Chapter 10: Policing

Police services account for 4.5% of total Government expenditure, and yet, like Defence, measures have been unable to mature past the ‘inputs = outputs’ stage, with the result that productivity growth is automatically zero. This is due to a number of inter-locking challenges: data availability, the complexity of the range of outputs which policing delivers, the attributability of crime outcomes to solely police activities, the preventative nature of many policing activities, and the way the Classification of Functions of Government (COFOG) groups policing with both Immigration services and the Criminal Justice services.

In 2017, the Office for National Statistics (ONS) published new methods for measurement of the Criminal Justice System, Quality adjustment of public service criminal justice system output: experimental method (2017) as a first step to developing this area, separately publishing these from the rest of the Public Order and Safety COFOG. Policing and Immigration are therefore published in the ONS statistics within the ‘Other’ grouping.

This Review made policing one of its highest priorities, particularly in the positive light of the National Police Chiefs’ Council led Policing Productivity Review published in 2023. This identified new data sources and encouraged the Home Office to create a Centre for Police Productivity (see Chapter 6), whom the Review has collaborated with in developing new methods.

10.1 Core methodological and data challenges identified

A complicating factor in identifying an appropriate methodology for measuring police service productivity is that the relevant COFOG category also encompasses a range of Immigration & Citizenship services, as discussed in Chapter 11. Consequently, substantial changes in expenditure in these services make spending in the overall category appear volatile and hard to relate to policing experience in practice. The Review therefore prioritised separating spending and output related to Immigration & Citizenship services to create a separate productivity measure (see Chapter 11). This enabled a clear line of sight on policing inputs.

Following the decision to separate Policing from Immigration and Criminal Justice, and in the light of new data becoming available, the remaining major challenges are:

• Identifying and weighting the wide range of policing activities in what is an operationally conceptually complex service to deliver an output metric.
• The attributability of crime outcomes to police activity in terms of quality adjusting that activity.
• The preventative nature of police activity, and how best to capture and weight this within the measurement of output.

The Review has, drawing on wider practice, divided police activity into three broad areas: ‘crime and criminal investigation’, ‘crime prevention’ and ‘public safety and welfare’. The Review has prioritised ‘crime and criminal investigation’ as this area has the strongest data sources. However, it does not underestimate the conceptual challenges and data limitations in bringing together all the aspects of police activity into a single output measure. In other services, cost weighted indices are standardly applied, but it may be necessary to consider alternative methods. The ONS continues to work with Home Office and the Centre for Police Productivity to consider options for weighting.

The Review explored improvements to inputs estimates but focused on producing direct measures of output. The Review has considered quality adjustment, but implementation is dependent on the exact design of output measures. This is clearly an area for further work which may benefit from considering analogous services elsewhere: comparing police custody suites with prisons, potentially in respect to measures of safety and decency such as deaths in custody.

10.2 Improvements to inputs estimates

Unlike Defence, where the Review explored moving labour measures from indirect to direct (see Annex F), direct measures for Policing were already available. The degree of detail achievable has been examined.

As in almost all services, labour accounts for the greatest contribution to inputs, and most of this is directly measured using workforce data and associated average salary information. However, the Review recognised that the granularity and coverage of these data may be improved. The Review investigated a number of labour measurement improvements:

• Granularity of police salary data: Improving granularity of direct police labour data by disaggregating officer salaries by their respective ranks.
• Accounting for the Police Uplift Programme (2019 to 2023): Accounting for different skill-levels of police labour arising from the Uplift programme.
• Accounting for reduced capacity: Accounting for different quality of police labour arising from the increasing proportion of officers on limited duties.
• Improving police labour data coverage for the devolved governments: Improving the comprehensiveness of devolved police labour data. This area is covered under the ‘devolved governments’ subsection.

Full-time Equivalent (FTE) police staff numbers (police and civilians) by rank in England and Wales is shown in Figure 4. Overall numbers have fluctuated over time, rising to a peak in 2010 of 244,497 from a low of 198,375 in 2003, with material variation in the share in specific categories.

Figure 4: Number of full-time equivalent police staff by rank, England & Wales, 2003 to 2021

Granularity of police salary data

The Review has used newly acquired average salary data from the Home Office to split the consolidated groups of salaries to the same level of granularity as the full-time equivalent data used in the ONS direct labour measure from 2014 onwards. Prior to this, the average salary information used was grouped for (i) Constables and Sergeants, and (ii) Inspectors & Above.

Average salary data for Police Community Support Officers (PCSOs) are also available from ONS Annual Survey of Hours and Earnings (ASHE) data and the Review proposes using these to disaggregate PCSOs in the measure from 2011 onwards. This will result in the most accurately measured police direct labour index to date.

While the indicative impact on annual input growth rates may be small, the improved granularity ensures the measure is more reflective of changes in the level of labour input at each rank, as even a small impact on the inputs measure can reflect a considerable amount of underlying expenditure. This impact is also made more visible when policing is presented as a service distinct from Immigration & Citizenship services, which will also enhance the useability of these data for practical purposes.

Accounting for Police Uplift Programme

The Police Uplift Programme was a police recruitment programme running from October 2019 to March 2023. Its goal was to recruit 20,000 new officers within this period, which was achieved, with approximately 21,000 officers recruited. The implication of such a recruitment regime for trainees on productivity measurement is that it has the potential to cause an over-estimation of the value of the inputs being used in a service, and hence under-estimate productivity.

Figure 5: Number of police officers recruited under the Police Uplift Programme, to end March 2023

Source: Home Office: Accredited official statistics: Police officer uplift, final position as at March 2023

Given the availability of distinct salary rates for staff on uplift, the Review concluded, as discussed in Chapter 3 that further intervention would ‘double-count’ this effect: the lower salary rate of staff when on the Uplift scheme is assumed, as with all other services, to reflect the lower contribution made by these staff at this time. These data were used to derive separate weights to apply to FTE growth rates to account for the impact of the uplift on labour input.

Accounting for reduced capacity through officers on limited duties

Limited duties include officers on recuperative and adjusted duties, as well as those on restricted duties. Recuperative duties typically occur when an officer has been injured whilst on duty and requires a period of recuperation with less or no deployment to either regain their full duties or move onto adjusted duties. Typically, this lasts for up to six months, however, there are exceptions where this can be increased to 12 months. During this time the officer is paid their full wage. Restricted duties are for officers who have not met the standards required for policing, typically by poor behaviour.

The Review considered whether an adjustment should be applied to account for reduced capacity arising from an increased proportion of staff on limited duties. The Review attempted to identify an appropriate discount factor to apply to this group of staff.

The Winsor Review (2012) calculated the risks that are encountered by fully operational police relative to other public sector workers: this “x-factor” equates to 8% of pay. The Police Remuneration Review Body 2022 report presents a potential replacement for the “x-factor”: the “p-factor”, valued at 13%, which is defined as “element of police basic pay that reflected the responsibilities, obligations, constraints, expectations and terms and conditions that are unique to a police officer’s work”. Therefore, the Review investigated the impact of incorporating an FTE reduction factor of both 8% and 13% in the direct police labour measure to account for the impact of limited duties. This had a negligible impact on the measure, with the 13% discount factor only reducing the 2021 annual inputs growth rate by 0.01 percentage points, as the share of staff on limited or restricted duties changes.

However, the lack of impact was not the reason for the Review deciding not to recommend incorporation to the ONS measure. The Review concluded that, while effectiveness at delivering frontline service has been impacted, the system is designed to accommodate such staff within the wider range of services and processes delivered by policing, and hence is ‘priced-into’ the operating model of the service. Unless there is a material change in the number of such staff through time causing the operating model to change, the ‘average input’ of a police officer has not changed and as such adjustment is unnecessary.

Input deflators

The Review conducted a full audit of the input deflators currently used in the compilation of estimates (labour, intermediate consumption and capital consumption). The central government intermediate consumption deflator (GDP implied deflator) will be replaced with a bespoke police and immigration deflator, based on detailed expenditure from the Online System for Central Accounting and Reporting (OSCAR) from Spring 2025. The Review also developed separate bespoke intermediate consumption deflators for Police and Immigration & Citizenship services, in preparation for the intended disaggregation of estimates for these two areas.

Devolved governments

Current input data sources include labour data from England, Wales, and Scotland. The Review was able to source labour data from Northern Ireland for the first time, adding a time series from 2000 onwards. There is a need for more disaggregated labour data (FTE by rank) for Scotland for earlier in the time series to fill in current gaps.

10.3 Development of outputs

On the output side, the Review has divided police outputs into three categories to simplify working through the diversity of services delivered:

• ‘Crime and crime investigation’ relating to services addressing crimes which have already occurred. Crime refers to a deliberate act that causes physical or psychological harm, damage to or loss of property, and is against the law (police.uk).
• ‘Crime prevention’ relating to services to prevent the occurrence of crimes which are yet to occur. Crime prevention is activity taken by the police that actively prevents a crime incident from occurring (such as visible patrols).
• ‘Public safety and welfare’ refers to police activity other than that related to crimes, such as that related to road traffic or anti-social behaviour incidents.

For ‘crime and criminal investigation’, the Review developed a proposed method using police recorded crime and outcomes data, weighting the activity associated with these by the police resource used in investigating different crimes and achieving different outcomes. For ‘public safety and welfare’, the Review investigated a range of potential metrics. ‘Crime prevention’ is particularly challenging, for the same reasons faced by multiple services and discussed in depth in Chapter 3.

A direct volume output measure is required to construct a measure for police productivity. This entails:

• Defining the activities to be included as output (across the three different categories of crime, public safety and welfare and crime prevention).
• Collecting data on these activities completed.
• Developing weights for each category of activity, and if necessary within each category.

In February 2023 the Home Office undertook a Police Activity Survey (PAS), a currently unpublished time-use study covering England and Wales. In collaboration with Home Office, the Review assessed the utility of this dataset in developing output metrics for policing activity and found this to be favourable. If such a data source were to be available on an ongoing basis this would facilitate the development of policing output estimates. Moreover, there could be potential in extending the survey to Scotland and Northern Ireland in the context of developing UK wide estimates.

The PAS collects police activities by crime and criminal investigation, public safety and welfare and ‘other’ activities and can therefore be used for the cost weighting of much of police outputs. ‘Other’ activities are activities that are not related to a specific crime or public safety and welfare incident. However, as these activities relate to those which support the delivery of crime and public safety and welfare services (and would not normally be considered as service output), only crime and public safety and welfare were considered for police output.

Using the PAS, the Home Office has estimated police ‘resource unit costs’ by considering time spent on activities to be representative of the population, and weighting by the rank pay of the officer carrying them out. The time weighting is calculated by comparing the survey activity percentages with the annual total using Police Recorded Crime (PRC) for the crime incident groups available. Note that these weights therefore only consider the relative labour costs of different activities, although usage of capital and intermediate consumption inputs can be assumed to be correlated with labour inputs.

Whilst the PAS provided a ‘snapshot’ in time, the Review understands that Home Office intend to repeat the survey in financial year ending 2026 subject to available funding. Initial exploration has begun of ONS adding the survey to its existing survey portfolio.

Crime and criminal investigation

Police Recorded Crime and Outcomes data provides a time series of police criminal investigation data from all England & Wales police forces on the number of recorded offences committed and reported. These are produced quarterly by the Home Office and are disaggregated by force, type of offence, and the outcome for each case. Full outcomes data has been published since financial year ending 2016. As each crime recorded in the data is associated with police activity, this can serve as a time series for police crime and criminal investigation activity.

The PRC classifies crimes into 10 categories, which can be broken down into a further 23 sub-categories. PAS uses the same categories for reporting the breakdown of staff resource costs for crime activity. The Review, therefore, used these two data sources to produce a ‘proof of concept’ cost weighted activity index for crime and criminal investigation activity, albeit one where the costs do not include non-staff inputs and constant weights are applied, where these should change through time. This would require further waves of the PAS.

Whilst PAS may not cover crimes which are not reported to the police, retaining a focus on the crimes police are made aware of feels sensible in terms of measuring police activity. One potentially important factor affecting how much resource is deployed is whether the crimes included in PRC are uncovered through proactive policing or reported by victims or witnesses. To date, no data has been uncovered that would facilitate this, but this omission means changes in the extent to which crimes are uncovered through proactive policing is not captured in the output measure.

As Action Fraud is not included in the PAS, the proof-of-concept work produced excluded the fraud crime category meaning that only nine categories were used: Criminal damage and arson; Theft offences; Violence against the person; Miscellaneous crimes against society; Sexual offences; Drug offences; Possession of weapons offences; Public order offences; and Robbery.

Police crime output can be further refined by disaggregating the weights allocated to each crime category by the outcome achieved, using the 12 outcomes categories captured in PRC. There is clearly an overlap here with the concept of quality adjustment: identifying an offender and bringing them to justice is clearly preferable to closing a case without identifying a suspect and would normally be considered within a quality adjustment factor. However, in this case, the input variation is also suggestive that these outcomes represent a different quantity of output. The Review has therefore considered how to reflect differences in the police resource used to achieve an outcome in the quantity of output but further development reflecting differences in the social value of different outcomes should be investigated for the quality adjustment.

Data on average time taken to reach a particular outcome by crime category are also available. By considering what outcome is associated with each crime, as well as time taken to reach that outcome, it could be possible to get a more accurate idea of how much activity went into a particular crime event.

The Review developed a method for sub-dividing the weights for each crime category to take into account both outcomes and time taken to reach the outcome. The outcomes can be split into high, medium and low-value outcomes, each given their own weight according to the number of days taken to reach the outcome, with high-value outcomes taking more time to achieve than low-value outcomes and thus reflecting a greater quantity of police activity. The identification of which outcome falls into each group has been completed following expert consultation.

The outcomes are weighted in three levels as shown in Table 3:

Table 3: Crime outcome by proposed weighting level used to subdivide weights for each crime category

Public safety and welfare activity

In contrast to crime and criminal investigation activity there is no single source for all public safety and welfare activities, and the individual sources capturing these activities are not always comprehensive.

The PAS splits public safety and welfare activities into 37 categories. However, as many of these take up very little police time and some do not lend themselves to measurable activity (“police generated resource activity”, “unknown”, “messages”, “other public safety and welfare” and “other incident”), our investigation has focussed on a refined list of seven categories, comprised of 12 of the 37 public safety and welfare categories in PAS:

• Missing persons.
• Concern for safety.
• Anti-Social Behaviour (ASB), comprising three categories: ‘Nuisance’, ‘Personal’ and ‘Environmental’.
• Domestic incident.
• Transport comprising four categories: ‘Road related offence’, ’Road traffic collision – death or injury’, ’Road traffic collision – damage only’, and ’Other transport’).
• Sudden death.
• Pre-planned events.

Cost weighting public safety and welfare activity is more challenging than for crime and criminal investigation. While the PAS provides resource unit costs for all crime activities, there are only limited resource unit costs available for public safety and welfare activities. These are: anti-social behaviour, transportation, administration, and public safety and welfare. Of these, only anti-social behaviour and transport can easily be mapped to activity, with the other two categories being too generic, but further work is encouraged here, particularly if future waves of the PAS can be refined to better map onto the desired categories.

One source that contains data on multiple categories is the Police Efficiency, Effectiveness and Legitimacy (PEEL) data collected by HM Inspectorate of Constabulary and Fire & Rescue Services (HMICFRS). HMICFRS use this to assess the performance of all 43 police forces in England and Wales. Data collected includes information on missing persons, anti-social behaviour and domestic incidents, and therefore provides potential activity data for three of the seven public safety and welfare activities. However, the annual statistical report from the National Crime Agency Missing Person Unit contains more comprehensive data on missing persons and could be used in place of or in conjunction with the PEEL data for that activity.

The broad category of transport comprises of four sub-categories: Road related offence, road traffic collisions – death or injury, road traffic collision – damage only and other transport. Data taken from the Department for Transport on collisions provides potential data for road traffic collisions – death or injury. Useable data have not currently been identified for any of the other categories within transport.

While activity data could potentially exist for the remaining categories, this has yet to be identified. This means that concern for safety, sudden death and pre-planned events could not currently be included in police public safety and welfare output.

Crime prevention activity

As addressed in Chapter 2 and Chapter 3, accurately measuring prevention has its challenges. This is particularly the case with crime where it is not always clear what crimes would be prevented from particular activities. For example, having visible patrols may prevent certain crimes but it would be difficult to allocate this prevention measure to an existing group of activities for which the ONS has measures. Similarly, estimation requires evaluation evidence on the effectiveness of prevention activities and the probabilities of prevention which would beneficially be of equivalent academic quality as those applied to other services, such as Healthcare.

In addition, to ensure different aspects of crime prevention are addressed broadly consistently, data on different interventions are required. To support this, the Review recommends a literature review to identify data.

Weighting outputs to derive an aggregate output index

The core issue identified by the Review is the lack of a time-series for weighting both within crime and criminal investigation or public safety and welfare and, importantly, between crime and public safety and welfare over time. In order to combine these two activity types into a single police output, an appropriate time-varying weight would need to be sourced.

The Review has explored the Police Objective Analysis data from the Chartered Institute of Public Finance and Accountancy (CIPFA). These split funding by different categories which it may be possible to align to the crime and public safety and welfare categories currently being explored. More importantly, this is a time-series from financial year ending 2015 and could therefore allow for a time-varying cost weight. Alternatively, future runs of the PAS would be beneficial to enabling a Laspeyres index to be constructed.

10.4 Development of quality adjustment estimates

The Review has considered potential quality adjustments (i.e. outcomes) of crime investigation outputs (explored for England and Wales in the first instance) to better reflect the value citizens receive from these functions, and how this is translated through to citizens in terms of the quality of the activity undertaken. The following were considered:

• Police legitimacy.
• Victim (of crime) satisfaction.
• Casefile quality.

Police legitimacy

Police legitimacy or ‘policing by consent’ is key to the nine principles developed by the founder of the Metropolitan Police, Robert Peel. This consent is dependent upon carrying out the ‘law fairly, impartially and by using minimal force’ (Policing in the UK, House of Commons Library). It is important to reflect this as it determines to what extent crimes and criminality are reported to the police: a police force which is not made aware of crimes cannot be fulfilling the function it is designed to deliver to citizens, both at the national and local levels.

The views of the general population are likely a useful gauge of public sentiment around police legitimacy (for example, as collected in the Crime Survey of England & Wales (CSEW) or other surveys such as YouGov Monthly Tracker, Ipsos Veracity Index, HMICFRS Public Perceptions of Policing Surveys, IPSOS Global Trustworthiness Index, IOPC Public Perceptions Tracker, or the ONS’ Opinions and Lifestyle Survey). Although there is an argument it may prove a more informed quality adjustment to use responses from members of the public who have had direct contact with the police through being victims of crime, this misses the universal aspect of the police offering which should ensure all members of society feel safe and protected.

The CSEW is a large sample survey (approx. 31,000 in 2023) that asks household residents about their experience of a range of crimes in the 12 months prior to being surveyed. The survey provides an estimate of the level of crime committed, covering both reported and unreported crimes. It is considered the most reliable indicator for long term trends, as it is unaffected by changes in police recording, crime reporting rates or police activity and the Review considers it the best data source for this purpose. While not exhaustive, potential indicators are shown in Table 4.

Table 4: Potential Police legitimacy indicators from the CSEW

Victim satisfaction

One of the most straightforward ways of adjusting crime investigation outputs would be to utilise data on victims of crime and their experience of the police at different stages of the police investigation. Whilst the CSEW is the most likely data source, the Review also investigated the potential of other victim related surveys (e.g. Victims’ Commissioner’s Victim Survey Reviews and reports – Victims Commissioner, Home Office Annual Data Requirement on forces to interview victims of a range of crimes).

Further research is required to demonstrate feasibility of CSEW data, including around different crime types or different aspects of the victim experience, as shown in Table 5, specifically in relation to whether samples yielded enough victims to allow valid comparisons through time.

Table 5: Victim satisfaction indicators from the CSEW

Case file quality

Using administrative data recorded in the criminal justice system, the timeliness and quality of investigation as evidenced through the case file, could act as a quality adjustment, through for example, the share of case files that are returned to police by the Crown Prosecution Service (CPS) due to problems with case notes. This can result in delays in prosecution. The Police Productivity Review (2023) noted that 47% of case files were not accepted at first triage by the CPS and recommended that case file timeliness and quality be added to its proposed Model Process tool.

Wider quality adjustments

In relation to ‘public safety and welfare’ and ‘crime prevention’ how one may quality adjust depends on the final output measures which are agreed on, but this is clearly an area for future work.