Evidence-informed police reform
What the research evidence says about changing large organisations, and how we can learn more
Matt Bland, Chief Operating Officer, Society of Evidence Based Policing.
Every organisational reform is a bet. That bet is usually along the lines of 'make sweeping changes and better outcomes will follow'. More resilient forces, more consistent standards, specialist capability where it is needed, money saved on duplication, and so on. Those are reasonable things to want. They are also measurable claims.
That is the spirit in which I want to approach police reform in England and Wales in this article, which accompanies the 2026 National Evidence-Based Policing Conference. What works in police reform is this year's focus. Reform is widely seen in policing as a real opportunity to improve the service, and it is also an opportunity to test some important premises properly. In this piece I do not intend to make a case for or against reform. The purpose is to bring evidence into focus at a point when the UK is about to enter possibly one of the largest organisational experiments in public sector history. The public sector has been here before. In 1999, as NHS trusts merged at speed, a BMJ editorial asked "on what basis do we judge whether the merger of two or more NHS organisations is successful", and whether there was any evidence base for how to run one.46 Nearly thirty years on, policing can do better than repeat the question.
This piece addresses two questions:
First, what does the research evidence actually say about large scale organisational reform, in policing and in other sectors?
Second, what role should research evidence play in the police reform agenda, given that in many respects this is uncharted territory?
This isn't an academic article, but I have tried to keep the claims proportionate with the norms of one.
The footnotes are there so readers can check my workings out if they wish. There's a note at the end on how this article was put together for those interested.
Where we are
The government's January 2026 white paper1 proposes big reforms which are largely seen as necessary and positive. A National Police Service to replace the many bodies at the centre of policing, a significant reduction in the number of forces over this Parliament and the next, 13,000 additional neighbourhood policing personnel, a new performance framework and a Licence to Practice for officers. It commits to reviewing the funding formula again.
The resulting independent review of force structures was asked to design a model "built around fewer, larger police forces", to identify pathfinder mergers and to support its recommendations with evidence and cost analysis.2 It was due to report by the end of the summer but at the time of writing, it has not been published. The new Prime Minister, in his first week back at the despatch box, told the Commons on 2nd September that the funding formula is being reviewed "as part of a wider programme of police reform".3 HM Chief Inspector's latest annual assessment says it would be hard to argue that anyone designing policing from scratch today would land on 43 forces, and warns that creating a National Police Service is "exceptionally complex". It also says the inspectorate's evidence on force size is mixed, that it has not identified an optimal size, and that leadership, governance and workforce capability "have a much greater effect on performance than the size of a force".4
A summary then: the direction was set in January, and a change of Prime Minister has not, so far, altered it. The detail is pending, public finances are tight to say the least, and the demands on policing are unusually heavy. That is the backdrop to the fundamental decision about how many police forces there should be. The available evidence cannot tell us what the number should be, and it would be wrong to pretend otherwise. What it can do is tell us something about what tends to happen when large public organisations are restructured, and where the risks and the benefits are likely to lie. That at least may help to navigate the "exceptional complexity".
What the evidence says about police reform
Several countries near us have done large scale police reform like this recently. Scotland and the Netherlands both created single national forces in 2013. Denmark cut 54 districts to 12 in 2007, Sweden merged 21 authorities into one in 2015, and Norway went from 27 districts to 12 in 2016.48 Most of these reforms have been evaluated, some of them at length, and a small but serious body of peer reviewed research exists, from which a few themes stand out.
Bigger is better has not been proven yet. The main published review of the evidence on force size and performance, by Jonathan Mendel, Nick Fyfe and Garth den Heyer, concludes that the relationship is mixed and inconclusive (I know, sigh, but that's how it is). They find no compelling evidence of any simple cause and effect relationship between force size and efficiency, local policing, protective services or governance. The one strand of quantitative work they highlight suggests medium sized forces were the most efficient and the largest and smallest less so, but that work is a quarter of a century old. Given the gaps in empirical evidence, they argue that the political and symbolic dimensions of mergers deserve serious attention in their own right.5 Lars Holmberg's comparison of the Scandinavian reforms makes a similar point from a different angle: Nordic policing has been restructured for fifty years on a doctrine that districts must be big enough to handle any contingency, without that doctrine ever being clearly evidenced.6
The clearest reported benefit is national and specialist capability. This needs emphasis because the literature is often read as uniformly gloomy and it is not. Scotland's official evaluation indicated that access to national specialist resources and the response to major incidents were successes of the 2013 reform.7 Sweden's official evaluation found the single authority had improved information sharing and the flexibility to move resources around, and that its ability to act against the most serious incidents "has certainly been enhanced, but their ability remains unchanged for other crimes".8 The Dutch evaluation likewise found "relatively quickly visible and tangible benefits" of scale in mobilising for major incidents and mutual aid, and nobody it spoke to wanted to go back to force by force IT.16 Much of this is qualitative and self-reported by the organisations involved, but if the case for fewer, larger forces rests on resilience and specialist capability, that is the part of the case with the best support. It has a cost that shows up locally though. Officers in Scotland and the Netherlands described specialist teams that now arrive from fifty miles away to areas they do not know, reached through an electronic form rather than a phone call to someone they knew.10,11
Local presence and local knowledge need protecting. Jan Terpstra and Nick Fyfe describe an "implementation gap" in both Scotland and the Netherlands between what was promised for local policing and what was put in place. The causes differed. In Scotland the legislation required local policing at council level, but the new force was built around fourteen divisions and an enforcement led style that departed from it. In the Netherlands the law said little about local policing beyond a ratio of one community officer per 5,000 residents, and the same reform strengthened local policing in some municipalities and stripped it out of others, depending largely on the views of local police leaders.9 Their later work with Renze Salet sketches the "abstract police", an ideal type rather than a measured outcome: national organisations that become more formalised, more dependent on systems and system information, and more distant from local knowledge and from citizens.10 Fieldwork with local policing teams in four Scottish communities found local resources stretched not by any single decision but by the cumulative effect of several: officers redeployed to specialist units and not replaced, civilian posts cut and the work picked up by constables, stations and custody suites closed so that beats grew and travel times lengthened. None of these decisions was intended to weaken local policing. Together, they had that effect.11 The Dutch evaluators saw the same and gave the credit for keeping the service running to the loyalty and adaptability of officers: during the rebuilding, they said, the shop stayed open.16 Andrew Wooff's work on rural custody in Scotland shows how a nationalised function can drift away from the communities it serves in ways nobody intended.12Scotland's international comparison heard the same from the Netherlands, where one interviewee described local policing as "the main source of police officers for lots of other issues that are nationally", and similar concerns in Norway, though some there thought the change was in visibility more than substance.13
None of this makes poor local policing an inevitable feature of a large force. The reviewers of the size literature find the opposite argument convincing, provided local teams are well resourced and well led, and Scotland's fire service, nationalised on the same day, was felt by the same communities to have kept its localism.5,7 The challenge is that adequate local resourcing is what most often gives way when a merger is also asked to save money.
Public confidence follows contact, not structures. The proportion of Scottish adults rating their local police as good or excellent fell from 61 per cent in 2012/13 to 57 per cent by 2017/18 and to 45 per cent by 2023/24.13,14 Two qualifications matter. The same survey shows that confidence in the police's ability to do specific things, such as investigate incidents or respond quickly, dipped slightly after 2012/13, then stabilised above its 2008/09 level, and the auditors record that trust in the police as an institution remains high.13,14 And the Netherlands, which merged in the same year, saw no fall at all. Satisfaction with the police rose slightly between 2012 and 2016, trust held steady, and the Dutch evaluators concluded that public appreciation is shaped by concrete experiences and personal contact with local officers, not by structural reform, which most citizens barely notice.16 Sweden's evaluation found no definite improvement in access to the police, and at local level police staff and municipalities felt availability had in many cases worsened.8 In England and Wales, without a merger, the proportion rating their local police as good or excellent fell from 62 per cent in 2015 to 49 per cent in 2025.39 So one national time series does not establish a general rule, and it would be wrong to lay all of the Scottish fall at the door of reform. But it would be equally wrong to assume a restructured service will earn public confidence simply by being better organised. What the Scottish evidence does suggest is where to look: the auditors say that since reform "greater emphasis has been placed on the development of a national force with specialist capabilities such as public protection rather than community policing", and that visibility and connectedness are the main drivers of local confidence, which is lowest in deprived areas.14
Savings need to be demonstrated, not assumed. Policing in Scotland reports an estimated £290 million a year in efficiency savings against the old eight force structure. The joint Best Value audit published this year records that figure as policing's own estimate, accepts the historical reporting as valid, and, separately, finds that policing does not clearly demonstrate progress towards its strategic outcomes, so the rate and scale of improvement cannot be assessed. It also notes that policing struggled to balance its budget for its first eight years, lost the ability to hold reserves or borrow when it moved from local to central government, and has "failed to realise or demonstrate intended business benefits" from its transformation programmes.14 The Dutch reform was meant to save EUR 230 million a year. Non-operational staff fell by 19 per cent between 2012 and 2016, but running costs rose by 6 per cent, the administrative burden did not fall and probably grew, the one off reorganisation budget was doubled from EUR 230 million to EUR 460 million, and four years in the evaluators found the police "certainly not yet in financial control".16 In Sweden, the National Audit Office reported last year that, a decade after the merger, person based clearance rates were lower than in 2015 and the cost per cleared crime had risen sharply, with some improvement in the most recent year, even though funding had nearly doubled and staffing had grown.15 The auditors point to two causes: crime that has become more complex, and internal governance that has been uncoordinated, weak and insufficiently focused on results. The Auditor General put it plainly: "With more employees available per reported crime, investigation results should reasonably improve, but the figures show the opposite." Administrative savings are logical and often reported, but translating scale into better frontline results is a separate job that does not happen automatically. Magdalena Dominguez and Tom Kirchmaier make the underlying point plainly: in policing the dominant cost is people, and pay and staffing structures barely move when boundaries do, so the classic economies of scale argument "may have limited force".43 The Danish local government researchers put the same thing another way: it is schools that teach children, not councils, so merging the council does not change the size of the school.25 In policing the equivalent units are the response team and the neighbourhood team.
Implementation is often underestimated. Every statutory evaluation hints at the same thing in different words. The Dutch evaluation found the merger had been "seriously underestimated": too many changes were launched at once within tight deadlines, and an approach it summarised as "we arrange it centrally and uniformly, unless it really cannot be otherwise" bred rigidity and resistance until the pace, the degree of centralism and the financial framework were all reset in 2015. The evaluators note that the previous Dutch restructuring, in 1993, took about eight years to bed in, and that reforms of this kind "take ten years rather than five".16 Sweden's evaluation found insufficient staff to populate the new local organisation, weak performance management and slow integration of support functions, and observed that neither the implementation committee nor the new authority had ever analysed what resources the new organisation would need.8 Scotland's found early strain on internal communication and career development and called for a renewed commitment to localism.7 Stable leadership, workforce buy in, realistic timescales and not trying to change everything at once come up again and again as the conditions for success. Scotland's evaluators add that culture takes far longer to change than structures, and that Scotland chose to fix structures first and people and skills later, where Norway did the reverse.13 The size review reaches the same place from the literature: the risks with the strongest evidence behind them are loss of skills and competence, damage to relations with the workforce, underestimating the cost of change, and inadequate planning.5
Accountability moves in both directions. Centralisation can weaken local democratic scrutiny, and research on Scotland's local scrutiny arrangements argues that the demand for localism is not nostalgia but a real requirement for recognition and responsiveness.17 Yet national scrutiny can intensify at the same time. Kath Murray and Diarmaid Harkin show how the creation of Police Scotland moved policing from a "cool" political climate to a "hot" one, in which mass non statutory stop and search, tolerated for two decades regionally, was dismantled within a couple of years.18 Their explanation is that reform tied the reputation of the new force to the government that created it, which gave opposition politicians and the press a reason to look harder. The Dutch evaluation found the mirror image: a minister and parliament drawn into local incidents that made the news, while democratic control of policing at local level "remains weak" and had not improved under the national force.16 A national force is easier to hold to account nationally, but the attention is uneven and it can cut across local judgement.
The limits of this evidence
The evidence base is thin, it has to be acknowledged.
Most of these studies are observational, many were conducted in the first few years after reform, and they come from a fairly small network of researchers. There is no randomised evaluation of a police force merger, which is hardly surprising, and no strong quasi experimental study of a merger between separate forces. The closest thing is a 2026 LSE paper by Dominguez and Kirchmaier on the Met's consolidation of 32 borough commands into 12: senior posts fell, frontline numbers rose slightly, recorded calls fell (most in more deprived areas, and the pattern of which calls fell, burglary, civil disputes and criminal damage, points to less contact rather than less crime), and public attitudes did not worsen on average, though any gains were weaker or absent in the more deprived boroughs.43 It is a study of internal restructuring within one force, and the authors are careful about what it shows. Beyond that, the literature is qualitative studies and official evaluations, and there is no meta analysis to give us a quantifiable understanding of the average effects (nor is one realistic). Government evaluations supply most of the quantitative data and are themselves cautious about attributing outcomes to reforms. Scotland's evaluators noted early on that they had more information about structures and processes than about public outcomes.7 The evidence on equity of service, on whether restructuring affects some communities differently from others, is thinnest of all. Scotland's case studies deliberately contrasted urban, rural, affluent and deprived areas and found more negative views of officers in deprived areas, more positive ones in rural and affluent areas, and the largest gains in access to specialists outside the former Strathclyde force; the Dutch evaluators found rural teams whose capacity "regularly falls short" and response times not met because of distance; Scotland's auditors note lower confidence in deprived areas; and the Met study found the gains in public attitudes weaker or absent in more deprived boroughs. But none of these sources measures whether restructuring widened or narrowed those differences.7,12,14,16,43
None of that means the evidence is useless. It means it tells us about risks and mechanisms to be accounted for rather than giving us an exact formula of what to do.
Learning from other sectors
Policing is not the first public service to reach for structure as the lever of improvement. Health, education and local government have all run large restructuring programmes in the past two decades, and several have been studied with methods the policing literature lacks. Their findings rhyme with the policing themes to a striking degree. A caution first: hospital mergers, council amalgamations and school conversions are not the same intervention as a police merger, and the studies below offer analogies and hypotheses rather than direct evidence about policing.
Restructuring on its own does not reliably deliver the outcomes promised for it. Martin Gaynor, Mauro Laudicella and Carol Propper studied a wave of English hospital mergers between 1997 and 2006 in which around half of acute hospitals were involved. Across finances, productivity, waiting times and clinical quality, they found little evidence of gains beyond a reduction in activity. Deficits grew, waiting times rose and spending on agency staff went up by around a third.19 An earlier BMJ study of London trust mergers found management cost savings well short of the £500,000 a year that had been targeted, senior management becoming more remote, and service developments delayed by at least eighteen months through loss of managerial focus.20 A 2022 systematic review of hospital mergers found inconsistent effects on quality and no strong evidence of improvement or deterioration, though what signal there was pointed towards worse outcomes for heart attack and stroke patients.21 Australian teaching hospitals that restructured showed no efficiency gain afterwards, and the four that restructured twice became significantly less efficient, which the authors suggest may say more about the frequency of change than about any particular structure.22
Economies of scale are real, but mostly in the back office, and transition costs can eat them. English local government research does find that larger authorities carry lower central administrative costs.23 But when English counties were actually amalgamated into unitaries in 2009, the restructuring costs were not recouped and the merged authorities performed poorly on reserves and budget balance.24 Denmark's 2007 municipal reform, which cut 271 municipalities to 98, is one of the best studied natural experiments we have: it found no overall effect on spending and no change in cost per unit of service, with administrative savings offset by rising costs elsewhere.25 A systematic review published this year concludes that amalgamations frequently fail to achieve economies of scale, have limited effects on services, and reduce political representation in the smaller places inside merged units.26
Scale can carry a democratic cost. The same Danish reform caused a sizeable and measurable fall in citizens' sense that they could understand and influence local decisions.27 That is the local government version of the "abstract police" finding, produced with a causal design the policing literature does not have.
Fit between the problem and the intervention seems to matter. Education gives us a suggestive contrast. When failing secondary schools in disadvantaged areas were converted to academies in the 2000s, pupil performance improved significantly, but only where the school gained a large increase in autonomy, and the main changes were a new head and a new curriculum.28 When already successful primary schools were allowed to convert after 2010, they changed how they operated, received more money and spent it mostly on administration rather than teaching, and pupil performance did not improve.29 Same structural lever, different starting problem, different result. The two studies differ in more than targeting, so this is a hypothesis rather than a rule. But it points the same way as the policing literature: a merger with a specific problem to solve and a credible mechanism for solving it has a better chance than a merger justified by the general proposition that bigger is better.
Proximity has operational value. Jordi Blanes i Vidal and Tom Kirchmaier, using Greater Manchester Police data and a clever boundary design, found that a 10 per cent increase in response time reduced the likelihood of clearing a crime by nearly five percentage points.30 It is a study about response, not mergers. But it is a reminder that distance and time are not administrative details. The same paper estimates that moving response stations closer to where crime happens would pay for itself in some divisions and not others, so the right footprint differs from place to place, and changes to deployment geography have consequences that should be measured rather than assumed away. An Institute for Fiscal Studies working paper on London's closure of more than 70 per cent of its police stations between 2008 and 2018, with officer numbers held constant, found that losing proximity persistently increased violent crime, cut clearance rates, lengthened median response times from 13 to 19 minutes, reduced reporting of lower level offences and lowered trust. It also found that a different pattern of closures could have reduced crime overall, which is the point: it is the design of consolidation, not the fact of it, that decides the outcome.44
Expertise at the top matters. Amanda Goodall's research on leading American hospitals found those led by physicians ranked significantly higher on a widely used quality index than those led by professional managers, and her later work with Benjamin Artz and Andrew Oswald found that having a technically competent boss was, in their American and British survey data, a stronger predictor of a worker's job satisfaction than pay or any other factor they measured.31,32The hospital finding is an association, not proof of cause, (as Goodall says). But it is a useful counterweight to the assumption that a bigger organisation mainly needs more generic management.
Implementation is often where reforms succeed or fail. Implementation science, a field that grew up in health precisely because good evidence kept failing to change practice, has produced many frameworks that disagree on words but agree on substance: what happens depends on the change itself, its setting, the people involved and on the receiving end, and the process used to introduce it.33 Trisha Greenhalgh and colleagues, studying technology supported change in health and care, concluded that programmes which were complicated proved difficult but not impossible to implement, while those that were complex in several domains at once rarely if ever became mainstream. The most common mistake they saw was treating a complex problem as if it were merely complicated: knowable, predictable and controllable.34 The analogy to a national police reform is not exact, but the pattern is worth a look. The updated Consolidated Framework for Implementation Research now gives explicit attention to equity and to the people on the receiving end of change.35
Of course, none of this is a recipe. It is a set of questions that a reform programme can ask of itself before, during and after the event, informed by evidence.
What role should evidence play?
Research evidence cannot determine the structure of policing on its own, and I would be deeply suspicious of anyone who told you it could. There is not enough of it, and what exists cannot tell you how many forces England and Wales should have or where the boundaries should go. Those are decisions for government, informed by professional judgement in conditions no study can reproduce, and by the legitimate views of communities about the service they want. Rob Briner, David Denyer and Denise Rousseau's account of evidence based management gets this balance right: good decisions draw on four sources of evidence:
1. the best available research,
2. data from the organisation itself,
3. professional expertise, and
4. the perspectives of those affected.36
Research is one of four legs, not the whole chair. What it can do is constrain the claims made for a structure and shape how those claims are tested.
But there are three things evidence can and should do in a reform programme, and there is still time to do all three well.
First, test the promises. Every reform proposal contains causal claims. Fewer forces will improve resilience. National standards will raise consistency. Concentrating specialist functions will improve learning.1 The white paper introduces that last claim with "the evidence suggests" without yet citing it. Those claims deserve to be written down as claims, with the supporting evidence and the important uncertainties beside them. That is what Briner and colleagues mean by making the justification for a decision explicit, so that the assumptions can be appraised even where the evidence is thin.36 Where the evidence is strong, say so. Where it is weak, say that too and act anyway if the judgement is that action is needed. What we should not do is enter expected savings or expected improvements into the business case as though they had already happened. The white paper already says "The money saved will be reinvested in local frontline policing", and the review's terms of reference require the merger programme's costs to be offset "through efficiencies in later years".1,2 Scotland's auditors looked at nine public body mergers in 2012, before the police one, and found that "performance measures relating to the expected benefits of mergers were absent or underdeveloped in all of the mergers we examined", and that costs for the four they studied closely had reached £42 million against a £30 million forecast. Their guidance asked for measurable benefits, baseline data from the predecessor bodies and a full evaluation within two years. A year later Scotland merged its police forces without a baseline.45 The hospital and local government literature is a long list of organisations whose expected savings were written into the plan and never turned up, sometimes because merging the organisation did nothing to the size of the units that deliver the service, and sometimes because the cost of the merger itself was, in one NHS director's words, "very hidden but very colossal".20,25
Second, plan for the evidenced risks. The evidence identifies a credible risk to local presence and local knowledge. It suggests particular exposure for rural and some deprived communities. It shows, repeatedly, that implementation takes longer and costs more than planned, and that governance and workforce planning are where reforms most often come unstuck. Sweden's auditors add that government demands for quantitative follow up "led to a focus on numbers rather than quality and effect", HM Chief Inspector warns that national targets must not "encourage perverse behaviour that worsens outcomes for the public", and one expert told the Cordis Bright researchers that pressure to "count stuff rather than do stuff" means "we may hit the targets but miss the point".15,4,38 A new performance framework is part of the reform package. It deserves the same scrutiny as the boundaries. And the evidence gives no reason to assume public confidence restores itself. None of these is a reason not to reform. Every one of them is a reason to build specific protections and specific measures into the design from day one, rather than discovering the problem in year four.
Third, learn while doing. This is the one I care most about. The white paper envisages pathfinder mergers this Parliament, and the independent review was asked to say which forces are best suited to them. Neither document says how the pathfinders will be evaluated; the white paper describes them as steps "towards the end-state".1,2 A pathfinder is, by definition, an opportunity to find out whether something works before doing it everywhere. That only happens if somebody has set the baseline before the boundaries move, agreed what success looks like, arranged for comparison with places that have not yet changed, and given the findings a route into the decisions that follow. That last skill is scarce. In the Cordis Bright survey, only 2 per cent of the 48 analysts said they had tested an activity's outcomes against an unaffected comparison in the previous twelve months.38 Scotland deserves credit for commissioning a four year externally conducted evaluation of its reform.7,13 But it began in 2015, nearly two years after the merger, and its authors noted how little baseline evidence existed for the period before reform. The Dutch evaluators record that "no baseline measurement was carried out before the reorganisation", that no counterfactual was available, and that four years in nobody in the system could answer the question "how much safety does society get for the money invested?"; their central recommendation was a monitoring and evaluation unit at the top of the organisation.7,16 Thirteen years after the Scottish merger, the auditors are still recommending that policing "set clear performance and financial baselines" before the next stage of reform.14 England and Wales can do at least as well as Scotland, and has the chance to go further by starting earlier.
Two examples of what this could look like
Children and violence. The Youth Endowment Fund published its practice guidance on policing, children and violence this month (I should say that I sat on its expert panel, SEBP's Chair sat on its strategic advisory group, and SEBP was one of the organisations interviewed for the supporting research), with six recommendations covering everyday interactions with children, safeguarding, problem oriented policing, stopping harmful and unevidenced approaches, the quality of stop and search, and fair out of court resolutions.37 The accompanying study by Cordis Bright, based on a survey of 528 police respondents across 44 forces, interviews and case studies, is a candid read on how evidence is currently used. Fewer than half of senior leaders (inspector and above, in the survey's definition) felt they had the capacity to support evidence based practice. Decisions were informed primarily by internal data, professional judgement and operational demand rather than research. Nearly half of respondents believed prison awareness programmes were effective, when the YEF's guidance lists them among approaches likely to increase violence. Only 17 per cent reported routinely collecting and analysing data on how interventions affect children from Black, Asian and minority ethnic backgrounds.38 The survey was distributed largely through evidence based policing networks, including SEBP's own mailing list, so people already engaged with evidence are probably over represented; this is a picture of the more engaged end of the workforce, not a census of forces.
Reform could make this better or worse. Consolidation might bring more consistent diversion and out of court resolution across a larger area; the YEF's safeguarding review found children receiving different outcomes for knife possession "with the only differentiating feature being force boundaries", and its drug markets review found trafficking investigations "the first casualty" of working across force lines. Or it might centralise a policy while stretching the local relationships through which diversion actually works; the same reviews describe an "internal postcode lottery" inside forces, and officers "frequently redeployed; no continuity". Children told the researchers what they valued: "Everybody knew them, and they knew us."37 The way to know is to agree a small set of shared measures now, with youth justice partners and with children themselves: who is offered support, who receives it, what happens afterwards, and how those experiences differ by area and by ethnicity. Give a named operational leader responsibility for acting on what the numbers say. That is what using evidence in reform could look like in practice.
The Licence to Practice. The white paper proposes a licence to renew the office of constable for the twenty first century and says the government will consult extensively with the service on how to introduce it.1 The Police Leadership Commission, set up by the College of Policing, recommended a route towards it in July: a national database of professional digital passports, offered as "a light touch non-bureaucratic way of achieving the Government's stated aim". The whole idea has been called a gimmick, though on grounds of capacity rather than principle.39,40 So traction is not yet clear, but if a licence is to mean anything beyond a database entry, it needs to say something about what a licensed professional is expected to know and to keep knowing. Neither the white paper, which frames the licence around training completion, wellbeing checks and career pathways, nor the Commission, which frames it around performance reviews and a record of learning, yet says anything about research evidence. For a constable, that might mean knowing where to find trustworthy guidance and understanding its limits; in one large force studied recently, fewer than four in ten operational officers did.47 For a supervisor, checking whether the team's practice reflects it. For a chief officer, being able to tell whether an intervention or an organisational change is producing the benefit claimed for it. This should be proportionate, it should be supported with time and access to material, and it should be piloted and evaluated before it is rolled out. A requirement without the means to meet it is just another burden.
Sowing the seeds now
If reform is a hypothesis (that bigger = better), the test has to be designed before the experiment starts. I would ask those leading it to consider five things, none of which requires a large research budget and all of which are much harder to retrofit.
Name the better outcomes and capture the baselines before anything moves. For each major change, a short statement of the problem, the expected benefit, the supporting evidence and the main uncertainties. Measure where we are now, in the places that will change first, before the systems and boundaries alter. The London trust merger researchers said as much in 2002: their study was commissioned after the mergers, and they concluded that ideally such research should begin before.20
Make evaluation part of the pathfinders, not a report on them. Plan comparisons with areas that have not yet changed. Use trials where they fit, for example on training and practice changes. Use comparison areas, trend analysis and close study of implementation where they do not. A before and after snapshot on its own cannot tell you what reform caused.
Give practitioners and communities a stake in the questions and protect the independence of the answers.Officers, staff, partners and residents should help decide what gets measured and help interpret what comes back. Researchers would ideally control the analysis and the conclusions, with publication agreed at the outset. Involvement should improve relevance; it should not confer a veto on inconvenient findings.
Build a route from findings to decisions. Reform boards could receive short, regular accounts of what is improving, what is getting worse and what remains uncertain, and should record what they did about it and when they will look again. New Zealand's police, when they wanted their measures to reflect the aims of reform, cut around 2,000 performance indicators to four outcome measures and gave local commanders more room to decide how to reach them.13 Protect longer term evaluation too, because some effects, on confidence especially, will take years to show up.
Resource the ability to respond. This is partly money, but mostly access to data, analytical time, and the permission for managers to change course if and when the evidence says they should. The new £50 million UKRI Safer Streets programme, including work on trust and confidence, violence against women and girls and knife crime, is an obvious place to connect reform's learning agenda to the research community.41,42 It will not be quick. The £2 million trust and confidence consortium starts in February 2027 and does not expect its new measures to be ready for piloting until the end of that year, and it will not fund large scale surveys. Baselines for the first pathfinders will have to come from the surveys we already have. Learning that nobody is allowed to act on is an expensive way of being right too late.
Why this matters to us
The Society of Evidence Based Policing exists to help policing use, produce and communicate evidence. UK policing has come a long way on that in the past fifteen years: a national What Works Centre in the College of Policing, a growing network of university partnerships, a professional society with thousands of members, and a conference that this year brings the Policing Minister and the Chair of the National Police Chiefs' Council into the same room as the researchers. The theme of this year's National Evidence-Based Policing Conference is what works in delivering police reform, and the truth is that we know less than we would like about structures, a good deal about the risks, and almost everything about how to find out more. Scotland's evaluators put the gap bluntly: "a culture of evaluation is only weakly embedded in police and fire and rescue services in Scotland and internationally".13
Reform will be judged by what changes for the public: whether victims get a better service, whether children are safer, whether communities with fragile trust see reasons to trust more, and whether officers and staff are better supported to do the job. Structure may help with all of that. It will not deliver it on its own, and it will not tell us whether it has worked unless we ask. Now is the time to kick on and make asking part of the plan.
Notes and sources
This is a narrative synthesis of English language peer reviewed research and independent official evaluations published mainly in the past twenty years, with one Dutch language statutory evaluation read in the original. It is not a systematic review. Research findings, official policy commitments and my own proposals are distinguished in the text. Sources were checked against full texts on 15 to 17 September 2026.
1. Home Office (2026) From local to national: a new model for policing, CP 1489. https://www.gov.uk/government/publications/from-local-to-national-a-new-model-for-policing Policy intentions, not evidence of achieved outcomes. Paragraphs 144 (specialist concentration), 150 (reinvestment of savings), 157 (pathfinders) and 268 to 274 (Licence to Practice).
2. Home Office (2026) Independent review of police force structures: terms of reference. https://www.gov.uk/government/publications/independent-review-of-police-force-structures-terms-of-reference/independent-review-of-police-force-structures-terms-of-reference
3. HC Deb, 2 September 2026, Prime Minister's Questions, response to Jo Stevens MP. https://www.theyworkforyou.com/debates/?id=2026-09-02c.159.9
4. HMICFRS (2026) State of Policing: The Annual Assessment of Policing in England and Wales 2025/26. https://www.gov.uk/government/publications/state-of-policing-2025-to-2026 Force size and structures, pages 6 to 7; national targets, page 11.
5. Mendel, J., Fyfe, N.R. and den Heyer, G. (2017) Does police size matter? A review of the evidence regarding restructuring police organisations. Police Practice and Research, 18(1), 3 to 14. https://doi.org/10.1080/15614263.2015.1135399 Narrative review; heterogeneous and partly dated evidence base; English language sources only; written for the Scottish Government's 2011 reform process. The cost curve finding is from Drake, L. and Simper, R. (2000) Productivity estimation and the size-efficiency relationship in English and Welsh police forces. International Review of Law and Economics, 20(1), 53 to 73. https://doi.org/10.1016/S0144-8188(00)00021-1
6. Holmberg, L. (2019) Continuity and change in Scandinavian police reforms. International Journal of Police Science and Management, 21(4), 206 to 217. https://doi.org/10.1177/1461355719889461 The fuller Danish account is Holmberg, L. and Balvig, F. (2013) Centralization in disguise: the Danish police reform 2007 to 2010, in Fyfe, N.R., Terpstra, J. and Tops, P. (eds) Centralizing Forces? The Hague: Eleven, 41 to 57.
7. Scottish Government (2017) Evaluation of Police and Fire Reform: Year 2 Report. SIPR, What Works Scotland and ScotCen. https://www.gov.scot/publications/evaluation-police-fire-reform-year-2-report/ Qualitative case studies in four contrasting areas; no causal estimate of merger effects. Fire service localism, sections 4.3.2 and 5.1; limited baseline evidence, section 2.2.
8. Statskontoret (2018) Reorganisation to create a unified police authority: final report, 2018:18 (English summary). https://www.statskontoret.se/in-english/publications/2018/reorganisation-to-create-a-unified-police-authority.-final-report-201818/
9. Terpstra, J. and Fyfe, N.R. (2015) Mind the implementation gap? Police reform and local policing in the Netherlands and Scotland. Criminology and Criminal Justice, 15(5), 527 to 544. https://doi.org/10.1177/1748895815572162 Early implementation study; 15 interviews in each country, October 2013 to June 2014.
10. Terpstra, J., Fyfe, N.R. and Salet, R. (2019) The Abstract Police: a conceptual exploration of unintended changes of police organisations. The Police Journal, 92(4), 339 to 359. https://doi.org/10.1177/0032258X18817999 Conceptual paper drawing on prior fieldwork. The empirical study of Dutch local teams behind it is Terpstra, J. (2021, online 2018) Local policing in a nationalized police force: a study on the local teams of the Netherlands' National Police. Policing: A Journal of Policy and Practice, 15(1), 251 to 262. https://doi.org/10.1093/police/pay037
11. Fyfe, N.R., Anderson, S., Bland, N., Goulding, A., Mitchell, J. and Reid, S. (2021) Experiencing organizational change during an era of reform: Police Scotland, narratives of localism, and perceptions from the 'frontline'. Policing: A Journal of Policy and Practice, 15(1), 263 to 276. https://doi.org/10.1093/police/pay052 Qualitative fieldwork with local policing teams in four communities, 2016.
12. Wooff, A. (2024) Police custody in rural Scotland: negotiating boundaries, risk and organisational change. Policing: An International Journal, 47(3). https://doi.org/10.1108/PIJPSM-01-2024-0007 Qualitative; fieldwork 2016 to 2018.
13. Scottish Government (2019) Evaluation of Police and Fire Reform Year 4: International Perspectives on Police and Fire Reform, and the companion Year 4 Summary of Key Findings and Learning Points. https://www.gov.scot/publications/evaluation-police-fire-reform-year-4-international-perspectives-police-fire-reform/ International report: 29 elite interviews, no outcome data. The companion Summary reports Scottish Crime and Justice Survey trends (page 8), the key learning points (page 12) and the culture of evaluation finding (page 18). New Zealand indicators, International report section 5.1.2.
14. Audit Scotland and HMICS (2026) Best Value in Policing: Joint Best Value Audit of Policing in Scotland, January 2026. https://audit.scot/uploads/2026-01/nr_260122_bv_policing.pdf The savings figure is policing's own estimate, which the auditors accept as valid historical reporting (paragraph 49). Trust and local confidence, key message 8 and case study 3; baselines, key message 9 and recommendation 2.
15. Riksrevisionen (Swedish National Audit Office) (2025) The 2015 reform of the Swedish Police Authority: intentions not yet fulfilled, RiR 2025:26. English summary: https://www.riksrevisionen.se/en/news-archive/nyhetsarkiv-eng/2025-11-13-unclear-governance-and-poor-performance-after-the-police-reform.html Audit, not a causal evaluation; cannot isolate the reform from crime trends.
16. Commissie Evaluatie Politiewet 2012 (2017) Evaluatie Politiewet 2012: Doorontwikkelen en verbeteren. The Hague. https://www.eerstekamer.nl/overig/20171116/evaluatie_politiewet_2012/document Statutory evaluation; Dutch language, translations by Claude. Formative evaluation four and a half years in, based on five commissioned sub studies and about 350 interviews; the commission said a final judgement was not yet possible and recommended a further system evaluation after five years. Public appreciation, section 6.5; savings and costs, sections 4.2 and 4.4; baseline, section 6.1; governance, chapter 5.
17. Henry, A., Malik, A. and Aydin-Aitchison, A. (2019) Local governance in the new Police Scotland: renegotiating power, recognition and responsiveness. European Journal of Criminology, 16(5), 573 to 591. https://doi.org/10.1177/1477370819856528
18. Murray, K. and Harkin, D. (2017) Policing in cool and hot climates: legitimacy, power and the rise and fall of mass stop and search in Scotland. British Journal of Criminology, 57(4), 885 to 905. https://doi.org/10.1093/bjc/azw007 Documentary, statistical and interview study of one policy area.
19. Gaynor, M., Laudicella, M. and Propper, C. (2012) Can governments do it better? Merger mania and hospital outcomes in the English NHS. Journal of Health Economics, 31(3), 528 to 543. https://doi.org/10.1016/j.jhealeco.2012.03.006 Quasi experimental event study; England specific. Deficit, waiting time and agency staff figures are from the working paper version (NBER 17608).
20. Fulop, N., Protopsaltis, G., Hutchings, A., King, A., Allen, P., Normand, C. and Walters, R. (2002) Process and impact of mergers of NHS trusts: multicentre case study and management cost analysis. BMJ, 325, 246. https://doi.org/10.1136/bmj.325.7358.246 Small case study in one region, two years after merger; the authors call the findings interim. Savings were £178,700 in year one and £346,800 in year two against a £500,000 target.
21. Mariani, M. et al. (2022) Impact of hospital mergers: a systematic review focusing on healthcare quality measures. European Journal of Public Health, 32(2), 191 to 199. https://doi.org/10.1093/eurpub/ckac002 Sixteen studies, searches to January 2020; strength of evidence rated insufficient for every indicator. See also Beaulieu, N.D., Dafny, L.S., Landon, B.E. et al. (2020) Changes in quality of care after hospital mergers and acquisitions. New England Journal of Medicine, 382, 51 to 59. https://doi.org/10.1056/NEJMsa1901383 246 acquired US hospitals against 1,986 controls: modest decline in patient experience, no change in mortality or readmissions.
22. Braithwaite, J., Westbrook, M.T., Hindle, D., Iedema, R.A. and Black, D.A. (2006) Does restructuring hospitals result in greater efficiency? An empirical test using diachronic data. Health Services Management Research, 19(1), 1 to 12. https://doi.org/10.1258/095148406775322016Twenty hospitals; Australian context.
23. Andrews, R. and Boyne, G.A. (2009) Size, structure and administrative overheads: an empirical analysis of English local authorities. Urban Studies, 46(4), 739 to 759. https://doi.org/10.1177/0042098009102127 Cross sectional; overheads only.
24. Andrews, R. (2013) Local government amalgamation and financial sustainability: the case of England and Wales. Public Finance and Management, 13(2), 124 to 141. https://doi.org/10.1177/152397211301300205 Short post reform window.
25. Blom-Hansen, J., Houlberg, K., Serritzlew, S. and Treisman, D. (2016) Jurisdiction size and local government policy expenditure: assessing the effect of municipal amalgamation. American Political Science Review, 110(4), 812 to 831. https://doi.org/10.1017/S0003055416000320Difference in differences over eight post reform years; spending and cost per unit, not service quality. The "plant versus firm" point is at pages 4 and 17. The companion study finding that administrative costs per inhabitant did fall is Blom-Hansen, J., Houlberg, K. and Serritzlew, S. (2014) Size, democracy, and the economic costs of running the political system. American Journal of Political Science, 58(4), 790 to 803. https://doi.org/10.1111/ajps.12096
26. Zheng, T., Sun, B. and De Witte, K. (2026) Do municipal amalgamations work? A systematic literature review. Urban Affairs Review, 62(5), 1571 to 1603. https://doi.org/10.1177/10780874251410080
27. Lassen, D.D. and Serritzlew, S. (2011) Jurisdiction size and local democracy: evidence on internal political efficacy from large-scale municipal reform. American Political Science Review, 105(2), 238 to 258. https://doi.org/10.1017/S000305541100013X Measures perceived efficacy, not participation or outcomes.
28. Eyles, A. and Machin, S. (2019) The introduction of academy schools to England's education. Journal of the European Economic Association, 17(4), 1107 to 1146. https://doi.org/10.1093/jeea/jvy021 Early sponsored academies only; effects confined to conversions from community schools.
29. Eyles, A., Machin, S. and McNally, S. (2017) Unexpected school reform: academisation of primary schools in England. Journal of Public Economics, 155, 108 to 121. https://doi.org/10.1016/j.jpubeco.2017.09.004 Voluntary converters only, up to four years after conversion.
30. Blanes i Vidal, J. and Kirchmaier, T. (2018) The effect of police response time on crime clearance rates. Review of Economic Studies, 85(2), 855 to 891. https://doi.org/10.1093/restud/rdx044 Single force; clearance is a police recorded outcome. The estimate is a local average treatment effect and larger than the simple correlation.
31. Goodall, A.H. (2011) Physician-leaders and hospital performance: is there an association? Social Science and Medicine, 73(4), 535 to 539. https://doi.org/10.1016/j.socscimed.2011.06.025 Cross sectional association in the top 100 ranked US hospitals in three specialties; no causal claim.
32. Artz, B.M., Goodall, A.H. and Oswald, A.J. (2017) Boss competence and worker well-being. ILR Review, 70(2), 419 to 450. https://doi.org/10.1177/0019793916650451 Competence is worker reported; US and British survey data, 1979 to 2000.
33. Nilsen, P. (2015) Making sense of implementation theories, models and frameworks. Implementation Science, 10, 53. https://doi.org/10.1186/s13012-015-0242-0 Conceptual taxonomy, not an empirical synthesis.
34. Greenhalgh, T. et al. (2017) Beyond adoption: a new framework for theorizing and evaluating nonadoption, abandonment, and challenges to the scale-up, spread, and sustainability of health and care technologies (NASSS). Journal of Medical Internet Research, 19(11), e367. https://doi.org/10.2196/jmir.8775 Six UK health and care technology case studies followed for up to three years.
35. Damschroder, L.J., Reardon, C.M., Widerquist, M.A.O. and Lowery, J. (2022) The updated Consolidated Framework for Implementation Research based on user feedback. Implementation Science, 17, 75. https://doi.org/10.1186/s13012-022-01245-0 User survey and literature review; mostly US healthcare.
36. Briner, R.B., Denyer, D. and Rousseau, D.M. (2009) Evidence-based management: concept cleanup time? Academy of Management Perspectives, 23(4), 19 to 32. https://doi.org/10.5465/amp.23.4.19 Conceptual, not empirical. On the professional judgement leg see also Fleming, J. and Rhodes, R.A.W. (2018) Can experience be evidence? Craft knowledge and evidence-based policing. Policy and Politics, 46(1), 3 to 26. https://doi.org/10.1332/030557317X14957211514333
37. Youth Endowment Fund (2026) Policing, Children and Violence, practice guidance, 3 September 2026. https://youthendowmentfund.org.uk/reports/policing-children-and-violence/ The guidance draws on four YEF funded reviews published alongside it: Cordis Bright on evidence based practice (note 38); Allnock, Beckett, Hibbin and others (Bedfordshire and Lancashire) on child centred policing; Beckett, Allnock, Richardson Foster and others (Lancashire, Bedfordshire, Open University and VKPP) on safeguarding; and Macdonald, Tudor, Cosgrove and others (Durham) on children's exploitation in drug markets. All are mixed method and descriptive; none estimates causal effects.
38. Farrell, J., Collins, A., Bhambri, S., Tomkins, J., Kaur, K. and Brennan, I. (Cordis Bright) (2026) Review of evidence-based practice in policing to prevent children from being involved in violence: exploratory study report. Youth Endowment Fund. https://youthendowmentfund.org.uk/wp-content/uploads/2026/08/EBP_Report_SEP_2026.pdf Mixed method exploratory study; survey October to December 2025; respondent percentages are not percentages of forces. Comparison group finding, section 3.6; performance framework quotations, page 53.
39. Police Leadership Commission (2026) Professionalism and performance: police leadership for the future, 6 July 2026, Recommendations 10 and 11 and paragraph 59 (Crime Survey for England and Wales confidence trend). https://www.college.police.uk/article/police-leadership-commission-report-published The Commission was established by the College of Policing with Home Office support and co chaired by the College's Chair. Ministerial statement: HC Deb, 6 July 2026.
40. Police Federation of England and Wales, Norfolk branch (2026) Federation brands licence to practice for police officers as a gimmick, 28 January 2026. https://polfed.org/norfolk/news/2026/federation-brands-licence-to-practise-for-police-officers-as-a-gimmick/
41. UKRI (2026) £50m research and innovation programme launched to make Britain's streets safer for everyone, 7 April 2026. https://www.ukri.org/news/50m-research-and-innovation-programme-launched-to-make-britains-streets-safer-for-everyone/ Programme commitments, not results. The £50 million spans five challenges, including prison security technology.
42. UKRI (2026) Trust and confidence in policing: Research and Innovation Consortium, funding opportunity opened 1 September 2026. https://www.ukri.org/opportunity/trust-and-confidence-in-policing-research-innovation-consortium/ Total fund £2 million; start by 1 February 2027; tools for piloting by December 2027; conventional large scale surveys excluded.
43. Dominguez, M. and Kirchmaier, T. (2026) Restructuring a police force. Centre for Economic Performance Discussion Paper 2148, London School of Economics, January 2026. https://cep.lse.ac.uk/pubs/download/dp2148.pdf Working paper; staggered difference in differences on the Metropolitan Police's 2017 to 2018 consolidation of 32 borough units into 12; internal restructuring within one force, not a merger of separate forces; data run to March 2019, so effects are measured over roughly one to two years after each unit merged. On scale economies the paper cites Finney, M. (1997) Scale economies and police department consolidation: evidence from Los Angeles. Contemporary Economic Policy, 15(1), 121 to 127, which found no evidence of increasing returns.
44. Facchetti, E. (2025 revision) Police infrastructure, police performance and crime: evidence from austerity cuts. IFS Working Paper W24/16. https://doi.org/10.1920/wp.ifs.2024.1624 PDF: https://ifs.org.uk/sites/default/files/2025-12/WP202511-Police-infrastructure-police-performance-and-crime-evidence-from-austerity-cuts2.pdf Working paper, not yet peer reviewed; stacked difference in differences across 4,701 census blocks; analysis stops at 2016 because of the Met's later restructuring.
45. Audit Scotland (2012) Learning the lessons of public body mergers: good practice guide, and Review of recent mergers, June 2012. https://audit.scot/docs/central/2012/nr_120614_public_body_mergers_guide.pdf and https://audit.scot/docs/central/2012/nr_120614_public_body_mergers.pdf Nine mergers reviewed, four in depth; not police bodies.
46. Garside, P. (1999) Evidence based mergers? BMJ, 318, 345 to 346. https://doi.org/10.1136/bmj.318.7180.345 Editorial; commentary, not evidence.
47. Selby-Fell, H. and Newton, A. (2024) Embedding evidence-based policing (EBP): a UK case study exploring organisational challenges. The Police Journal, 97(1), 73 to 91. https://doi.org/10.1177/0032258X221128404 Single metropolitan force, 2016/17; 327 survey respondents; 49 per cent overall knew where to find research evidence, 86 per cent of senior officers and 38 per cent of operational officers.
48. District counts: Sweden, Statskontoret 2018:18 (21 county authorities, the National Police Board and the National Laboratory of Forensic Science into one authority, 1 January 2015); Norway, Ministry of Justice and Public Security, Prop. 60 L (2015 to 2016) (27 districts to 12, 1 January 2016); Denmark and Norway, Scottish Government (2019) Year 4 International Perspectives, Appendix A. Scotland and the Netherlands 2013, note 9.
A note on methods
Artificial Intelligence tools played a large role in the preparation of this article. Here's how they were used:
• An initial list of papers was compiled using author specification of the time, language and keyword parameters of the research papers sought.
• Claude Fable 5.1 was used to prepare an initial tabulated summary of those papers.
• From this summary, the author dictated an article framework from which Claude Fable 5.1 prepared a first full draft.
• The author used Chat GPT 6 Astra to retrieve original texts for all but two of the cited papers and Claude Fable 5.1 to translate foreign language papers. These papers were used to validate and correct the initial full draft.
• The author then rewrote the draft article over four iterations, in dialogue with Claude Fable 5.1 and Chat GPT 6 Astra which played the roles of researcher and reviewer respectively.
• A final draft was prepared by the author with no AI use, based on version 4 and the 48 source documents collated.
• Chat GPT 6 Astra prepared the accompanying summary based on the article and the social media content associated.
Why was AI used? This is not primary research, nor does it masquerade as a peer-reviewed literature review. It is a weblog article prepared as a discussion note. It has been thoroughly checked for provenance and the core ideas have been assembled and developed by a human, but it is entirely possible errors have been retained (or introduced by that human). It would not be possible to assemble an article of this scope in the time available (four days from start to finish, while working on other tasks) without AI and at SEBP we are keen to learn how AI can support the use, production and communication of evidence. We also acknowledge the importance of accuracy. We see articles like this as part of natural experimentation with the use of new technology and seek earnest feedback on its effectiveness. This article is not peer-reviewed but feedback from readers, particularly those cited, is welcomed. The article will be updated periodically with necessary corrections. All feedback to the author at coo@sebp.police.uk.
