessor scores all responses to the first question first, while another assessor scores all responses to the final question first. Randomisation mitigates ordering effects where assessors may score the same response differently depending on the sequence that they review it (Whillans and Polzer, 2021[15]). Randomisation within sequence can support greater consistency and accuracy in grading written application tasks, especially when paired with case conferences.

• Anonymisation to de-identify each written response. Anonymisation makes it easier for the assessors who grade each written response to ignore superfluous characteristics of the applicant and instead focus only on the quality of their written response. Anonymising written responses is distinct from CV anonymisation (see above).

Public administrations may then invite candidates to a structured interview based on the scores of their written tasks. These interviews employ pre-written interview questions which are the same for each candidate. This prevents the interviewer from unwittingly posing easier or harder questions to certain candidates, supporting consistency and fairness in the interview (Levashina et al., 2013[21]). Candidates' interview responses are scored according to a competency-based rubric (Whillans and Polzer, 2021[15]). Structured interviews involve multiple assessors who independently assess the candidate’s interview performance which minimises the noise in any one assessors’ score (see case conferences, above).

All or some of these behaviourally informed practices are used now, or have been in the past, to hire civil servants in the UK, Singapore, and Australian public administrations (Nolan-Flecha, 2018[22]). However, there is little independent evaluation of their combined effect. Public administrations may wish to explore how they could adapt these hiring techniques in their own recruitment, but ongoing evaluation would be invaluable to understand their effect.

Motivate all applicants to do their best throughout the recruitment process

Certain groups may be under-represented in public sector workforces, not only due to recruiter behaviour, but also due to the risk of applicant attrition during recruitment. For example, the UK government has found that applicants to its Fast Stream graduate programme for future leaders are disproportionately likely to attrit during the application process if they come from a lower socio-economic background (Cabinet Office, 2016[23]). Public administrations may wish to mitigate this risk by making reasonable adjustments and sending empirically validated messages encouraging all candidates to do their best. For example, public administrations can develop standard, simple and supportive emails that are sent automatically to all candidates who pass the initial screen-in.

Research suggests that these messages which encourage all candidates to do their best may disproportionately help under-represented candidates to succeed. For example, in Avon & Somerset in the United Kingdom, 6.8% of the region's population identified as non-white but only 2.5% of its police force identified as such. One contributing factor was that non-white candidates to the police force were disproportionately likely not to pass the Situational Judgment Test, an online test of how they would react to scenarios they would face as a police officer. Researchers addressed this by applying behavioural science to the email sent to all applicants before the test (Linos, Reinhard and Ruda, 2017[24]). The researchers simplified the email, removed anxiety-inducing language (e.g. removed the sentence that said “there is no appeals process for this stage”), and added the sentence:

“Before you start the test, I'd like you to take some time to think about why you want to be a police constable. For example, what is it about being a police constable that means the most to you and your community?” (Linos, Reinhard and Ruda, 2017[24])

The researchers found that this reminder, sent to all candidates regardless of their background, increased the proportion of non-white applicants who passed the test by 21%-points (up from 40.6% in the control group, p < .05). White candidates who received the same message performed no better or worse on the test and maintained a high pass rate overall (Linos, Reinhard and Ruda, 2017[24]). Public administrations seeking to help under-represented groups succeed in their applications may wish to develop similar messages that are sent to all candidates, regardless of background, encouraging them to do their best. This will help more candidates meet the high standard expected of all applicants to public sector positions.

Encourage unsuccessful applicants to apply again in future

Even when a candidate for a given role is unsuccessful, their willingness to apply for future roles is a behavioural challenge. This willingness may be influenced by the peak-end rule, a tendency to judge a process by the experience at its peak and its end (Kahneman, 2000[25]). Unsuccessful candidates may feel unjustifiably discouraged to re-apply because the end of the process (rejection) looms large in their memory. Public administrations may mitigate this bias by developing standard protocols to guide and encourage unsuccessful candidates to apply for future roles. This may take the form of standardised emails and structured phone calls to unsuccessful candidates.

The evidence for such an approach is promising. For example, the New South Wales Government in Australia found that men applied for nearly 15% more roles than women over a five-year period, and women who applied unsuccessfully for senior leadership roles were 1.4%-points less likely than men to reapply in the following six-months. While small, this discrepancy compounded over time and contributed to men outnumbering women nearly 2:1 in senior public sector leadership roles.

To address this, researchers ran an RCT which tested sending an email and phone-call to candidates who were interviewed for, but not offered, senior roles. The email congratulated candidates for progressing as far as they did, emphasised that most people apply for multiple roles before they’re successful, and directed them to future opportunities on the NSW Government’s jobs portal. A subsequent phone call reiterated the email’s message and specifically encouraged female candidates to apply for future roles.

“I’m emailing you because you made it through to the final round of assessment and we’d like to encourage you to apply for future vacancies across the NSW Government. On average, fewer than 1 in 5 applicants get through to the final stages of our recruitment process and most people apply for multiple jobs before they are successful, particularly at the senior levels." (NSW Behavioural Insights Unit, 2020[26]).

The researchers found that the interventions significantly reduced the re-application gender gap. Over a six-month period, 59.5% of men in the control group applied for another role compared to 41% of women, an 18.5%-point gender gap (p < 0.001). However, the gap shrank to 5.5%-points when candidates were sent only the email (45.4% of women vs 50.9% of men, p = 0.343). Among those sent the email and phone call, the gender-gap was just 2.2%-points (51.9% of women vs 54.1% of men, p = 0.853) (NSW Behavioural Insights Unit, 2020[26]). Public administrations can explore closing re-application gaps in their own recruitment by sending similar, encouraging messages to all unsuccessful candidates.

How to improve learning and development

Professional development has a major impact on the long-term effectiveness of public administration. An under-appreciated facet of learning is that it is behavioural and, therefore, amenable to a behaviourally informed approach. For example, cognitive science suggests that learning can be more effective by employing multiple learning settings, interweaving instruction on separate topics, spacing content, as well as using quizzes to measure and reinforce learning (Bjork and Bjork, 2011[27]). These insights could be applied to make learning and development in public administration more effective.

Establish whether training is the right behaviour change tool

Under the COM-B Model, a common behaviour change framework, people will change their behaviour (B) when they have the capability (C), opportunity (O), and motivation (M) to do so (Michie, van Stralen and West, 2011[28]). Public administrations can use learning and development as an effective tool to change behaviour by building capability (C); however, training may be less suited to improve opportunity (O) and motivation (M). Before commissioning formal training, civil service managers may wish to ask whether the problem that the training solves is:

• Motivation. Does the person have the right incentives to perform the behaviour? If they lack the right incentives, it may be more beneficial to create them before engaging learners in training.

• Opportunity. Do the person’s circumstances enable the person to perform the behaviour? If there are contextual blockers preventing them from performing as desired, it may be more effective to remove these blockers before engaging them in training.

• Capability. Is the person capable of performing the desired behaviour? If so, then training may be appropriate, but it still may not be the best solution. For example, if civil servants are not using a new briefing template correctly, then training them in the new template may help, but redesigning the template to be self-explanatory may be equally effective – checklists, templates and guides can help novices perform at the level of trained experts in some settings.

Public administrations can assess whether formal training is an appropriate solution through a needs analysis – a process to identify the organisation, team or individual’s training needs and to align a training program with those needs. One meta-analysis suggests that needs analyses were associated with more effective training, on average, because they helped tailor the training content to real workplace challenges (Lacerenza et al., 2017[29]).

Prioritise training based on evidence-informed practices

How training is structured can enhance how participants learn. For example, a rapid-evidence review for the UK government suggested that training facilitated by a trainer, rather than self-administered by the learner, was often more effective (Schuster, 2025[30]). Public administrations may wish to prioritise courses that:

• Tailor training to learners' work. Training can often be tailored to the specific problems, methods and ways of working of participating civil servants. Tailored training may be, on average, more effective than one-size-fits-all training because it aligns with the public administration’s and the participant’s work (Lacerenza et al., 2017[29]). Public administrations may find they achieve better learning outcomes by commissioning tailored training that addresses existing capability gaps and incorporates examples of real work that learners face.

• Model skills and practices, especially online. Research suggests that training may be more effective if it incorporates behavioural modelling: emphasising well-defined behaviours, guiding learners to practice those behaviours, and providing feedback on their performance in training (Taylor, Russ-Eft and Chan, 2005[31]). Research suggests that learners apply more knowledge from leadership training when their course was face-to-face compared to online because online courses tended to have fewer opportunities for modelling, demonstration and practice (Lacerenza et al., 2017[29]). Civil servants who commission training may wish to ensure participants have sufficient opportunities to practice their learning and receive feedback in the course.

• Deliver training over multiple sessions. Learning tends to be more effective when participants recall its content over many sessions rather than one intense session. The time between recall helps the brain to form long-term memories (Latimier, Peyre and Ramus, 2020[32]). Research suggests that this spaced repetition “is one of the most general and robust effects from across the entire history of experimental research on learning and memory” (Bjork and Bjork, 2011[27]). A meta-analysis of leadership training found that programs spaced across multiple days or weeks were associated with better learning outcomes (Lacerenza et al., 2017[29]). Public administrations may wish to prioritise training which, when logistically possible, is spaced across multiple sessions and provides opportunities to refresh learners’ knowledge over time.

Ensure there is a behaviour change to embed training outcomes

The long-term impact of training relies on practices outside the classroom. However, research suggests that professional development training often does not transfer into learners’ behaviour at work (Grossman and Salas, 2011[33]). Public administrations can facilitate better knowledge transfer by prioritising evidence-informed training (above) as well as embedding learning in work environments, including:

• Test knowledge early and often. Effective training often embeds multiple quizzes, tests and opportunities to assess participants’ knowledge. This helps to measure impact and to reinforce learning because of the testing effect, the tendency for knowledge tests to reinforce the tested material (Roediger and Karpicke, 2006[34]). Repeat tests can also act as a form of spaced repetition to reinforce learning, see above. Civil servants may wish to prioritise formal training programs that use regular testing during and after the course to help participants embed the learning material.

• Create a post-training implementation plan between learners and their managers. Some learners lack opportunity and motivation to apply their learning to their work (Hagger and Luszczynska, 2013[35]). Managers can address these barriers by helping learners develop an implementation plan to apply their learning. This plan may include implementation intentions and goal setting, which are structured plans for when and how learners will apply their learning in the days, weeks and months post-training (Blume et al., 2009[36]; Greenan, 2023[37]; Gollwitzer, 1999[38]). This plan may cover informal check-ins, such as coaching, which can guide and course correct learners as they apply their new knowledge (Jones, Woods and Guillaume, 2015[39]). Public administrations may wish to encourage or require that training participants create an implementation plan with their manager to apply and maintain their training.

• Assess the impact of training. Effective programs measure impact to understand whether training has, or has not, built participants’ capability. One popular evaluation framework is the Kirkpatrick Model that assess training in terms of: Reaction, whether participants are satisfied with a course; Learning, whether participants acquire the course’s intended knowledge; Behaviour, whether participants and their team apply their learning to change how they perform their role; and, Results, whether the team produces better results as a result of completing the training (Kirkpatrick, 1959[40]). Participant engagement and satisfaction are useful intermediary measures, but civil servants will benefit most from measuring the long-term knowledge and behavioural impact of training. Public administrations can use this data to scale-up effective training programs and improve or discontinue ineffective programs.

Create an environment of continuous learning

Training is only one setting that civil servants learn and develop. On the job learning is also critical. Public administrations can take a behaviourally informed approach to informal, peer-to-peer learning with:

• Performance evaluations as a tool for learning. Line management meetings and annual performance reviews are an under-used tool for learning. Traditional performance management often focuses on past achievements and compliance metrics, missing opportunities to identify learning needs and development priorities. Public administrations may benefit from encouraging managers to discuss performance with employees on a regular basis, instead of once per year. Managers may benefit from conversation guides to help discuss the topic in a psychologically safe manner.

• After Action Reports and team debriefs. Managers can facilitate team learning through After-Action Reports (AAR) and debriefs. These are short, structured conversations within a team about their work, the intended outcome, what happened, why, and what they might do differently in future. Research suggests that this type of structured, informal learning can improve team performance (Keiser and Arthur, 2021[41]; Tannenbaum and Cerasoli, 2012[42]). However, these debriefings need to be psychologically safe – managers are responsible for ensuring debriefings are run in the spirit of non-judgemental error correction, not blame (Edmondson, 1999[43]). Public administrations may benefit from creating AAR templates and recommended debriefing schedules throughout projects to capture learnings and continuously improve team performance.

• Communities of Practice (CoPs). Public administrations can foster informal knowledge sharing through CoPs dedicated to policies, methods and areas of interest across the public sector. CoPs help disseminate civil servants’ experiences and maintain learned knowledge (see spaced repetition, above). Research from health administrations have suggested that CoPs are most effective when they have a clear scope, are facilitated with a light-touch, are held regularly and work towards clearly stated objectives (Ranmuthugala et al., 2011[44]; Li et al., 2009[45]).

All these principles need to be balanced in practice. Public administrations may find it more feasible to schedule training in one large block than spread across multiple sessions, even if that reduces opportunities for spaced repetition. Tailored training is often more effective, but also more expensive than one-size-fits-all courses. These principles are not absolute, but a tenet of behavioural public administration is for civil servants to understand the evidence and make informed trade-offs.

Behaviourally informed insights

• Recruitment and people management work best when designed as an integrated system. Recruitment is fairer and more effective when it is treated as a multi-stage process – isolated interventions, such as anonymising CVs on their own, may lead to mixed results. In addition, many aspects of hiring and learning are prescribed within public administrations: many techniques described in this chapter may need to be adapted to suit administrative requirements.

• Learning is ongoing and is more effective when supported by measurement. Learning is most effective when it is ongoing rather than one-off and when it can be accurately measured. Formal training should be reinforced through informal mechanisms, such as performance reviews, after-action reports, and communities of practice. Training courses that measure learning before, immediately after and in the long-term following training should be prioritised. Robust measurement (beyond training attendance and satisfaction) is essential to reinforce learning and assess behavioural change.

For further reading on how to use behavioural science to improve recruitment, learning and development, see for example: How to Improve Workplace Equity Evidence-based Actions for Employers (Employment and Social Development Canada, 2022[46]); How to use skill-based assessment tasks (BIT, 2021[47]); How to run structured interviews (BIT, 2021[48]).

References

[11] Åslund, O. and O. Skans (2012), “Do Anonymous Job Application Procedures Level the Playing Field?”, ILR Review, Vol. 65/1, pp. 82-107, https://doi.org/10.1177/001979391206500105.

[9] Behaghel, L., B. Crépon and T. Barbanchon (2012), Do anonymous resumes make the field more even? Evidence from a randomized field experiment..

[48] BIT (2021), How to run structured interviews, https://www.bi.team/wp-content/uploads/2021/07/BIT_How_to_improve_gender_equality_guide_RSI.pdf.

[47] BIT (2021), How to use skill-based assessment tasks, https://www.bi.team/wp-content/uploads/2021/07/BIT_How_to_improve_gender_equality_guide-SBAT.pdf.

[27] Bjork, E. and R. Bjork (2011), Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning.

[36] Blume, B. et al. (2009), “Transfer of Training: A Meta-Analytic Review”, Journal of Management, Vol. 36/4, pp. 1065-1105, https://doi.org/10.1177/0149206309352880.

[10] Bøg, M. and E. Kranendonk (2011), Labor Market Discrimination of Minorities? Yes, but not in Job Offers, https://mpra.ub.uni-muenchen.de/33332/.

[20] Bohnet, I., A. Geen and M. Bazerman (2016), “When Performance Trumps Gender Bias: Joint vs. Separate Evaluation”, Management Science, Vol. 62/5, pp. 1225-1234.

[23] Cabinet Office (2016), Socio-Economic Diversity in the Fast Stream, https://assets.publishing.service.gov.uk/media/5a80d8c7e5274a2e87dbc256/BG_REPORT_FINAL_PUBLISH_TO_RM__1_.pdf.

[16] Cabrera, M. and N. Nguyen (2001), “Situational Judgment Tests: A Review of Practice and Constructs Assessed”, International Journal of Selection and Assessment, Vol. 9/1-2, pp. 103-113, https://doi.org/10.1111/1468-2389.00167.

[1] Cantarelli, P., N. Belle and P. Belardinelli (2020), “Behavioral public HR: Experimental evidence on cognitive biases and debiasing interventions”, Review of Public Personnel Administration, Vol. 40/1, pp. 56-81.

[43] Edmondson, A. (1999), “Psychological Safety and Learning Behavior in Work Teams”, Administrative Science Quarterly, Vol. 44/2, pp. 350-383, https://doi.org/10.2307/2666999.

[46] Employment and Social Development Canada (2022), How to Improve Workplace Equity Evidence-based Actions for Employers, https://www.bi.team/wp-content/uploads/2023/03/How-to-Improve-Workplace-Equity.pdf.

[5] Eurostat (2024), Work from home, from an external site or on the move (2018), https://doi.org/10.2908/ISOC_IW_HEM (accessed on 29 October 2025).

[38] Gollwitzer, P. (1999), “Implementation intentions: Strong effects of simple plans.”, American Psychologist, Vol. 54/7, pp. 493-503, https://doi.org/10.1037/0003-066x.54.7.493.

[8] Government of Canada (2018), Anonymized Recruitment Pilot Project.

[37] Greenan, P. (2023), “The impact of implementation intentions on the transfer of training from a management development program”, Human Resource Development International, Vol. 26/5, pp. 577-602, https://doi.org/10.1080/13678868.2023.2174976.

[33] Grossman, R. and E. Salas (2011), “The transfer of training: what really matters”, International Journal of Training and Development, Vol. 15/2, pp. 103-120, https://doi.org/10.1111/j.1468-2419.2011.00373.x.

[35] Hagger, M. and A. Luszczynska (2013), “Implementation Intention and Action Planning Interventions in Health Contexts: State of the Research and Proposals for the Way Forward”, Applied Psychology: Health and Well-Being, Vol. 6/1, pp. 1-47, https://doi.org/10.1111/aphw.12017.

[19] 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.

[12] IZA (2024), “Anonymous job applications and hiring discrimination”, IZA World of Labor, https://doi.org/10.15185/izawol.48.v3.

[39] Jones, R., S. Woods and Y. Guillaume (2015), “The effectiveness of workplace coaching: A meta‐analysis of learning and performance outcomes from coaching”, Journal of Occupational and Organizational Psychology, Vol. 89/2, pp. 249-277, https://doi.org/10.1111/joop.12119.

[25] Kahneman, D. (2000), “Evaluation by Moments: Past and Future”, in Choices, Values, and Frames, Cambridge University Press, https://doi.org/10.1017/cbo9780511803475.039.

[18] Kahneman, D., O. Sibony and C. Sunstein (2021), Noise: A flaw in human judgment, Little, Brown Spark.

[7] Kanninen, O., S. Kiviholma and T. Virkola (2023), Anatomy of an Anonymous Hiring Pilot, https://labore.fi/wp-content/uploads/2023/09/Tyopapereita-338.pdf.

[41] Keiser, N. and W. Arthur (2021), “A meta-analysis of the effectiveness of the after-action review (or debrief) and factors that influence its effectiveness.”, Journal of Applied Psychology, Vol. 106/7, pp. 1007-1032, https://doi.org/10.1037/apl0000821.

[40] Kirkpatrick, D. (1959), “Techniques for Evaluating Training Programs”, Journal of the American Society for Training and Development, https://www.scirp.org/reference/referencespapers?referenceid=1735231.

[13] Krause, A., U. Rinne and K. Zimmermann (2012), “Anonymous job applications in Europe”, IZA Journal of European Labor Studies, Vol. 1/1, https://doi.org/10.1186/2193-9012-1-5.

[29] Lacerenza, C. et al. (2017), “Leadership training design, delivery, and implementation: A meta-analysis.”, Journal of Applied Psychology, Vol. 102/12, pp. 1686-1718, https://doi.org/10.1037/apl0000241.

[14] Ladders (2018), Eye-Tracking Study, https://www.theladders.com/static/images/basicSite/pdfs/TheLadders-EyeTracking-StudyC2.pdf.

[32] Latimier, A., H. Peyre and F. Ramus (2020), “A Meta-Analytic Review of the Benefit of Spacing out Retrieval Practice Episodes on Retention”, Educational Psychology Review, Vol. 33/3, pp. 959-987, https://doi.org/10.1007/s10648-020-09572-8.

[21] Levashina, J. et al. (2013), “The Structured Employment Interview: Narrative and Quantitative Review of the Research Literature”, Personnel Psychology, Vol. 67/1, pp. 241-293, https://doi.org/10.1111/peps.12052.

[45] Li, L. et al. (2009), “Use of communities of practice in business and health care sectors: A systematic review”, Implementation Science, Vol. 4/1, https://doi.org/10.1186/1748-5908-4-27.

[24] Linos, E., J. Reinhard and S. Ruda (2017), “Levelling the playing field in police recruitment: Evidence from a field experiment on test performance”, Public Administration, Vol. 95/4, pp. 943-956, https://doi.org/10.1111/padm.12344.

[6] Londakova, K. et al. (2021), Encouraging employers to advertise jobs as flexible, https://www.gov.uk/government/publications/encouraging-employers-to-advertise-jobs-as-flexible-with-jobs-site-indeed/encouraging-employers-to-advertise-jobs-as-flexible.

[28] Michie, S., M. van 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.

[22] Nolan-Flecha, N. (2018), “Next generation diversity and inclusion policies in the public service: Ensuring public services reflect the societies they serve”, OECD Working Papers on Public Governance, Vol. 34.

[26] NSW Behavioural Insights Unit (2020), Behavioural Insights Unit: Update Report 2020, https://www.nsw.gov.au/sites/default/files/2021-05/2020-BIU-Report_2.pdf.

[2] OECD (2023), “Strengthening the attractiveness of the public service in France: Towards a territorial approach”, OECD Public Governance Policy Papers, No. 28, OECD Publishing, Paris, https://doi.org/10.1787/ab9ebe85-en.

[44] Ranmuthugala, G. et al. (2011), “How and why are communities of practice established in the healthcare sector? A systematic review of the literature”, BMC Health Services Research, Vol. 11/1, https://doi.org/10.1186/1472-6963-11-273.

[34] Roediger, H. and J. Karpicke (2006), “Test-Enhanced Learning”, Psychological Science, Vol. 17/3, pp. 249-255, https://doi.org/10.1111/j.1467-9280.2006.01693.x.

[30] Schuster, C. (2025), What do we know about the design of effective management training in government? A rapid evidence review to inform training design and evaluation, https://www.gov.uk/government/publications/effective-management-training-in-government/what-do-we-know-about-the-design-of-effective-management-training-in-government-a-rapid-evidence-review-to-inform-training-design-and-evaluation.

[42] Tannenbaum, S. and C. Cerasoli (2012), “Do Team and Individual Debriefs Enhance Performance? A Meta-Analysis”, Human Factors: The Journal of the Human Factors and Ergonomics Society, Vol. 55/1, pp. 231-245, https://doi.org/10.1177/0018720812448394.

[31] Taylor, P., D. Russ-Eft and D. Chan (2005), “A Meta-Analytic Review of Behavior Modeling Training.”, Journal of Applied Psychology, Vol. 90/4, pp. 692-709, https://doi.org/10.1037/0021-9010.90.4.692.

[4] Timewise (2023), The Timewise Flexible Jobs Index 2023, https://timewise.co.uk/wp-content/uploads/2023/11/Timewise-Flex-Jobs-Index-2023.pdf.

[3] Timewise (2019), Flexible Working: A Talen Imperative, https://timewise.co.uk/wp-content/uploads/2019/06/Flexible_working_Talent_Imperative.pdf.

[17] Whetzel, D. and M. McDaniel (2009), “Situational judgment tests: An overview of current research”, Human Resource Management Review, Vol. 19/3, pp. 188-202, https://doi.org/10.1016/j.hrmr.2009.03.007.

[15] Whillans, A. and J. Polzer (2021), Applied: Using Behavioral Science to Debias Hiring, https://www.hks.harvard.edu/sites/default/files/centers/wappp/921046-PDF-ENG.pdf.