Series details: CRDA/2

The British Crime Survey (BCS) is one of the most important methods used by the Home Office to gather information on crime in England and Wales. The survey - conducted annually from 2001 - asks a representative sample of the adult population about their experiences as victims of household crimes and personal crimes during the previous year. Its main purposes have been defined as being to "provide an alternative measure of crime to offences reported by the police; provide information on crime risks; provide a picture of the nature of crime; [and] take up other crime related issues".1

The data gathered in the BCS is intended to complement the Home Office's published statistics on notifiable offences, i.e. crimes reported to and recorded by the police which the police are required to notify to the Home Office. The BCS captures household and personal crimes which were not reported to the police, or not recorded by the police, or not in those categories of crimes which the police were required to notify to the Home Office. It therefore provides a truer estimate of the extent of household and personal crimes. BCS data is also less prone than the statistics on notifiable offences to changes in police practices in recording crimes, and to changes in the willingness of the public to report crimes, so it can provide a more accurate indication of trends in crime over time. The survey provides demographic profiles of the victims of crime (highlighting those groups which are most at risk of victimisation), and also provides information about the nature of crime, the circumstances in which crime takes place, and the impact of crime. Respondents are also asked for their opinions on related topics such as sentencing, the police, and the efficacy of the criminal justice system. The results of the BCS have been published by the Home Office in a variety of statistical and research publications (see Publications produced by the originating department), and are used to measure trends in crime, to provide a point of comparison for statistics on notifiable offences, and to inform policy making on crime and punishment.2

It should be noted that the BCS does not cover all crimes, only those classed as household crimes (e.g. burglary, thefts of and from vehicles, vandalism and theft from the home), and personal crimes (e.g. assaults, robberies, thefts from the person and other personal thefts). The distinction reflects the fact that "for some crimes the household is a natural unit of analysis, whereas for others the individual is a better choice".3 The BCS does not cover crimes against organisations (e.g. fraud, shoplifting), "victimless" crimes (e.g. drug abuse) or crimes where the victim is no longer available for interview (e.g. murder, manslaughter). As the core sample in each sweep of the BCS covers adults aged 16 and over, the BCS does not record crimes against children (the exception is the 1992 survey, which included an additional sample of 12-15 year olds from households where an adult was also interviewed). The BCS sample also excludes adults in institutions (e.g. prisons).4 Further details on the limitations of the BCS are given in Constraints on the reliability of the data.

The first sweep of the BCS was conducted in 1982 by the Home Office and the Scottish Office. It covered 10,905 respondents in England and Wales and 5,031 respondents in Scotland. Subsequent sweeps of the survey were conducted in 1984, 1988, 1992, 1994, 1996, 1998, 2000, and annually from 2001. The 1988 survey was also extended by the Scottish Office to a sample of about 5,000 respondents in Scotland, though as in 1982 the results of the Scottish survey were reported separately. Scottish data is not included in the 1982 and 1988 BCS datasets. The 1984 sweep, and all sweeps of the BCS since 1992, have focused solely on England and Wales. Scotland is covered by a separate Scottish Crime Survey, which was conducted by the Scottish Office in 1993 and 1996, and by the Scottish Executive in 2000. In 1996 the Northern Ireland Office commissioned its own Northern Ireland Crime Survey, using the questionnaire developed for the 1996 BCS. A further Northern Ireland Crime Survey was conducted in 2001.5

The BCS was conducted biennially from 1992 until 2000. In 2001 the BCS became an annual survey in which data was gathered through continuous interviewing: for details of the significant changes which occurred at that stage, see How data was originally captured and validated.

Responsibility for the management of the BCS and the reporting of the survey's results rested (in 2002) with the Crime Surveys Section, which was part of the Crime and Criminal Justice Unit within the Research Development and Statistics Directorate (RDS) of the Home Office. The 1982-1996 surveys were carried out by a predecessor body, the Home Office's Research and Planning Unit, which became part of the Home Office's Research and Statistics Department (another of RDS's predecessors) in 1989-1990. For further information on the history of the Research and Planning Unit and its successors, see the Administrative History of the Home Office.

Although overall control of the BCS rests with the Home Office, certain tasks - fieldwork, piloting, sample design, question coding and some question development - have always been delegated to one or more outside contractors (in some surveys, two contractors have acted as a consortium). The contractors have included:

• 1982: Social and Community Planning Research (SCPR).
• 1984: NOP Market Research Limited.
• 1988: SCPR and NOP Market Research Limited.
• 1992: SCPR and the British Market Research Bureau (BMRB).
• 1994: Social Survey Division of the Office of Population Censuses and Surveys.
• 1996 and 1998: SCPR.
• 2000: National Centre for Social Research (SCPR was renamed the National Centre for Social Research in 1999) and the Social Survey Division of the Office for National Statistics.
• 2001- : BMRB Social Research (part of BMRB International Limited) has been awarded a contract to carry out the annual BCS survey from 2001 until 2004.6 Some fieldwork and development work for the 2001 survey was subcontracted to Ipsos-RSL (part of the Ipsos Group).

The questions asked in the BCS have evolved over time. Most questions relating to the experiences of crime victims have remained the same, but questions on other topics have varied with different sweeps of the BCS.7 The following is a summary of the types of information gathered in the different sweeps of the survey:

• Demographic information about respondents and their households, to allow analysis of the types of people who have and have not been victims of crime: e.g. size of household, numbers of children, sex and age of members of the household, health and employment experiences of the respondent, type of accommodation, household income.
• Information about crimes experienced by the respondent and their household. This includes (1) general information about crimes (e.g. numbers and types during the reference period, whether incidents were isolated or part of a series), gathered through "screening" questions intended to identify crime victims; and (2) details of specific incidents or related series of incidents gathered from those respondents who answered the screening questions positively: e.g. when and where the incident occurred, details of property stolen, the nature of any relationship between the victim and the offender, and whether the incident was racially motivated. These details are used to determine whether the incident or series counts as an offence, and to classify the offence.
• Information about fires which have occurred in the respondent's home (a feature of BCS sweeps in the periods 1988-1996 and 2000-2002).
• Information about the attitudes of the respondent on topics such as fear of crime, the police, the Criminal Justice System, and the punishment and sentencing of offenders, the respondents' perceptions of the prevalence and seriousness of crime, attitudes towards Neighbourhood Watch schemes, etc. The nature of these attitude-related questions has varied with the different sweeps of the survey.
• Information on other topics, dealt with on an ad hoc basis, which may change with different sweeps of the survey: e.g. experiences of drug use, domestic violence, contacts with the police, self-reported offending, stalking, sexual victimisation, mobile phone theft, and technology crime.8

More detailed information about the contents of datasets derived from the BCS is given in the catalogues of individual datasets: see Links to dataset catalogues. It should be noted that the BCS datasets do not include the names or contact details of respondents to the survey. To maintain the confidentiality of respondents, this information has never been transferred to the Home Office by the contractors carrying out the field work.

Until 2001 interviews for the BCS were conducted during the first six months of the year, with the bulk of the interviews in the first quarter. Those questions which asked respondents about whether they or their households had been victims of crime covered the previous calendar year and the present year up to the date of the interview. This meant that the data gathered on victimisation could relate to periods of up to 18 months prior to the interview. In the 2001 survey the reference period for questions on victimisation was changed to the 12 months prior to the interview, for that part of the 2001 sample which was gathered through the new methodology of continuous interviewing (see How data was originally captured and validated).9 Questions on topics other than victimisation might have no reference period (e.g. attitudinal questions) or different reference periods (e.g. in the 1996 survey respondents were asked about their experiences of household fires since January 1994).10 Further information about the reference periods of questions in different sweeps of the BCS is given in the catalogues of individual datasets: see Links to dataset catalogues.

Datasets for the 1982-2001/02 surveys have been transferred to NDAD from the Home Office. A programme of work was agreed with The National Archives, whereby datasets were processed in stages over a period of years. See Links to dataset catalogues.

The BCS datasets and related dataset documentation are public records under the Public Records Acts 1958 and 1967. The Public Record Office has assigned these datasets and documents the class reference HO 400.

The BCS datasets are subject to access restrictions which are described in greater detail in the catalogues of individual datasets (see Links to dataset catalogues. Dataset documentation relating to the BCS is open to users of NDAD: see the Dataset Documentation Catalogue.

The Home Office recommends that users should contact Research Development and Statistics Directorate for advice (email: bcsinfo.rds@homeoffice.gsi.gov.uk) before undertaking analysis of BCS data.

The BCS datasets and related dataset documentation are Crown Copyright. Copies may be made for private study and research purposes only.

The BCS datasets do not contain identifying information on individual respondents, and are therefore not subject to registration under the Data Protection Act. The contractors who have carried out the fieldwork for the BCS do not pass on details of respondents' names and addresses to the Home Office.11

The language of the materials is English.

A number of items of dataset documentation have been transferred to NDAD, including Home Office publications relating to the BCS, Technical Reports prepared by the survey contractors, papers of the British Crime Survey Advisory Group, training notes for BCS users, and other material. Further details are available in the Dataset Documentation Catalogue.

Copies of the BCS datasets for 1982-2002 have been deposited by the Home Office in the UK Data Archive, reference numbers 1869 (for 1982), 2077 (for 1984), 2706 (for 1988), 3202 (for 1992), 3591 (for 1994), 3832 (for 1996), 4081 (for 1998), 4463 (for 2000), 4786 (for 2001) and 4787 (for 2001-2002).

Publications relating to the BCS have been published by the Home Office's Research Development and Statistics Directorate, and its predecessors, in a number of publication series: Research Bulletins, Statistical Bulletins, Home Office Research Studies, Research and Planning Unit Papers, Research Findings, Occasional Papers and Policing and Reducing Crime Unit Briefing Notes. In 2002 RDS introduced a new annual publication series, Crime in England and Wales, to combine the reporting of police recorded crime statistics and results from the BCS.12 Publications relating to the Scottish section of the 1982 and 1988 surveys were published by the Scottish Office.

The Home Office has produced a list of BCS publications (1999) - including Home Office and non-Home Office publications - which can be consulted via the Dataset Documentation Catalogue, reference CRDA/2/DD/5/1. A more recent version of this list (2001) is included in training notes for users of the BCS (Dataset Documentation Catalogue, reference CRDA/2/DD/3/8). The Dataset Documentation Catalogue includes links to many of the more recent Home Office publications (1996 onwards) which are held by NDAD. Users are also advised to consult the web site of Research Development and Statistics Directorate for details of its current publications, including those available on-line.

NDAD has received copies of the Technical Reports on the 1982-2001 surveys, which were produced by the contractors who carried out the fieldwork: see the Dataset Documentation Catalogue, references CRDA/2/DD/2/1-9. These and other non-Home Office publications relating to the BCS are included in a list of BCS publications which has been transferred to NDAD (see Publications produced by the originating department).

From 1982 until 1999 the Home Office held BCS data on a VAX 6410 mainframe computer manufactured by Digital Equipment Corporation. In 1999 the VAX was taken out of service and replaced by PCs linked to a server.

A VMS operating system manufactured by Digital Equipment Corporation was used with the Home Office's VAX mainframe. VMS version 5.5 was in use in 1998; version 4.0 or 4.1 may have been used at earlier dates. Microsoft Windows NT was adopted as an operating system after the VAX was decommissioned (see Hardware).13

From the introduction of Computer Assisted Personal Interviewing in 1994 (see How data was originally captured and validated), until 2000, data from the BCS was gathered by the Home Office's contractors using BLAISE software developed by Statistics Netherlands. According to one of its distributors, BLAISE offers the following features:

• Computer-assisted personal interviewing (CAPI), Computer-assisted telephone interviewing (CATI), and Computer-assisted self interviewing
• Interactive data editing
• Forms-based data entry
• Structured questionnaire design
• Complex routing and checking
• Top-down editing with interactive tables
• Survey management
• Data manipulation and tabulation
• Interactive coding
• Export to statistical and database packages14

Data gathered in BLAISE has been exported to the Home Office by the survey contractors, using SPSS database software manufactured by SPSS Inc. Until 1999, data sent to the Home Office in SPSS was exported into P-STAT (manufactured by P-STAT Inc), which was the primary software used by the Home Office to hold and manipulate BCS data. P-STAT "was begun in the early 1960's for analysing psychopharmacological data. The original authors, Roald and Shirrell Buhler, left the Princeton University Computer Center in 1979 to form P-STAT, Inc. Today, P-STAT offers survey analysis, statistical analysis, data and file management, data presentation, and report-writing capabilities". P-STAT version 2.14 was used by the Home Office from 1992 (earlier versions may have been used before then).

In 1999, as part of a move to decommission a VAX mainframe (see Hardware), Research Development and Statistics Directorate migrated all of the BCS datasets to SPSS version 8.0. P-STAT output files continued to be held to allow previous work to be replicated in SPSS. Microsoft Excel was also used by the Home Office for certain calculations and analyses.15

From 2001, the software used to gather the survey data was changed to Quancept CAPI, also produced by SPSS, Inc. The related Quanquest module was used to author the questionnaire, which was designed to replicate exactly the earlier BLAISE surveys; Quancept controlled every aspect of the data gathering.

Training notes produced by the Home Office for users of BCS data, both inside and outside Research Development and Statistics Directorate, indicate the type of interface that was available to Home Office staff after the migration of the BCS datasets to SPSS (see Application software). The datasets were held in a shared directory in the Home Office's directory tree, and were accessed via a Windows-type interface (with windows, icons, pull-down menus etc). Data files were opened and manipulated using SPSS Data Editor, in which data was displayed in a tabular format. Dialogue boxes could be used for basic data manipulation; separately created syntax files (text files containing commands or instructions written in SPSS syntax) were used for more complex manipulations. SPSS Output Viewer was used to automatically generate statistics and charts.16

"Rates" files containing syntax used by the Home Office to analyse the BCS datasets have been transferred to NDAD: see the Dataset Documentation Catalogue, reference CRDA/2/DD/8/1/1-7.

Before the migration of the BCS datasets from P-STAT to SPSS (see Application software), the datasets held by the Home Office consisted of separate tables which corresponded, broadly, to the main sections of the survey questionnaire. The principal exception was the self-completion sections, which were apparently held as a single table. A different structure was adopted following the conversion of the datasets to SPSS. Since 1999 each dataset has primarily consisted of two tables: one for data from the Victim Forms section (known as the "vf" table), the other for data from all other sections of the questionnaire (the "nonvf" table). This reflects the fact that the Victim Form section is hierarchical (one or more victim form records can relate to the same interviewee), whereas the data from the other parts of the questionnaire can be held as a single flat data file, with one record per interview. It also appears that separate files have been used to hold data from "open-ended" fields (fields which record a response to a question in the respondent's own words - e.g. where the answer was coded as "other" in a related field).

Other changes in the BCS datasets can be traced to the introduction of Computer Assisted Personal Interviewing and Computer Assisted Self Interviewing in 1994 (see How data was originally captured and validated). From 1994 onwards the BCS datasets have field names which normally correspond to the fields used in the interviewing software, and which match the fields recorded in the BCS questionnaires in the Technical Reports (for copies of the BCS Technical Reports, see the Dataset Documentation Catalogue). However, for the 1984-1992 sweeps a different convention was used. Field names have prefixes indicating the section of the survey questionnaire to which they relate; these include: M (Main Questionnaire), V (Victim Forms), FA (Follow-up A), FB (Follow-up B), D (Demographics) and SC (Self-completion). Often these prefixes are not included in the field names in the Technical Reports. Another difference between the Technical Reports and the pre-1994 datasets is that 'MR' is used in field names in the datasets to indicate "multiple-response" variables, where more than one answer to a question was possible, and so separate fields were created for each possible answer (e.g. "M45MR1" indicates the first possible answer to question 45 in the Main Questionnaire section). In the 1994 and later datasets multiple response fields are indicated by a numeric suffix at the end of each field name (e.g. "PCWHO1"), which will match the field names in the Technical Reports.17

Further details of the file structure of the BCS datasets are contained in the catalogues of individual datasets: see Links to dataset catalogues.

The BCS datasets are closed, in the sense that once data from each sweep of the survey has been input, checked and validated, the data is preserved and is not subsequently overwritten by data from later surveys.

To ensure consistency between the different sweeps of the BCS, many features of the design of the survey have remained relatively constant over time. This section summarises how the methods of data capture and validation used in the survey have evolved since 1982. More specific information about the design of particular sweeps of the survey are given in the catalogues of individual BCS datasets (see Links to dataset catalogues).

Sampling technique

BCS data is gathered from a "core" sample of adults aged 16 and over, plus various "booster" samples which have been employed for particular purposes in certain sweeps of the survey. The first four surveys (1982-1992) achieved successful interviews with a core sample of between 10,000 and 11,000 respondents in England and Wales: 10,905 in 1982, 11,030 in 1984, 10,392 in 1988 and 10,059 in 1992. About 5,000 adults were interviewed in Scotland in the separate sections of the BCS conducted there in 1982 and 1988.18 In 1994 it was decided to increase the core sample to around 15,000 adults in England and Wales. A core sample of 14,520 interviews was achieved in 1994, 16,348 in 1996 and 14,947 in 1998. The core sample was further increased to 19,411 for the 2000 survey, and 37,170 for the 2001/02 survey.19 In each sweep of the survey the sample of achieved interviews represents a fraction of the original number of households that were approached for interview. The BCS has tended to achieve a positive response rate of around 80%, excluding addresses which were empty, demolished, untraceable or ineligible for interview for other reasons.20

The method of selecting the core sample has changed with different sweeps of the BCS. In the first three surveys electoral registers were used as a sampling frame. In 1984 and 1988 this involved selecting a sample of constituencies from the total number of constituencies in England and Wales, then selecting two wards within each constituency, a polling district within each ward, and a set number of addresses within each polling district, with the final selection of the respondent for each address being done by the interviewer in the field. The sample for the 1982 BCS was chosen on broadly the same basis except that addresses were drawn from either one ward or two polling districts per selected constituency (this was done to allow comparison of the results from each method of selection).21

By 1988 it was acknowledged that electoral registers were flawed as a sampling frame, chiefly because they "do not cover all private households- about 4% are omitted, in particular those with young people, the unemployed, ethnic minorities and those in rented accommodation".22 Consequently the 1992 survey and all subsequent sweeps of the BCS have based their sampling frame on the Postcode Address File (PAF), a "listing of all postal delivery points in the country, with almost all households having one delivery point or letterbox".23 The BCS has employed the "small" user section of the PAF, which corresponds most closely to private households.24 In the 1994, 1996 and 1998 surveys the sample design required the selection of a set number of postcode sectors from a list of postcode sectors, stratified according to factors such as whether inner city or not, standard region, population and social class of the household head (this was known as a "stratified multi-stage random probability design"). A set number of addresses were then randomly selected from each postcode sector. Random sampling techniques were used by the interviewer to make the final selection of the household to be approached, if more than one dwelling unit was located at the same address; and the final selection of the adult to be interviewed, if more than one adult was resident in a household. The 1992 BCS operated on similar lines, the main difference being that the initial sampling was done by constituency, with two postcode sectors being randomly selected per constituency, and then a set number of addresses within each postcode sector. An innovation introduced with the 1998 survey was the use of a "rotated" sample of postcode sectors: half of the addresses selected for interview were taken from postcode sectors used in the 1996 survey, while the other half were taken from a fresh sample of postcode sectors. This was done "to reduce the variance of round-on-round measures of change".25

Further changes to the methods of sampling were introduced in the 2000 BCS. Primary stratification by inner city and non inner city was abandoned; instead, primary sampling was done by Police Force Area (PFA). The postcode sectors within each PFA were stratified by population density and social class of the household head. Once postcode sectors had been selected, addresses were ordered in alpha-numeric order of postcode within each sector, prior to the systematic selection of addresses. Stratification by PFA was retained in the 2001 BCS. 26 It should also be noted that in response to the higher incidence of crime in inner city areas, every BCS from 1982 until 1998 oversampled these areas by a factor of about two.27 In the first survey the classification of parliamentary constituencies, developed by the Planning and Applications Group (PRAG) of the Centre for Environmental Studies, was used to identify inner city areas. In subsequent surveys parliamentary constituencies or (from 1992 onwards) postcode sectors were classified as "inner city", using Census data, with the key factors being population density, level of owner-occupied tenure and social class profile. The 2000 BCS abandoned the over-sampling of inner city areas in favour of an oversampling of the smallest PFAs, in order to ensure a minimum number of achieved interviews per PFA. However, the "inner city marker", attached to the data that classified postcode sectors, was used as a post-stratifying factor, with adjustments being made to account for lower levels of response in inner city areas.28

In addition to core samples, the 1988-1996 and 2000-2001/2 sweeps of the BCS employed "booster" samples of ethnic minorities to gather information on how the experiences of these groups differed from the rest of the population, in regard to victimisation and attitudes to crime. Booster samples were identified using "focused enumeration": i.e. by interviewers asking respondents from the core sample whether anyone of an ethnic minority lived nearby, and then attempting to interview those households which were identified within a set number of adjacent addresses.29 The 1996 and 2000 surveys coupled this with an additional sampling of addresses from areas known to have large ethnic minority populations.30 These methods produced booster samples of 1,349 ethnic minority respondents in 1988, 1,654 in 1992, 2,030 in 1994, 2,608 in 1996, 3,874 in 2000, and 3,329 in 2001/2. The 2001/2 BCS covered a different set of ethnic minorities from the previous sweeps. Earlier surveys had drawn their booster samples from Black and Asian respondents (Asian respondents were those who classified themselves as Indian, Pakistani or Bangladeshi). In the 2001 /2survey all non-white groups were eligible for inclusion in the booster sample. 31

The 1992 BCS also employed a one-off booster sample of children aged 12-15, from households where an adult had already been interviewed as part of the core sample or ethnic minority booster sample. All children aged 12-15 were eligible for interview, not just one per address as with adults. The 1996 survey employed a booster sample of adults in the Thames Valley Police Force area, which was commissioned and paid for by the Thames Valley Police Force. The 2001/2 BCS included a "youth boost" of 1,669 respondents aged 16-24, gathered from households where an older respondent had been already been selected for interview.32

In 2000 the Home Office announced that, with effect from 2001, the BCS would be held on an annual rather than a biennial basis and would aim for a considerably larger sample of around 40,000 adults (including an ethnic minority booster sample). The following reasons were advanced for these changes:

• A greater number of interviews was needed, to monitor crime trends to the necessary degree of accuracy. In particular, a larger sample would enable better tracking of violent offences.
• A growing number of performance indicators could best be delivered by the BCS: e.g. indicators of fear of crime and satisfaction with the police.
• An enlarged BCS could be used to monitor some Best Value indicators for police forces at Police Force Area level.
• An enlarged sample would enable some parts of the questionnaire to be administered to a subset of the full sample.33

In preparation for the enlarged survey the Home Office commissioned a review of the BCS's methodology which recommended a number of changes to the design of the survey (for a copy of the report arising from the review, see the Dataset Documentation Catalogue, reference CRDA/2/DD/7/1). The most significant of the recommendations were implemented in the 2001 BCS. These were:

• That data should be gathered through continuous interviewing throughout the year, rather than by concentrating most of the interviews in the first quarter of the year as had previously been the case.
• That the recall period for questions relating to victimisation should be the 12 months prior to the interview, rather than the previous calendar year.
• That in order to measure the effect of the changes, the 2001 BCS should employ a "spliced design", involving a mixture of the old and new methodology.

The "spliced design" approach meant that 8,985 core sample interviews in the first half of 2001 were conducted using the old BCS methodology and the traditional recall period. The remainder of the core sample interviews and all interviews for the ethnic minority booster sample and the "youth boost", conducted in the second half of 2001, employed the continuous sampling design and the new recall period.34

Survey forms

The sections of the BCS questionnaire and the order in which questions are asked have remained broadly similar across different editions of the survey. The following table summarises how the forms used in the BCS have evolved:35

Paper forms were used in the BCS until the 1994 sweep, when Computer Assisted Personal Interviewing (CAPI) using BLAISE software (see Application software) was introduced. In CAPI, answers to survey questions are logged by the interviewer onto a laptop computer. CAPI has been used for all sections of the BCS questionnaire except for the Self-completion sections, where Computer Assisted Self Interviewing (CASI) has been used (i.e. the answers are entered directly onto the laptop by the respondent). The computer program "specifies the questions, range and structure of permissible answers and routing instructions".38 Under CAPI/CASI, the entire interview consists of one electronic questionnaire, although this questionnaire has been broken down into various sections, corresponding to the paper forms which were used before 1994.

Coding and classification

The input, coding and checking of answers to BCS questions has been done by coders employed by the contractor(s) carrying out each particular sweep of the survey. This has included coding the occupation, social class, etc. of the respondent and the head of household; and coding the offences recorded on Victim Forms, according to a standard scheme for classifying offences which has remained broadly consistent across different sweeps of the BCS. The BCS's instructions for classifying offences were originally developed by the Home Office in consultation with the statistical officers of a number of police forces.

The BCS's method of classifying offences differs from that of the police, in that the BCS includes any incident which is technically criminal (in 1998, only 62% of BCS crimes fell into offence types that could be compared to police recorded crime). Also some types of offences (e.g. common assault) have always been included in the BCS, but were not included in the recorded crime count until April 1998. Further details of the classification schemes employed by the BCS are given in the Technical Reports for each sweep of the BCS (see the Dataset Documentation Catalogue), and in the catalogues of individual BCS datasets (see Links to dataset catalogues).

After the introduction of CAPI/CASI in 1994, post-interview coding involved the editing and checking of data input as part of the original interview. This included checking "open-ended" questions, where a response was recorded in the respondent's own words: e.g. to assign a classification to the response, to determine whether refinements were needed to the coding frame, or to determine whether an answer should in fact have been recorded in a related encoded field (a process known as "back-coding"). In 1992, and from 1996 onwards, it also involved checking the offence classifications which were automatically assigned by the computer program (see below). Another change which accompanied the introduction of CAPI/CASI was in the coding of "don't know" and "refusal" responses (where the respondent did not know the answer to the question, or refused to answer). From 1994 onwards "don't know" answers were coded as '9' or '99', and "refusals" were coded as '8' or '98'; before 1994 the opposite was the case, with '9' or '99' indicating "refusals" or questions not completed in error, and '8' or '98' indicating "don't know".39

In the 1992 survey the contractor developed a computer program which suggested the most likely classification for an offence, based on a range of variables recorded in the Victim Forms. Similar programs were used in the 1996, 1998, 2000 and 2001 sweeps, as part of the edit version of the interviewing software (about 82% of Victim Forms were initially classified in this way in 1998). The classifications suggested by the computer were confirmed or modified by the coders, who also assigned codes to offences not classified by the program. Coders referred to the Home Office for guidance in the case of less common offences, or where there was doubt about a classification. An innovation introduced in the 2000 BCS and continued in 2001 was the procedure that coders should indicate whether their offence classification was "certain". A 10% quality check of "certain" classifications was made by the Home Office, while all uncertain classifications were checked by Home Office researchers before a final code was assigned.40

Weighting

In each BCS survey data has been weighted at the data processing stage in a number of ways, to correct for imbalances in the sampling and imbalances created by the design of the interview. The nature of the weights which have been applied has varied with changes in the nature of the survey. Every BCS sweep has used a series weight to account for the fact that series incidents are recorded on a single Victim Form, and an inner-city weight to correct for the oversampling of inner city areas (up to 1998) and the difference in response rates between inner city and other areas. Other weights which have been developed at various times include:

• A weight to account for the mismatch between electors and persons at addresses, when data was used to represent the adult population (this was applied to the 1982-1988 surveys, in which electoral registers were used as the sampling frame).
• A weight for household crimes, where the victim is the household rather than the individual respondent (applied to the 1982-1988 surveys), to correct for the fact that in these sweeps larger households had a disproportionately high chance of selection.
• A dwelling unit weight (applied to surveys from 1992 onwards), to correct for cases where more than one household was covered by an entry on the Postcode Address File.
• An individual weight (applied to surveys after 1992), to account for the fact that as the Postcode Address File generates a sample of addresses, individuals living in larger households tend to be under-represented.
• Ethnic minority weights (used in the 1992 and 1996 BCS), to down-weight the ethnic minority booster sample.
• The 2000 and 2001 surveys employed a police force area weight, to correct for disproportionate sampling by Police Force Area.
• The first two BCS surveys included weights to account for the fact that in these sweeps, not all respondents were required to complete Follow-up Questionnaires.

Other weights have been derived from the above weights, e.g. for individual, household or incident-based analysis.41 Additional details about the weights which have been applied in different sweeps of the BCS are given in the catalogues of individual datasets (see Links to dataset catalogues).

In 2001 the Home Office announced that it intended to introduce into the BCS the calibration weighting technique, which had been recommended by the Government Statistical Service for use in all government social surveys. The technique "iterates from initial survey weights to generate weights that match the sample to the distribution of demographic and other factors (such as region) with those in the population by means of a single household weight . . . [This] can protect against non-response bias arising due to differential response weights in grouping on which calibration is applied".42

The BCS is subject to a number of technical limitations, some of which are particular to the BCS, while others are common to similar surveys:

• Like any sample survey, the BCS faces difficulties in accurately representing the entire population. This is particularly true of the first three sweeps, which used electoral registers as a sampling frame (see How data was originally captured and validated). The Postcode Address File was adopted to overcome some of the problems associated with electoral registers, but the BCS still faces the problem of selected respondents who cannot be traced, or who decline to be interviewed. A mini-survey of respondents who refused to take part in the full survey in 1996 suggested that non-respondents face slightly lower crime risks.43 As previously stated (see Aim and purpose), the BCS excludes certain social groups: in particular, children under 16 and adults in institutions. The survey may also tend to over-represent older age groups at the expense of younger respondents, and may be particularly prone to sampling error where rare crimes (e.g. robbery and serious assault) are concerned.
• Because of the small number of incidents reported to the survey, estimates of sexual offences derived from the BCS are not considered to be reliable.
• The BCS tends to undercount crimes where the victim and offender know each other: e.g. domestic and "non-stranger" violence. Respondents may not regard incidents as crimes or may be reticent with interviewers.
• It has been suggested that there may be differential undercounting between social groups: e.g. some types of offences are more likely to be reported by better educated respondents.
• The BCS is limited by the ability of respondents to remember and report incidents accurately and truthfully. There is evidence that trivial offences tend to be under-reported while serious incidents are over-counted. A follow-up study of 35 respondents to the 1996 survey has explored the issue of the accuracy with which respondents report crime: see C. White and J. Lewis, Following up the British Crime Survey 1996: a Qualitative Study (London: Social and Community Planning Research, 1998).44

Details of the content and transformation validation checks performed by NDAD on each BCS dataset are contained in the catalogues of individual datasets: see Links to dataset catalogues.

Dataset catalogues provide more detailed information about individual datasets, and are currently available for the following dataset(s):

1. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 1.

2. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, pp. 1-2; Pat Mayhew, Natalie Aye Maung and Catriona Mirrlees-Black, The 1992 British Crime Survey, Home Office Research Study no.132 (London: HMSO, 1993), pp.1-4.

3. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 70; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 7-8.

4. Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp. 3-5, 7, 149, 154; Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 2; CRDA/2/DD/1/2000/2, p. 2.

5. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 69; CRDA/2/DD/1/2000/2, p. 109; CRDA/2/DD/1/2001/11, pp. 80-81; Mike Hough and Patrick Mayhew, The British Crime Survey: First Report, Home Office Research Study no.76 (London: Home Office, 1983), pp. 4, 39; Data Archive, dataset description for British Crime Survey 1982 (reference no.1869), consulted on 30 March 1998 (http://biron.essex.ac.uk/); Scottish Executive web site, page on "The Scottish Executive Statistical Plan 2000-2001" (http://www.scotland.gov.uk/consultations/sesp-05.asp) consulted on 6 June 2000.

6. BMRB International website (http://www.bmrb.co.uk/), page on "British Crime Survey 2001-2004" consulted 20 December 2002.

7. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 69.

8. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 69; CRDA/2/DD/1/1998/1, p. 48; CRDA/2/DD/2/7/3; CRDA/2/DD/1/2001/11 p. 81; Hough, Anxiety about Crime: Findings from the 1994 British Crime Survey, p. 52; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp. 7-8, 149; Pat Mayhew, David Elliott and Lizanne Dowds, The 1988 British Crime Survey, Home Office Research Study no.111 (London: HMSO, 1989), pp. 6-7, 89; Mike Hough and Pat Mayhew, Taking Account of Crime: Key Findings from the Second British Crime Survey, Home Office Research Study no.85 (London: HMSO, 1985), pp. 7, 77; Hough and Mayhew, The British Crime Survey: First Report, p. 5.

9. Dataset Documentation Catalogue, reference CRDA/2/DD/2/6/2, pp. 1, 10; CRDA/2/DD/1/2001/11, pp. 79, 81-82.

10. Dataset Documentation Catalogue, reference CRDA/2/DD/2/6/8, p. 2.

11. Home Office web site, page entitled "The British Crime Survey 2000" (http://www.homeoffice.gov.uk/bcs2000/bcs2000.htm) consulted on 1 June 2000.

12. Research Development and Statistics website, "Crime in England and Wales 2001/2002" (http://www.homeoffice.gov.uk/rds/crimeew1.html) consulted 20 December 2002.

13. Notes of telephone conversations between NDAD and the Home Office on 11 and 12 November 1998; email from the Home Office to NDAD, 19 June 2000; dataset transfer form for the 1996 BCS dataset; letter of 25 February 1999 from the Home Office to NDAD.

14. Website for Westat, Rockville, Maryland, USA (http://www.westat.com/blaise/index.htm) consulted on 12 November 1998.

15. Website for P-STAT Inc (http://www.pstat.com/lookat.html) consulted on 12 November 1998; note of telephone conversation between NDAD and the Home Office on 12 November 1998; emails from the Home Office to NDAD, 19 June 2000.

16. Dataset Documentation Catalogue, reference CRDA/2/DD/3/6, pp. 1-9.

17. Letter of 28 January 1998 from the Home Office to NDAD; letter of 18 November 1999 from the Home Office to NDAD; email of 19 June 2000 from the Home Office to NDAD; Dataset Documentation Catalogue, reference CRDA/2/DD/3/3, p. 16; CRDA/2/DD/3/8, pp. 20-22.

18. Hough and Mayhew, The British Crime Survey: First Report, pp. 4, 39; Hough and Mayhew, Taking Account of Crime: Key Findings from the Second British Crime Survey, p. 80; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, pp. 6, 93; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 154.

19. Dataset Documentation Catalogue, references CRDA/2/DD/2/5/2, pp. 1, 15; CRDA/2/DD/2/6/2, p. 12; CRDA/2/DD/2/7/1, pp. 10-11; CRDA/2/DD/1/2000/2, p. 1; CRDA/2/DD/1/2001/11, p. 79; CRDA/2/DD/10/5, p. 2; copy of presentations by Home Office staff at an RDS seminar on the British Crime Survey, 13 March 2000.

20. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 72; CRDA/2/DD/1/2001/11, p. 85.

21. Hough and Mayhew, The British Crime Survey: First Report, p. 37-38; Hough and Mayhew, Taking Account of Crime: Key Findings from the Second British Crime Survey, p. 78-79; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, pp. 90-92.

22. Mayhew, Elliott and Dowds, The 1988 British Crime Survey, p. 90.

23. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 88.

24. Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 150.

25. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/1, pp. 47-48; CRDA/2/DD/2/7/1, pp. 6-8; CRDA/2/DD/2/6/2, pp. 6-8; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp. 150-153; Hough, Anxiety about Crime: Findings from the 1994 British Crime Survey, p. 51.

26. Dataset Documentation Catalogue, reference CRDA/2/DD/7/1, p. 28; CRDA/2/DD/1/2001/11, p. 84.

27. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 71.

28. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 71; CRDA/2/DD/2/6/2, p. 6; CRDA/2/DD/2/7/1, p. 6; CRDA/2/DD/7/1, pp. 10, 28; CRDA/2/DD/1/2000/2, pp. 111-112; Hough and Mayhew, The British Crime Survey: First Report, p. 37-38; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 151.

29. Dataset Documentation Catalogue, reference CRDA/2/DD/2/6/2, p. 1; CRDA/2/DD/7/1, p. 30; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, pp. 6, 92-93.

30. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1997/4, p. 69; CRDA/2/DD/2/6/2, p. 1; CRDA/2/DD/1/2000/2, p. 112.

31. Dataset Documentation Catalogue, reference CRDA/2/DD/2/6/2, p. 1; CRDA/2/DD/1/1998/5, p. 71 n. 4; CRDA/2/DD/1/2000/2, pp. 1, 112; CRDA/2/DD/1/2001/11, p. 79; CRDA/2/DD/10/5, p. 2; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, p. 93-94; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p.7; Hough, Anxiety about Crime: Findings from the 1994 British Crime Survey, p. 52.

32. Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp.149, 154; email of 11 May 2000 from the Home Office to the crim-bcs-users@mailbase.ac.uk mailing list; Dataset Documentation Catalogue, reference CRDA/2/DD/1/2001/11, p. 79; CRDA/2/DD/10/5, p. 2.

33. Dataset Documentation Catalogue, reference CRDA/2/DD/6/1, p. 1; CRDA/2/DD/1/2000/2, p. 60; talk by Chris Kershaw to the Research, Development and Statistics Directorate Research Conference, Queen Elizabeth II Conference Centre, London (8 February 2001).

34. Dataset Documentation Catalogue, reference CRDA/2/DD/1/2001/11, pp. 1-2, 79, 84.

35. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/1, pp. 48, 70; CRDA/2/DD/2/6/2, pp. 2-4; CRDA/2/DD/2/7/1, pp. 3-5; CRDA/2/DD/3/3, pp. 6-8; CRDA/2/DD/6/1, p. 4; CRDA/2/DD/1/2001/11, pp. 80-83; Hough and Mayhew, The British Crime Survey: First Report, p. 5; Hough and Mayhew, Taking Account of Crime: Key Findings from the Second British Crime Survey, pp. 7, 77; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, pp. 6-7; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp. 7-8, 149-150; Hough, Anxiety about Crime: Findings from the 1994 British Crime Survey, p. 52; letter of 25 February 1999 from the Home Office to NDAD.

36. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/1, p. 70; CRDA/2/DD/2/6/2, p. 3; CRDA/2/DD/3/3, pp. 9-12.

37. Dataset Documentation Catalogue, reference CRDA/2/DD/10/5.

38. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, pp. 70, 86; CRDA/2/DD/2/6/2, pp. 1-2.

39. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, pp. 17, 73; CRDA/2/DD/2/6/2, pp. 17-18; CRDA/2/DD/3/3, p. 13; CRDA/2/DD/2/7/1, p. 16; CRDA/2/DD/3/8, p. 22; Hough and Mayhew, The British Crime Survey: First Report, pp. 40-41; letter of 25 February 1999 from the Home Office to NDAD.

40. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, p. 73; CRDA/2/DD/2/4/1, pp. 19-20; CRDA/2/DD/2/6/2, p. 18; CRDA/2/DD/2/7/1, p. 17; CRDA/2/DD/1/2000/2, p. 114; CRDA/2/DD/1/2001/11, pp. 86-87; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 157-158.

41. Dataset Documentation Catalogue, reference CRDA/2/DD/2/6/2, p. 21; CRDA/2/DD/1/1998/5, p. 73; CRDA/2/DD/3/3, pp. 5-6; CRDA/2/DD/1/2000/2, p. 114; CRDA/2/DD/1/2001/11, pp. 85-86; Hough and Mayhew, The British Crime Survey: First Report, pp. 39-40; Mayhew, Elliott and Dowds, The 1988 British Crime Survey, p. 95-96; Hough and Mayhew, Taking Account of Crime: Key Findings from the Second British Crime Survey, pp. 80-81; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, p. 156-157; Hough, Anxiety about Crime: Findings from the 1994 British Crime Survey, p. 53.

42. Dataset Documentation Catalogue, reference CRDA/2/DD/10/3, p. 6.

43. The results of this study were reported in P. Lynn, "Collecting data about non-respondents to the British Crime Survey", unpublished report to the Home Office (London: Social and Community Planning Research, 1997).

44. Dataset Documentation Catalogue, reference CRDA/2/DD/1/1998/5, pp. 75-76; CRDA/2/DD/3/5; CRDA/2/DD/1/2000/2, p. 2; Mayhew, Aye Maung and Mirrlees-Black, The 1992 British Crime Survey, pp. 5-6, 159.