Measure and geographical coverage
The Home Office records arrests for notifiable offences under its counting rules. A person may appear again after a separate arrest. Where several notifiable offences are involved on one occasion, the principal offence rule applies. An arrest does not establish guilt. These observations are not unique people, crimes committed, charges or convictions.
The historical series covers England and Wales from April 2020 to March 2023, with 43 police force areas and two national aggregate views. Scotland and Northern Ireland are outside its coverage. The England and Wales count includes City of London arrests; City of London has no published local population rate.
Verified source and extraction
The original Ethnicity Facts and Figures CSV contains 109,350 data rows. All fields retained in the supplied workbook match the official rows, but the workbook omits the legacy gender field. In the original CSV, 2020/21 uses gender for sex categories; 2021/22 and 2022/23 use sex. The Cartulary restores this dimension from the original official file. The workbook’s added formulas and cached dashboard are not used.
Select one financial year, sex, age category, geography and ethnicity classification. The five broad groups and nineteen detailed groups are alternative views; they are never added together. Unknown ethnicity occurs in both source classifications with identical values and is stored once. The all-ethnicity headline is an original source observation, not a sum of overlapping rows.
Checks
All 109,350 dimensional keys are unique after restoring the year-specific sex field. There are 4,050 selectable year/sex/age/geography contexts. In every context the five broad groups plus unknown ethnicity reconcile to the source total. National totals of 643,832 in 2020/21, 654,597 in 2021/22 and 668,979 in 2022/23 reconcile to all 43 forces. All available sex and age splits also reconcile when Other/Unknown categories are included.
All 9,636 numeric source rate cells match arrests divided by the supplied population and multiplied by 1,000 within one-decimal rounding tolerance. The 2,700 unavailable arrest cells are Roma observations in the first two years. They remain unavailable, not zero.
Population rates and interpretation
Only source-published rates are displayed. They use 2021 Census resident populations for the matching ethnicity, sex and police force area; the same denominator is used for all three years. Rates are available only for all ages and All, Female or Male sex selections. No rate is calculated for unknown ethnicity, individual age categories, Other/Unknown sex or City of London. A resident population field does not by itself authorise a rate.
Rates are not age-standardised, do not track population change after 2021 and may include people arrested outside their area of residence. Ethnicity is self-defined; missing ethnicity is retained. Recording practices, demography and use of police powers affect comparisons. These figures do not establish a causal effect of ethnicity or an individual likelihood of offending.
For small resident groups, a few additional arrests can change a rate sharply. Repeat arrests and arrests of non-residents mean the rate does not measure the share of that population arrested.
Dates, revisions and continuing publications
The historical page was published 3 July 2024 and retrieved 9 September 2026. The source incorporates the March 2024 North Wales correction to primary-reason arrest counting. Other versions may revise historical years and need separate reconciliation.
March 2023 is the end of this selected extract, not the end of Home Office collection. The Home Office publication of 6 November 2025 includes arrests and ethnicity through March 2025. Later years have not been silently merged into this historical display. ONS supplies the census population basis; the Home Office publishes the arrests.
Original file
Download the original official CSV. Source SHA-256:
61db99c554a75f8f8c030b74b9988f53261186ee2378283b85501d7e1c49e61b
Contains public sector information licensed under the Open Government Licence v3.0.
