Jump to ContentJump to Main Navigation
Political Choice in Britain$

Harold D. Clarke, David Sanders, Marianne C. Stewart, and Paul Whiteley

Print publication date: 2004

Print ISBN-13: 9780199244881

Published to Oxford Scholarship Online: November 2004

DOI: 10.1093/019924488X.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 24 February 2017

(p.329) Appendix A. 2001 BES Data Used in this Book

(p.329) Appendix A. 2001 BES Data Used in this Book

Source:
Political Choice in Britain
Publisher:
Oxford University Press

Several major data collections were conducted in the 2001 British Election Study (BES). Data used for the analyses in this book are as follows:

I. MAIN SURVEYS

Appendix A. 2001 BES Data Used in this Book

(p.330) II. ROLLING CROSS-SECTION CAMPAIGN SURVEY AND POST-ELECTION PANEL SURVEY

Appendix A. 2001 BES Data Used in this Book

GUIDE TO THE DATA

The file names listed above match the names of the files submitted to the University of Essex Data Archive. All of these files can also be accessed via the BES website (www.essex.ac.uk/bes). Note that data sets submitted to the UK Data Archive at the University of Essex are SPSS system files (with a.SAV suffix). Data files on the BES website are in SPSS portable (.POR) format. An Excel file (ROSETTASTONE.XLS) cross-references survey questions and variables. Anyone with questions about the data should contact Harold Clarke (email address: hclarke@utdallas.edu) or David Sanders (email address: sanders@ essex.ac.uk) for additional information.

I. MAIN SURVEYS

Wave 1

Wave 1 was a pre-election benchmark survey. The fieldwork was conducted by NOP, during the period 3 March–14 May 2001 (95.7 per cent of the interviews were completed in March and April). Face-to-face interviews were conducted in the respondents' homes, via CAPI (computer-assisted personal interviews). The average length of an interview was 29.8 minutes. The following table summarizes (p.331) the response rate to Wave 1:

N

% issued

% in-scope

Number of addresses issued

6460

100.0

Vacant/non-residential

366

5.7

Number of addresses in-scope

6094

94.3

100.0

Interview obtained

3219

49.8

52.8

Interview not obtained

2875

44.5

47.2

Refusal

1408

21.8

23.1

Non-contact

910

14.1

14.9

Other

557

8.6

9.1

Wave 2

Wave 2 was a face-to-face (CAPI) interview conducted with as many respondents as possible to the pre-election benchmark survey. Fieldwork was conducted by NOP, during the period 8 June–30 July 2001. The average length of an interview was 42.4 minutes, and a self-completion questionnaire was left with respondents. The response rate is summarized in the following table:

N

% issued

% in-scope

Number of addresses issued

3223

100.0

 Moved awaya

40

1.2

Number of addresses in-scope

3183

98.8

100.0

 Interview obtainedb

2354

73.0

74.0

With self-completion

1628

69.2

Without self-completion

726

30.8

Interview not obtained

829

25.7

26.0

Refusal

409

12.7

12.8

Non-contact

213

6.6

6.7

Other

168

5.2

5.3

Moved

39

1.2

1.2

(a) Notes: Of the seventy-nine respondents who moved between Waves 1 and 2, forty were designated ineligible by the fieldwork agency.

(b) During the construction of the panel, 2315 definite panel matches were identified. The N of 2354 reported in the table therefore includes thirty-nine respondents designated as ‘unlikely panel matches’. These respondents are assigned a value of ‘3’ on the pre–post panel filter variable, and so users can decide whether or not to include them in analyses. If these thirty-nine are assumed not to be part of the panel, then the response rate falls to 72.7%.

(p.332) In order to adjust for the impact of panel attrition, and to ensure that the post-election survey yielded a representative sample of the electorate comparable with previous BES post-election cross-sections, the post-election survey included an additional ‘top-up’ sample of respondents, to constituencies that were under-represented in the panel component of Wave 2. NOP conducted the top-up fieldwork over the period 23 July–30 September 2001. The response rate for the top-up sample is summarized below:

N

% issued

% in-scope

Number of addresses issued

1548

100.0

Vacant/non-residential

72

4.7

Number of addresses in-scope

1476

95.3

100.0

Interview obtained

681

44.0

46.1

With self-completion

379

55.7

Without self-completion

302

44.3

Interview not obtained

795

51.4

53.9

Refusal

407

26.3

27.6

Non-contact

229

14.8

15.5

Other

154

9.9

10.4

With its panel and top-up components, the post-election survey was designed to yield a representative sample of the British electorate. Initially, a random sample of 128 constituencies was drawn, stratified by standard region. (Constituencies north of the Caledonian Canal were excluded.) Then, from within each of these constituencies, two wards were randomly selected. This provided the 256 clusters for sampling. Respondents within these wards were sampled from the Postcode Address File. The survey incorporates booster samples for both Scotland and Wales. Users wishing to analyze just the Scottish or the Welsh samples should select from the data set cases on which REGNALL = 9 (Scotland) or REGNALL = 4 (Wales). (Alternatively, separate data sets containing these Celtic sub-samples are available via the BES website at www.essex.ac.uk/bes/data.html.)

Weighting

There is a variety of weighting variables included in the main BES data sets. These variables offer the possibilities of weighting by region, by gender, or by age within gender. The recommended option is to use the overall weighting variables, thereby weighting by region, gender, and age within gender simultaneously. The names of the various weighting variables are listed below.

(p.333) Data Files

PREPOST.SAV—an SPSS data file containing data from the first two waves of the main British Election Study survey. Variables from the pre-election survey begin with the letter ‘a’; those to the post-election survey begin with ‘b’, while those from the self-completion mail questionnaire begin with ‘c’. Note that the structure of the BES is such that several sub-data sets are nested within this overall file. The filter variables to select these subsets, and the weight variables that they require, are set out in the following table:

To use sub-data sets

Pre-election

Pre1Post panel

Post-election cross-section

Mailback

Filter variable

TOPUP = 0

PREPOST = 1

POSTRESP = 1

MAILFILT = 1

Unique respondent ID labela

BUNIQIDR

BUNIQIDR

BUNIQIDR

BUNIQIDR

Weight variables

Region

REGWGTHC

BREGNWGT

REGOCTWT

MREGNWGT

Gender

GENWGTHC

BGENWGT

GENOCTWT

MGENWGT

Age/w/gender

AGEWGTHC

BAGEWGT

AGEOCTWT

MAGEWGT

Region × gender × age/w/gender

AWGTGB

BPANWGT

POSTOCTW

MAILWGT

Unweighted N

3219

2315

3035

2191b

Weighted N

3223

2303

3025

2195

(a) Notes: The variable ‘buniqidr’ links respondents through each stage of data collection: it matches the ‘arespid’ to the pre-survey, it is the principal ID for the post-survey (including the top-up), and it matches the ‘crespid’ to the mailback survey. (The variable ‘brespid’ does not uniquely identify respondents.)

(b) These overall Ns include mailbacks to the ‘unlikely panel match’ respondents. They also include 184 mailbacks to respondents involved in the pre-but not the post-interview. (These are respondents who were unwilling to be re-interviewed, but were persuaded by the interviewer to fill out the self-completion supplement in order to avoid a complete refusal and loss of information.)

PREPOSTAGG.SAV—an SPSS data file, identical to PREPOST.SAV, but also containing constituency-level information for aggregate analysis. The first 157 variables in the data set are taken from the British Parliamentary Constituency database, 1992–2001 (may be obtained via http://ksghome.harvard.edu/~pnorris.shorenstein.ksg/data.htm).

NOPPRE.DOC and NOPPRE.PDF—WORD and ADOBE versions of the questionnaire used for Wave 1 of the main BES.

(p.334) NOPPRECARDS.DOC and NOPPRECARDS.PDF—WORD and ADOBE versions of the show-cards used in conjunction with the pre-election survey.

NOPPRE.LST—an ASCII file containing summary tables of frequencies for variables generated by the pre-election wave (Wave 1) of the 2001 BES main survey.

NOPPOST.DOC and NOPPOST.PDF—WORD and ADOBE versions of the questionnaire used for the face-to-face interview in the post-election wave (Wave 2) of the 2001 BES main survey administered to panel and top-up respondents.

NOPPOSTCARDS.DOC and NOPPOSTCARDS.PDF—WORD and ADOBE versions of the show-cards used in conjunction with the post-election survey.

NOPPOST.LST—an ASCII file containing summary tables of frequencies for variables generated by the post-election wave (Wave 2) of the 2001 BES main survey (contains data for both panel and top-up respondents).

NOPMAILBACK.DOC and NOPMAILBACK.PDF—WORD and ADOBE versions of the self-completion mailback questionnaire left with panel and top-up respondents in the post-election survey.

NOPMAILBACK.LST—an ASCII file containing summary tables of frequencies for variables generated by the self-completion questionnaire left with panel and top-up respondents in the post-election survey.

ROLLING CROSS-SECTION CAMPAIGN SURVEY

Wave 1

Freestanding from the main BES surveys described above, the rolling cross-section campaign survey was conducted by the British Gallup organization during the election campaign. Fieldwork took place between 8 May and 6 June, with 4810 telephone interviews being completed during the campaign. Approximately 160 interviews were conducted each day. The average interview time was approximately 10.5 minutes. Information on response rates is summarized below:

N

%

Numbers called

16620

100.0

Successful

4810

28.9

Unsuccessful

11810

71.1

Screened

1874

11.3

No reply/engaged/answerphone

6744

40.6

Refused

2789

16.8

Other

403

2.4

(p.335) Wave 2

Respondents interviewed in the rolling cross-section campaign survey were re-interviewed by telephone immediately after the election (8–18 June 2001). The average interview time was just over 3 min. Seventy-eight per cent (N = 3751) of those interviewed prior to the election were re-interviewed.

Sampling

Interviewing was conducted by random digit dialling to ensure that ex-directory households were included. Interviewers then asked to speak to the person aged 18 or over in the house who had had the most recent birthday to ensure that the sample was random within households.

Weighting

The variable WEIGHT is included in the data sets. It is calculated on the basis of five factors: gender by social class, gender by age, household car ownership, housing tenure, and population by region.

Files

GALLUP.SAV—an SPSS system file containing data to the election campaign rolling cross-sectional survey (unweighted N = 4810). The weight variable is WEIGHT (weighted N = 4810). The file also contains data gathered from post-election interviews with a panel of the pre-election respondents (unweighted N = 3751). To select only panel respondents for analysis, select those cases on which the filter variable PANELIST = 1. All post-election variables start with the letter ‘Z’. Note that filters and weights are included so that analysts may construct each of the 7-day rolling cross-sectional campaign surveys. The rolling cross-sectional filters are FILT0514–FILT0606 and the respective weights are WGT0514–WGT0606.

GALLUPAGG.SAV—an SPSS data file, identical to GALLUP.SAV, but also containing constituency-level information. The first 157 variables in the data set are taken to the British Parliamentary Constituency database, 1992–2001. These aggregate data were obtained to Professor Pippa Norris' website (http://ksghome.harvard.edu/~pnorris.shorenstein.ksg/data.htm). An additional constituency-level population density variable (to Butler and Kavanagh 1997: Appendix 1, Table A1.3) also is included. Filters and weights to construct the 7-day rolling cross-sectional campaign surveys are omitted.

(p.336) GALLUPPRE. DOC and GALLUPPRE.PDF—WORD and ADOBE versions of the questionnaire used for the election campaign rolling cross-sectional survey.

GALLUPPRE.LST—an ASCII file containing summary tables of frequencies for variables generated by the campaign rolling cross-sectional survey.

GALLUPPOST.DOC and GALLUPPOST.PDF—WORD and ADOBE versions of the post-election survey questionnaire administered to respondents in the campaign rolling cross-sectional survey.

GALLUPPOST.LST—an ASCII file containing summary tables of frequencies for variables generated by the post-election survey of respondents in the campaign rolling cross-sectional survey.