- Comment on Nazaretian and Merolla
In the April 2013 issue of the Canadian Journal of Criminology and Criminal Justice, Zavin Nazaretian and David Merolla question the current approach to capping criminal incidents within the Canadian General Social Survey: Victimization (GSS-V). Statistics Canada would like to clarify this approach and correct some interpretation errors in their article, “Questioning Canadian Criminal Incidence Rates: A Re-analysis of the 2004 Canadian Victimization Survey.”
The authors correctly note that measures of victimization based on capping have been adopted by leading statistical agencies and are in keeping with international practice. The authors specifically examine the Canadian example and criticize the capping procedure adopted by the GSS-V. In particular, Nazaretian and Merolla suggest that the victimization rates published by Statistics Canada are capped at three incidents per victim, and that this capping is carried out to limit inflation due to telescoping. This assumption, however, is not accurate.
The GSS-V uses two capping approaches, one to address victim sensitivity issues and the other to limit the impact of extreme cases (i.e., ‘outliers’). The first capping takes place during the actual collection of data when respondents are being interviewed. Each respondent is limited to completing a maximum of 5 incident reports per offence type (up to an overall total of 20 incident reports per respondent). This cap was introduced primarily out of respect for the victims, since each incident report requires a great deal of their time and can be emotionally difficult, in some cases. While Statistics Canada strives for completeness of data, we also have to be mindful of both respondent burden and victims’ needs.
The second cap is implemented at the incident report level. Each incident report can represent up to 10 incidents with similar circumstances (generally in cases of repeat victimization or when a victim has difficulty recalling the details of the incidents). It is this number of incidents per [End Page 137] report that is capped at 3 for official victimization rates. Thus, given that a victim can complete up to 20 incident reports and each of those reports can represent up to 3 incidents after capping, the total number of incidents that may be counted for a single victim is 60, not 3. In addition, incidents of spousal violence are counted separately.
The main reason for this capping procedure is not to limit the effects of telescoping but to limit the impact of “outliers.” As the authors point out, uncapped rates could be considered at the national level, especially since, as they demonstrate, rates for some of the most serious offences, including sexual assault, are most likely to be affected by this form of capping. However, while this option would be possible for national rates, it would not be feasible for smaller geographical units or specific population groups, as the impact of “outliers” on analyses would be too great.
In closing, despite technical errors in interpreting the capping performed in the GSS-V, the authors correctly note that there is more than one way of measuring victimization, with quite different results, and that more than one measure should be considered. Furthermore, their compelling arguments about the effects of various capping practices contribute to the discussion on the ways to measure and present rates of victimization.
As always, Statistics Canada will continue to ensure that victimization data are presented as accurately as possible and will maintain its commitment to robust analysis by considering all possible approaches to analysis, including complementing the current measure of victimization with other indicators. [End Page 138]