MAINTAINING BASIC COMMAND UNIT AND CRIME AND DISORDER PARTNERSHIP FAMILIES FOR COMPARATIVE PURPOSES Gerard Sheldon, Rebecca Hall, Chris Brunsdon, Martin Charlton, Seraphim Alvanides and Nikos Mostratos July 2002 The views expressed in this note are those of the authors, not necessarily those of the Home Office (nor do they reflect Government policy). Background In January 2000 the Home Office first published local-level crime statistics for the 318 Basic Command Units (BCUs) covering England and Wales. In July 2000 work was carried out to list the 376 Crime and Disorder Reduction Partnerships (CDRP) areas in England and Wales in groupings, or families (Leigh et al, 2000). These families were based on socioeconomic and demographic factors that were shown to correlate geographically with the level of crime and disorder within a locality. These families were intended to be used: To provide a basis for the national publication of crime statistics at a local level ; To help forces and police authorities undertake best value reviews by enabling them to compare local-level performance across a range of functions or processes; To help CDRPs identify which Partnerships in their family have the lowest crime rates and, over time, are most successful at reducing crime, so that they can learn lessons from them; and To assist HMIC inspections at BCU level. In July 2000, the Home Office published crime statistics organised by CDRP family (Povey et al 2000). A need for a similar grouping was identified for BCUs, so that publications of crime statistics at this level could also be organised by family. However, BCU boundaries are chosen according to operational policing needs, and do not necessarily coincide with the CDRP boundaries. Thus, in July 2001, Home Office Briefing Note 4/01, Family Ties: Developing Basic Command Unit Families For Comparative Purposes (Harper et al. 2001) was published, listing the families of the 318 Basic Command Units. A more detailed report is available on the Home Office RDS website www.homeoffice.gov.uk/rds (Harper et al 2002). In this report, details of the work on CDRP families are also included. Since the initial publication of BCU-level crime data organised by family, a number of BCU boundaries have been changed, and some areas have also seen changes in their socio-economic conditions. Some BCU areas were entirely re-organised, usually resulting in a reduction in the number of BCUs, with several smaller areas being rearranged into a smaller number of larger BCUs. Whilst CDRP boundaries remain unchanged, there have some socio-economic changes. Hence the need to update the family membership. However, the families are constructed mostly from Census data. Currently only the 1991 Census data is available, and the 2001 Census data will become available at a later date (later in 2002 and in 2003). Therefore, it was decided not to carry out a full update of the families at this stage, but simply to maintain them as an interim step. The overall objective of this work, then, is to update boundaries, socio-economic data and family membership of the BCUs in the light of changed boundaries and to evaluate and adjust CDRP family membership in the light of any major changes to their circumstances. How the families were maintained The original families (briefing notes 3/00 and 4/01) were created using Cluster
Analysis (Hartigan et al 1979 and Kohonen 1989) a statistical method which groups entities into a number of different families on the basis of a number of variables. Individual entities (here BCUs and CDRPs) are assigned to a family into which they best fit in terms of these variables. The assignment process may require several iterations, since when a family gains a new member, the overall characteristics of that family change slightly. The aim of the analysis is to find a stable set of families, with distinct characteristics, allowing meaningful comparisons to take place between family members. (i) Variables used The variables used in the updating process were the same as those used in the original BCU and CDRP analyses. These are listed in Box 1. Box 1. Variables used for the cluster analyses 1. % young males (16-24) 2. % minority ethnic population 3. % single adult households 4. % single parent households 5. % student households 6. % residents living at a different address one year before the 1991 Census 7. % local authority housing, plus new town renters and housing associations 8. % terraced housing 9. % overcrowding 10. Population sparsity 11. Population density 12. Length of A, B and minor roads per head of population 13. % daytime population (leisure and retail employment) 14. % young male claimants (of unemployment-related benefits) 15. % long-term claimants (of unemployment-related benefits) 16. Index of homogeneity 17. Motorway junctions per 1000 population 18. Population per square kilometre 19. % claiming income support (for CDRPs only) 20. % claiming family credit (for CDRPs only) (ii) Maintaining the variables The values of the variables used remained the same as those used originally (Briefing Notes 3/00 and 4/01), unless the police reported changes to BCU boundaries and /or changes in socio-economic environment. Where boundaries changed, existing variables were re-aggregated to the new spatial units. In cases where police forces and authorities had provided evidence of socio-economic changes, appropriate adjustments were made to relevant variables prior to the cluster analysis. (iii) Running the cluster analysis The CDRP families were updated using Cluster analysis. A CDRP could only move families if there had been a significant change in the socio-economic environment. Cluster analysis was also applied to BCUs. However there were two main constraints: the family membership of the BCU remained unchanged (compared with its membership in Briefing Note 4/01), if there were no significant changes to its boundaries and socio-economic environment; and the family membership of a BCU was fixed to its corresponding coterminous CDRP only applicable if there exists a coterminous CDRP. Allowing for re-organisations, there are 280 BCUs in England and Wales to be considered (compared with the 318 analyses in Briefing Note 4/01). There were 30 newly created BCUs. Taking account of the constraints above, the cluster analysis could assign 21 newly created BCUs to families, and was permitted to re-assign 6 existing ones to new families. (iv) Verification of the clustering method Having produced new family memberships using the approach outlined above, the results were compared with those obtained using a related technique - the neural network based adaptive resonance theory (ART) method (Grossberg and Carpenter, 1998). The comparison
showed the results to be robust. ART is a commonly used alternative technique for this kind of work. Results (i) Family membership The following table shows the average, minimum and maximum family sizes for the BCUs and CDRPs. For BCUs, the corresponding statistics for the July 2001 analysis are shown in brackets for comparison purposes. Families BCU CDRP Average family size 20 (24) 29 Smallest family 6 (6) 3 Largest family 35 (42) 62 The family membership for each of the fourteen BCU families (including the three airports) is shown in Box 2. The letters in bold following each BCU name are explained as follows with the number of relevant BCUs in square brackets: N means that a BCU is a newly created one (and in some cases, the new BCU takes on the name of an old one, even if the boundaries are very different) [30]; R means that an existing BCU has been reassigned to a new family [2]. U means that an existing BCU was free to move to another family but nevertheless remained in the same one [4]. The BCUs listed in this note are those on which crime statistics are reported in the July 2002 Statistical Bulletin (07/01) (Simmons et al 2002). For the CDRPs, the only change is for Thanet, which is now in family 10. For the remaining CDRPs, family membership is the same as in Briefing Note 3/00. The full listing is therefore omitted in this study to preserve space.
BCU Family 1 (6) BCU Family 3 (17) BCU Family 4 (28) BCU Family 5 (29) City of London Bishopsgate Snowhill Camden Westminster Newcastle Central Millgarth BCU Family 2 (16) Central Brent Ealing Greenwich Hackney Haringey Islington Lambeth Lewisham Newham Southwark Tower Hamlets D1 - Erdington / Aston / Nechells / Saltley / Ward End E3 - Acocks Green / Sparkhill / Sparkbrook / Edgbaston / Balsall Heath / Selly Park / Moseley F1 - Birmingham City Centre / Digbeth F3 - Soho / Handsworth / Sandwell / Perry Barr /Aston (part) Central Bristol Bedfordshire Luton Barnet Croydon Enfield Hammersmith and Fulham Harrow Hillingdon Hounslow Kensington and Chelsea Kingston upon-thames Merton Redbridge Richmond-upon -Thames Waltham Forest Wandsworth Slough and District Cleveland Middlesbrough GMP North Manchester South Manchester Humberside Kingston upon Hull Liverpool North (N) Liverpool South (N) Gateshead East Newcastle East Newcastle West Sunderland City Sunderland West Wallsend Nottinghamshire Nottingham (N) South Yorkshire Sheffield Central (N) D3 - Stechford / Shard End / Bordesley Green / Bromford / Sheldon E1 - Bournville / Bartley Green / Longbridge / Selly Oak / Frankley E2 - Kings Heath / Billesley / Kings Norton F2 - Soho / Winson Green / Harbourne / Ladywood / Quinton G1 Wolverhampton Town Centre / Wolverhampton West / Whitmoreanes /Tettenhall / Penn G2 - Wednesfield / Bilston H2 - Willenhall / Bloxwich / Brownhills / Darlaston K1 - West Bromwich / Wednesbury / Tipton / Great Barr K2 - Smethwick / Oldbury / Old Hill / Langley / Warley M3 - Radford / Foleshill / Bell Green / Wyken Bradford North (N) Bradford South (N) Keighley (R) Killingbeck Northern Warrington Derbyshire Alfreton (A Division) Chesterfield (C Division) Derby (D Division) Durham Darlington Durham Thurrock GMP Bury Stockport Trafford Wigan Havant Swale Southern West Sefton (N) North Wales Eastern Northamptonshire Eastern Northern Nottinghamshire Mansfield/Sutton in Ashfield (N) South Nottinghamshire (N) South Wales Bridgend Vale of Glamorgan Redditch Telford Wakefield (N) Weetwood Wiltshire Swindon
BCU Family 6 (22) BCU Family 7 (17) BCU Family 8 (35) BCU Family 9 (32) Cleveland Hartlepool Langbaurgh Stockton Durham Derwentside Easington Sedgefield Wear-Tees Basildon Gwent Blaenau Gwent (U) Caerphilly St Helens Wirral (N) Barking and Dagenham South-East Northumberland South Tyneside Washington South Wales Merthyr Tydfil Neath & Port Talbot Rhondda Cynon Taff South Yorkshire Barnsley (N) Doncaster (N) Rotherham (N) Crewe Cumbria Barrow and Kendal Workington and Whitehaven Tendring Gwent Torfaen and Monmouthshire Isle of Wight Humberside East Riding of Yorkshire North Lincolnshire Lincolnshire West Lincolnshire Norfolk Norfolk Central / Southern Area North Wales Central Nottinghamshire Bassetlaw/Newark & Sherwood (N) Suffolk Suffolk Eastern Senlac (U) Southern Kidderminster Worcester Bath and North-East Somerset North Somerset Congleton and Vale Royal Macclesfield Dorset Bournemouth Poole Chelmsford Colchester Rayleigh Southend Gloucestershire Cheltenham and Tewkesbury Forest and Gloucester Fareham New Forest North East Hertfordshire (N) Canterbury West Northern South Bexley Bromley and Orpington Havering Sutton Staffordshire Chase North Staffordshire Surrey East Surrey (N) West Surrey (N) East Downs Highdown Chiltern Vale Thames Forest J2 - Halesowen / Stourbridge / Lye/Cradley / Kingswinford L1 - Solihull / Chelmsley Wood / Shirley Somerset East Somerset West Central Southern Derbyshire Buxton (B Division) Area 1 (N) Area 3 (N) Area 4 (N) Dorset Eastern Braintree Gloucestershire Cotswold and Stroud Andover Central Weald Lincolnshire East Lincolnshire South Lincolnshire Norfolk Norfolk Eastern Area Norfolk Area North Yorkshire North Yorks - Eastern North Yorks - Northamptonshire Suffolk Suffolk Weald Aylesbury Vale Northern Oxfordshire Southern Oxfordshire West Berkshire Shrewsbury Wiltshire Chippenham (C) Salisbury (A)
BCU Family 10 (26) North Bristol South Bristol Area 2 Gosport Portsmouth Southampton Central Thanet (R) Central Knowsley Northamptonshire Northampton Gateshead West Newcastle North Tynemouth South Wales Cardiff Swansea South Yorkshire Sheffield North Sheffield South Staffordshire Stoke on Trent Brighton Hove and Shoreham Milton Keynes Oxford D2 - Sutton Coldfield/ Castle Vale / Kingstanding H1 - Walsall North and South / Walsall Town Centre / Aldridge J1 - Brierley Hill / Dudley / Sedgeley / Gornal Chapeltown BCU Family 11 (22) South Gloucester Bedfordshire Bedford Dunstable Chester and Ellesmereport Harlow Basingstoke Eastleigh Hertfordshire Central (N) Eastern (N) Maidstone North South-East East North North Yorkshire North Yorks - Central Staffordshire Trent Valley Suffolk Suffolk Southern Surrey North Surrey (N) North West Surrey(N) Forest Reading and Wokingham Northern BCU Family 12 (9) Cumbria Carlisle and Penrith Dyfed-Powys Carmarthenshire (N) Ceredigion (N) Pembrokeshire (N) Powys (N) North Wales North Northumberland South-West Northumberland Hereford BCU Family 13 (18) Halton GMP Bolton Oldham Rochdale Salford Tameside Gwent Newport (U) Humberside North-East Lincolnshire Medway (U) Eastern Pennine M1 - Coundon / Hillfields / Tile Hill / Coventry City Centre M2 - Fletchamstead / Stivichall / Willenhall - Coventry / Stoke Calderdale (N) Dewsbury Holbeck Huddersfield Pudsey BCU Family 14 (3) Stansted Airport Heathrow Airport Gatwick Airport
References Carpenter, G. and Grossberg, S., (1988) The ART of Pattern Recognition by a Self- Organizing Network, Computer, 21:77-88 Harper, G., Williamson, I., See L.,Emmerson, K., Clarke, G. (2001) Family Ties: Grouping Basic Command Unit Families for Comparative Purposes, Home Office Briefing Note, 4/01 Harper G, Williamson I, Clarke G and See L (2002), Family Origins: Developing groups of Crime and Disorder Reduction Partnerships and Police Basic Command Units for comparative purposes, Home Office RDS Website (www.homeoffice.gov.uk/rds) Hartigan J.A. and Wong, M.A. (1979) A k- Means Clustering Algorithm, Applied Statistics, 28:100-108 Kohonen, T. (1989) Self Organisation and Associative Memory, Springer-Verlag, Berlin Leigh, A., Arnott, J., Clarke, G. and See L. (2000) Family Values: Grouping Similar Policing and Crime Reduction Areas For Comparative Purposes, Home Office Briefing Note, 3/00 Povey D, Cotton J, and Sisson S (2000), Recorded Crime Statistics, England and Wales, Home Office Statistical Bulletin, 12/00. Simmons J and Colleagues (2002), Crime in England and Wales 2001/02, Home Office Statistical Bulletin, 07/02.