On Monday 14 October 2019, the RSS Social Statistics Section held this meeting at the Royal Statistical Society to discuss the effect of uncapping the number of repeat/multiple crime victimisation incidents reported to the Crime Survey for England and Wales (CSEW) on estimates of national and gender- and age- specific crime rates. It was organised and chaired by Professor Andromachi Tseloni (Nottingham Trent University) who, along with invited speakers Joe Traynor (Centre for Crime and Justice, ONS), Sylvia Walby (City University of London) and Brian Francis (Lancaster University), presented the relevant research evidence.
Participants included the proponent of the importance of measuring and recognising repeat victimisation for effective crime prevention, Ken Pease (University of Derby), and a good mixture of social survey methodology professionals, crime prevention policy practitioners, academics and PhD candidates.
Andromachi opened up the meeting with a short history of previous research and measurement in social surveys of:
- multiple victimisation (respondent reporting at least two crime types in any given period, usually a year, for example, an assault and a burglary),
- repeat victimisation (respondent reporting more than one incident of the same crime type in any given period, for example, four unrelated assaults) and
- series victimisation (respondent reporting two or more incidents of the same crime type that occurred under similar circumstances and possibly by the same offender(s) in any given period, for example, two assaults in or outside the local pub perpetrated by members of the same gang).
Joe presented recent methodological developments by the ONS to include all incidents reported by the same victim up to the 98th percentile (rather than up to five, as was the case to date) per crime type in estimates of national crime rates. The revised ONS crime counting methodology did not affect the crime drop in England and Wales since 1995 but introduced volatility to violence trends. Multiple/repeat victimisation fell faster than single victimisation. However, 'the most victimised (>5 incidents) [still] account for a high proportion of the crime burden'.
Sylvia and Brian argued that capping is a form of one-sided winsorisation, which is known to produce bias. They demonstrated that removing the five incident and 98th percentile respective previous and current ONS caps in violent crime estimates results in considerable rise of domestic violence against women estimates. They purported that a three-year rolling average could be used for addressing the volatility of the CSEW-based crime rates.
Following a short break for tea, coffee and discussion between participants, Andromachi showed that most uncapped non-domestic violence trends follow changes in concentration (mean number of crimes per victim) while capped ones follow victimisation risk (victims per 100 adults) trends. Removing the five incident per series cap resulted in statistically significant different trends in stranger violence and stranger violence against men. Men also experienced more (uncapped) violence by strangers after the crime drop and relative to women.
In the discussion among participants that followed the early contribution of late American criminologists Richard F. Sparks on multiple/ repeat victimisation research in the late 1970’s and 1980 and by contemporary scholars, Graham Farrell (University of Leeds) and Dainis Ignatans (University of Huddersfield), was noted.