The Scottish Government

« Previous | Contents | Next »

Listen

THE 2000 SCOTTISH CRIME SURVEY: OVERVIEW REPORT

Appendix B - Survey Design and Methods

Sample Design

Main Sample

Analyses of the 2000 SCS data are based on responses relating to both households and individuals. The aim of the sample design was to secure data which would be representative of the Scottish adult population residing in private households which would allow comparison with previous survey data, in particular the 1993 and 1996 SCS. As in previous surveys, the institutional population was excluded. The basic design approach was the same as that of the two previous surveys; that is, a probability sample selected according to a multi-stage stratified design.

In the first sweep of the BCS in Scotland, the Electoral Register (ER) was used as the sampling frame, while in the second sweep the ER was used in combination with the Postcode Address File (PAF). Its is now generally accepted that the ER is flawed as a listing of individuals - a study by the Office of Population Censuses and Surveys (OPCS) in 1991 estimated that the level of non-registration in Scotland was 6.6%. In 1993, 1996 and 2000 it was decided to use only the PAF as the sampling frame. The PAF provides details of all postal delivery addresses rather than a listing of individuals and is generally considered to be a much more complete listing than the ER.

One disadvantage of the PAF is that it contains a small number of properties which are not private residences. This was not a serious problem as these could be screened out during the survey. However, it remained a factor affecting the issuing of the sample and fieldwork control. A more significant problem was that the number of residences at each property listed was not always correctly identified by the number of delivery points listed in the PAF. The most common inaccuracy here was tenemental property. This constituted a potential bias in that a randomly taken sample within each sample point could under-represent a particular property type (e.g. tenements). This was likely to affect the survey results, as property type is almost certainly a discriminatory variable.

A field check operation was therefore undertaken before the sample was drawn. Fifty-nine sample points containing tenement housing, where problems were expected, were identified using the Scottish *ACORN classification, together with local knowledge. Each of these areas were checked in the field and amendments to the PAF were made where appropriate. The field check in the 1993, 1996 and 2000 surveys suggested that undetected inaccuracies would not significantly affect the accuracy of the survey results.

Selection of Sample Points

In order to achieve reasonably economical fieldwork, the sample was clustered. However, to minimise the adverse design effects of clustering, a large number of sample points were specified with correspondingly few interviews in each. Five thousand interviews were targeted, with 15 in each of 334 sample points. The sample points were spread across the whole of Scotland including the Highlands and Islands. While the 1993, 1996 and 2000 SCS included Grampian and Highland Region and the largest islands in the survey area, previous sweeps of the BCS (Scotland) had not. The sample points were pseudo-enumeration districts: that is groupings of Output Areas (normally three), as defined by OPCS. Pseudo-enumeration districts containing fewer than 20 addresses were excluded from the sample frame. These accounted for only 0.1% of the Scottish population.

The sample point selection was stratified by geographic area and Scottish *ACORN group. Appendix 1 gives a brief outline of The Scottish *ACORN classification system. This stratification was the same as that used in the 1996 SCS.

The number of sample points within each stratum was proportional to the adult household population. The geographic areas used in the stratification were Scottish Regions, excluding the smallest islands. Most of these Regions no longer have any administrative function following local government reorganisation. Specifically they were:

• Borders Region
• Dumfries and Galloway Region
• Lothian Region
• Strathclyde Region (mainland only)
• Fife Region
• Central Region
• Tayside Region
• Grampian Region
• Highland Region (mainland only)
• Islands of Skye, Mull, Uist, Lewis/Harris, Islay, mainland Orkney and mainland Shetland

This proportion stratification would give the most efficient use of the sample size in providing the most reliable data for Scotland as a whole.

Selection of Addresses and Dwelling Units

Eighteen addresses were randomly selected at each sample point. These were issued to interviewers as the 'A List'. Four further addresses were then randomly selected at each sample point and issued to the interviewer. This 'B1 reserve list' was used only if the 18 'A' addresses had been exhausted and 15 successful interviews had not been achieved. The remaining addresses in each sample point were listed randomly - the 'B2 reserve list' - and issued by MVA staff to interviewers only if their 'B1 reserve list' was used up. When an address was transferred from the 'B' to 'A' lists it then became liable to full call back procedures.

In few cases the PAF address listing included more than one household. Where this was the case, interviewers selected a household at random using a 'Kish' grid.

Selection of Respondents Aged 16 and Over

After selecting the dwelling unit and obtaining household co-operation, the interviewer listed the adults (aged 16 and over) resident at the time of the survey. Then an individual was selected for interview according to who had the next birthday. No substitution was permitted and, if necessary, the interviewer had to make at least three calls back to contact and interview the selected adult.

Young Person (12-15) Sample

At the start of the interview, the interviewer also identified whether there were any young people between the ages of 12 and 15 resident at the address and sought parental permission to place a self-completion questionnaire with all of these. In cases in which an interview was successfully completed with an adult member of the household, but it was not possible to place the Young Person's questionnaire because the targeted 12-15 year old was out, at least one further call back was made.

Ethnic Minority Booster Sample

For the first time, the 2000 SCS included an ethnic minority 'booster' sample in addition to the main sample. Unlike the main sample, the central concern was not representativeness and comparability, but rather, achieving a sufficiently sized sample from which relatively reliable data could be secured.

The system of ethnic boosting used in the BCS (focused enumeration) was not considered to be suitable for the SCS because Scotland does not have sufficient concentration of ethnic minorities to make the method practicable. The ethnic minority booster sample was therefore drawn by searching for appropriate names in the Electoral Register. This was done using the 1999 Electoral Register information from the 334 sample points used in the main sample and the two pseudo-enumeration districts closest to each of these sample points. Using the information from these areas, addresses were screened into the sample on the basis of the surname of the occupant listed in the Electoral Register . The selected sample was then checked against the main sample addresses to ensure no repetition had occurred.

The ethnic minority booster sample consisted of 782 addresses selected from 192 of the 334 sample points used in the main sample and their adjoining areas. The other sample points did not yield any suitable names. Due to the nature of the ethnic minority population, the selected addresses were mostly in the urban, lowland areas of Scotland. All regions used in the main sample apart from Dumfries & Galloway and Islands were represented in this sample.

As discussed above, there are a number of benefits and drawbacks to be considered when using Electoral Register information as a sampling frame. Given the limited information available on Scotland's ethnic minority population, the Electoral Register was the only viable sampling frame available. However, it is vulnerable to under-registration and the information was somewhat out of date (at 19% of sampled addresses, adults of a non-white ethnic origin were no longer resident). The method used to select addresses also meant that only those persons with 'ethnic-sounding' surnames were selected. Effectively this excluded many Afro-Caribbean residents and those non-white residents, who because of marriage or parentage do not have an 'ethnic-sounding' surname, from the sample. However, the main ethnic groups in Scotland, Asian and Chinese, would usually be identified.

Ethnic minority addresses were given to interviewers in addition to their main addresses when any were located within their assigned sample points and adjacent areas. Initially, interviewers were assigned a maximum of eight additional addresses per main sample point. The interviewers were asked to obtain as many ethnic interviews as possible from their addresses, using the same call back procedures as in the main sample. Although a boosted ethnic minority sample of 500 was initially targeted, this was not achieved due to the data inaccuracies discussed above.

The nature of the sample used meant that the dwelling unit selection procedures used in the main sample were not employed for the ethnic minority booster. The selection of respondents for both the Adult and Young Persons Self-Completion questionnaire was similar to that used in the main sample, excepting that additional screening was required to ensure that ethnic minorities were resident in the sampled household and that the person selection process included only those of non-white ethnic origin.

The SCS Questionnaires

Interview Structure

Prior to attempting each interview, information relating to the address was recorded on a Survey Control sheet. This listed the building type, level of address (e.g. first floor), number, time and date of contact attempts, the expected and actual number of dwelling units at the address and the eventual outcome (i.e. successful interview, no contact, or refusal). For each address issued, a Survey Control form was returned. This allowed the calculation of response rates, and weighting factors for the expected/actual number of dwelling units and any property type response bias in the main sample.

The questionnaire was divided into sections. This made the interview easier for interviewers to manage. In addition, two versions of the main questionnaire were administered, each to half the sample. Interviewers administered Main A or Main B questionnaires according to whether the serial number (issued with the address) was odd or even. This served to expand the range of data that could be collected within an interview of reasonable duration. The questionnaire sections were:

• Main questionnaires A and B: the victimisation screening and socio-demographic data were common to both versions
• Victim Forms: these were filed in for each (in-scope) crime up to a maximum of five
• Adult Self-Completion: these were filed in by respondents aged 16-59 inclusive
• Young Person's Self-Completion: children aged 12-15 inclusive in respondents' households were asked to fill in one of these

The questionnaires' contents are described in more detail in the following sections.

Main Questionnaires

As in previous surveys, the main questionnaires were divided into stages. These were:

• introductory questions on fear of crime and general social issues
• screening questions on victimisation experiences
• a check for eligibility for Victim Forms
• a follow-up section, version A or B as appropriate
• socio-demographic questions

The introductory, screening and socio-demographic questions were the same for both Main A and Main B questionnaires. Section 2, the follow up questions, covered different areas.

Main A Section 2 covered the criminal justice system, specifically:

• contact with the police
• attitudes towards the police
• the police complaints system
• attitudes towards sentencing and the prison system

Main B was concerned with:

• attitudes towards personal safety in the local area
• security features of both home and car (if appropriate)
• experience of housebreaking and violent crime
• experience of nuisance telephone calls
• experience of incidents not normally thought of as offences and possible racial motivations for these

The victimisation screening questions set out in the main questionnaires were designed to identify crimes that had affected the respondent since 1 January 1999. A typical example of a 'household' incident is housebreaking, while an example of a personal incident is assault. Unlike previous surveys, an extra 'screener question' was included which asked whether any violent offences suffered by the respondent had been committed by another member of the household. This was to help raise disclosure rates of incidents of domestic violence.

As in previous surveys, the theft of milk bottles or newspapers from the doorstep was identified at the screening stage. Victim Forms were not completed for these offences because previous surveys had indicated that this resulted in the collection of a considerable volume of uninformative data. The 2000 SCS, like the 1996 SCS also collected data for all incidents of criminal damage, whereas these incidents had only been recorded for respondents completing the Main B questionnaire in the 1993 SCS.

Where a respondent reported a series of similar and connected incidents, carried out under the same circumstances and probably by the same people, only one Victim Form was completed for each series of crimes, although a record was kept of the number of incidents involved. Respondents were asked only about the most recent incident in the series, which avoided potentially repetitive questioning of 'series' victims and Victim Forms being 'used up' on very similar incidents.

Victim Forms

The Victim Form collected details of each incident - it identified the number of incidents, whether it was part of a series of incidents, and when these occurred in the year. The Victim Form also explored the victim's perception of any racial motivation to the offence and the impact of each particular crime/series of crimes on the victim in terms of emotional effects and financial loss. It identified whether the police were notified of the incident and what help victims wanted or received from external organisations such as Victim Support. The respondent was also asked what sentence they thought the offender should receive.

Prior to 1992, the BCS (Scotland) had an upper limit of four Victim Forms per respondent. For the 1992 BCS (England and Wales) and 1993 SCS the limit of four Victim Forms per respondent was increased to five, though the third, fourth and fifth Victim Forms asked only key questions. For the 1996 SCS, a maximum of five Victim Forms were completed in full. The 2000 SCS returned to the format used in the 1993 SCS, allowing a maximum of five victim forms, three of which were completed in summary. Although only a very few respondents completed 5 Victim Forms, this allowed the interview to be conducted more quickly, which the pilot study had indicated would be beneficial.

Limiting the number of completed Victim Forms to five avoided excessive interview length, and, by prioritising crimes at the screener questions, the survey covered the most important crimes. There was therefore no increase in the maximum number of Victim Forms that could be completed per respondent. However less information was collected in Victim Forms three to five than in 1996.

Self-Completion Questionnaires

Unlike the 1996 SCS, the 2000 SCS Adult Self-Completion questionnaire did not include any questions relating to self-reported offending. Instead, the questions focused more upon exposure to and use of illegal drugs, attitudes towards illegal drug use and experience of domestic violence, which was included for the first time in 1996. It was felt that it would be unreasonable and unproductive to ask questions about drug use of elderly respondents, so the Adult Self-Completion questionnaire was given only to respondents under the age of 60.

The Young Person Self-Completion questionnaire was administered to all people aged 12-15 resident in the respondent's household where the permission of a parent or guardian was granted. It was felt that the 1993 and 1996 versions had not always been completed well by young people, and that the design had been rather confusing. In the 2000 SCS, this questionnaire was designed in a workbook format, making it more user-friendly for younger respondents. It included questions about victimisation when out of the home including bullying and theft, contacts with and attitudes towards the police, 'confessions' of petty crimes, knowledge of and attitudes towards illegal drugs.

Piloting

Prior to the 2000 SCS, a considerable amount of developmental work was put into producing the questionnaires. Although much of the survey remained the same, the structure, content and detailed design of the 2000 questionnaires differed in some important respects from the 1996 versions. The main changes were:

• an extra 'screener question' in the Main questionnaire relating specifically to domestic violence
• additional questions within the demographic section of the Main questionnaire asking respondents about their health and their willingness to participate in future crime surveys
• changes to the Follow-up A questionnaire, which included omitting some questions on the specific details of police contacts, and on awareness of the Children's Hearings, Procurator Fiscal and jury service and adding some new questions on experience of police complaints procedures, public priorities for the police, and attitudes towards prisons
• changes to the Follow-up B questionnaire including minor amendments to questions on vehicle security and the addition of new questions relating to experience of violent crimes, violence at work and racially motivated harassment
• the inclusion in the Victim Form of a summary checklist (for Victim Forms 3-5), questions on racial motivation and more questions on victim support
• the omission from the Victim Form of detailed questions on the outcome of cases, in terms of progress to Court, conviction and sentencing
• the omission from the Adult Self-Completion questionnaire of questions on self-reported offending and the inclusion of more detailed questions on domestic violence, illegal drug use and attitudes towards illegal drugs
• a change in the presentation of the Young Person's questionnaire to a workbook format, incorporating questions on victimisation as integral parts of the questionnaire rather than as separate sheets and adding a separate question on bullying
• the omission of questions from the Young Person's Questionnaire about illegal drug use, and an increased emphasis on knowledge of and attitudes towards illegal drugs

These changes to the questions were thought to be significant enough to require field piloting. This took place in December 1999 with a field staff of three, each of whom were asked to obtain 10 interviews. The pilot study raised very few issues and it was therefore decided to proceed with the survey using these versions of the questionnaires.

The questionnaires were colour coded to allow easy identification.

Fieldwork

Although the 2000 SCS was conducted in exactly the same way as the 1993 and 1996 surveys, and with many of the same interviewers it proved to be a much more difficult exercise. It took longer to complete than planned, and much longer than previous surveys, and achieved a lower response rate. This reflects wide experience in the social and market research industry at present, where falling response rates and difficulties in interviewer recruitment and retention are major concerns. Also, anecdotal evidence strongly suggests that the Scottish field force was under considerable pressure during the survey period, with intense competition for the available field force. The following discussion reflects the difficulties experienced.

Briefing

Six day-long briefing sessions took place in early to mid January - three in Glasgow, two in Edinburgh and one in Aberdeen. Three MVA researchers conducted briefings, one of whom had undertaken these tasks in 1996, along with the fieldwork manager. A representative from the Scottish Executive Central Research Unit was present.

A total of 141 interviewers attended briefings. A substantial number of these had worked on the 1996 SCS, and so were already familiar with the survey methodology and questionnaire content. Interviewers who had worked on the 1996 survey undertook over half the interviews (53%).

Fieldwork, Supervision and Quality Checks

By 17 January 2000 all the interviewers had been briefed and all sample points had been issued.

Interviews and contact sheets were normally returned to the office by post. All contact sheets returned, regardless of whether they were successful interviews or not, were entered into a survey monitor database. All interviews were manually checked at this stage for consistency and completeness. Where mistakes occurred, interviewers were contacted as soon as possible.

The survey monitoring system developed specifically for the 1993 SCS enabled a close check to be kept on interviewer productivity. When this lapsed in an individual case the interviewer was contacted and, when necessary, the work reassigned. This system was modified slightly and utilised for both the 1996 and 2000 sweeps of the SCS.

The first batch of questionnaires sent in by each interviewer was subject to 100% manual check by the fieldwork supervision staff to ensure that they had fully understood all aspects of the briefing. Interviewers were contacted by telephone if they had made any mistakes.

Approximately half way through the survey period, telephone backchecks were carried out. In all, 274 backchecks (5.4%) were undertaken for the main sample and 20 (5.3%) were undertaken for the ethnic minority booster sample. These showed a high level of satisfaction with the interviewers and did not reveal any significant data problems.

Response Rate

The target completion date for the 2000 SCS was 24 March 2000. This date was set to minimise the recall period for which respondents had to remember victimisation. However, due to a combination of factors, the data collection stage of the main survey extended well beyond the intended length of time originally planned. The ethnic minority booster sample extended even further, due to prioritisation of the main sample and the need to re-assign addresses in order to achieve a sufficient sample size. All main sample interviews were completed and returned to the office by 5 June 2000.

The total number of questionnaires returned in the main and ethnic minority booster samples by questionnaire type are shown in Tables B1 and B2 respectively.

Table B1 Number of completed Returns in main sample - 1993, 1996 and 2000 SCS

Table B2 Number of Completed Returns (Ethnic Minority Booster Sample)

In the main sample, 7,760 addresses were issued to interviewers. Of these, 472 were found to be out-of-scope, leaving a total of 7,179 in-scope addresses. The total number of interviews completed was 5,059. The response rate therefore was 70.5%. This was a decrease from the 1996 SCS response rate of 77.0% from 5,045 interviews and the 1993 SCS response rate of 76.5% from 5,030 interviews.

In the Ethnic Minority Booster sample, 782 addresses were issued. Of these, 231 were found to be out-of-scope, 157 because there was no adult of a non-white ethnic origin resident within the household. This left 549 in-scope addresses, from which 380 interviews were completed; a response rate of 69.2%.

Of the 2,998 Adult Self-Completion questionnaires returned in the main sample, 112 were completed by adults aged 60 or over. Strictly speaking, these were out-of-scope. There was thus a total of 2,886 in-scope Adult Self-Completion questionnaires; a response rate of 90.7%. All 260 Adult Self-Completion questionnaires returned in the ethnic minority booster sample were completed by adults aged less than 60 years.

In addition to the 380 ethnic minority interviews, those questionnaires within the main sample completed by ethnic minority respondents were added into the ethnic minority booster sample. This increased the total number of main questionnaires in this sample to 424 (198 Version A questionnaires and 226 Version B questionnaires). Likewise, the number of Victim Forms increased to 225, the number of Adult Self-Completion questionnaires to 278 and the number of Young Persons Self-Completion questionnaires to 48.

In the main sample, 403 Young Person Self-Completion questionnaires were collected from a total in-scope child population of 576. This is a response rate of 70%. The equivalent figure for the ethnic minority booster sample prior to adding duplicated interviews from the main sample was 39.7% (44 completed questionnaires from a total in-scope child population of 116). This reduced to 38.7% when duplicated interviews had been added (48 completed questionnaires from a total in-scope child population of 124).

The planned 15 interviews were achieved at all except five of the 334 sample points in the main sample. Table 4.7 shows the distribution of returns by sample points. More than 15 interviews were obtained at 38 sample points. This was because interviewers initiated the interview process for a higher number of addresses than was turned out to be necessary. The uneven distribution is accounted for in the weighting process described below.

Coding and Data Preparation

Offence Classification

The purpose of the offence classification was to identify a single offence code per Victim Form from which victimisation rates could be calculated. This would allow comparison with other statistics (e.g. police statistics and past crime surveys).

The framework for offence classification had been developed over the series of crime surveys. Prior to the 1993 SCS, offence coding instructions consistent with previous sweeps of the BCS were drawn up. These instructions were used for the 1996 and 2000 SCS without significant amendment.

All offence classification was undertaken by one individual who was closely supervised by both MVA professional staff and a representative from The Scottish Executive Central Research Unit. All cases where there was any doubt were referred to the supervisors.

Once the initial offence classification had taken place, a computer check was run. This check used the answers to key questions (e.g. question 30 'was anything actually stolen or damaged?') to determine the type of crime committed. The program was able to identify logical inconsistencies within the data and offence codes which were not permitted. These were flagged as potential errors and warnings were manually checked using the free language description in the original Victim Forms.

In very few cases, the offence classification had been incorrectly coded. This was often because responses could be ambiguous, for example in relation to the exact location of the crime, particularly when only the closed questions were considered. Once the free language description of the offence was considered, which the computer program was unable to do, the correct offence code could usually be clearly determined.

For some cases, in particular in the description of the incident, questions were sometimes answered ambiguously by the respondents. This was specifically the case with question 27, regarding damage. Occasionally, in the verbatim description (question 4) damage was clearly indicated, yet questions 27, 28 and 29 indicated that no damage was done. The reason for this was not always clear, but it appeared that respondents did not always define minor acts of vandalism, especially those committed in conjunction with other offences, as damage.

Other Coding

Coding frames for all open-ended questions other than offence classification were drawn from the 1993 and 1996 SCS coding frames. New coding frames were developed where necessary for new questions. A small team of coders then allocated codes.

Social class coding was undertaken by one individual, using information on the status of the respondent or head of household where appropriate, the job and industry in which they worked and the number of people for whom they were responsible.

Data Preparation

All data were keyed in using a package specifically designed for this purpose (SPSS-DE). Each variable had range checks to ensure that data outside a valid range could not be entered. Automatic skip and fill routines ensured that the correct routing was followed.

Once all data had been input, it was examined at an aggregate level for each questionnaire type using basic frequency counts. As the data entry stage involved skip and fill rules as well as range checks, only very minor errors were found.

Contents of Data Files

Seven data files were produced, one for each questionnaire type:

• Main Questionnaire - version A
• Main Questionnaire - version B
• Victim Form
• Adult Self-Completion
• Young Person Self-Completion
• Main Survey Control Form
• Ethnic Minority Booster Survey Control Form

The Technical Report details the contents of these files.

Data Weighting

Weighting Process

Weighting was applied to correct possible sampling and response biases and to correct imbalances created by the survey method.

Weighting was conducted in four main stages:

• Household weighting (Main Sample only)
• Individual weighting
• Victimisation weighting
• Non-response weighting for self-completion questionnaires

Each of these is discussed in turn below.

Household Weighting

The first stage in the weighting process was to weight the data to adjust for factors relating to household selection. Three types of bias were possible.

Bias originating from inaccuracies in the PAF would occur if additional or fewer dwelling units than indicated in the PAF were found during the survey at a delivery point. Completed Survey Control forms returned from the field showed that at only two sample points (0.6%) or for two addresses (0.03%) there was a different number of delivery points to that listed in the amended PAF. The small extent of this aspect of bias was undoubtedly the result of the preliminary investigations and revisions made to the PAF. For this reason, it was decided not to weight to adjust for differences between the issued PAF and what was found in the field.

Property type bias would be indicated if the distribution of property types for completed interviews differed from those of the in-scope addresses sampled.

Area bias would occur where more or fewer than the target number of interviews (15) were achieved within a cluster. This occurred at 43 of the sample points (12.9%).

The computation of the household weighting factor to adjust for property type bias involved the calculation of a weighting factor (W _house) was a combination of the property type bias and area weighting factors.

Individual Weighting

All household data (e.g. theft from the home) had to be weighted by the household weighting factor (W _house) prior to analysis. Data which apply to individuals (e.g. attitudinal data or offences against the person) had to be weighted by a combined household and individual weighting factor.

This further weighting of the individual level data was necessary to account for the fact that each individual's chance of selection for interview was inversely proportional to the adult household size. For example, in a single person household that person was automatically selected for interview, whereas in a two person household each person had a one in two chance of being selected, and so on. Data weighted by the household weighting factor only were biased towards those living in small sized households.

Bias could also have occurred in terms of the respondent's characteristics, such as age or sex. In this case, despite a requirement that interviews were randomly selected within households, an examination of the data showed a bias towards interviews with females. Also, males aged between 16-24 and 25-59 were under represented.

Firstly, to overcome age/sex bias, interviews were weighted to reflect the age/sex composition within households. Secondly, interviews were weighted to their household size to adjust for household size bias. These two elements of the weighting process were combined to produce an individual weighting factor (W _person) for each household size age/sex band.

This formula was then repeated for each household size so that interviews within each age/sex category had the appropriate value of W _{d e f.} In the main sample, all data relating to the respondent were weighted by the individual and household weighting factor. Household data were weighted by the household weighting factor only. Within the ethnic minority booster sample, only the individual weight could be calculated and data were therefore weighted by that alone.

Victim Form Series Weighting

For certain analysis, for example when calculating victimisation rates, a series weight needed to be applied to the Victim Form data to reflect the fact that some Victim Forms refer to two or more incidents.

A weight was calculated (W _series)which referred to the number of incidents in the series, up to a maximum of five multiplied by the appropriate weight, depending on the type of offence.

Non-Response Weighting

The final stage in the weighting process was to examine the weighted data, to decide if adjustment was needed for any non-response bias in the completion of the Adult and Young Person Self-Completion questionnaires. For the Adult Self-Completion data the age, sex and social class of the respondents who completed Self-Completion questionnaires was compared with those who did not. Data weighted by the individual weight was used for this purpose.

A 90.7% response rate was achieved for the Adult Self-Completion forms in the main sample. The response rate distribution between age and sex groups was similar. For this reason, it was felt that no weighting of the Adult Self-Completion data was necessary.

A 74.7% response rate was achieved for the Adult Self-Completion forms in the ethnic minority booster sample. Again the response rate distribution between age and sex groups was similar. For this reason, it was felt that no weighting of the ethnic minority booster Adult Self-Completion data was necessary.

The same exercise was undertaken for the Young Person's questionnaire using weighted data. In the main sample, of a possible total of 576 interviews with children aged 12 to 15 inclusive, 403 interviews were achieved. This represents a response rate of 70%. There were small differences of sex profile between respondents and non-respondents. It was decided to weight the Young Person's data to reflect these differences by calculating weighting factors for each sex category. A comparable analysis was not conducted for the ethnic minority booster sample due to the absence of an initial household weighting factor.

Tables B3 and B4 show an overall summary of the weighting factors used for each questionnaire in the main and ethnic minority booster samples respectively.

Table B3 Summary of Weighting Factors Used in Main Sample

Table B4 Summary of Weighting factors Used (Ethnic Minority Booster Sample)

In the data files, the weight for an individual (W _house*W _person) was expressed as a single variable (W _indiv).

Census Comparisons

The weighted and unweighted data were compared with 1991 Census data and other information. The following key household variables are analysed below:

• building type
• Scottish *ACORN group
• household size
• household composition
• tenure

There were some fairly marked differences between the household weighted data and the 1991 census data. Particularly, there is a notable difference between the percentage of one person households, which can largely be attributed to the under-representation of single female households in the main sample. Though an initial decision was made to weight for this under-representation, the effect of this weighting on other variables was unpredictable. It was therefore decided that it would be best not to weight to external sources in this case.

It is not surprising also that the 2000 SCS weighted tenure data is not similar to the 1991 Census data. Between 1991 and 1996 there was a 7.1% increase in the total number of owner occupied dwellings and a similar fall in the number of dwellings rented from Local Authorities.

Fewer households with children were interviewed than would be expected given the 1991 Census data. An investigation did not reveal any reasons why this should be the case.

To examine the effect of individual weighting, the following key variables were analysed and compared with 1991 Census data:

• age
• sex
• household size

An examination of the unweighted respondent data showed a bias towards interviews with females and against respondents aged 16-24 when compared with the age/sex household distribution.

Data relating to the respondent only was therefore weighted further to reflect the age/sex structure of the Scottish population resident in households. Due to its age, it was thought to be inappropriate to weight to the 1991 Census data as in the 1992 and 1996 surveys. The 1999 mid-year population estimate based on the Census data was used instead. The effect of individual weighting in the ethnic minority booster sample was not examined due to the lack of suitable external data.

The final adjustment was for household size bias. As persons in single person households had a greater chance of selection for interview than adults in larger households, weighting of individual level data was applied to adjust for this bias. Comparisons between Census data and the weighted and unweighted data suggest that weighting to adjust for age/sex and household size bias adjusted the survey data to become broadly representative of the Scottish distribution of persons in each household size band.

Calculation of Rates and Standard Errors

Victimisation Rates

In order to make comparisons between police recorded crime statistics and the 2000 SCS data it was necessary to make the following adjustments to the data:

• all incidents which had occurred outwith the calendar year 1999 were excluded;
• incidents which occurred outside Scotland were excluded; and
• single incident Victim Forms were given a value of one, and series incident Victim Forms were given the value of the number of incidents in the series but up to a maximum of five.

For each interview, numbers of incidents were then summed across all Victim Forms to give the number of incidents of each offence type for that individual/household. The Victim Forms for non-victims were given a value of zero.

Victimisation rates were then calculated as a ratio per 10,000 households, individuals, motor vehicle or bicycle owners. The final calculations were based on weighted data. Victimisation rates were not calculated for the ethnic minority booster sample because of the small sample size.

Standard Errors

The calculation of standard errors is a measure of the precision of estimates derived from a sample. Since the 2000 SCS was not a simple random sample (SRS), it was necessary to calculate complex standard errors (CSEs) taking the sample design into account. The methodology used to calculate the CSEs was the generally accepted variance formula for a ratio in sampling error estimation (calculated using the Taylor-Series Expansion Method). Confidence limits were calculated at 95% and 90% intervals for each victimisation rate. The design factor, which represents the ratio of the CSE to the SRS standard error, was also calculated for each rate.

Confidence limits, SRS standard errors, CSEs and design effects for key victimisation variables in the 2000 SCS are included in Appendix C.

« Previous | Contents | Next »