Transparency data

Police National Computer proxy for knife-enabled offenders

Published 7 April 2026

Applies to England and Wales

1. Introduction

The Home Office collects police recorded crime (PRC) data on selected violent and sexual offences involving a knife or sharp instrument (knife-enabled crime), and these are published quarterly by the Office for National Statistics (ONS), with the underlying data published by the Home Office. These include offences where a knife or sharp instrument has been used to injure a victim, attempt to injure, or used as a threat. This includes offences where the weapon may not have been seen, but was believed to be present, at the time of the offence by the victim or another witness (ONS Crime in England and Wales).

The National Data Quality Improvement Service (NDQIS) uses a computer-assisted classification tool to determine whether an offence involved a knife or sharp instrument (methodology published by the ONS). There are 7 offences that can be ‘flagged’ as a knife-enabled offence (assault with injury and assault with intent to cause serious harm, robbery, threats to kill, homicide, attempted murder, rape, sexual assault), with the majority consisting of knife-enabled assault and knife-enabled robbery.

The police recorded crime data is ‘victim-oriented’ and therefore contains limited data on suspects. The Police National Computer (PNC) database, which is a large administrative database containing information about police cautions and court convictions held by individual offenders in England and Wales, contains much more comprehensive information on offenders. This dataset captures cautions and convictions for knife possession offences (as reported in Ministry of Justice’s Knife and Offensive Weapons Sentencing statistics - these capture offenders cautioned and convicted for specific possession and threat offences, most of which are possession.

These are not the same as knife-enabled offences, and there is no crossover between these offences and PRC knife-enabled offences). Unlike police recorded crime data, there is no knife-enabled flag for the crime types in the PNC, so accurate offender data for knife-enabled crime is not available.

2. Description of measure

This analysis explores an experimental method for estimating information about offenders cautioned and convicted for offences that could be knife-enabled using the PNC dataset. This estimate is based on analysis of PNC data, and selects offenders who have been cautioned or convicted for:

• an offence of possession of a knife, bladed article, or offensive weapon (henceforth collectively referred to in this report as knife possession offences)
• any of the 7 offences that appear in the published police recorded crime knife-enabled offences data (robbery, assault with injury and assault with intent to cause serious harm, threats to kill, homicide, attempted murder, rape, sexual assault)
• that these offence types occurred on the same date

The possession offences are used in the absence of a knife-enabled flag on PNC and instead create a ‘proxy’ flag used to identify possible knife-enabled offences. Where a knife possession offence has been recorded for the same date as one of the 7 PRC knife crime offence types, it is inferred that this could have been a knife-enabled version of that offence.

This proxy for knife-enabled offenders will differ from the knife-enabled police recorded crime statistics published by ONS as only a minority of offences will result in an offender being brought to justice, and so volumes will be much smaller^{[footnote 1]}. Both this proxy knife-enabled offender measure and the Ministry of Justice’s knife and offensive weapon sentencing statistics are derived from the PNC. However, only a subset of those with knife and offensive weapons offences will have a concurrent violent or sexual offence and be included in this measure; therefore the proxy knife-enabled offender cohort will be a subset of this group and so volumes will be smaller.

3. Key findings

The data shows:

• in 2023 (the last complete year for which data is currently available), 9,106 offenders were convicted of, or cautioned for, a knife possession offence and one of the 7 knife crime offences committed on the same day (henceforth referred to as proxy knife-enabled offenders)

Across the 6-year period analysed:

• 50% of proxy knife-enabled offenders were cautioned or convicted for knife-enabled assault, and 40% for knife-enabled robbery, together making up 90% of the cohort
• most proxy knife-enabled offenders (95%) were male; this was similar for knife-enabled robbery (97%) and knife-enabled assault (94%)
• a quarter (25%) of proxy knife-enabled offenders were juveniles (aged 17 and under); this was a higher proportion for knife-enabled robbery (31%) than for knife-enabled assault (22%)
• about two-thirds of proxy knife-enabled offenders were White (68%), but Black offenders were disproportionately over-represented (21%) compared to the general population in the Census; this was similar for knife-enabled robbery and knife-enabled assault

4. Data notes and limitations

The data used for this analysis comes from a static subset of PNC data, known as the National Policing – Home Office Research and Development (NPHORD) dataset, created in August 2024. It contains records of cautions and convictions in the UK, but this analysis only includes offences recorded in England and Wales. The PNC dataset is an administrative dataset and was not collected for statistical analysis purposes. Administrative datasets can involve revisions to the data over time and errors in data processing. The dataset has been processed and quality assured for analytical purposes to mitigate these risks as far as possible.

The use of possession offences as a proxy for the knife flag has limitations. It is not known how well the proxy aligns with actual knife-enabled crime. It is possible to result in false positives – cases where the offender had a knife on their person and this was found during arrest but was not used in the commission of the offence. These would still be included as a ‘proxy’ knife-enabled offender within this analysis. These would overestimate instances of knife crime and incorporate non-knife crime offenders. It is also possible to result in false negatives – cases where the offender has not been cautioned or convicted of a possession offence, but one was used in the commission of their offence. These would not be included as a ‘proxy’ knife-enabled offender within this analysis, and therefore may underestimate instances of knife crime and miss knife-enabled crime offenders.

The possession offences used in the analysis include both offences for possession of an article with a blade or point, and possession of offensive weapons (the latter of which can include knives such as zombie knives and ninja swords but will also capture other offensive weapons). Therefore, these offences may include offences that would not be knife-enabled under police recorded crime.

Additionally, as recording of knife possession offences is influenced by police activity such as Stop and Search, which is known to be used disproportionately for certain demographic groups (Home Office Police powers and procedures statistics), using knife possession offences as a proxy flag for knife-enabled crime may also bias the demographics in this analytical approach; for example, higher proportions of young people, males, and people from ethnic minority groups.

This means that, whilst able to provide estimates for knife-enabled offender data, these findings should be treated with caution as they are not officially recorded knife-enabled offender cases.

Due to the way in which the proxy knife-enabled offenders are identified, it is possible for an offender to have more than one of the 7 offence types that can be flagged as a police recorded knife crime on the same date. These offenders will only be counted once in Table 1, but in order to capture all possible types of knife-enabled offence, they may be captured in more than one category in Table 2. This means total volumes in Table 1 and Table 2 may not match.

There is missing demographic data in the dataset. Whilst all offenders will have a date of birth (and therefore a recorded age), other data such as ethnicity and sex may be incomplete. For ethnicity, the tables include where the information is unknown. For sex, unidentified data has been excluded from the table, and therefore volumes may not match those in other tables.

5. Tables

Table 1: Offender data by year for all possession offenders and proxy knife-enabled offenders

Year	All possession offenders	Proxy knife-enabled offenders
2018	20,016	1,832
2019	19,967	1,774
2020	18,312	1,467
2021	16,120	1,354
2022	16,734	1,434
2023	15,323	1,245
Total	106,472	9,106

Table 2: Proxy knife-enabled offender data by type and by year

Year	Proxy knife-enabled assault offenders	Proxy knife-enabled robbery offenders	Proxy knife-enabled other 5 offence types (low volume offences) offenders
2018	898	861	150
2019	828	844	170
2020	779	613	137
2021	735	517	158
2022	785	519	183
2023	666	466	156
Total	4,691	3,820	954

Table 3: All proxy knife-enabled offenders - by sex by year (where sex identified)

Year	Male	Female
2018	1,739	87
2019	1,679	92
2020	1,393	71
2021	1,273	74
2022	1,353	73
2023	1,175	67
Total	8,612	464

Table 4: All proxy knife-enabled offenders - by ethnicity by year (including unknown ethnicity)

Year	White	Black	Asian	Unknown	Other
2018	1,256	369	139	51	17
2019	1,224	365	118	44	23
2020	992	311	113	35	16
2021	920	287	84	45	18
2022	981	275	126	31	21
2023	834	274	93	28	16
Total	6,207	1,881	673	234	111

Table 5: All proxy knife-enabled offenders - by age by year

Year	10 to 17	18 to 24	25 to 29	30 to 39	40 to 49	50+
2018	472	496	253	363	167	81
2019	441	488	245	374	161	65
2020	357	394	212	292	130	82
2021	346	341	185	266	143	73
2022	332	347	190	322	145	98
2023	329	273	142	274	156	71
Total	2,277	2,339	1,227	1,891	902	470

Table 6: Proxy knife-enabled robbery offenders - by sex by year (where sex identified)

Year	Male	Female
2018	831	27
2019	810	32
2020	587	24
2021	503	14
2022	507	11
2023	447	17
Total	3,685	125

Table 7: Proxy knife-enabled robbery offenders - by ethnicity by year (including unknown ethnicity)

Year	White	Black	Asian	Unknown	Other
2018	575	192	65	25	4
2019	592	166	57	22	7
2020	419	139	32	16	7
2021	357	117	30	9	4
2022	366	108	30	9	6
2023	314	107	27	9	9
Total	2,623	829	241	90	37

Table 8: Proxy knife-enabled robbery offenders - by age by year

Year	10 to 17	18 to 24	25 to 29	30 to 39	40 to 49	50+
2018	277	200	113	182	68	21
2019	256	221	105	185	67	10
2020	199	153	84	101	59	17
2021	151	128	69	101	55	13
2022	147	134	71	105	47	15
2023	149	93	50	106	54	14
Total	1,179	929	492	780	350	90

Table 9: Proxy knife-enabled assault offenders - by sex by year (where sex identified)

Year	Male	Female
2018	843	53
2019	770	55
2020	735	42
2021	679	49
2022	732	48
2023	624	41
Total	4,383	288

Table 10: Proxy knife-enabled assault offenders - by ethnicity by year (including unknown ethnicity)

Year	White	Black	Asian	Unknown	Other
2018	622	172	69	23	12
2019	570	165	59	19	15
2020	526	154	72	18	9
2021	500	145	44	32	14
2022	539	138	77	17	14
2023	455	136	54	15	6
Total	3,212	910	375	124	70

Table 11: Proxy knife-enabled assault offenders - by age by year

Year	10 to 17	18 to 24	25 to 29	30 to 39	40 to 49	50 to 59
2018	194	268	130	170	88	48
2019	178	235	127	159	86	43
2020	159	209	124	176	58	53
2021	177	187	105	140	77	49
2022	164	185	103	185	82	66
2023	157	149	80	146	94	40
Total	1,029	1,233	669	976	485	299

1. The Home Office also publish other ad hoc statistical releases on knife-enabled robbery for a subset of forces. This uses Management Information from the police recorded crime dataset, and is unrelated to this ad hoc statistical release. ↩