|
The labelling of 'bq5' on the post-election survey datafile and summary tables confused Labour and Conservative responses. Labour should have the value label "1", while the Conservatives should be "2". Hence, around three times as many voters regarded Labour as best able to handle the issue they deemed most important.
Occasionally, the datafiles contain quaint abbreviations that may be confusing for non-British users. On the final versions, a constituency label such as 'Hants NE' will be expanded to 'Hampshire NE'.
Two new variables have been added to some of the data files. Both are additional constituency variables. The first, labelled ddcode, provides the codes used in the 2001 census (where '1'=Barking, '2'=Battersea etc). The second, labelled pacode, provides the Press Association identifiers (simply an alphabetical list of constituencies).
An error in File 3: due to errors in the interviewers' initial codings of constituency and ward, some of the respondents in wards 31, 32, 71 and 72 were wrongly listed as being in constituencies 31, 32, 71 and 72 (instead of 3 and 7). Please correct this in your CD-rom copies, or simply download the corrected file from this site.
In the campaign surveys, questions 30a-33a were ultimately omitted from the telephone interviews owing to time constraints. Hence they do not appear in Files 7 and 8. Please note that a battery of questions on party contacting is included in the main post-election survey (File 3).
During the vote validation exercise, it was discovered that 14 respondents resided in constituencies other than those designated in the constituency of residence variable ('aconstit'). Typically, these respondents resided very close to the border of the constituency in which they were originally coded. The corrected constituency data for these 14 respondents appears in the variable 'constrev'. Similarly, corrected census and press association constituency codes for these respondents appears in the variables 'ddrev' and 'parev' respectively.
There was a miscoding of variables bq36 and bq37 in the original dataset (due to different value labels being used in the top-up). The corrected file can be downloaded from this site, and the summary tables (File 22) have also been corrected.
|
