Each year, information from respondents to the Crime Survey for England and Wales is held across two datasets: the Victim Form and the Non-Victim Form datasets. Each row of the Non-Victim Form dataset contains information on an individual respondent, such as measures of their socioeconomic status. Each row of the Victim Form dataset contains detail on an individual instance of crime or a series of instances of the same crime, including, if the incident was violence and what type of violence it was– domestic, stranger, or acquaintance (Table 1 shows which variables relating to this work are contained in each dataset). In order to analyse details of a crime and its victim together, these datasets are merged. This is achieved by appending respondent characteristics (from the Non-Victim Form dataset) to the incident or crime series data (in the Victim Form dataset) via a unique ’Case identifier’ contained in each. Using this method, all records were matched accurately without duplication.

Further, given the relatively rare nature of violent events, a large sample was required in order to assure sufficient cases of violence for analysis. To this end, data were pooled from five years in order to increase the reliability and accuracy of any results. The final sample thus totalled at 174,178, including 1398 incidents (unweighted) of alcohol-related violence. Weighting variables used ensure sample is nationally representative, by (amongst other things) "[compensating] for unequal address selection probabilities" as well as "[adjusting] for differential non-response” [41 p. 97], and to create estimates of how many victims and incidents of each kind there were across the whole population. Further details of the full weighting procedure are available in the CSEW User Guide [41]. Previously, such weighting included a cap of five on the incidents of one kind that could be reported by a respondent, as a method to remove outliers. This led to undercounting of violence–particularly domestic violence–and a method developed by Walby, Towers, and Francis [27] was needed to remove this capping and more accurately count these incidents. However, this undercounting has since been addressed through changes to the weighting variables, using “the 98th percentile of victim incident counts for each crime type (calculated over several years)” to cap repeat victimisation reports, avoiding the undercounting problem potentially encountered by previous work of this kind [41 p. 14].

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.