Reconciliation Barometer 2003-2011

Version 1 of the data was received by DataFirst from the Institute for Justice and Reconciliation (IJR) in 2012.

DataFirst has identified the following data quality issues concerning this dataset:

Questionnaires
A questionnaire was supplied for each round of the survey, however:
The round 1 questionnaire does not match the variable labels in the dataset -- this probably means that this is an incorrect questionnaire or an incorrect earlier version of the questionnaire
The round 2 questionnaire is missing
The round 9 questionnaire was not available during the data cleaning process, potentially resulting in mismatches between the questions in the codebook and in the questionnaire -- in the event of mismatches the questionnaire takes precedence
Many questions were repeated in every round of the survey, although there were changes with every round. Researchers should be careful to be aware of changes in the wording of questions (or in the options available for selection).

Technical Reports
A technical report for each round of the survey was supplied by IPSOS/Markinor (sometimes in the form of powerpoint presentations) and are incomplete with respect to sampling processes etc. In the event of queries, users should be encouraged to submit them to datafirst, and it can be passed on to the IJR or IPSOS/Markinor for clarification if required.

The archived IJR version of the dataset has been substantially cleaned by DataFirst, mainly in the form of harmonising coding and adding metadata to the datasets.

The SARB datasets were cleaned and harmonised across rounds in the following ways:

1. Unique identifier variables ("id") were transformed to be unique between and within rounds by turning each id into a 7-digit number (with a leading zero) or 8-digit number of which the first two digits represent the round (e.g. the first respondent in the round 1 dataset is identified by 01000001 and the first respondent in the round 12 dataset by 12000001).

2. Demographic and general (i.e. non-SARB question) variables were renamed to a standard list of names (including "id", "weight", "gender", "race", etc...), given consistent variable labels and placed at the beginning of the dataset in a standard order. NOTE: While coding of these variables were often harmonised, we cannot guarantee complete compatibility: for many variables categories (e.g. dwelling type) categories changed several times. The rounds should therefore NOT be merged/appended together but rather analysed separately.

3. Substantive questions were renamed with a round prefix in the form "r01_" (for the first round), usually followed by the letter/number combination corresponding with the designation of questions on the questionnaire.

4. Substantive questions were recoded to match a standard coding scheme consistent across questions in a round and, as far as possible, consistent across rounds. Missing values were recoded to -9 for "Don't know", -5 for "Refused", -3 for "Not applicable" and -1 for "Missing" (the latter representing cases where there is no response recorded but the data producer specified no reason in the data for there being no response). Scales were transformed to run from "negative" to "positive" (e.g. "disagree" to "agree", "unjustified" to "justified", "never" to "often", etc.). Standard label schemes were applied to ensure value labels are consistently spelled.

5. All transformations of the data from the supplied form to the first release version can be traced by viewing the cleaning do-files released with the data. Researchers requiring the original files to re-run the synax files used in the data cleaning process should contact support@data1st.org

6. A codebook was produced for each round of the survey EXCEPT rounds 1 and 2 (owing to the incorrect and missing questionnaires). The codebooks are automatically generated from the post-cleaning data files, which have the full questions from the questionnaires inserted into the data as "notes". The codebooks show the variables names (altered during the cleaning process, but easy to match with variables in the original data files and with questions in the questionnaires), variable labels (usually truncated versions of the full question but sometimes more descriptive) and coding/value labels (which have been harmonised during data cleaning). For rounds 1 and 2 the best source of the full questions are the variable labels in the original, unaltered SPSS-version data files. Researchers requiring these files for their data analysis should contact support@data1st.org