Data quality

1.23 The quality of the uplift data is high. Most users we spoke to told us that data quality meets their needs, and that, generally, they have no concerns about the accuracy and reliability of the statistics.

1.24 To measure progress in recruiting additional officers, Home Office had to establish a baseline. The methodology is complex: Home Office took the headcount figure from the police workforce statistics at 31 March 2019 as a starting figure and adjusted this to account for people in post at the start of the recruitment drive and other recruitment planned prior to the uplift announcement. Further, smaller in-year adjustments are made to account for organisational restructuring (including posts transferring out of a territorial force to the National Crime Agency). Each force has its own baseline, and only when forces exceed the baseline level will newly recruited officers count towards their uplift allocations. This means that the actual number of new officers that forces are recruiting through the programme is much higher than 20,000, because they must backfill officers that have left, for example, because they resigned or retired. The National Audit Office (NAO) has estimated that forces will have to recruit approximately 50,000 officers across the three years of the programme.

1.25 We are impressed by the robustness of the methodology and the clarity around the baseline: it is well-explained and Home Office has been transparent with users about its development. The close engagement with police forces means there is a shared understanding of the baseline and which officers count towards the uplift target, which is vital for the successful delivery of the programme. The clear baseline also ensures that progress towards the target is reported in a consistent way, which supports trustworthiness of the statistics.

1.26 We commend Home Office, NPCC, College of Policing and police forces on their collaborative, joined up approach to collecting data and producing statistics on the uplift programme. The data supply chain is unique:

• Police forces send their data to an NPCC regional lead, who collates the data across the forces for which they are responsible.
• The regional leads send the data to the central NPCC uplift team, which collates and processes the data from all forces. The NPCC team produces a monthly management information report and dashboard on the performance of the programme.
• The NPCC team shares the raw data with the Home Office uplift statistics team, which processes the data for release as official statistics. Home Office publishes the statistics within seven days of receiving the data from NPCC.

1.27 This setup ensures that the data are fit for purpose and users can have confidence in data quality. For instance, the data are comprehensively quality assured (see 1.31) and the constructive working relationships at all levels allow data issues to be identified and resolved quickly. The regional lead is a new role, created specifically for the uplift programme. They act as a two-way communication channel between the NPCC uplift team and police forces, managing the needs of forces and providing guidance on data collection. The local context and knowledge they add helps the central NPCC team understand the performance of all forces.

1.28 As part of the initiative to improve data on the police workforce, Home Office, NPCC and the College of Policing have been collaborating in developing ‘National Standards for Workforce Data’. These data standards draw on existing harmonised standards set out by the Government Statistical Service (GSS) and aim to bring more standardisation within policing for the collection of data on protected characteristics. Prior to the uplift programme, many forces did not collect data on the ethnicity or other protected characteristics of their officers in a standardised way. The new standards ensure that protected characteristics data are collected in a consistent way across all forces, enhancing the coherence of the data and allowing for comparisons across forces. Many users were extremely positive about the data standardisation work, which will have long-term benefits for data quality and policy by allowing for better-evidenced decision making.

1.29 The completeness and reliability of the protected characteristics data varies by police force and characteristic; the information is self-reported by officers on police HR systems. It is complete or mostly complete for age, sex and ethnicity, but less complete and reliable for other characteristics. Due to the incomplete nature of the data on sexual orientation and disability status, Home Office publishes these breakdowns as experimental statistics. While it is good that this information is published, the experimental statistics label is not appropriate because the statistics are not going through development; they are processed in the same way as the ethnicity, sex and age data. Under the Code of Practice, the experimental statistics status label should only be used for newly developed or innovative official statistics undergoing evaluation. Instead, we encourage Home Office to be clear about the lower quality of these data and the factors users need to consider when interpreting these statistics. Home Office has agreed to remove the experimental statistics label for future releases of the statistics.

1.30 Home Office and NPCC are continuing efforts to increase the completeness of these data. For instance, NPCC has written to individual forces to try to improve their confidence to provide the information or to investigate system-based issues. In addition, police forces, in collaboration with the College of Policing, ran ‘Safe to Say’ campaigns to encourage officers to declare their protected characteristics information. The bulletin and police workforce statistics user guide contain a brief summary of the completeness of protected characteristics data and the limitations of the data, but this should be expanded, for example, by explaining the completeness of data for all protected characteristics and assigning a quality rating for each characteristic (instead of an overall rating). Home Office told us it will explain the completeness of data for all protected characteristics by adding the percentage of unknowns for each characteristic In future releases of the statistics.

1.31 The workforce statistics user guide outlines the key characteristics and limitations of the workforce numbers as well as the implications of the limitations. Users told us that the limitations of the uplift data are well-presented, with changes and revisions clearly explained and caveated. While the bulletin and user guide are transparent about the nature and limitations of the uplift data, they contain no information about the nature and limitations of the census population estimates, which are used to compare the proportion of officers of certain ethnicity with those in the general population in England and Wales. It is important to explain the limitations of the population data because the current estimates are based on the 2011 Census and will soon be replaced with those from the 2021 Census. Also, the updated estimates may affect the narrative around the representativeness of new officers and police forces, and this should be communicated to users.

1.32 The quality assurance (QA) arrangements are rigorous and well-established. Checking and validation is carried out at every step of the process. For example:

• The senior responsible officer (at the chief officer level) in police forces checks the figures before they are sent to the NPCC regional lead.
• The NPCC regional lead compares figures with those from the previous month and identifies any errors or emerging issues.
• The NPCC central team queries unusual figures with police forces and keeps a log of all QA checks raised with forces.
• Home Office carries out completeness and consistency checks.

However, the summaries of the QA process in the bulletin and user guide do not capture all aspects of the process. Also, they do not explain the roles and responsibilities of the different organisations involved in collecting and processing data.

1.32 The statistics team has automated the production of the data tables through implementing a reproducible analytical pipeline (RAP). This reduces the risk of errors and contributes to robust quality management. The team is currently exploring applying RAP principles to other aspects of the statistics, including the charts in the statistical bulletin. We support the team’s ongoing work with RAP to enhance the quality of the uplift statistics.